Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Towards a plant-based method of guiding CO₂ enrichment in greenhouse tomato Edwards, Diane Roselyn 2008

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Notice for Google Chrome users:
If you are having trouble viewing or searching the PDF with Google Chrome, please download it here instead.

Item Metadata

Download

Media
24-ubc_2009_spring_edwards_diane.pdf [ 1.79MB ]
Metadata
JSON: 24-1.0066866.json
JSON-LD: 24-1.0066866-ld.json
RDF/XML (Pretty): 24-1.0066866-rdf.xml
RDF/JSON: 24-1.0066866-rdf.json
Turtle: 24-1.0066866-turtle.txt
N-Triples: 24-1.0066866-rdf-ntriples.txt
Original Record: 24-1.0066866-source.json
Full Text
24-1.0066866-fulltext.txt
Citation
24-1.0066866.ris

Full Text

     TOWARDS A PLANT-BASED METHOD OF GUIDING CO2 ENRICHMENT IN GREENHOUSE TOMATO   by  Diane Roselyn Edwards  M.Sc., The University of Guelph, 1993 B.Sc.(Agr), The University of Guelph, 1991    A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF  DOCTOR OF PHILOSOPHY   in   The Faculty of Graduate Studies  (Plant Science)     THE UNIVERSITY OF BRITISH COLUMBIA  (Vancouver)     December 2008   © Diane Roselyn Edwards  ii Abstract Atmospheric CO2 enrichment is employed by greenhouse tomato growers to increase fruit yields, and CO2 applications are managed according to atmospheric set points or CO2 injection rates.  These methods do not immediately focus on the targets of CO2 applications: plant performance and the regulation of plant carbon status.  This thesis explores several plant-based approaches that may have potential for use in the management of CO2 in greenhouse tomato production. Three plant-based approaches to CO2 management were explored in commercial and experimental tomato greenhouses.  These were: (1) simulation modeling, (2) non-destructive analysis of growth and (3) the status of plant carbon reserves.  A cost and benefit analysis (c/b) using simulation modeling was carried out using grower-collected greenhouse environment and yield data.  Simulation modeling was useful for retrospectively determining c/b of several CO2 scenarios.  The model was effective in predicting long term yields, but not short term yield variations, which limits its application for CO2 management.  Non- destructive measures of growth:  stem length and diameter, leaf area and fruit load were found to be too sluggish for daily CO2 dosing decision-making.  Finally, plants growing under CO2 enrichment can deposit substantial carbon as starch in their leaves.  Plant carbon status was evaluated by determining the spatial distribution of leaf starch in the shoot and by following its variation diurnally and after the onset of CO2 enrichment.  As starch is difficult to measure by a grower, leaf mass per unit area (LMA) was also monitored for assessment as a surrogate measure for starch.  Leaves in positions 7 to 9 were identified as the most meaningful in the shoot to sample.  Diurnal profiles indicated these leaves carryover substantial starch from one day to the next.  Monitoring starch at its peak time of  iii accumulation (14 h to 16 h), at sunset and sunrise will indicate how much the peak starch reserves are used overnight.  If starch remains high between peak and sunrise the following day, then the plants are in a carbon-surplus state and CO2 enrichment could be postponed. For upper canopy leaves LMA is substantially influenced by starch and thus is a promising surrogate.  iv Table of Contents Abstract.................................................................................................................................... ii List of Tables .......................................................................................................................... ix List of Figures.......................................................................................................................... x Symbols and Abbreviations ................................................................................................ xvi Acknowledgements ............................................................................................................... xx Co-authorship statement ..................................................................................................... xxi 1 Towards a plant-based method of guiding CO2 enrichment in greenhouse tomato........................................................................................................................... 1 1.1 INTRODUCTION .................................................................................................... 1 1.2 RESEARCH OBJECTIVES ..................................................................................... 7 1.3 LITERATURE CITED ............................................................................................. 9 2 An application of simulation modeling to explore CO2 dosing scenarios in greenhouse tomato production. ............................................................................... 11 2.1 INTRODUCTION .................................................................................................. 11 2.1.1 CO2 enrichment in greenhouses...................................................................... 11 2.1.2 Modeling greenhouse systems ........................................................................ 14 2.1.3 Models for CO2 management.......................................................................... 15 2.1.4 Objectives of chapter ...................................................................................... 16 2.2 MATERIALS AND METHODS ............................................................................ 17 2.2.1 Plant and greenhouse environment data. ........................................................ 17 2.2.2 Plant based models.......................................................................................... 21 2.2.3 Relating CO2 generated from greenhouse heating to greenhouse CO2 levels 24 2.2.4 Costs of CO2 ................................................................................................... 25 2.2.5 Price structure for fruit produced.................................................................... 25 2.2.6 CO2 enrichment scenarios............................................................................... 26 2.3 RESULTS ............................................................................................................... 26 2.3.1 Photosynthesis and yield models .................................................................... 26 2.3.2 CO2 generation and greenhouse CO2 usage.................................................... 31 2.3.3 Seasonal price of tomatoes.............................................................................. 35 2.3.4 Cost and revenue of CO2 enrichment scenarios.............................................. 39 2.4 DISCUSSION......................................................................................................... 43 2.4.1 Modeling greenhouse tomato fruit yield......................................................... 44 2.4.2 CO2 generation and usage in greenhouse........................................................ 47 2.4.3 CO2 enrichment scenarios............................................................................... 49 2.5 CONCLUSIONS..................................................................................................... 50 2.6 LITERATURE CITED ........................................................................................... 53  v 3 Assessing the potential for using non-destructive plant growth measurements in guiding CO2 dosing ................................................................................................... 57 3.1 INTRODUCTION .................................................................................................. 57 3.1.1 Effects of CO2 enrichment on plant growth.................................................... 57 3.1.2 Analysis of tomato plant growth using non-destructive measurements ......... 59 3.1.3 The management of greenhouse tomato growth............................................. 61 3.1.4 Objectives of chapter ...................................................................................... 62 3.2 MATERIALS AND METHODS............................................................................ 62 3.2.1 Greenhouse layout .......................................................................................... 62 3.2.2 Plant culture .................................................................................................... 64 3.2.3 Management and application of CO2.............................................................. 65 3.2.4 Experimental design........................................................................................ 66 3.2.5 Plant observations, growth and yield measurements ...................................... 66 3.2.6 Calculation of fruit load and leaf area............................................................. 69 3.2.7 Data analysis ................................................................................................... 70 3.3 RESULTS ............................................................................................................... 72 3.3.1 Greenhouse environment ................................................................................ 72 3.3.2 Stem length and diameter................................................................................ 76 3.3.3 The profile of leaf area expansion .................................................................. 78 3.3.4 Canopy fruit load ............................................................................................ 82 3.3.5 Fruit harvest, mass and number. ..................................................................... 84 3.4 DISCUSSION......................................................................................................... 84 3.4.1 The non-destructive measurement of plant growth under CO2 enrichment ... 88 3.4.2 Effects of CO2 enrichment on fresh fruit yield ............................................... 90 3.4.3 Comments on the experiments and non-destructive measurements ............... 92 3.5 CONCLUSIONS..................................................................................................... 94 3.6 LITERATURE CITED ........................................................................................... 96 4 Canopy profiles of starch and leaf mass per area in greenhouse tomato and the relationship with leaf area and fruit load. ............................................................ 100 4.1 INTRODUCTION ................................................................................................ 100 4.1.1 Leaf starch..................................................................................................... 101 4.1.2 Leaf mass per unit leaf area .......................................................................... 102 4.1.3 Objectives of chapter .................................................................................... 104 4.2 MATERIALS AND METHODS.......................................................................... 104 4.2.1 Environment and crop culture in commercial greenhouses.......................... 104 4.2.2 Collection of leaf length and fruit diameter data from plant canopies ......... 107 4.2.3 Calculation of leaf area and fruit load........................................................... 108 4.2.4 Leaf collection, tissue analysis, and calculation of LMA and leaf starch..... 108 4.2.5 Data analysis ................................................................................................. 110 4.3 RESULTS ............................................................................................................. 113 4.3.1 The influence of starch on leaf mass per area............................................... 113 4.3.2 Canopy profiles of LMA and starch ............................................................. 115 4.3.3 Profile of leaf area in the canopy .................................................................. 124 4.3.4 Canopy profiles of starch content per leaf .................................................... 126  vi 4.3.5 Relationships between fruit load and starch per leaf in the canopy.............. 130 4.4 DISCUSSION....................................................................................................... 133 4.4.1 Leaf starch and LMA values......................................................................... 133 4.4.2 Profile of leaf starch in the shoot canopy...................................................... 135 4.4.3 The profile of LMA in the shoot canopy. ..................................................... 138 4.4.4 The influence of leaf starch on LMA............................................................ 141 4.4.5 Source and sink relations in the shoot canopy. ............................................. 143 4.4.7 Assessment of LMA and starch as a management tool for CO2 dosing. ...... 148 4.5 CONCLUSIONS................................................................................................... 149 4.6 LITERATURE CITED ......................................................................................... 151 5 Temporal variations of leaf starch and mass ....................................................... 155 5.1 INTRODUCTION ................................................................................................ 155 5.1.1 The diurnal variability of leaf starch and LMA............................................ 155 5.1.2 The dynamics of leaf starch and LMA at the onset of CO2 enrichment. ...... 156 5.1.3 Objectives of chapter .................................................................................... 157 5.2 MATERIALS AND METHODS.......................................................................... 157 5.2.1 Greenhouse venues ....................................................................................... 158 5.2.2 Data collection at the onset of CO2 enrichment............................................ 159 5.2.3 Data collection for diurnal profiles ............................................................... 160 5.2.4 Data analysis ................................................................................................. 161 5.3 RESULTS ............................................................................................................. 162 5.3.1 The diurnal variability of leaf starch and LMA............................................ 162 5.3.2 Leaf starch and LMA after the onset of CO2 enrichment ............................. 170 5.4 DISCUSSION....................................................................................................... 179 5.4.1 Diurnal profiles of leaf starch ....................................................................... 180 5.4.2 Diurnal profiles of LMA............................................................................... 183 5.4.3 Leaf starch and LMA at the onset of CO2 enrichment.................................. 185 5.5 CONCLUSIONS................................................................................................... 187 5.6 LITERATURE CITED ......................................................................................... 189 6 General discussion .................................................................................................. 192 6.1 RESEARCH VENUES......................................................................................... 192 6.2.1 Plant-based modeling.................................................................................... 195 6.2.2 Non-destructive growth measures................................................................. 197 6.2.3 Starch and LMA............................................................................................ 199 6.3 ADDITIONAL IMPLICATIONS OF THIS RESEARCH................................... 201 6.4 CONCLUSIONS ON PLANT-BASED APPROACHES .................................... 202 6.5 LITERATURE CITED ......................................................................................... 205  vii Appendix 1.1 Leaf length verses leaf area for tomato cv Rapsodie.  Panel A is the raw response and Panel B is the mean response.  The equation of the line is  Y = e(-0.53+2.45(ln(X)), R2=0.996.  The equation was weighted with the inverse of the variance and fitted to the mean response.  Those data with a variance of 0 were not plotted in Panel B.. ............................................................... 207 Appendix 2.1 The CO2 levels (ppm) used for low, intermediate and high CO2 enrichment scenarios at several light regimes used for modelling in Chapter 2. ..........................................................................................................  ................................................................................................................... 208 Appendix 3.1 Leaf injury and CO2 enrichment........................................................... 209 Appendix 4.1 A summary of the environment from experiments conducted in commercial greenhouses in 2001 and 2002............................................... 214 Appendix 4.2 Dates of data collection in commercial greenhouses 2000 to 2002. .... 216 Appendix 4.3 The starch content and corresponding leaf mass per area (LMA) of tomato leaf tissue from South Alder (SA) and Gipaanda (GI) commercial greenhouses.  Panels A-C data collected in May 2002, D-F data collected in September 2002.  Data from all leaf positions are in panels A and D, panel B has data from leaf positions <=11, C has data from leaf positions >=12, E has data from leaf positions <=8 and F has data from leaf positions >=9.  Each sample was 0.000785 m2 of leaf tissue. ................... 217 Appendix 4.4 Leaf mass per area of leaves in tomato plant canopies from South Alder (SA) and Gipaanda (GI) commercial greenhouses.  Leaf position 1 is the leaf closest to the first flowering truss, - leaves are younger and + leaves are older.  Panel A and B, data collected in June 2001, Panel C and D, data collected June 2002, line equations are in Appendix 4.5................. 220 Appendix 4.5 Linear regression equations for leaf mass per area or leaf starch and canopy leaf position for Appendix 4.4....................................................... 222 Appendix 4.6 Leaf area profile of tomato plant canopies from South Alder (SA), CanAgro (CA) and Gipaanda (GI) commercial greenhouses.  Leaf position 1 is the leaf closest to the first flowering truss, - leaves are younger and + leaves are older.  Panel A, data collected in March 2001, equations for lines are:  SA:  Y = 0.032e[(0.43(1-e(-0.56X))/0.56],  CA: Y = 0.033e[(0.36(1-e(-0.34X))/0.34], Panel B, data collected in September 2001, equations for lines are:  SA:  Y = 0.0094e[(0.93(1-e(-0.70x))/0.70], CA:  Y = 0.021e[(0.43(1-e(-0.73x))/0.73].  Panel C, data collected in May 2002, equations for lines SA: Y = 0.026e[(0.32(1-e(-0.34x))/0.34]  and GI:  Y = 0.019e[(0.42(1-e(-0.33x))/0.33], Panel D, data collected September 2002, equations for lines are SA:.Y = 0.0045e[(2.0(1-e(-0.9x))/0.9] and GI: Y = 0.007e[(1.82(1-e(-0.85x))/0.85]. All regressions are weighted by the inverse of the variance.............................................. 223  viii Appendix 4.7 Starch content per leaf for tomato plant canopies from South Alder (SA) and CanAgro (CA) commercial greenhouses..  Leaf position 1 is the leaf closest to the first flowering truss, - leaves are younger and + leaves are older.  Panel A, data collected in March 2001, line equations are:  SA. Y = 0.34 + 0.11X - 0.19X2 + 0.0071X3, R2=0.80, CA. Y = 0.53 + 0.093X - 0.013X2 + 0.0004X3, R2=0.80.  Panel B, Data collect in June 2001, line equation SA. Y = 0.19 + 0.022X - 0.0058X2 + 0.00025X3, R2=076.  Panel C, data collected in September 2001, line equations are:  SA. Y = 0.084 + 0.014X - 0.0029X2 - 0.00012X3, R2=0.38, CA. Y = 0.14 - 0.019X -0.00075X2, R2=0.51. Note different scales for Y-axis. ................................................. 225 Appendix 5.1 The effect of CO2 enrichment on photosynthetic rates for tomato cv Rapsodie canopies at PARC-Agassiz in June 2002 and 2003. ................ 226 Appendix 5.2 The weekly profile of mean fruit load for tomato plants grown at PARC- Agassiz under CO2 enrichment and ambient CO2 from June 11 to August 30, 2002.  Plants had 7 trusses with 3 fruit.  Each symbol is the mean of 2.  227 Appendix 5.3 Canopy profiles of starch content and leaf mass per unit area (LMA) of leaves from tomato plant canopies treated by enriched or ambient CO2 at PARC-Agassiz.  Data was collected July 4, 2003, 58 days after CO2 enrichment commenced.  Leaf position 1 is the leaf closest to the first flowering truss, - leaves are younger and + leaves are older.  Panel A. leaf starch, line equations are Ambient: Y = 5.57 - 1.67X + 0.23X2 - 0.013X3 + 0.00024X4, R2=0.65 and Enriched Y = 6.69 - 1.21X + 0.083X2 - 0.0019X3, R2=0.62.  Panel B. LMA, line equations are Ambient:  Y = 40.22 - 0.48X, R2=0.40 and Enriched:  Y = 46.25 - 0.44X, R2=0.51. All regressions are weighted by the inverse of the variance .................................................... 228 Appendix 6.1 Quantification of leaf starch using enzymatic and near-infrared spectroscopic methods. ............................................................................... 229    ix List of Tables Table 2.1 Model diagnostics for comparison of grower measured and modelled tomato (cv Rapsodie) fruit yield for the 2002 and 2003 growing seasons. ................ 32 Table 2.2 The revenue from several CO2 enrichment scenarios for tomato production in British Columbia............................................................................................. 41 Table 3.1 A summary of the daily mean greenhouse environment for the experiments at PARC-Agassiz. ............................................................................................... 75 Table 3.2 A summary of the fresh tomato fruit harvest for the experiments at PARC – Agassiz............................................................................................................ 85 Table 4.1 Model II regression equations for the influence of leaf starch (X) on leaf mass area (Y) for upper (A) and lower (B) regions of tomato plant canopies.  Data were collected from commercial tomato greenhouses during 2001 and 2002. ....................................................................................................................... 118 Table 4.2 Linear regression equations for leaf mass or leaf starch per unit leaf and canopy leaf position for Figure 4.5. .............................................................. 123 Table 4.3 Canopy and individual leaf areas from tomato shoot canopies measured from commercial greenhouses for several months in 2001 (A) and 2002 (B). ..... 125 Table 4.4 The equations for the vertical profile of starch per leaf in the tomato canopies from two commercial greenhouses from Figure 4.6..................................... 128 Table 4.5 The mean and total leaf starch per shoot for tomato shoot canopies, and starch accumulation before measurable fruit load and at full leaf expansion from commercial tomato greenhouses................................................................... 129 Table 4.6 The size and distribution of fruit load in tomato shoot canopies for several months in 2002 from two commercial greenhouses. .................................... 132 Table 5.1 Leaf starch and leaf mass per unit area from leaves collected prior to and during the first seven days of exposure to CO2 enrichment at PARC-Agassiz, June 10 to 17, 2002. ...................................................................................... 173  x List of Figures Figure 2.1 A conceptual diagram of the components of the CO2 enrichment system used for the model development.  LAI is leaf area index and PAR is photosynthetically active radiation. ................................................................ 18 Figure 2.2 The response of the Acock et al. model for canopy photosynthetic rate to light (Panel A) and CO2 concentration (Panel B).  See the text for a full description of the model used. ........................................................................................... 27 Figure 2.3 The diurnal net canopy photosynthesis at several CO2 concentrations using the Acock et al. model, June 15, 2003.  Panel A, hourly canopy net photosynthesis and global radiation.  Panel B, the cumulative daily amount of CO2 assimilated for several CO2 concentrations and the percentage increase above the ambient level of 370 ppm for each level of CO2. ........................... 29 Figure 2.4 Measured and the model predicted fruit yield from a commercial beefsteak tomato greenhouse.  Panel A, 2002 and Panel B, 2003. The predicted yield was calculated from the greenhouse plant and environment data using the Acock et al.-Aikman models for photosynthesis and fruit growth................. 30 Figure 2.5 The contributions of sources of CO2 generation to the total amount of CO2 produced for each month in the production cycle of tomatoes.  Plant uptake (Plt Upt) through photosynthesis is indicated as negative.  NG is CO2 from the combustion of natural gas and Liquid is liquid CO2.  Panel A is the 2002 season and Panel B is 2003 season.  The gap for Plt Upt in February 2002 was caused by missing greenhouse environment data. .......................................... 33 Figure 2.6 The ratio of CO2 used by the plant through photosynthesis versus total CO2 generated by the greenhouse for the 2002 and 2003 growing seasons. The gap for February 2002 was caused by missing greenhouse environment data...... 34 Figure 2.7 The relationship between CO2 generated and the mean CO2 concentration in the greenhouse for the months of March to September.  The regression equation is Y = 76.29X - 25.88, R2=0.82.  NG+LG is CO2 from natural gas and liquid CO2 and NG is CO2 from natural gas.  Only data from months where all generated CO2 was directed into the greenhouse are shown........... 36 Figure 2.8 The monthly total CO2 generated from natural gas used for heating, March to September, 2002 and 2003.  The line equation is Y = -35.10 + 38.47X - 10.83X2 + 1.22X3 - 0.48X4, R2=0.95............................................................... 37 Figure 2.9 The weekly price per kg of beefsteak tomato fruit at the Vancouver food terminal from 1998 to 2004.  The equation of the line is Y = 5.31 - 0.19X + 0.00271X2 and is the weighted mean response with no adjustment for years. Weights were based on the number of observations per year.  Equation R2=0.50 and N=331. ....................................................................................... 38 Figure 2.10 The mean monthly CO2 concentration for a commercial greenhouse and the mean monthly high, low and Intermediate scenarios of CO2 enrichment.  The specific CO2 conditions for each scenario can be found in Appendix 2.1...... 40  xi Figure 2.11 The seasonal net revenue of three CO2 enrichment scenarios for an increasing cost of CO2 in a commercial greenhouse.  The 2002/2003 costs of CO2 from natural gas and liquid CO2 are indicated.  The specific CO2 condition for each treatment can be found in Figure 2.10. ........................................................... 42 Figure 3.1 The layout of the PARC –Agassiz experiment.  Shown (not to scale) is the 2002 arrangement for the CO2 treatments.  E1: atmosphere enriched with CO2, replicate 1, A1: ambient CO2, replicate 1, E2: enriched CO2, rep 2 and A2: ambient, rep 2.  The large rectangles represent 4 plants and the smaller ones 2 plants. The solid lines between compartments were glass walls and the dotted lines represent a plastic partition.  The arrows indicate the direction of training for the crop.  Crops in all compartments were trained the same way.  The direction of north is indicated in the top left corner........................................ 63 Figure 3.2 Schematic of a tomato plant (not to scale) with leaves, trusses with flowers and fruit indicated.  Leaf 1 is the leaf closest to the truss with open flowers, leaves above are identified as 0, -1, -2 etc. and leaves below are 2, 3, 4, 5, 6 etc.................................................................................................................... 68 Figure 3.3 The environment for the greenhouses receiving either ambient or enriched CO2 for the 2002 experiment at PARC-Agassiz, June 1 to August 30.  Panel A. mean daytime CO2, B. daily sum of global radiation, C. mean 24 hour temperature and D. mean 24 hour relative humidity. N=2. ............................ 73 Figure 3.4 The environment for the greenhouses receiving either ambient or enriched CO2 for the 2003 experiment at PARC-Agassiz, May 1 to June 28.  Panel A mean daytime CO2, B daily sum of global radiation, C. mean 24 hour temperature and D. mean 24 hour relative humidity. Gap in the data at 45 and 46 days was caused by a power failure.  N=2................................................. 74 Figure 3.5 Mean stem length for plants grown under CO2 enrichment and ambient CO2. Panel A is the 2002 experiment, June 11 to August 30 and Panel B is the 2003 experiment, May 6 to June 28.  Each symbol is the mean response of plants from each greenhouse compartment, N=2.  There were no statistical differences between treatment means at any sampling date. .......................... 77 Figure 3.6 Mean stem diameter for plants grown under CO2 enrichment and ambient CO2 at PARC-Agassiz.  Panel A is the 2002 experiment, June 11 to August 30 and Panel B is the 2003 experiment, May 6 to June 28. Each symbol is the mean response of plants from each greenhouse compartment, N=2.  There were no statistical differences between treatment means at each sampling date. ........ 79 Figure 3.7 Mean area per leaf for plants grown under ambient CO2 (Panel A) and CO2 enrichment (Panel B) for the 2002 experiment at PARC-Agassiz.  Data were collected weekly from May 7 to August 30.  The horizontal line is 50% of the final leaf area of the first leaf measured.  Each symbol is the mean of 14 leaves............................................................................................................... 80 Figure 3.8 Mean area per leaf for plants grown under ambient CO2 (Panel A) and CO2 enrichment (Panel B) for the 2003 experiment at PARC-Agassiz.  Data were collected weekly from May 6 to June 28.  The horizontal line is 50% of the  xii final leaf area of the first leaf measured.  Each symbol is the mean of 20 leaves............................................................................................................... 81 Figure 3.9 Fruit load increase for plants grown under CO2 enrichment and ambient CO2. Panel A is the 2002 experiment, June 11 to August 30 and Panel B is the 2003 experiment, May 6 to June 28.  In each experiment plants initially had 7 trusses with 3 fruit but only trusses formed after the onset of CO2 enrichment were measured.  Therefore, every 7 days a newly formed truss was added and measured until the fruit were harvested or the experiment ended. * indicates treatment means were statistically different at p< 0.001. ............................... 83 Figure 3.10 Cumulative harvested fresh fruit mass for plants grown under CO2 enrichment and ambient CO2 from the 2002 experiment, June 11 to August 30.  Panel A is the cumulative fresh fruit mass harvested. Treatment means from 24 days were significantly different at p<0.05, means at 36 days and beyond were significantly different at p<0.0001. Panel B is the cumulative fruit number harvested.  Treatment means from 38 days were significantly different at p<0.05, means at 45 days and beyond were significantly different at p<0.0001. ......................................................................................................................... 86 Figure 3.11 Cumulative harvested fresh fruit mass for plants grown under CO2 enrichment and ambient CO2 from the 2003 experiment, May 6 to June 28.  Panel A is the cumulative fresh fruit mass harvested and Panel B is the cumulative fruit number harvested. Treatment means were not significantly different at the p<0.05 level for any date. ............................................................................... 87 Figure 4.1 Starch as a percentage of leaf dry mass for tomato leaf tissue collected at three positions in the canopy from two commercial greenhouses during the 2000 growing season.  SA is South Alder and CA is CanAgro.  Leaves were collected from position 1 (next to the truss with flowers), position 7 (mid- canopy) and from leaf position 13 (lower canopy).  Bars above each column indicate standard errors of the means, N=12. ............................................... 114 Figure 4.2 The starch content per unit leaf area and corresponding leaf mass per unit leaf area from samples of 0.000785 m2 of tomato leaflet tissue collected from South Alder (SA) and CanAgro (CA) greenhouses.  Panel A: data collected in March 2001.  Panel B: data collected in September 2001.  Panel C is redrawn from B with data from leaf position <=7 for SA and <=6 for CA.  Panel D is redrawn from B with data from leaf position >=6 for SA and data for leaf position >=7 for CA...................................................................................... 116 Figure 4.3 The starch content and corresponding leaf mass area from samples of 0.000785 m2 of tomato leaflet tissue from South Alder (SA) and Gipaanda (GI) greenhouses.  Data were collected in June 2002. Panel A: upper canopy and Panel B, lower canopy.  Line equations for the Model II regression are in Table 4.1. ...................................................................................................... 117 Figure 4.4 The seasonal range of leaf starch (A) or leaf mass per unit leaf area (LMA) (B) collected from tomato cv Rapsodie canopies.  Plants were from three commercial greenhouses during the 2000 to 2001 growing seasons............ 119  xiii Figure 4.5 Leaf mass per unit leaf area (LMA) and starch per unit leaf area from leaves in tomato plant canopies from two commercial greenhouse, South Alder (SA) and CanAgro (CA).  Leaf position 1 is the leaf closest to the first flowering truss, - leaves are younger and + leaves are older.  Panel A and B, data collected in March 2001, Panel C and D, data collected in September 2001. Line equations are given in Table 4.2........................................................... 121 Figure 4.6 Starch content per leaf and the cumulative contribution of each leaf to total starch for shoot canopies of plants in two commercial greenhouses. Greenhouses were South Alder (SA) and Gipaanda (GI).  All data were collected in 2002, Panel A: May, B: June, C: July, D: August and E: September.  S is for starch and C is for cumulative.  Line equations for starch per leaf are given in Table 4.4. ..................................................................... 127 Figure 4.7 The modelled starch content per leaf and total fruit volume per truss from the shoot canopies of plants in two commercial greenhouses.  Greenhouses were South Alder (SA) and Gipaanda (GI).  All data were collected in 2002, Panel A: May, B: June, C: July, D: August and E: September.  S is for starch and F is for fruit.  Line equations for starch per leaf are given in Table 4.4. ......... 131 Figure 5.1 Diurnal and seasonal patterns of leaf starch concentration at leaf position 1 from tomato plants in two commercial greenhouses, South Alder (SA) and CanAgro (CA).  Data were collected dawn to dawn for selected dates in 2001, Panel A. SA June 4, CA June 12, Panel B. SA August 22, CA August 20, Panel C. SA September 19, CA September 6. Bars are 95 % confidence intervals of the mean. The dark horizontal bars indicate the night. .............. 163 Figure 5.2 Diurnal and season patterns of leaf mass per unit area (LMA) at leaf position 1 from tomato plants in two commercial greenhouses, South Alder (SA) and CanAgro (CA).  Data were collected dawn to dawn for selected dates in 2001, Panel A. SA June 4, CA June 12, Panel B. SA August 22, CA August 20, Panel C. SA September 19, CA September 6. Bars are 95 % confidence intervals of the mean. The dark horizontal bars indicate the night. ............. 164 Figure 5.3 The diurnal environment from two commercial greenhouses, CanAgro (CA) and South Alder (SA) in 2001.  Panel A. is global radiation -exterior to greenhouse, Panel B is hourly CO2 concentration, Panel C is hourly temperature and Panel D is hourly humidity deficit (SA) or relative humidity (CA).  Data were collected June 3 to 5 for SA and June 11 to 13 for CA.... 165 Figure 5.4 The diurnal environment from two commercial greenhouses, CanAgro (CA) and South Alder (SA) in 2001.  Panel A is global radiation, measured exterior to the greenhouse, Panel B is hourly CO2 concentration, Panel C is hourly temperature and Panel D is hourly humidity deficit (SA) and relative humidity (CA).  Data were collected August 21 to 23 for SA and August 19 to 21 for CA. ................................................................................................................ 166  xiv Figure 5.5 The diurnal environment from two commercial greenhouses, CanAgro (CA) and South Alder (SA) in 2001.  Panel A is global radiation exterior to greenhouse, Panel B is hourly CO2 concentration, Panel C is hourly temperature and Panel D is hourly humidity deficit (SA) and relative humidity (CA).  Data were collected September 9 to 11 for SA and September 5 to 7 for CA. ................................................................................................................ 167 Figure 5.6 Diurnal profiles of leaf starch in plants exposed to ambient or enriched CO2 at PARC-Agassiz.  Data were collected on August 24 and 25, 2002, 74 days after the onset of CO2 enrichment.  Panel A: leaves from position 1 (top of canopy), Panel B: leaves from position 7 (middle) and Panel C: leaves from position 13 (lower).  The horizontal dark bars indicate night.  Vertical bars are 95% confidence intervals of the mean, N=2.  Means statistically significantly different at p<0.05 and p<0.01 are indicated by * and **, respectively. ...... 169 Figure 5.7 Diurnal profiles of leaf mass per unit area (LMA) in plants exposed to ambient or enriched CO2 at PARC-Agassiz.  Data were collected on August 24 and 25, 2002, 74 days after the onset of CO2 enrichment.  Panel A: leaves from position 1 (top of canopy), Panel B: leaves from position 7 (middle), and Panel C: leaves from position 13 (lower).  The horizontal dark bars indicate night.  Vertical bars are 95% confidence intervals of the mean, N=2. ......... 171 Figure 5.8 The diurnal environment of the two treatments in the 2002 PARC-Agassiz experiments.  Ambient did not receive supplemental CO2 and Enriched was given supplemental CO2.  Panel A is global radiation -exterior to greenhouse. Panel B is mean hourly CO2 concentration.  Panel C. mean hourly temperature, and Panel D is mean hourly relative humidity.  Data were collected between August 23 and 25, 2002.  Except for global radiation, N=2. ....................................................................................................................... 172 Figure 5.9 Summary of the environmental data for greenhouse compartments treated with either Ambient or Enriched CO2 at PARC-Agassiz during the CO2 onset experiment in 2002.  Panel A is total global radiation per day measured external to the greenhouse.  Panel B is mean daily atmospheric CO2. Panel C is mean daily temperature and Panel D is mean daily relative humidity.  Data were collected between June 10 to 17 2002. N=2 except for global radiation and bars are 95% confidence intervals of the mean...................................... 174 Figure 5.10 Leaf starch per unit leaf area prior to and during the first seven days of exposure to Enriched or Ambient CO2 at PARC-Agassiz, May 6 to 13 2003. Data were collected from three leaf positions at 14 h.  Panel A is for leaf position 1 (top of canopy), Panel B is for leaf position 7 (middle) and Panel C is for leaf position 13 (lower).  Symbols are the means (N=2) and bars are 95% confidence intervals of the mean.  Means significant different at p<0.05 and p<0.01 are indicated by * and **, respectively...................................... 176  xv Figure 5.11 Leaf mass per unit area (LMA) prior to and during the first seven days of exposure to Enriched or Ambient CO2 at PARC-Agassiz, May 6 to 13 2003. Data were collected from three leaf positions at 14 h.  Panel A is from leaf position 1 (top of canopy).  Panel B is from leaf position 7 (middle) and Panel C is from leaf position 13 (lower). Symbols are the means (N=2) and bars are 95% confidence intervals of the mean.  Means significant different at p<0.05 and p<0.01 are indicated by * and **, respectively...................................... 177 Figure 5.12 Summary of the environmental data for greenhouse compartments treated with either Ambient or Enriched CO2 at PARC-Agassiz during the CO2 onset experiment in 2003.  Panel A is total global radiation per day measured external to the greenhouse.  Panel B is mean daily atmospheric CO2. Panel C is mean daily temperature and Panel D is mean daily relative humidity.  Data were collected between May 5 to 13 2003. N=2 except for global radiation and bars are 95% confidence intervals of the mean.  CO2 enrichment was started at 15 h on day 0; therefore day -1 data (two days before the experiment commenced) were shown to confirm no pre-treatment differences in CO2 concentration................................................................................................. 178   xvi Symbols and Abbreviations a…………………….. Asymptotic leaf area (m2) aa………………….... Constant in canopy photosynthesis model AGOP………..…….. Amyloglucosidase-glucose oxidase-peroxidase aleaf………………….. Leaf light use efficiency for tomato g(CO2)J-1 acanopy……………..… Canopy light use efficiency b…………………..… Leaf area at leaf position zero ANOVA…………..… Analysis of variance ba…………….……… Constant in canopy photosynthesis model BC……………….….. British Columbia c…………………….. Growth rate coefficient CA………………...… CanAgro Produce Ltd. cm2………………..… Square centimetre cm3………………….. Cubic centimetres CO2 or C………….… Carbon dioxide (ppm) d…………………….. Exponent DM………………….. Dry mass e…………………….. Exponential FLE…………………. Full leaf expansion FV…………………... Fruit volume (cm3) g…………………….. Gram GI…………………… Gipaanda Greenhouses Ltd.  xvii GJ…………………… Gigajoule GLM………………... General linear model Gnh…………………. Greenhouse GR………………….. Global radiation (W m-2) h…………………….. Hour ha…………………… Hectare I…………………….. Photosynthetically active radiation at top of canopy (W m-2) I ……………………. Average PAR  at the top of the canopy - previous week k…………………….. The extinction coefficient for light (0.94) kg………………….. kilogram L……………………. Leaf position ln…………………… Natural logarithm LA…………………... Leaf area (m2) LAI…………………. Leaf area index (m2 m-2) LCS………………… Lack of positive correlation weighted by standard deviations LL…………………... Leaf length (m) LMA………………... Leaf mass per unit area (g m-2) LS…………………… Leaf starch m2…………………… Square meter m……………………. Meter mm………………….. Millimeter m……………………. The leaf transmission coefficient for light  xviii MFL………………… Measurable fruit load MJ…………………... Megajoule ml…………………… Millilitre MSD………………… Mean square deviation MT………………….. Megatonne N or n……………….. Number of observations NIRS………………... Near-infrared spectroscopy p…………………….. Probability of falsely rejecting the null hypothesis when  true pCO2…………………. Density of CO2 at 20 C (kg m-3) Pg……………………. Hourly gross photosynthesis (g (CO2) m-2 h-1) Pn……………………. Hourly net photosynthesis (g (CO2) m-2 h-1) PAR………………… Photosynthetically active radiation (W m-2) PARC-Agassiz……... Pacific AgriFood Research Centre, Agassiz, BC pdiff………………… p-values for all possible pair-wise comparisons π…………………….. Pi ppm…………………. Part per million Proc…………………. Procedure rp……………………. Dark respiration (g (CO2) m-2s-1) r……………………... Radius R……………………. Correlation coefficient R2…………………… Coefficient of determination RMSE....…………… Root mean squared error  xix RRMSE……………. Relative root mean squared error RUBISCO..………... Ribulose-1,5-biphosphate carboxylase / oxygenase SA…………………… South Alder Greenhouses Ltd. SB…………………… Squared bias SD…………………… Standard deviation SLA…………………. Specific leaf area SLS…………………. Short leaf syndrome SPL…………………. Starch per leaf SPM………………… Stems per square meter τ……………………... Carbon dioxide stomatal conductance (m s-1) t……………………… Time (seconds) TOMGRO………….. Tomato growth model W……………………. Watt     xx Acknowledgements My heart-felt thanks to my thesis advisors Dr. Peter Jolliffe and Dr. David Ehret for giving me this opportunity, and providing the mentorship, support and encouragement I needed to complete this thesis. To committee members Drs. Mahesh Upadhyaya and Rob Guy thanks for your excellent advice.  Thanks to Dr. Kathy Baylis for help with the economic modeling, Dr. Mary-Lou Swift for help with the NIRS and Dr. Mary Peet of NC State for her encouragement. Funding from the BC Greenhouse Growers’ Association was instrumental to this project.  The interest and involvement of the local growers was crucial.  Special thanks to: Dave Ryall, Gordon Yakel, Gary Van Straalen, Robert Erwin, Megan Scarlett and Lynn Fu. Your generosity and patient instruction of the art and science of growing tomatoes was greatly appreciated.  Without the help of the staff at PARC-Agassiz this research would not have been possible.  Thanks to:  Tom Helmer, Brenda Frey, Mark Gross, June Dawson, Georgia Kliever, Heidi Remple, Marisa Demurs and Sandy Gillespie.  At UBC, Alain Boucher, Tara Moreau, Hardy Hall, Faride Unda, Sonya Chiu, Nyssa Temmel, Dr. Bob Copeman and Marcus Samuel were a great source of knowledge and friendship. This thesis, which went on and on and on was undoubtedly difficult for my family. Thanks for the moral and financial support and love from my parents John and Rosemary Edwards and my in-laws, Jack and Frances Swann.  Huge thanks to my husband John Swann, who has been with me every step of the way in this journey.  You have been an unwavering source of encouragement, maintained a positive outlook for both of us and were a decent lab assistant too.  To my son Cameron Swann, as you came along near the end of this journey, you provided me with both a needed break and impetus to complete this thesis.  xxi Co-authorship statement In this thesis the authorship is mine and I was advised mainly on experimental procedures, chapter structure and editorial matters.  In Chapter 2 I was advised by Drs. Peter Jolliffe, David Ehret and Kathy Baylis.  In Chapters 3, 4 and 5 I was advised by Drs. Peter Jolliffe and David Ehret.  In addition, Dr. Mary-Lou Swift provided assistance with the measurement of starch using near-infrared spectroscopy in Chapters 4 and 5 and Appendix 6.   1 1 Towards a plant-based method of guiding CO2 enrichment in greenhouse tomato 1.1 INTRODUCTION Enriching the greenhouse atmosphere with supplemental CO2 is a prevalent production technique used by commercial tomato growers to increase the yield of their plants and improve the economic viability of their businesses.  Carbon dioxide management practices have been developed which base dosing decisions on several main considerations: the potential ability of supplemental CO2 to boost photosynthesis, the economics of CO2 applications, and to some degree the visible appearance of the plants.  In commercial practice, none of these techniques rely on quantifiable aspects of the plants or measures of plant performance.  This research will place more focus upon the plant, to explore whether plant indicators can be used to help guide greenhouse CO2 management decisions. In the lower mainland of British Columbia, greenhouse tomato growers have been able to achieve remarkably high annual fruit yields (≈ 75 kg m-2).  This achievement reflects the favourable maritime climate of coastal BC, as well as the willingness of growers to apply advanced technologies.  Enrichment of the greenhouse atmosphere with CO2 has been identified as being beneficial to plant growth since the mid 1800’s.(Nederhoff 1994). However, commercial greenhouses did not adopt it as a conventional production practise until the 1960’s (Wittwer 1986; 1988; Mortensen 1987; Nederhoff 1994).  Numerous reviews have been published on plant responses to elevated CO2, and Kimball (1983), Slack (1986), Mortensen (1987), Nederhoff (1994), Heuvelink and Dorais (2005) and Peet and Welles (2005) have specifically covered greenhouse tomato plant responses.  Many greenhouses utilize CO2 that is provided as a by-product from the combustion of fossil fuels they use for heating.  When the boiler exhaust is low in impurities it can be diluted and injected into the  2 greenhouse.  If impurities are high, or CO2 is not adequately available from the heating system, pure compressed (liquid) CO2 can be purchased and used as the CO2 source for the greenhouse.  Technical aspects of CO2 dosing and energy management have been reviewed by Hicklenton (1988), Hanan (1998) and Nederhoff (1994; 2004a; 2004b). Several factors are motivating the local greenhouse industry to re-examine its CO2 dosing practises.  The economic status of the industry is being affected by increases in costs of heating and CO2 generation.  According to the BC growers, the cost of CO2 enrichment can amount to about 20% of the greenhouse’s natural gas bill.  In recent years the price of natural gas has increased several-fold, and it seems unlikely that prices will soon return to pre-year 2000 levels.  At the same time, local growers have faced lower prices for fruit as a result of fragmentation and expansion of the domestic industry and a recent expansion of production in the United States and Mexico. There are also environmental concerns associated with the industrial generation and release of significant quantities of CO2.  Carbon dioxide is a “greenhouse” gas that is playing a significant role in current and future global climate change.  All industrial sectors, including agriculture, have been encouraged by all levels of government to reduce the emissions of this gas.  British Columbia greenhouse growers appreciate that their high CO2 use can led to pollution.  They are endeavouring to be responsible global citizens and it is hoped that the present studies will help them to use CO2 efficiently. Finally, many greenhouses now have the capability to generate and supply significant doses of CO2 to their crops.  Carbon dioxide is a physiologically active gas and excessive CO2 is known to have the ability to cause plant damage (Madsen 1968; van Berkel 1984; Ehret & Jolliffe 1985; Mortensen 1987; Tripp 1990; Tripp et al. 1991; Nederhoff et al. 1992;  3 Nederhoff 1994; Hao et al. 2006).  Local growers have observed leaf damage which they feel may be caused by excessive CO2 and it is possible that they may be chronically overdosing their plants. Without the use of supplemental gas, CO2 can be substantially depleted from the greenhouse atmosphere when crop photosynthesis is active and greenhouse vents are closed. CO2 supplementation can overcome that depletion, replenishing CO2 as the gas is withdrawn by photosynthesis and escapes to the surroundings.  Enrichment of the atmosphere to concentrations above ambient can promote crop photosynthesis, ultimately furnishing additional carbon to other parts of the plant, and helping to promote fruit growth.  High CO2 concentrations above the crop promote the delivery of the gas to leaf surfaces by diffusion and bulk flow, and increase photosynthesis rate toward its maximum.  However, at high CO2 levels, the rate of escape of the gas from the greenhouse when the vents are open is higher. Current recommendations are that the greenhouse atmosphere be maintained at least at 380 ppm CO2 (near current ambient levels) but not more than 1000 ppm (Mortensen 1987).  That upper limit is based on a review of the literature between 1975 and 1986 (Mortensen 1987), although the achievement of 1000 ppm CO2 is uneconomical when the greenhouse requires ventilation, due to rapid CO2 escape. In contemporary commercial practice, growers tend to use the injection rate of CO2 rather than CO2 concentration to regulate dosing.  They find that CO2 injection is easier to monitor as part of their energy management practises, and is independent of variations associated with the operation of ventilation systems.  The range of recommended and applied rates of CO2 injection is wide.  In the past, recommended dosing rates have been 5 g m-2 h-1 (5 g of CO2 per square meter of greenhouse floor space per hour) to prevent the depletion of  4 CO2 in the greenhouse below the ambient outside level (Nederhoff 1994; Nederhoff 2004b). This was based on the estimated CO2 uptake of the crop through photosynthesis (Nederhoff 1994).  However, it is common in a modern greenhouse to inject 20 g m-2 h-1 (Nederhoff 2004b) while some growers have been known to inject as high as 50 g m-2 h-1 during the summer. It is notable that both atmospheric CO2 concentration thresholds and CO2 injection rates do not take into direct account either the plants’ ability to assimilate CO2 carbon or to utilize the assimilated carbon.  This opens the possibility that growers might either over- or under-dose their crops with CO2.  Overdosing is undesirable because it may be a wasteful expense for fossil fuel combustion, and because it contributes unnecessarily to industrial CO2 emissions into the environment.  Underdosing is undesirable because potential crop yield is not fully realized.  A more broadly based dosing approach, utilizing plant as well as environmental information, might allow growers to fine-tune CO2 applications to better match the plants’ needs. An effective plant-based method of CO2 management must use plant measures that are obtainable under greenhouse conditions and sufficiently responsive to guide CO2 dosing. Several candidate methods can be suggested.  One candidate would be some measure of photosynthesis rate, as it is a direct measure of CO2 use.  Photosynthesis is usually determined on single leaves by measuring the net carbon exchange of a leaf enclosed in a cuvette.  Such measurements are not feasible for a grower to carry out because they are labour intensive, highly variable and not suitable to plant canopies with numerous leaves. An assessment of photosynthesis by the shoot canopy would be preferable to single leaves, being more representative of whole crop performance.  Measuring photosynthesis by  5 enclosing whole shoots in chambers would not be very feasible, because of crop growth, cost of equipment and the requirement for constant supervision of the equipment.  That approach would also disturb the local climate of the enclosed plants, making the observations unrepresentative of the main crop.  However, the responses of leaf photosynthesis to light and CO2 have been used to develop a model for canopy photosynthesis (Acock et al. 1978). Using such a canopy photosynthesis model might yield meaningful information that growers could integrate into their environmental control systems for CO2 decision making. The measurement of plant growth is another potential approach, as the objective of CO2 enrichment is to obtain greater plant growth and yield.  Increased growth is often determined by a destructive analysis of the plant to determine leaf area and dry mass.  Since individual plants can contribute over 20 kg fruit per year, substantial removal of plant material for destructive analysis would be an unacceptable practise in a commercial greenhouse.  However, removal of small quantities of plant material, or a non-destructive analysis of growth, would be acceptable, allowing the grower to determine the response of the plants to CO2 enrichment without destroying them.  Ehret et al. (2001) have reviewed several non-destructive and automatable methods for monitoring the growth of greenhouse crops.  These were imaging cameras, linear variable displacement transducers and suspended load cells.  At this time, applications of these techniques for measuring plant growth are expensive and technologically challenging.  The early stages of implementation are only just beginning (Helmer et al. 2005) and it is probably too early to know how these data could be used to address specific questions surrounding CO2 dosing. Other potential candidates are measurements of sucrose or starch, the products of photosynthesis.  Both products increase in plants grown in enriched CO2, and starch has the  6 advantage of exhibiting higher stability (Körner et al. 1995; Bertin et al. 1999).  As well, leaf starch levels are reflective of both the source activity and sink capacity of the plant (Farrar 1996; Minchin & Thorpe 1996) and are considered an overflow product of photosynthesis. Presumably plants with high levels of starch have more fixed carbon than they can use at the moment and thus may also be an indicator of overdosed plants.  Currently, the analysis of leaf starch requires a laboratory, which makes its use by growers inconvenient.  However, the accumulation of leaf starch may have a strong effect on leaf mass, and the measurement of leaf mass per unit area (LMA) might be an alternative. In summary, simulation modelling of crop CO2 responses, non-destructive measurements of plant growth, and monitoring plant carbon balances through observations of leaf starch and mass are three potential plant-based approaches which might provide information helpful to decision making for greenhouse CO2 dosing.  These approaches were investigated in this work, and they focus on three hierarchies of organization in a greenhouse: the production system, the plant, and the leaf.  Chapter 2 will explore the use of canopy photosynthesis and partitioning models to track the seasonal profile of CO2 uptake by the crop.  These models were used in conjunction with greenhouse CO2 generation data and fruit prices to evaluate costs and benefits of CO2 dosing.  Chapter 3 will consider the use of some non-destructive plant measures to monitor plant growth over time, comparing growth under enriched or ambient CO2 treatments.  In Chapters 4 and 5 a more detailed approach is taken, where plant carbon status is evaluated by the measurements of starch and the mass of individual leaves.  In those studies diurnal and spatial profiles of leaf starch and LMA are evaluated in an attempt to identify when and where leaf data should be collected.  7 These investigations were carried out in commercial greenhouses, whenever possible, and in research greenhouses at Agriculture and Agri-Food Canada, Pacific Agri-Food Research Centre (PARC) in Agassiz, BC.  In both venues the main source of CO2 was from the combustion of natural gas and the investigations used the commonly grown tomato (Lycopersicon esculentum) beefsteak cultivar, Rapsodie.  Throughout the research my focus was on approaches that are useable, or have the potential to be used by commercial tomato growers.  In the commercial greenhouses it was not possible for me to influence the application of CO2, which prevented me from comparing different dosing strategies under those circumstances.  At PARC – Agassiz, however, four greenhouse compartments (experimental units) were available, and these allowed two replicates of an ambient CO2 treatment plus two of an enriched CO2 treatment. 1.2 RESEARCH OBJECTIVES The overall goal of this research was to examine several plant-based approaches that could be used to improve the management of CO2 in commercial tomato greenhouses.  The following objectives were pursued as candidates for inclusion in such a method of CO2 enrichment. 1. To investigate the costs and benefits of CO2 enrichment using plant-based simulation models (Chapter 2). 2. To characterize the growth of mature fruiting plants under ambient and enriched CO2 using non-destructive measures of growth (Chapter 3). 3. To elucidate the distribution of starch and leaf mass per unit area by the development of their profile in the plant canopy and responsiveness to the influences of fruit load and leaf area (Chapter 4).  8 4. To examine the temporal dynamics of leaf starch and leaf mass per unit area by an investigation of (i) their diurnal variability and (ii) their responsiveness at the onset of CO2 enrichment (Chapter 5).  9 1.3 LITERATURE CITED Acock, B., Charles-Edwards, D.A., Fitter, D.J., Hand, D.W., Ludwig, L.J., Warren Wilson, J., and Withers, A.C. 1978. The contribution of leaves from different levels within a tomato crop to canopy net photosynthesis:  An experimental examination of two canopy models. Journal of Experimental Botany 29: 815-827. Bertin, N., Tchamitchian, M., Baldet, P., Devaux, C., Brunel, B., and Gary, C. 1999. Contributions of carbohydrate pools to the variations in leaf mass per area within a tomato plant. New Phytologist 143: 53-61. Ehret, D.L. and Jolliffe, P.A. 1985. Leaf injury to bean plants grown in carbon dioxide enriched atmospheres. Canadian Journal of Botany 63: 2015-2020. Ehret, D.L., Lau, A., Bittman, S., Lin, W., and Shelford, T. 2001. Automated monitoring of greenhouse crops. Agronomie 21: 403-414. Farrar, J.F. 1996. Sinks-integral parts of a whole plant. Journal of Experimental Botany 47: 1273-1279. Hanan, J.J. 1998. Greenhouses.  Advanced Technology for Protected Horticulture. CRC Press , Boca Raton. Hao, X., Wang, Q., and Khosla, S. 2006. Responses of a greenhouse tomato crop to summer CO2 enrichment. Canadian Journal of Plant Science 86: 1395-1400. Helmer, T., Ehret, D.L., and Bittman, S. 2005. CropAssist, an automated system for direct measurement of greenhosue tomato growth and water use. Computers and Electronics in Agriculture 48: 198-215. Heuvelink, E. and Dorais, M. 2005. Crop growth and yield. In Heuvelink, E., ed. Tomatoes. CABI Publishing, Cambridge, MA, pp 85-144. Hicklenton, P.R. 1988. CO2 enrichment in the greenhouse:  principles and practise. Timer Press, Portland, Oregon. Kimball, B.A. 1983. Carbon dioxide and agricultural yield:  an assemblage and analysis of 430 prior observations. Agronomy Journal 75: 779-788. Körner, Ch., Oelaez-Riedl, S., and van Bel, A.J.E. 1995. CO2 responsiveness of plants:  a possible link to phloem loading. Plant, Cell and Environment 18: 959-600. Madsen, E. 1968. Effect of CO2 - concentration on the accumulation of starch and sugar in tomato leaves. Physiologia Plantarum 21: 168-175. Minchin, P.E.H. and Thorpe, M.R. 1996. What determines carbon partitioning between competing sinks? Journal of Experimental Botany 47: 1293-1296.  10 Mortensen, L.M. 1987. Review:  CO2 enrichment in greenhouse. Crop Responses. Scientia Horticulturae 33: 1-25. Nederhoff, E.M., De Koning, A.N.M., and Rijsdijk, A.A. 1992. Leaf deformation and fruit production of glasshouse grown tomato (Lycopersicon esculentum Mill.) as affected by CO2, density and pruning. Journal of Horticultural Science 67: 411-420. Nederhoff, E.M. 1994. Effects of CO2 concentration on photosynthesis, transpiration and production of greenhouse fruit vegetable crops. Wageningen. Nederhoff, N. 2004a. Carbon dioxide enrichment and risks of noxious gases part 2. Practical Hydroponics & Greenhouses September/October: 46-55. Nederhoff, N. 2004b. Carbon dioxide enrichment - Fuels and Figures. Practical Hydroponics & Greenhouses May/June: 50-59. Peet, M.M. and Welles, G. 2005. Greenhouse tomato production. In Heuvelink, E., ed. Tomatoes. CABI Publishing, Cambridge, MA, pp 257-304. Slack, G. 1986. CO2 enrichment of tomato crops. In CRC Press, Boca Raton, Florida, pp 151-163. Tripp, K.E. 1990. Physiological aspects of tomato foliar deformation at elevated CO2 concentration. North Carolina State University. Tripp, K.E., Peet, M.M., Pharr, D.M., Willits, D.H., and Nelson, P.V. 1991. CO2-enhanced yield and foliar deformation among tomato genotypes in elevated CO2 environments. Plant Physiology 96: 713-719. van Berkel, N. 1984. Injurious effects of high CO2 concentrations on cucumber, tomato, chrysanthemum and gerbera. Acta Horticulturae 162: 101-112. Wittwer, S. 1986. Worldwide status and history of CO2 enrichment - an overview. In Enoch, H.Z. and Kimbal, B.A., eds. Carbon Dioxide Enrichment of Greenhouse Crops Volume I Status and CO2 Sources. CRC Press Inc., Boca Raton, Florida, pp 3-15.  11 2 An application of simulation modeling to explore CO2 dosing scenarios in greenhouse tomato production1,2 Carbon dioxide management in greenhouse tomato production involves an interplay at several levels of organization. At one level, there is the CO2 concentration of the greenhouse atmosphere and how that concentration reflects CO2 delivery into the greenhouse from supplemental CO2 sources and losses to the external atmosphere via vents and leaks. Another level involves the ability of the crop to absorb CO2 and utilize the enhanced carbon intake to boost tomato fruit production.  Finally, there is the connection between the economic yield increase and the additional costs associated with CO2 dosing. This chapter exploits commercial greenhouse climate data, a crop production simulation model, and uses cost-benefit analysis to explore this interplay in three steps.  The first step is to test and apply a plant-based simulation model that predicts tomato yield responses to greenhouse environmental conditions, including atmospheric CO2 levels.  The second step is to use greenhouse data to establish an empirical relationship between supplemental CO2 supply and the CO2 concentrations occurring in the greenhouse atmosphere.  In the final step these results were then used in a cost benefit analysis of several CO2 enrichment scenarios. 2.1 INTRODUCTION 2.1.1 CO2 enrichment in greenhouses Greenhouses produce high quality commodities but at an appreciable cost: 1 ha has a direct capital cost of $2.5 million (Canadian) dollars with construction cost of approximately  1 A version of this chapter will be submitted for publication.  Edwards, D., Jolliffe, P., Ehret, D. and Baylis, K. An application of simulation modeling to explore CO2 dosing scenarios in greenhouse tomato production.  2 A version of this chapter has been published.  Edwards, D., Jolliffe, P., Ehret, D. and Baylis, K.  Towards a plant-based method of CO2 management.  Acta Horticulturae. 797:273-278.  12 $600 000 per ha (BC Greenhouse Growers’ Association, pers. com.).  To keep their operations profitable greenhouse growers readily adopt technologies that increase the efficiency and scale of production.  One such technology has been enrichment of the greenhouse atmosphere with CO2.  It is well known that photosynthesis responds strongly to CO2 increase, and estimates are that increasing the CO2 concentration of the greenhouse atmosphere up to 1000 ppm can increase tomato fruit yield by about 25% (Nederhoff 1994). Even lower levels of CO2 enrichment can be helpful since they guard against CO2 depletion by plant photosynthesis, that would otherwise restrict plant growth when CO2 levels get very low ( Hand 1984; Nederhoff 2004).  Carbon dioxide depletion can occur not only when ventilation is low or not needed for temperature control but also when ventilators are wide open and plants are photosynthesizing rapidly (Ohyama et al. 2005).  To prevent depletion the general recommendation for growers has been to add at least 5 g of CO2 per m2 per hour to the greenhouse atmosphere (Hand 1984; Nederhoff 2004;). It became possible for greenhouses to dose large amounts of CO2 with the development of gas analyzer technology to control CO2 levels and when inexpensive sources of low contaminant CO2 became available (Hand 1984; Nederhoff 1994).  In modern greenhouses, enrichment of the greenhouse atmosphere with CO2 occurs by two means, pure liquid CO2 and combustion of low contaminant petroleum products.  In the lower mainland of British Columbia most commercial tomato greenhouse growers couple their CO2 and heat production; by using a central boiler fired by natural gas for heat production and then by injecting the CO2-rich flue gases into the greenhouse.  However, for most of the year heat production and CO2 usage are not synchronized, as the greatest need for heating is usually when plant photosynthesis is low.  With the advent of the heat buffer, heat and CO2  13 production became asynchronous.  The heat buffer is a silo-like structure that enables water to be circulated between the boiler and buffer instead of the greenhouse.  Therefore, the boiler can be operated during the day for CO2 and the heat can be stored until needed. During the summer growers often must use liquid CO2 for atmospheric enrichment.  This is because heat storages become full with only short nights to exhaust them, and boilers may not have the capacity to produce the necessary amount of CO2 when venting is high. The use of supplemental CO2 is one of the technologies that has enabled the BC tomato producers to remain competitive under conditions of stiffer competition in the North American tomato market.  However, several recent factors have caused growers to reexamine this production practice.  These factors are high fuel costs, changing market conditions and the Kyoto protocols. Up to the early 2000’s growers had enjoyed very low natural gas prices, being less than $2 per gigajoule (GJ).  This ended in early 2001 with prices spiking up to $14 per GJ caused by a variety of national infrastructure and international factors.  Although natural gas prices are now lower it is forecast that they will be at least $6 per GJ for the foreseeable future (Anonymous 2006a).  In the early 2000’s competition became fiercer in the North American tomato market with large areas in Arizona and later Mexico entering the market, and uncertainties occurring in the tomato export market to the United States.  Finally, Canada has ratified the Kyoto protocol which necessitates CO2 emission reductions of 6% below the 1990 level of 596 MT of greenhouse gas equivalents (Anonymous 2006b).  Since emissions have continued to climb, by 2012 Canada will have to reduce current emissions by 24% to meet the 1990 target (Anonymous 2006b).  Thus all sectors, including agriculture, have been encouraged to reduce emissions of greenhouse gas equivalents.  14 These factors have caused growers to reexamine many of their production practices, including their practices of CO2 enrichment.  At present, growers manage their CO2 dosing either on the basis of the CO2 concentration measured in the greenhouse atmosphere or according to the rate of CO2 injection from supplemental CO2 sources (g m2 hr-1).  These methods do not explicitly take into account the CO2 assimilation by the plants.  One possibility is to use a simulation model of canopy photosynthesis to aide in the management of CO2 enrichment. 2.1.2 Modeling greenhouse systems Models provide a framework to understand complex phenomena, commonly by providing a mathematical representation of the real world.  Models break complex systems into their components; thereby summarizing what is known and what insights can be gained from an existing system (Lentz 1998).  Models also have a role in teaching and training, and if a model is effective it can be used to simulate outcomes with changing input conditions, which is useful for decision making (Gary et al. 1998; Lentz 1998; Marcelis et al. 1998). In greenhouse production the main application of models has been for environmental control.  Typically, such models operate using environmental set points, established by the grower, which are used to trigger heating or ventilation events.  More sophisticated control- based models consider several environmental parameters and strive to optimize the environment for least cost and/or best plant growth.  At best these approaches only indirectly consider plant performance.  Recently some researchers have focused on the direct measurement of plant performance with the objective that this crop information can be incorporated into models as part of a plant centered approach (Ehret et al. 2001; Ewert 2004; Heuvelink et al. 2005).  15 Greenhouse production should lend itself well to modeling because of the high degree of control of the greenhouse environment and the strict management of the crop(s) (Challa 2002).  Greenhouse-grown plants are largely protected from extremes of radiation, drought and nutrient levels, and from biotic stresses such as weeds, insects and diseases.  Greenhouse vegetable production is further controlled in that it is mostly a synchronous monoculture, sometimes with only one cultivar grown.  Plant spacing and leaf area development are consistent, and growers routinely collect and save large amounts of environmental and plant data that could be utilized in a model. 2.1.3 Models for CO2 management Enrichment of the greenhouse atmosphere with CO2 has always been a significant cost to greenhouse vegetable production.  Growers and scientists have recognized the paradox that the best time to enrich the greenhouse atmosphere with CO2 for photosynthesis is the most difficult time to maintain high CO2 levels in the greenhouse with the need to ventilate for temperature control.  In the last 20 years a significant amount of work has been dedicated to optimizing CO2 dosing for a sensible boost in plant performance through the development of models.  Researchers have investigated mathematical models with the aim to find the optimal CO2 set point under changing conditions and for energy savings (Challa & Schapendonk 1986; Aaslyng et al. 1999; Dieleman et al. 2005; Aaslyng et al. 2006).  With the advent of increased computing power, attempts were made to develop real-time control using an expert system and neural networks (Linker et al. 1998).  The recognition of the need for time scaling (photosynthesis can respond to minutely changing CO2 conditions, while it takes 6 to 8 weeks for a fruit to develop) led to more sophisticated models that contained simulation sub-models for predicting plant growth (Aikman 1996; Chalabi et al. 2002a;  16 Chalabi et al. 2002b).  Advantages of simulation models are that they can be used to gauge crop responses when greenhouse climate parameters are changed, without incurring the costs and delays of running full-scale experiments (Lentz 1998).  Some simulation models (Alscher et al. 2001b; 2001c) combine mechanistic and empirical models as not all greenhouse functions have fully developed models (Alscher et al. 2001a).  Such models can be validated later as data becomes available.  Most of these simulation models include an economic component because the economic threshold for CO2 enrichment is often reached before the plant response threshold. To my knowledge, none of the above approaches have been implemented by the Canadian greenhouse industry.  Barriers to this technology transfer are:  poor accessibility of the scientific literature to the growers, the models have not been validated in commercial conditions, the models may be overly complicated for the growers’ needs, and the potential utility of such approaches have not been adequately demonstrated to the greenhouse industry. In fact, Challa (2002) pointed out the lack of implementation of models and he cautions that more implementation is needed or future research in this area will stagnate. 2.1.4 Objectives of chapter The goal of this chapter is to use a plant-based simulation model of tomato production, in conjunction with environmental data from commercial greenhouses, to explore the costs and benefits of several CO2 enrichment scenarios for greenhouse tomato production.  In order to attain this goal several objectives were developed: 1. To test the accuracy of a simple yield model for predicting seasonal fruit production in commercial greenhouses.  17 2. To develop an empirical relationship between the amounts of supplemental CO2 applied in a commercial greenhouse and the CO2 concentrations occurring in the greenhouse atmosphere. 3. To develop a relationship between costs associated with CO2 enrichment and the revenues from fruit yield. 4. To investigate several CO2 dosing strategies to assess greenhouse CO2 usage in relation to its economic value. 2.2 MATERIALS AND METHODS According to Lentz (1998) the first step in model development is to characterize the system - determining the important and controllable input and output factors (Figure 2.1). Subsequent steps involve the development of a conceptual model and the translation of that conception to an explicit model.  In this study, a canopy photosynthesis/fruit yield model will be used to provide a bridge between greenhouse CO2 dosing and the seasonal revenues and costs of CO2 (Figure 2.1).  For the CO2 dosing scenarios, the CO2 concentration in the greenhouse atmosphere was altered with the resulting change in CO2 produced and fruit production determined. 2.2.1 Plant and greenhouse environment data. Plant data.  Data were for tomato (Lycopersicon esculentum (Mill.) cv Rapsodie) plants grown in commercial greenhouse cultivation.  Rapsodie is a beefsteak cultivar and it was grafted onto the rootstock cultivar Maxifort to improve vigour.  The crop was maintained and fruit were harvested according to commercial practices for growing beefsteak tomatoes in the lower mainland of British Columbia (Portree 1996).  Data were obtained for two years of production.  In 2002, the planting was late with the crop not set into the  18 Aikman Model Gross Canopy Photosynthesis (Respiration) Net Canopy Photosynthesis Carbon for Fruit Growth Fruit Yield Plant PAR Canopy PAR Historical Greenhouse Light PAR Outside Environment LAI CO2CO2 Flue Gas Liquid  CO2 Supplemental CO2 Fruit Price  per Week Economic Cost of Natural Gas Cost of Liquid CO2 CO2 Scenarios Fruit Revenue Cost of CO2 Net Revenue      Acock et al. Model   Figure 2.1 A conceptual diagram of the components of the CO2 enrichment system used for the model development.  LAI is leaf area index and PAR is photosynthetically active radiation.  19 greenhouse until week 8 (from the beginning of the calendar year) with the main harvest period occurring from weeks 17 to 46.  In the 2003 cropping year the plants were set into the greenhouse starting at week 50 of 2002 with the main period of fruit harvest being between weeks 10 and 46. Leaf area index.  Leaf area of the crop was managed to take advantage of the seasonal changes in light conditions.  Accordingly, the leaf area index (m2 m-2) was increased and later decreased during the production cycle.  For tomato production leaf area index depends on the number of stems per m2 and the leaf area of each stem.  Leaf area per stem is relatively stable during the season, but the number of stems was adjusted during the season by controlling the growth of side shoots.  In 2002 the stem density (stems per m2) was 2.5 until week 15.  From week 15 to 20 density was 3.25, from week 20 to 34 density was increased to 3.5, and after week 34 the density was reduced to 2.5.  A similar pattern was used for 2003 with some of the dates being earlier as the planting was earlier.  In 2003, stem density (stems per m2) was 2.5 until week 8, 3.25 between week 8 and 13, 3.75 from week 13 to 34 and 2.5 after week 34. In a commercial greenhouse setting it is not possible to directly measure the leaf area of the plants, as this measurement would destroy those plants and cause unacceptable crop losses.  Leaf area was indirectly determined from a function that related leaf area to blade length for the cultivar Rapsodie (Appendix 1.1).  Using this function leaf area per stem for Rapsodie was determined to be 0.91 m2.  Leaf area index was calculated as: Equation 2.1 LAI = 0.91·SPM where LAI is leaf area index is in m2 m-2 and SPM is stems per m2.  20 Greenhouse data.  South Alder Greenhouses Ltd (SA) was the commercial greenhouse that participated in this study.  It is a Venlo greenhouse located in Surrey, BC, 4 ha in area and 5 m in ridge height.  This greenhouse, like most commercial vegetable production greenhouses in the lower mainland of British Columbia, possessed a recovery system for extracting CO2 from the flue gas of the boiler and a large tank for heat storage (estimated at 600 m3).  This system enables the greenhouse to run the boilers during the day, to generate CO2 for enrichment and stores the hot water produced in a large tank (buffer) until it is needed.  The environmental control system collected environment data every 5 minutes for each growing season and these data were stored on a computer.  To reduce the amount of data, the night time data were removed and hourly means were calculated for the daylight readings prior to use in the models.  Variables obtained were:  greenhouse air temperature, relative humidity and CO2 concentration (in ppm).  In addition, total global radiation in W m-2 was obtained from the weather station positioned outside of the greenhouse.  Further modifications were applied to these data before its use in the models, according to Equation 2.2.  Photosynthetically active radiation (PAR) was calculated by multiplying the global radiation by 0.5 (Monteith 1973; Thimijian & Heins 1983; Jones 1992).  The outside PAR was multiplied by 0.7, which was used to account for transmission losses and to estimate the PAR that would reach the top of the plant canopy.  This reduction has previously been used by Bailey (2002) and Nederhoff & Vegter (1994a) for modeling work in a Venlo greenhouse in a northern latitude. Equation 2.2 I = GR·0.5·0.7  21 where I is PAR at the top of the plant canopy in W m-2, GR is global radiation measured outside the greenhouse in W m-2.  The variables I, LAI and CO2 in ppm were used in the model for determining canopy photosynthesis. 2.2.2 Plant based models Canopy photosynthesis.  Hourly net canopy photosynthesis was determined using the method of Chalabi et al. (2002b).  Their method exploited the canopy photosynthesis model initially developed by Acock et al. (1978) and modified by Nederhoff (1994).  To calculate canopy photosynthesis this model requires the LAI of the crop, the PAR in W m-2 at the top of the canopy and CO2 concentration.  The model consists of four parts which determine: stomatal CO2 conductance, canopy light (PAR) use efficiency, gross canopy photosynthesis and net canopy photosynthesis.  Using the equations, nomenclature and constants from Chalabi et al. (2002b) these sub models are: Stomatal CO2 conductance in m s-1 Equation 2.3 ⎟⎟⎠ ⎞ ⎜⎜⎝ ⎛ −+ −+= − )1( )1( )( meIb mkIbln kb a kL a a a aτ where I  is the average PAR (W m-2) at the top of the canopy during the previous week. The following are constants:  k =0.94 is the extinction coefficient for PAR, m=0.10 is the leaf transmission coefficient for PAR, aa is 8.9x10-5 m3 J-1 and ba is 2.1x10-2 m2 s J-1. Canopy PAR use efficiency Equation 2.4 ⎟⎟⎠ ⎞ ⎜⎜⎝ ⎛ −−= − m eaa kL leafcanopy 1 1 )(   22 where k and m are as above, aleaf is leaf PAR use efficiency for tomato of 2.11 x10-5 g (CO2) J-1 and L is leaf area index (m2 m-2). Hourly gross photosynthesis in g (CO2) m-2 h-1 (ground area basis) Equation 2.5 t CIa CIa P cocanopy cocanopy g ⎥⎥⎦ ⎤ ⎢⎢⎣ ⎡ += 2 2 ρτ τρ  where I is PAR (in W m-2) at the top of the canopy, pCO2 is 1.83x10-3 kg m-3 (the density of CO2 at 20 C), C is the volumetric CO2 concentration in parts per million by volume and t is 3600 seconds per 1 hour. Hourly net photosynthesis in g (CO2) m-2 h-1 (ground area basis) Equation 2.6 tLrPP gn −= where r is the dark respiration rate 4.0x10-5 g (CO2) m-2 s-1, t is as above. Fruit yield model.  Fruit yield was also determined using the method of Chalabi et al. (2002b), based on the model of Aikman (1996).  In this method 1 g of assimilated CO2 (net photosynthesis) produces 4.5 g of fruit fresh mass.  On a mole basis, 1 mole of assimilated carbon results in 0.0675 moles of carbon in the fruit.  The hourly rate of net CO2 fixation was determined from the canopy photosynthesis models and converted to grams of fruit fresh mass.  The yield in kg per m2 was determined by summing the fruit mass for a week. Weekly yields were obtained from the grower for 2002 and 2003 in order to validate the fruit yield determined by the models. Model validation for yield.  A difference-based procedure was used to assess the accuracy of the model (or simulation) to predict the measured yield.  Model output was compared to measured yield using the mean square deviation method of Kobayashi & Salam  23 (2000).  Using Excel software version 2003 (Microsoft Corp., 2003) mean square deviation (MSD) was calculated, along with the components of MSD, squared bias (SB), the standard deviation of the simulation and the measurements (SDsSDm) and lack of positive correlation weighted by standard deviations (LCS).  The equations are: Equation 2.7  MSD = SB +SDSD + LCS Equation 2.8  ( )2∑ −= n YXMSD nn Equation 2.9  ( )2YXSB −= Equation 2.10 ( ) ( ) ⎥⎦ ⎤⎢⎣ ⎡ −−⎥⎦ ⎤⎢⎣ ⎡ −= ∑∑ == n i i n i ims yyn xx n SDSD 1 2 1 2 11 Equation 2.11 ( )RSDSDLCS ms −= 12 Equation 2.12 ( )( ) ( )sm n i ii SDSD yyxx nR ∑ = −− = 1 1  Where R is the correlation coefficient, n is the number of observations, and subscripts s and m indicate simulation and measured, respectively.  Note that values for the simulation are represented by x and the measurements by y, which is the convention for the analysis (Kobayashi & Salam 2000; Gauch Jr. et al. 2003).  A final related statistic was also used to quantify the goodness of fit, the relative root mean squared error (RRMSE) (Equation 2.13) (Quilot et al. 2002; Gibert et al. 2005). Equation 2.13 RRMSE = y RMSE  24 where RMSE is the root mean squared error and y is mean of the measured output.  This statistic, like R and R2, is independent of the units of measurement, which allows for comparison to other studies. 2.2.3 Relating CO2 generated from greenhouse heating to greenhouse CO2 levels The CO2 produced by the greenhouse was calculated from the monthly use of natural gas and liquid CO2.  The natural gas was measured in GJ and converted to kg of CO2 using the conversions of 26.88 m3 of natural gas in 1 GJ and 1.73 kg of CO2 in 1 m3 of natural gas (1 GJ of natural gas contains 47.3 kg of CO2).  The natural gas and liquid CO2 data were then expressed as kg m-2 of CO2 by dividing the total kg of CO2 by the area of the greenhouse in m2.  The CO2 in the greenhouse atmosphere was measured in ppm by the greenhouse environment control system.  Linear regressions were used to describe several relationships. The relationship between the monthly mean concentration of CO2 in the greenhouse atmosphere and CO2 generated by the greenhouse heating system was empirically determined by two functions.  The first function was for the months March to September where all of the CO2 generated from natural gas for heating would be used in the greenhouse. The second function was for October to February, the months where more CO2 was generated for heating than was needed for enrichment of the greenhouse atmosphere with CO2.  Data not included in the analysis, as plants were either not present or very young, were for the months of January, February and December of 2002 and December of 2003.  The March 2002 data were also excluded from the analysis because there was an atypical relationship between CO2 dosing and CO2 generation.  This was caused by higher than normal heating to accelerate the development of the late planting of the crop and as the crop was young the requirements for CO2 were low.  25 A third function was fitted to the CO2 generated from natural gas for March to September.  This function describes the maximum amount of CO2 for enrichment generated during heat generation.  The assumptions are:  all the CO2 generated from heating was used in the greenhouse and all the natural gas used was necessary for heat production.  That is, no heat was sacrificed for CO2 enrichment.  The grower verified this assumption. 2.2.4 Costs of CO2 The costs of natural gas and liquid CO2 were obtained from the grower for 2002 and 2003.  Natural gas was priced in Canadian dollars per GJ but converted to dollars per kg of CO2.  The cost of CO2 from natural gas was the same for both years being $0.114 per kg CO2 ($5.37 per GJ of natural gas) from November to March and $0.101 per kg CO2 ($4.77 per GJ of natural gas) from April to September.  In 2002 liquid CO2 was $0.17 per kg and in 2003 $0.21 per kg. 2.2.5 Price structure for fruit produced The weekly wholesale price of greenhouse-grown tomato fruit from 1998 to 2004 for Vancouver, BC was obtained from Agriculture and Agri-Food Canada Online service (InfoHort 2006).  These prices were in Canadian dollars per pound and were converted to price per kg.  These data were then plotted against the week of the year to determine the general trend in tomato prices.  The price structure was determined by multivariate regression using SAS software version 8.2 (SAS Institute 2003).  In the analysis:  week, week squared and each year from 1998 to 2004 were used as variates in the model.  As well, an upward “tooth” in the price consistently occurred between week 29 and 31, so this variate was also added to the model.  26 2.2.6 CO2 enrichment scenarios After consultation with a commercial grower, three CO2 enrichment scenarios were selected and tested.  These scenarios were based on designating a CO2 concentration to match the season and several light levels.  The CO2 concentrations were based on the grower’s current practise, as well as low and high CO2 usages.  The specifics of each scenario are presented in Appendix 2.1.  Using SAS version 8.2 (SAS Institute 2003), the scenarios were programmed and the resulting weekly yield determined.  Using the relationship between CO2 generated and the mean CO2 in the greenhouse (developed in section 2.2.3) the amount of CO2 needed to achieve that greenhouse concentration was determined for March to September.  Using the relation developed from monthly total CO2 generated from natural gas (also section 2.2.3), the amount of CO2 for each month above the amount generated for heat production was determined.  The seasonal cost of the CO2, above that for heat production, could then be determined using either natural gas combustion or liquid CO2.  The weekly revenue from the fruit was determined from the price function (section 2.2.5) and summed for the season.  Net revenues from CO2 enrichment for each scenario were determined for both natural gas and liquid CO2 by subtracting the CO2 dosing cost from the fruit revenue.  Analysis was conducted between the net revenue (fruit revenue- minus cost of CO2) and the cost of CO2 for each scenario. 2.3 RESULTS 2.3.1 Photosynthesis and yield models Modified Acock et al. canopy photosynthesis model.  Model output for a tomato canopy under the range of light and CO2 conditions found in a commercial greenhouse is illustrated in Figure 2.2.  Unlike leaf photosynthesis, canopy photosynthesis does not exhibit saturation at high levels of light and CO2.  Modeled canopy photosynthesis was primarily  27  Figure 2.2 The response of the Acock et al. model for canopy photosynthetic rate to light (Panel A) and CO2 concentration (Panel B).  See the text for a full description of the model used. 0 2 4 6 8 10 12 14 16 18 20 400 500 600 700 800 900 1000 1100 1200 CO2  (ppm) Ca no py  n et  p ho to sy nt he sis  (g  C O 2 m -2  h r-1 ) 100 200 300 400 500 600 700 800 900 1000 External global radiation (W m-2) -2 0 2 4 6 8 10 12 14 16 18 20 0 100 200 300 400 500 600 700 800 900 1000 External global radiation (W m-2) C an op y ne t p ho to sy nt he si s ( g C O 2  m -2  h r-1 ) 400 500 600 700 800 900 1000 1100 1200 CO2 (ppm) A B  28 affected by light level and secondarily by CO2 concentration (Figure 2.2).  Thus at low light levels the photosynthetic response to CO2 concentration is very low; however, as light level increases the response to CO2 increases (Figure 2.2B).  As the CO2 concentration increases the canopy photosynthesis also increases, but the magnitude of the response diminished with each additional unit of added CO2 (Figure 2.2A).  This is illustrated in Figure 2.2A, where the canopy net photosynthesis between adjacent CO2 concentration lines at 1000 W m-2 declines as the CO2 concentration increases. The model output of canopy photosynthesis at several CO2 concentrations was plotted for a sunny day in June (Figure 2.3A).  Net canopy photosynthesis was responsive to increased CO2 between 7 am and 6:30 pm –all but 3.5 hours of daylight.  The positive affect of CO2 enrichment on photosynthesis is clearly illustrated by determining the total amount of CO2 fixed during a day.  Enrichment with CO2 above the ambient atmospheric level of 370 ppm increased carbon fixed per day by 6, 29, 44 and 53% for 400, 600, 800 and 1000 ppm, respectively (Figure 2.3B). Modeling yield.  Patterns of weekly fruit yield, modeled from the amount of carbon assimilated were compared with the fruit harvested by the grower over the same time period (Figure 2.4).  At times, the model over and underestimated the actual yield but generally gave a good approximation of the measured yield for both years.  For the entire harvest, the model overestimated the measured fruit yield by 8% in 2002 (64.9 versus 59.7 kg m-2) and 10.6% in 2003 (71.2 versus 63.6 kg m-2). A more formal assessment of the goodness-of-fit of the model was made using difference-based evaluations as per the recommendation of Donatelli et al. (2004). Following the recommendations of Kobayashi & Salam (2000) the mean square deviation  29  Figure 2.3 The diurnal net canopy photosynthesis at several CO2 concentrations using the Acock et al. model, June 15, 2003.  Panel A, hourly canopy net photosynthesis and global radiation.  Panel B, the cumulative daily amount of CO2 assimilated for several CO2 concentrations and the percentage increase above the ambient level of 370 ppm for each level of CO2. 0 2 4 6 8 10 12 14 16 1 4 7 10 13 16 19 22 Time of Day (hour) C an op y ne t p ho to sy nt he si s (g  C O 2 m -2  h r-1 ) 0 100 200 300 400 500 600 700 800 900 1000 G lo ba l r ad ia tio n (W  m -2 ) 370 400 600 800 1000 Radiation 0.0 20.0 40.0 60.0 80.0 100.0 120.0 140.0 370 400 600 800 1000 CO2 (ppm) C O2  a ss im ila te d (g  m -2 ) 0 10 20 30 40 50 60 70 80 90 100 Pe rc en ta ge  in cr ea se Assimilated Percentage CO2 (ppm) A B  30  Figure 2.4 Measured and the model predicted fruit yield from a commercial beefsteak tomato greenhouse.  Panel A, 2002 and Panel B, 2003. The predicted yield was calculated from the greenhouse plant and environment data using the Acock et al.- Aikman models for photosynthesis and fruit growth. 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 10 15 20 25 30 35 40 45 50 Fr ui t y ie ld  (k g m-2 ) 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 10 15 20 25 30 35 40 45 50 Week Fr ui t y ie ld  (k g m-2 ) Measured Model A B  31  (MSD) and its components squared bias (SB), squared distance between standard deviations (SDSD) and lack of correlation weighted by standard deviations (LCS) were determined (Table 2.1).  The MSD was about 15% higher in 2003 then 2002, which was mainly caused by a higher SDSD (Table 2.1).  A higher SDSD indicates the standard deviation of the model and the observed values are not similar (Gauch Jr. et al. 2003).  Overwhelmingly the greatest contributor to MSD was the LCS with the SB and SDSD being smaller components (Table 2.1).  A high LCS indicates that the model is having problems simulating the pattern of variation in the actual measurements (Kobayashi 2004).  A further goodness of fit test and useful statistic for comparing results to other studies is the relative root mean square error (RRMSE).  The RRMSE was similar between years being 0.245 and 0.295 for 2002 and 2003, respectively (Table 2.1). 2.3.2 CO2 generation and greenhouse CO2 usage The annual generation of CO2 in 2003 (a more typical tomato production year) was 125.5 kg per m2 of greenhouse area with the plants fixing about 17.4 kg m-2 or 14% of the generated CO2 (Figures 2.5 and 2.6).  Throughout the year the main source of CO2 for dosing was from the combustion of natural gas used for heat production (Figure 2.5).  In the summer months, when less heating was required, liquid CO2 was used to supplement the CO2 supply (Figure 2.5).  The ratio of CO2 used by the crop versus total CO2 generated exhibited a parabolic-like shape during the year (Figure 2.6).  The ratio increased from about 5% in January to a maximum of 40% in July and then decreased from August to December.  From January to February and from October to December heating requirements were high and ventilation and crop requirements were low; thus, not all the CO2 generated was needed for CO2 enrichment.  From March to September heating requirements were low, ventilation  32 Table 2.1 Model diagnostics for comparison of grower measured and modelled tomato (cv Rapsodie) fruit yield for the 2002 and 2003 growing seasons.  2002 2003 Mean Square Deviation (MSD)1 0.300 0.357 Squared Bias (SB) 0.030 0.027 SDSD2 0.001 0.025 LCS3 0.269 0.291 Relative Root Mean Square Error 0.245 0.295 1. MSD=SB+SDSD+LCS. 2. Standard deviations of measured and modelled data. 3. Lack of correlation weighted by standard deviations. N=30.  See text for full description.  33 Figure 2.5 The contributions of sources of CO2 generation to the total amount of CO2 produced for each month in the production cycle of tomatoes.  Plant uptake (Plt Upt) through photosynthesis is indicated as negative.  NG is CO2 from the combustion of natural gas and Liquid is liquid CO2.  Panel A is the 2002 season and Panel B is 2003 season.  The gap for Plt Upt in February 2002 was caused by missing greenhouse environmental data. -3.0 0.0 3.0 6.0 9.0 12.0 15.0 18.0 Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Month C O2  g en er at ed  (k g m- 2 ) NG LCO2 Plt Use Series3 -3.0 0.0 3.0 6.0 9.0 12.0 15.0 18.0 Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Month C O2  g en er at ed  (k g m- 2 ) NG Liquid Total Plt Upt A B  34  Figure 2.6 The ratio of CO2 used by the plant through photosynthesis to total CO2 generated by the greenhouse for the 2002 and 2003 growing seasons.  Canopy photosynthesis was determined using the Acock et al. model.  The gap for February 2002 was caused by missing greenhouse environment data. 0 10 20 30 40 50 60 70 80 90 100 Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Month C O2  u se d by  p la nt  (% ) 2002 2003  35 requirements and crop photosynthesis were higher, resulting in all of the CO2 generated being used for the CO2 dosing requirements of the greenhouse atmosphere.  This assumption, that all the generated CO2 was directed into the greenhouse allowed the development of a linear function to predict the mean CO2 concentration in the greenhouse from the CO2 produced from natural gas and liquid CO2 specifically for these months (Figure 2.7).  This function indicated that there was a direct relationship between the CO2 generated through heating and supplied through liquid CO2 and the concentration of CO2 in the greenhouse atmosphere (Figure 2.7). A quartic function was fitted to the CO2 generated from natural gas for the months of March to September (Figure 2.8).  The shape of this function is caused by the steady decline in heating from March to June as the days became longer, after which heating increased as the days shortened (Figure 2.8).  This function indicates the maximum amount of CO2 generated from heat production in the summer.  The underlying assumption is that the heat is needed for greenhouse production and no heat is generated for the sake of CO2 enrichment. 2.3.3 Seasonal price of tomatoes The price obtained for tomato fruit ($ Canadian per kg) exhibited a quadratic response with the week of the year.  Peak prices occurred early in the season and the lowest prices in mid-summer (Figure 2.9).  Multivariate analysis indicated that significant variates (p<0.0005) were week, week squared and years 1998 to 2002 (Figure 2.9).  As well, plots of the raw data indicated there was an upward “tooth” in the price between week 29 and 31. This variate was significant (p=0.0001) and was added to the model (Figure 2.9).  The following function was determined from the multivariate regression:  36 Figure 2.7 The relationship between CO2 generated and the mean CO2 concentration in the greenhouse for the months of March to September.  The regression equation is Y = 76.29X - 25.88, R2=0.82.  NG+LG is CO2 from natural gas and liquid CO2 and NG is CO2 from natural gas.  Only data from months where all generated CO2 was directed into the greenhouse are shown. 500 550 600 650 700 750 800 850 900 950 1000 7.0 8.0 9.0 10.0 11.0 12.0 13.0 CO2 generated (kg m-2) C O2  in  g re en ho us e (p pm ) 2002 NG+L NG 2003 NG+L NG  37 Figure 2.8 The monthly total CO2 generated from natural gas used for heating, March to September, 2002 and 2003.  The line equation is Y = -35.10 + 38.47X -10.83X2 + 1.22X3 - 0.48X4, R2=0.95. 0.0 2.0 4.0 6.0 8.0 10.0 12.0 14.0 1 2 3 4 5 6 7 8 9 10 11 12 Month CO 2  f ro m  n at ur al  g as  (k g m -2 ) 2002 2003 Jan      Feb     Mar     Apr    May    Jun      Jul     Aug     Sep      Oct     Nov   Dec  38 0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 0 4 8 12 16 20 24 28 32 36 40 44 48 52 Week Pr ic e ($  k g- 1 ) Obs Model  Figure 2.9 The weekly price per kg of beefsteak tomato fruit at the Vancouver food terminal from 1998 to 2004.  The equation of the line is Y = 5.31 - 0.19X + 0.00271X2 and is the weighted mean response with no adjustment for years.  Weights were based on the number of observations per year.  Equation R2=0.50 and N=331.  39 Price = 4.36 -0.19 X + 0.0027X2 + 0.68T + 1.2y1998 +0.81y1999 + 0.66y2000 + 1.05y2001 + 1.33y2002, R2=0.50.  With X and X2 being week and week-squared, respectively, T being the tooth at week 29 to 31 and y being year (y1999 was year 1999 and so on).  The function was simplified to a weighted mean quadratic response Price = 5.31 +-0.19X + 0.00271X2 with X and X2 being week and week-squared, respectively.  The intercept of this function was adjusted upwards by weighting (by the number of observations per year) and scaled to account for the coefficients not used (y1998 to y2001). 2.3.4 Cost and revenue of CO2 enrichment scenarios Several CO2 enrichment scenarios were assessed after discussions with the local growers (Appendix 2.1).  The monthly average CO2 concentrations for each scenario are given in Figure 2.10, and the amount of CO2 actually used in the greenhouse for the 2003 season was included for comparison.  For all scenarios, CO2 levels were lowest during the summer months when less CO2 is available from heating and venting is highest. The net revenue for the CO2 enrichment scenarios was investigated.  For this analysis net revenue was the price of the fruit minus the cost of CO2; it did not take into account all costs of the greenhouse operation, such as costs of the labour, transportation, marketing etc. At the current prices for CO2 the greatest net revenue was for the High CO2 enrichment scenario (Table 2.2).  The greatest amount of CO2 was needed but the revenue from the additional kg of fruit offset this expense.  However, when the cost of CO2 is increased the net revenue of the High CO2 scenario, and eventually the Intermediate scenario become lower (Figure 2.11).  As evident in Figure 2.11, once CO2 costs more than $0.50 per kg ($23.25 per GJ – natural gas) the Low CO2 usage scenario exhibited higher net revenue (Figure 2.11).  40   Figure 2.10 The mean monthly CO2 concentration for a commercial greenhouse and the mean monthly High, Low and Intermediate scenarios of CO2 enrichment.  The specific CO2 conditions for each scenario can be found in Appendix 2.1 300 400 500 600 700 800 900 1000 1100 1200 1300 Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Month M ea n C O2  in  g re en ho us e (p pm ) Com. GH Low Intermediate High  41 Table 2.2 The revenue from several CO2 enrichment scenarios for tomato production in British Columbia. Yield1 CO2 needed2 Cost of CO2 Enrichment3 Net Revenue4 Liquid NG Liquid NG  CO2 Scenario5  kg m-2  $ m-2  kg m-2 $ m-2 $ m-2 $ m-2 $ m-2 Com GH 72.05 100.74 34.52 7.25 3.55 93.49 97.19 Low  62.40 86.76 12.26 2.47 1.24 84.29 85.52 Intermediate 69.36 97.33 31.87 6.69 3.28 90.64 94.05 High  77.36 107.98 53.71 11.28 5.53 96.70 102.45 1. Yield was calculated from week 10 to 46, using the Acock et al. and Aikman models. 2. CO2 above the amount needed for heat production, 3. In 2002 and 2003 the cost of liquid CO2 was $0.21 per kg, cost of CO2 from natural gas was $0.114 per kg for March and $0.101 per kg, for April to September, 4. Net revenue is revenue from yield minus respective cost of CO2, 5. See Figure 2.11 for Low, Intermediate and High CO2 regimes.  42 0.00 10.00 20.00 30.00 40.00 50.00 60.00 70.00 80.00 90.00 100.00 110.00 120.00 0.00 0.10 0.20 0.30 0.40 0.50 0.60 0.70 0.80 0.90 1.00 Cost of CO2 ($ kg-2) N et  re ve nu e ($  m -2 ) Low Intermediate High Mean yearly cost of CO2 from natural gas Cost of liquid CO2  Figure 2.11 The seasonal net revenue of three CO2 enrichment scenarios for an increasing cost of CO2 in a commercial greenhouse.  The 2002/2003 costs of CO2 from natural gas and liquid CO2 are indicated.  The specific CO2 condition for each treatment can be found in Figure 2.10.  43 2.4 DISCUSSION This study has attempted to use plant-based simulation models to evaluate some relationships between CO2 dosing and greenhouse profitability. If successful, such an approach would aid growers in decision-making as they manage CO2 dosing in their greenhouses.  The modeling approach was investigated because of the complexities of CO2 enrichment, where economic thresholds are often reached before the maximum plant response is attained.  As well, optimizing CO2 dosing is difficult because photosynthesis responds to short-term variations in the greenhouse environment, while the revenues from plant and fruit growth are not realized until 6-8 weeks later for a fruit or 9 months for the duration of the crop (Aikman 1996).  Chalabi et al. (2002b) further expounded that it is not feasible to establish the optimal CO2 concentration experimentally because there are so many interrelated factors, such as the variation in the effectiveness of CO2 enrichment with light levels, greenhouse ventilation rates, the cost of CO2 and the expected price of the fruit The models investigated here were mechanistic for photosynthesis and carbon partitioning, and empirical for those linking CO2 generation to greenhouse atmospheric CO2 and seasonal fruit price.  Ideally all the models would be mechanistic, as this would better reflect underlying functional processes.  However, this is not always possible and should not be a barrier to providing a useful approach (Alscher et al. 2001a; 2001b; 2001c).  The simulation approach is advantageous because “what if” scenarios can be developed and evaluated.  These scenarios could include when to dose with CO2 and could also be used for choosing a CO2 source.  44 2.4.1 Modeling greenhouse tomato fruit yield In selecting the plant-based models, a main consideration was to find a model that a grower could use, with model inputs that would be easily available to a grower.  The simplest approach found was the method of Bailey & Chalabi (1994), Bailey et al. (1997) and Chalabi et al. (2002a; 2002b) where models for canopy photosynthesis and fruit partitioning were used.  Their models were based on the Acock et al. (1978) model modified by Nederhoff (1994) for canopy photosynthesis.  Many yield models have a photosynthesis sub-model. For example, TOMGRO (Jones et al. 1999) is a popular research model that also uses the Acock et al. (1978) model for canopy photosynthesis.  The Acock et al. (1978) model itself uses just three inputs:  leaf area index (LAI) in m2 m-2, CO2 concentration in ppm and PAR in W m-2 with several constants for tomato.  The partitioning model used was by Aikman (1996), who used literature values to estimate the efficiency of conversion of glucose to plant biomass, carbon lost to respiration and carbon allocated to fruit (harvest index) in tomato.  The simplicity of these models contrasts with that of the TOMGRO models, where version 1.0 uses 69 state variables, version 3.0, uses 574 and the compact version uses five state variables (Jones et al. 1999; Ramirez et al. 2004). Chalabi et al. (2002a; 2002b) strictly used their models for simulation with no apparent validation.  The reliability of simulations depends on the goodness of the models used for predicting photosynthesis and the partitioning of carbon to the fruit.  Validation of canopy photosynthesis is troublesome because it is so difficult to measure.  Chalabi & Fernandez (1994), Nederhoff & Vegter (1994) and Zekki et al. (1999) used a mass balance method to validate the Acock et al. canopy photosynthesis model for crops grown in a research greenhouse.  In their approach the greenhouse structure is treated as a cuvette where the CO2 influx is measured against the CO2 outflux (Chalabi & Fernandez 1994; Nederhoff  45 & Vegter 1994a; Zekki et al. 1999).  However, this approach has problems, because CO2 losses caused by greenhouse leakage and venting can impair the assessment of CO2 uptake by plants through photosynthesis. As Nederhoff (1994) has indicated, one is left wondering which method is less susceptible to error: the model or the mass balance approach? As it was not feasible to measure canopy photosynthesis in a commercial greenhouse, model validation was carried out using the harvested fruit.  At that level it cannot be known which models (canopy photosynthesis or carbon partitioning) are responsible for errors in estimation.  The general impression is that the present model output appeared to give a reasonable estimate of the seasonal fruit production in a commercial tomato greenhouse (Figure 2.4).  However, closer inspection of these data revealed the model overestimated the yield by about 10% for the season (Figure 2.4).  In general terms, in northern locations a one percent increase in light increases fruit yield by one percent (Heuvelink & Dorais 2005).  In equation 2.2, I used 0.5 as the conversion between global radiation and PAR (Monteith 1973; Jones 1992).  In later work, Monteith & Unsworth (1990) have provided a range of 0.4 to 0.5 for the global radiation to PAR conversion.  Had I used the midpoint value of 0.45 in equation 2.2, the modeled yield would have decreased by 10% and for the production season would have been similar to the measured yield. Following the advice of Kobayashi & Salam (2000) a more detailed analysis of the causes of the variation was carried out (Table 2.1).  These data indicated that the lack of correlation weighted by standard deviations (LCS) was the major contributor to the mean squared deviation (MSD).  What LCS indicates is that the model is having problems simulating the pattern of variation in the measurements and inspection of the plots reveals this (Figure 2.4).  On the other hand, the RRMSE (0.2 in Table 2.1) is similar to that reported  46 in other studies, with values of 0.17 to 0.50 for various fruit models (Gibert et al. 2005) and from 0.004 to 0.492 for peach (Quilot et al. 2002). The inability of the model to replicate the pattern of fruit yield is a shortcoming of this approach likely caused by the model’s simplicity.  In Aikman’s original work he specified additional empirical models to map current photosynthate to future fruit growth using a thermal scale and a Gompertz function for fruit relative growth rate (Aikman 1996). Incorporating these models would be impractical, however.  A tomato canopy has 7 to 9 trusses at various stages of development with 4 to 5 fruit per truss also at different developmental stages.  Hence, potentially the first and last fruit on consecutive trusses could be at roughly the same stage or even ahead in development.  If Chalabi et al. (2002a; 2002b) had incorporated Aikman’s models they did not include explicit details on how it was carried out.  Most likely, a more involved model such as TOMGRO would improve weekly yield estimates. Compared to the complexity of the TOMGRO model it is very surprising that the simple and straightforward approach used here was as effective as it was.  The models used here treated the canopy photosynthesis as a short-term dynamic response, with the light, CO2 and LAI for a week being responsible for the fruit yield for that week.  In reality the maximum fruit growth rate peaks around day 21 after anthesis and the growth rate of a fruit is low for the 7 days prior to harvest (Ho et al. 1982; Aikman 1996).  Possibly, a cancellation of opposite errors has helped the models to bridge short term (photosynthesis) and longer- term (fruit yield) performance.  Another possibility follows from the fact that the models are mainly driven by light level, and light available for any given week tends to be positively  47 correlated with the light conditions of a previous week.  Thus, the model outcomes are generally following the seasonal pattern of light level. 2.4.2 CO2 generation and usage in greenhouse Current greenhouse technology strives to separate daily temperature control and CO2 dosing, so that CO2 can be dosed during the day and some daytime heat production is stored for later use.  This has been found to be the most economical means of CO2 generation (Nederhoff 2004).  However, in the larger (seasonal) sense, the amount of CO2 generated by this greenhouse is not synchronized with the CO2 needs of the crop.  The CO2 generation profile from a typical commercial greenhouse follows the crop production cycle and its demand for heat (Figure 2.5).  Generation of CO2 is highest in the winter, when the needs for heat are high but the crop is small and the light levels are low.  Thus, the plants do not have the capacity to assimilate all the CO2 that is generated (Figure 2.5).  Additionally, the need to ventilate in winter is low, so only a small amount of CO2 is needed to maintain the greenhouse atmosphere at levels as high as 1000 ppm.  Therefore, much of the CO2 generated exits the greenhouse without being applied to the crop.  By the late spring and summer (March to September) the crop is more fully developed, is producing fruit, and light levels are high.  Accordingly, the crop has a greater need for, and response to, CO2 enrichment.  The disconnect arises that with increasing light levels greenhouse temperatures are higher.  Consequently not as much heat is needed, which reduces CO2 generation and there is a greater need for ventilation.  Employing liquid CO2 or dissipating heat to extend the capacity of the heat buffer can counteract the lower rates of CO2 generation in the summer.  It is difficult to reduce ventilation without adversely affecting the crop.  Dieleman et al. (2005) suggested that some loss of quality caused by higher daytime temperatures  48 (caused by reduced venting) may be offset by savings in lower nighttime heating costs (and thus maintenance of similar 24 hour temperature = higher day and lower night temperature). Despite losses of CO2 to venting, the efficiency of CO2 usage is higher in the summer as less CO2 is generated and crop photosynthesis is higher (Figure 2.5).  From April to August the crop utilized up to 40% of the generated CO2, while for the entire production season CO2 utilization was 14% (Figure 2.6).  It follows that CO2 release from greenhouses to the surrounding environment could be reduced if reasonably priced technology was available to store CO2 from the flue gases in the winter for later use in the summer.  At the very least, liquid CO2 would not be needed for greenhouses that use this technology.  Additionally, carbon sequestration in the crop would be increased because the conditions for optimum CO2 assimilation are present. One of the major hurdles during this work was the difficulty in linking CO2 input (from greenhouse heating and liquid CO2) to the CO2 concentration detected in the greenhouse atmosphere.  Chalabi et al. (2002a; 2002b) used a mechanistic model to describe the energy balance of a greenhouse and thus was able to estimate CO2 generation and CO2 lost through venting and leakage.  However it was not possible to use that model, as the required information (e.g. rate at which natural gas was burned or rate at which CO2 was generated, ventilation rate and leakage of greenhouse structure to CO2 with vents closed) was not available.  As well, natural gas usage and liquid CO2 data were only available on a monthly basis.  More precise work could be done if daily or hourly generation of CO2 was known.  In my work an empirical model was used to link the CO2 generated to the concentration of CO2 in the greenhouse structure (Figure 2.7).  Use of this function was restricted to the months between March and September (inclusive), the time when according  49 to the grower all the CO2 generated is directed into the greenhouse.  For the remaining months only a portion of the CO2 generated is used in the greenhouse and there was not enough information available to determine what portion was used for enrichment and what was emitted.  A straight line fit these March to September data reasonably well (R2=0.82); indicating there is a strong cause and effect between CO2 generated and mean monthly CO2 concentration in the greenhouse (Figure 2.7).  Had the plants assimilated a large amount of CO2 I would have anticipated a curvilinear relationship, with the line sweeping up as the amount of CO2 generated increased.  This is because, as indicated in Figure 2.2B, canopy net photosynthesis exhibits a saturation-type response as CO2 concentration increases.  A curvilinear response was not found as the amount of CO2 absorbed by the plants was very small compared to what exited the greenhouse through ventilation. A second empirical model was used for the same time span to determine the amount of CO2 that was generated from natural gas combustion.  The assumption here (verified by the grower) was that all the heat produced was needed and none was wasted for the sake of CO2 enrichment.  Therefore, this function describes the maximum amount of CO2 that is attainable or is free during heat generation.  This function is dependent on the local weather data and the size of the heat buffer. 2.4.3 CO2 enrichment scenarios Simulations of low, intermediate and high CO2 application were developed in consultation with a local grower.  Using current costs of CO2 enrichment ($0.10 per kg) and fruit prices (Figure 2.9), a high CO2 dosing regime, from natural gas combustion and liquid CO2, resulted in the highest revenue (Table 2.2).  However when the price of fruit is fixed and the cost of CO2 enrichment increases, the profitability of these scenarios re-rank  50 themselves (Figure 2.11).  When the cost of CO2 rises to $0.50 per kg, the net revenue from the low and high CO2 application become similar (Figure 2.11).  Under this scenario, if the generated CO2 was solely a by-product from natural gas combustion for heating, then the cost of natural gas would increase from $4.77 per GJ to $23.85 per GJ.  This increase would make natural gas prohibitively expensive for greenhouse usage under current fruit revenues. However, if the growers were using a low cost, impurity rich fuel, such as waste wood, and using liquid CO2, then a cost of $0.50 per kg is 2.4 times higher then the current price of liquid CO2 ($0.21 per kg). 2.5 CONCLUSIONS The goal of this chapter was to use a plant-based simulation model of tomato production, in conjunction with environmental data from commercial greenhouses, to explore the costs and benefits of several CO2 enrichment scenarios for greenhouse tomato production.  In the first part of this chapter, weekly tomato fruit yield was modeled using the Acock et al. and Aikman models and compared to the grower recorded yield.  In the second part, empirical linkages were established between the CO2 produced by the greenhouse and the amount of CO2 in the greenhouse atmosphere.  The third part was the development of an empirical model for costs associated with CO2 enrichment and revenue derived from fruit yield.  In the final part of this work the above models were used to simulate the fruit yield and CO2 costs under several CO2 enrichment scenarios and explored changes in revenue. The conclusions of this chapter are: 1. With current costs of CO2 of about $0.10 per kg and expected fruit prices (Figure 2.9), a high CO2 application was found to generate the greatest revenue (Figure 2.11).  51 However, when the cost of CO2 increases from $0.10 to $0.50 per kg the revenue of the low and high CO2 applications were similar with no change in fruit price. 2. Two simple and compact models, Acock et al. and Aikman models were used to estimate seasonal tomato fruit yield.  Validation of these models with grower- collected data indicated they overestimated the yield by only 10%, but had some trouble simulating the week-to-week variation in yield.  As these models only require greenhouse PAR and CO2 concentration and plant LAI, I believe they would not be difficult for a grower to implement. 3. A typical greenhouse in the Lower Mainland of BC produces 125.5 kg per m2 of CO2 per production season and the crop uses 17.4 kg per m2, 14% of the produced CO2. During the summer, CO2 usage by the plant increased to 25% (average) caused by increased photosynthesis and decreased CO2 generation through reduced heating. 4. On a seasonal basis CO2 generated is largely out of sync with CO2 needs of the crop. The majority of CO2 is produced as a byproduct of heat generation, with the greatest need for heat in the winter when CO2 utilization by the crop is low.  The crop can best utilize CO2 enrichment in the summer but CO2 generation is less, as heating requirements are less. 5. The lack of direct and detailed information on CO2 injection into the greenhouse necessitated the development of empirical models.  A direct linear relationship between CO2 generated and the mean concentration of CO2 in the greenhouse between March and September was found.  Over this period CO2 usage by the plant was low compared to CO2 lost through ventilation.  52 6. This approach could be substantially strengthened with an improved yield model, more detailed CO2 generation data, validation and tuning under commercial conditions. 7. The approach investigated here is useful for the broader aspects of CO2 management, its ineffectiveness at describing the short-term patterns limits its ability to be used in day-to day fine tuning of CO2 dosing.  53 2.6 LITERATURE CITED  Aaslyng, J.M., Andreassen, A., Korner, O., Lund, J., Jakobsen, L., Pedersen, J., Ottosen, C., and Rosenqvist, E. 2006. Integrated optimization of temperature, CO2 screen use and artificial lighting in greenhouse crops using a photosynthesis model. Acta Horticulturae 711: 79-88. Aaslyng, J.M., Ehler, N., Karlsen, P., and Rosenqvist, E. 1999. Intelligrow:  A component- based climate control system for decreasing greenhouse energy consumption. Acta Horticulturae 507: 35-40. Acock, B., Charles-Edwards, D.A., Fitter, D.J., Hand, D.W., Ludwig, L.J., Warren Wilson, J., and Withers, A.C. 1978. The contribution of leaves from different levels within a tomato crop to canopy net photosynthesis:  An experimental examination of two canopy models. Journal of Experimental Botany 29: 815-827. Aikman, D.P. 1996. A procedure for optimizing carbon dioxide enrichment of a glasshouse. Journal of Agricultural Engineering and Research 63: 171-184. Alscher, G., Krug, H., and Liebig, H. 2001a. Optimisation of CO2 and temperature control in greenhouse crops by means of growth models at different abstraction levels.  I. Control strategies, growth models and input data. Gartenbauwissenschaft 66: 105- 114. Alscher, G., Krug, H., and Liebig, H. 2001b. Optimisation of CO2 and temperature control in greenhouse crops by means of growth models at different abstraction levels.  II. Growth models and parameter generation for lettuce crops. Gartenbauwissenschaft 66: 153-163. Alscher, G., Krug, H., and Liebig, H. 2001c. Optimisation of CO2 and temperature control in greenhouse crops by means of growth models at different abstraction levels.  III. Simulation and optimisation with the combined models. Gartenbauwissenschaft 66: 213-218. Anonymous. 2006a.  Canadian Natural Gas Review of 2004 & Outlook to 2020.  Ottawa, ON, Canada, Natural Resources Canada. Anonymous. 2006b.  Canada's Greenhouse Gas Inventory 1990-2003. Bailey, B.J. 2002. Optimal control of carbon dioxide enrichment in tomato greenhouses. Acta Horticulturae 578: 63-69. Bailey, B.J. and Chalabi, Z.S. 1994. Improving the cost effectiveness of greenhouse climate control. Computers and Electronics in Agriculture 10: 203-214. Bailey, B.J., Chalabi, Z.S., Aikman, D.P., and Cockshull, K.E. 1997. Improved strategies for controlling CO2 enrichment in tomato greenhouses. Acta Horticulturae 443:  155-  54 162. Chalabi, Z.S., Biro, A., Bailey, B.J., Aikman, D.P., and Cockshull, K.E. 2002a. Optimal control strategies for carbon dioxide enrichment in greenhouse tomato crops - Part II. Using the exhaust gases of natural gas fired boilers. Biosystems Engineering 81: 323- 332. Chalabi, Z.S., Biro, A., Bailey, B.J., Aikman, D.P., and Cockshull, K.E. 2002b. Optimal control strategies for carbon dioxide enrichment in greenhouse tomato crops - Part 1: Using pure carbon dioxide. Biosystems Engineering 81: 421-431. Chalabi, Z.S. and Fernandez, J.E. 1994. Estimation of net photosynthesis of a greenhouse canopy using a mass balance method and mechanistic models. Agricultural and Forest Meteorology 71: 165-182. Challa, H. 2002. Crop models for greenhouse production systems. Acta Horticulturae 593: 47-53. Challa, H. and Schapendonk, A.H.C.M. 1986. Dynamic optimalization of CO2 concentration in relation to climate control in greenhouses. In  Enoch, H.Z. and Kimball, B., eds. Carbon Dioxide enrichment of Greenhouse Crops. I Status and CO2 Souces. CRC Press, Boca Raton, Florida, pp 147-160. Dieleman, J.A., Meinen, E., Marcelis, L.F.M., de Zwart, H.F., and van Henten, E.J. 2005. Optimisation of CO2 and temperature in terms of crop growth and energy use. Acta Horticulturae 691: 149-154. Donatelli, M., Acutis, M., Bellocchi, G., and Fila, G. 2004. Statistics:  New indices to quantify patterns of residuals produced by model estimates. Agronomy Journal 96: 631-645. Ehret, D.L., Lau, A., Bittman, S., Lin, W., and Shelford, T. 2001. Automated monitoring of greenhouse crops. Agronomie 21: 403-414. Ewert, R. 2004. Modelling plant responses to elevated CO2:  How important is leaf area index. Annals of Botany 93: 619-627. Excel. 1985-1999. Microsoft Corporation.  Redmond, WA, Microsoft Corporation. Gary, C., Jones, J.W., and Tchamitchian, M. 1998. Crop modelling in horticulture:  state of the art. Scientia Horticulturae 74: 3-20. Gauch Jr., H.G., Hwang, J.T.G., and Fick, G.W. 2003. Model evaluation by comparison to model-based predictions and measured values. Agronomy Journal 95: 1442-1446. Gibert, C., Lescourret, F., Genard, M., Vercambre, G., and Perez Pastor, A. 2005. Modelling the effect of fruit growth on the surface conductance to water vapour diffusion. Annals of Botany 95: 673-683.  55 Hanan, J.J. 1998. Greenhouses.  Advanced Technology for Protected Horticulture. CRC Press , Boca Raton. Hand, D.W. 1984. Crop responses to winter and summer CO2 enrichment. Acta Horticulturae 162: 45-63. Heuvelink, E., Bakker, M.J., Elings, A., Kaarsemaker, R., and Marcelis, L.F.M. 2005. Effect of leaf area on tomato yield. Acta Horticulturae 691: 43-50. Heuvelink, E. and Dorais, M. 2005. Crop growth and yield. In Heuvelink, E., ed. Tomatoes. CABI Publishing, Cambridge, MA, pp 85-144. Ho, L.C., Sjut, V., and Hoad, G.V. 1982. The effect of assimilate supply on fruit growth and hormone levels in tomato plants. Plant Growth Regulation 1: 155-171. InfoHort. 2006. Weekly Wholesale to Retail Market Prices.  13/03/2006. Jones, H.G. 1992. Plants and microclimate.  A quantitative approach to environmental plant physiology.  1-428.  Cambridge, UK, Cambridge University Press. Jones, J.W., Kenig, A., and Vallejos, C.E. 1999. Reduced state-variable tomato growth model. Transactions of the ASAE 42: 255-265. Kobayashi, K. 2004. Comments on another way of partitioning mean square deviation proposed by Gauch et al. (2003). Agronomy Journal 96: 1206-1208. Kobayashi, K. and Salam, M.S. 2000. Comparing simulated and measured values using mean squared deviation and its components. Agronomy Journal 92: 345-352. Lentz, W. 1998. Model applications in horticulture:  a review. Scientia Horticulturae 74: 151-174. Linker, R., Seginer, I., and Gutman, P.O. 1998. Optimal CO2 control in a greenhouse modeled with neural networks.  Computers and Electronics in Agriculture 19: 289- 310. Marcelis, L.F.M., Heuvelink, E., and Gourdriaan, J. 1998. Modelling biomass productuion and yield of horticultural crops:   a review. Scientia Horticulturae 74: 83-111. Monteith, J. 1973.  Principles of Enviromental Physics. London, UK, Edward Arnold Ltd., pp 1-243. Monteith, J. and Unsworth, M.H. 1990.  Principles of Environmental Physics. London, Edward Arnold Ltd., pp1-291. Nederhoff, E.M. and Vegter, J.G. 1994. Photosynthesis of stands of tomato, cucumber and sweet pepper measured in greenhouses under various CO2-concentrations. Annals of Botany 73: 353-361.  56 Nederhoff, E.M. 1994. Effects of CO2 concentration on photosynthesis, transpiration and production of greenhouse fruit vegetable crops. Wageningen. Nederhoff, N. 2004. Carbon dioxide enrichment - Fuels and Figures. Practical Hydroponics & Greenhouses May/June: 50-59. Ohyama, K., Kozai, T., Ishigami, Y., and Ochi, Y. 2005. A CO2 control system for a greenhouse with a high ventilation rate. Acta Horticulturae 691: 649-524. Portree, J. Greenhouse Vegetable Production Guide.  1996. Victoria, BC, Extension Systems Branch, BCMAFF. Quilot, B., Genard, M., Kervella, J., and Lescourret, F. 2002. Analysis of genotypic variation in fruit flesh total sugar content via an ecophysiological model applied to peach. Theoretical and Applied Genetics 109: 440-449. Ramirez, A., Rodriguez, F., Benreguel, M., and Heuvelink, E. 2004. Calibration and validation of complex and simplified tomato growth models for control purposes in the southeast of Spain. Acta Horticulturae 654: 147-154. SAS Institute. 1999-2001.  Statistical Analysis Software.  Cary, NC, SAS Institute Inc. Thimijian, R.W. and Heins, R.D. 1983. Photometric, radiometric, and quantum light units of measure:  a review of procedures for interconversion. HortScience 18: 818-822. Zekki, H., Gary, C., Gosselin, A., and Gauthier, L. 1999. Validation of a photosynthesis model through the use of the CO2 balance of a greenhouse tomato canopy. Annals of Botany 84: 591-598.  57 3 Assessing the potential for using non-destructive plant growth measurements in guiding CO2 dosing3,4 3.1 INTRODUCTION Since the late 1800’s growers have recognized that increasing the greenhouse atmosphere with CO2 leads to larger plants, greater yields and faster time to maturity. Greenhouse technology and environmental control have greatly advanced since the 1800’s and growers now have the ability to dose substantial amounts of CO2.  Indeed, greenhouse heating systems can at times produce more CO2 than the crop can use. However, the technology to dose CO2 is not matched by the growers’ ability to determine the carbon needs of the crop.  In fact, the growers could be unwittingly overdosing their plants with CO2, reducing the efficacy of CO2 enrichment, its profitability, and potentially resulting in plant damage.  The measurement of plant growth over time may be a useful tool for gauging the carbon needs of the crop.  To be useful in a commercial greenhouse these measurements would have to be non-destructive and cause a minimum of interference to the plants, or complication for the grower.  The main goal of this chapter is to examine several non-destructive measures of plant growth and assess if they can be used as part of a plant-based approach to improve the management of CO2. 3.1.1 Effects of CO2 enrichment on plant growth In simple terms, elevated CO2 increases the production of assimilates which over time leads to increased plant growth.  More specifically, increasing the concentration of CO2 in the greenhouse atmosphere leads to an increase in the rate of diffusion of CO2 into the photosynthetic tissues.  This increases the availability of the substrate (CO2) for the  3 A version of this chapter will be submitted for publication.  Edwards, D., Jolliffe, P and Ehret, D. Assessing the potential for using non-destructive plant growth measurements in guiding CO2 dosing.  4 A version of this chapter has been published.  Edwards, D., Jolliffe, P., Ehret, D. and Baylis, K.  2008. Towards a plant-based method of CO2 management.  Acta Horticulturae.  797:273-278.  58 enzyme ribulose-1,5-biphosphate carboxylase/oxygenase (RUBISCO), increasing the rate of carboxylation and suppressing photorespiration.  Assimilated carbon is then partitioned to the sinks in proportion to sink strengths and proximity to photosynthetic sources (Heuvelink 1996).  However, as Heuvelink & Dorais (2005) indicate the mechanisms governing the partitioning of assimilates among the different organs are still poorly understood, and increasing the potential for source activity does not always result in increased sink activity. A number of literature reviews (Kramer 1981; Kimball 1983; Cure & Acock 1986; Slack 1986; Mortensen 1987; Bowes 1993; Rogers & Dahlman 1993; Murray 1995; Drake & Gonzàlez-Meler 1997; Centritto et al. 1999; Pritchard et al. 1999; Poorter & Navas 2003) and more recently meta-analyses (Ainsworth et al. 2002; Ainsworth & Long 2005) which have summarized the effects of elevated CO2 on plant growth.  For tomato, these studies reported that enriched CO2 can result in larger, heavier and more numerous fruit, thicker stems, greater stem length, heavier leaves, shorter leaves, longer leaves and better flowering (Atherton & Harris 1986; Picken et al. 1986; Slack 1986; Mortensen 1987; Gruda 2005).  There have also been thorough reviews that have discussed the botany and physiology of the greenhouse tomato plant and the production systems commonly employed (Wittwer & Honma 1979; van de Vooren et al. 1986; Peet & Welles 2005).  While the preponderance of studies have indicated beneficial effects of CO2 enrichment, the practice has not been perfected. Some of the potential positive effects of CO2 enrichment on plant growth are likely unrealized by the current practises of tomato crop management used in greenhouses.  59 Much of the early research on CO2 enrichment in greenhouse tomato production focused on identifying atmospheric concentrations that maximize yield.  Later research such as that by Ho and colleagues (Ho & Shaw 1977; Ho 1978; Ho et al. 1982; Ho et al. 1983; Ho 1995; Ho 1996) focused on source – sink relations and Nederhoff and colleagues (Nederhoff et al. 1992; Nederhoff 1994; Nederhoff & Vegter 1994a; 1994b) focused on some of the side-effects of CO2 enrichment on plant growth.  More recent research, mentioned in Chapter 2, has focused on maximizing the economic returns from CO2 enrichment.  None of these studies have examined tomato plant responses to CO2 with the goal of exploring whether these responses can be used in decision-making for dosing CO2. 3.1.2 Analysis of tomato plant growth using non-destructive measurements Most growers track the weekly growth of their crop using non-destructive measurements.  This helps them to compare current crop performance with previous years’, and helps them to gauge the balance between generative (flowering, fruit setting and fruit growth) and vegetative (stems, leaves and roots) growth.  These measurements may have potential to contribute to decision-making in CO2 dosing. For plant scientists the preferred method to quantify growth in response to elevated CO2 is through destructive measurements.  In a conventional analysis of plant growth, entire plants are removed and partitioned into leaves, flowers, fruit, stem and roots.  The fresh mass and leaf area are often determined and then the tissues are oven dried to determine their dry mass.  Plants are harvested at the beginning and the end of an experiment and often at intervals in between.  With these measures the partitioning of carbon within the plant can be determined, as well as growth rates.  In order to determine  60 the effects of CO2 enrichment on plant growth, control plants growing under ambient CO2 are also needed. A conventional analysis of plant growth is difficult to carry out in a commercial greenhouse.  To the growers, it is unacceptable to remove plants for a destructive analysis, as this would reduce yield and revenue.  Greenhouse researchers also face challenges when using destructive measurements to quantify growth.  Compared to a commercial greenhouse, research greenhouses tend to be limited in availability and are small in size, so can only contain a small number of plants.  Removing plants for analysis alters the population density and thus changes the circumstances of an experiment, although replacement plants can sometimes be used.  Therefore, techniques for a non- destructive analysis would also be useful for greenhouse research.  A further use of non- destructive growth measures would be to fine-tune plant growth models, currently an active field of research as indicated in Chapter 2. Several non-destructive measures are obtainable that may have potential for utilization in a plant-based approach to managing greenhouse CO2 enrichment.  These measures are: stem length, stem diameter, leaf area, fruit load and yield.  Stem length gives an approximation of the growth rate, and stem diameter is currently used by growers to track the vegetative – generative balance.  Leaf area and fruit load are two measures that can be used to quantify source and sink strength, respectively.  Yield is the most important indicator of plant response to CO2 enrichment.  In this work yield is represented by the mass of fruit and by the number of fruit produced per plant.  61 3.1.3 The management of greenhouse tomato growth Leaf pruning and fruit removal are two crop management practises that affect source and sink strengths and thus the need for supplemental carbon.  Greenhouse tomato growers increase the CO2 concentration of the greenhouse atmosphere with the objective of increasing the yield and sometimes the size of marketable fruit.  Fruit size is of particular importance in beef steak tomato production, and the number of fruit allowed to set is restricted by pruning to ensure that the remaining fruit will be sufficiently large and uniform in size to command a premium price.  The desirable size for a beef steak fruit is approximately 200 g.  Removing some of the major sinks for carbon seems at odds with increasing source activity through CO2 enrichment, and removal may in fact reduce CO2 assimilation (Heuvelink & Dorais 2005).  A second reason for fruit pruning is to help maintain a balance between vegetative and generative growth.  Developing fruit are the strongest sinks, and might so strongly attract assimilates that the growth of weaker sinks, such as the roots, stem and leaves, might be restricted.  Such reduction of vegetative growth would hamper future fruit production.  On the other hand, when the plant is too vegetative, fruit production is less than could be supported by the available assimilates from the sources. In addition to fruit pruning, leaf removal also is performed for reasons associated with fruit quality and plant balance.  Lower leaves are removed from the three oldest trusses to facilitate fruit picking.  Also, the growers believe that the lower leaves can contribute to cracking of the fruit cuticle, and lower leaves are thought to make a low contribution to plant carbon gain.  Shading may restrict the photosynthetic capabilities of lower leaves, but excessive leaf pruning would be at odds with attempts to promote carbon gain by increasing the CO2 in the greenhouse atmosphere.  62 3.1.4 Objectives of chapter Research undertaken in this chapter will follow the growth of plants raised in ambient (unenriched) CO2 conditions, or raised in an atmosphere enriched with CO2.  Specific objectives of this chapter are to: 1. Compare the weekly growth of plants under CO2 enrichment or in ambient CO2 using non-destructive measures. 2. Compare the harvestable fresh fruit yield from plants receiving CO2 enrichment to those grown under ambient CO2. 3. Assess the usefulness of non-destructive plant growth measurements for inclusion into a plant-based approach to CO2 management. 3.2 MATERIALS AND METHODS Experiments were carried out in the production greenhouses at the Pacific Agri- Food Research Centre of Agriculture and Agri-Food Canada at Agassiz, British Columbia (PARC-Agassiz) during the summers of 2002 and 2003. A similar experimental design was used for both years and will be outlined below with the differences noted. 3.2.1 Greenhouse layout The production greenhouses at PARC-Agassiz were a Venlo design, cladded with glass and the ridges were oriented in a north-south direction.  Two adjacent compartments were modified to make four smaller compartments of 37.5 m2 each.  The two compartments were divided by erecting a temporary wall made of plastic and sealing it to the greenhouse floor, sides, and roof (Figure 3.1).  In each greenhouse compartment a row of ten 20 L “pillow bags” were filled with sawdust (yellow cedar [Chamaecyparis  63 Figure 3.1 The layout of the PARC –Agassiz experiment.  Shown (not to scale) is the 2002 arrangement for the CO2 treatments. E1: atmosphere enriched with CO2, replicate 1, A1: ambient CO2, replicate 1, E2: enriched CO2, rep 2 and A2: ambient, rep 2. The large rectangles represent 4 plants and the smaller ones 2 plants. The solid lines between compartments were glass walls and the dotted lines represent a plastic partition.  The arrows indicate the direction of training for the crop.  Crops in all compartments were trained the same way.  The direction of north is indicated in the top left corner. Greenhouse Corridor E 2 A 2 E 1 A1 Outside W all G reenhouse G reenhouse Guard p lants Treatment plants N G reenhouse G reenhouse  64 nootkatensis]) and were placed in the centre of each compartment in a north-south direction (Figure 3.1).  In these bags the 40 plants used in this experiment were placed, with 4 plants per bag.  These plants were trained onto twine attached to one of two over- head wires, 2.2 meters above the ground.  Attachment of every other plant to the left or right wire enabled plants to be trained to grow in a circle (Figure 3.1).  A guard row of 54 plants was planted around the perimeter of each compartment.  These plants were also trained to an over-head wire around the perimeter of each greenhouse compartment (Figure 3.1).  This arrangement resulted in a plant density of 2.5 plants per m2 for each compartment. The greenhouse environment was monitored and controlled separately for each compartment.  The concentration of CO2 in each compartment, as well as temperature and relative humidity, were monitored and these data were stored every 15 minutes using Argus software (Argus Control System Ltd, White Rock, BC).  Global radiation from the weather station mounted on the greenhouse exterior was also collected.  In 2002 the roof and external walls received an application of whitewash to help control summer temperatures.  In 2003 only, the external walls received whitewash. 3.2.2 Plant culture The tomato beefsteak cultivar Lycopersicum esculentum Mill. cv. Rapsodie was used for these experiments.  In 2002, grafted tomato plants were purchased from a local nursery (Houwelling Nurseries, Delta, BC).  The scion, Rapsodie (Syngenta Seeds Inc, Boise, ID) was seeded January 23 and the rootstock, Beaufort (DeRuiter Seeds Inc, Tecumseh, ON) was seeded January 9.  Grafting occurred on February 8 and plants were transplanted on March 7 into the greenhouse compartments at PARC – Agassiz.  In 2003,  65 grafted plants were not used, instead seeds were sown into rockwool cubes on December 10, 2002 and grown in a nursery on station until January 14, 2003, at which time plants were transferred to the greenhouse compartments.  The plants were maintained as per commercial practices for beefsteak tomatoes (Portree 1996) by the PARC-Agassiz staff. Flowers were pollinated by hand three times per week, all axillary shoots were removed, trusses were pruned to three fruit and the crop was grown at ambient CO2 until the commencement of the experiment. 3.2.3 Management and application of CO2 Carbon dioxide was produced on site from a boiler and delivered to the greenhouse compartments as needed.  Each greenhouse compartment had separate dosing, measurement and control of CO2 and was managed using environmental control software (Argus Control Systems Ltd, White Rock, BC).  Ductwork was installed to transfer the CO2 from a boiler to the compartments and perforated tubes were used to deliver the CO2 to the crop.  A small natural gas fuelled boiler (Lennox Industries, Calgary, AB, boiler capacity 158 MJ per hour, maximum CO2 input rate to compartments of 102 g CO2 m-2 hour-1) was used to generate the CO2.  A system of tubing, fans and valves was used to carry the CO2 from the compartments to the single infrared gas analyzer (IRGA) for measurement.  Before every CO2 measurement the air in the tubing was purged to prevent cross contamination between compartments. Carbon dioxide enrichment was delivered daily from 6 h to 21 h with a dosing set point of 1000 ppm.  As there was a need to ventilate the greenhouse compartments to reduce the temperature, the mean CO2 concentration of the enriched compartments was usually less than 1000 ppm but about double that of the ambient compartments.  In both  66 years CO2 dosing sporadically ceased, causing a sudden drop in the CO2 concentration to ambient levels.  This was caused by the boiler shutting down when the heat produced from the generation of CO2 could not be exhausted.  Enrichment usually resumed within two hours when the system had cooled. 3.2.4 Experimental design With four greenhouse compartments available the experimental design consisted of two replicates for two treatments of CO2.  In this arrangement an experimental unit was a greenhouse compartment with the plants being a sub-replicate.  The treatments were:  exposure to enriched CO2 and exposure to ambient CO2 (no CO2 enrichment).  To guard against possible gradients or biases caused by the arrangement of the greenhouse compartments, CO2 enrichment was applied to the atmospheres of the first and third compartments in 2002 and the second and fourth greenhouse compartments in 2003 (Figure 3.1).  In each year the remaining two compartments did not receive intentional supplemental CO2.  At times the CO2 concentration in the ambient greenhouses was slightly higher then the ambient air outside the greenhouse.  This was attributed to the transfer of some CO2 from the enriched to the ambient compartments.  In 2002, CO2 enrichment was started at 6 h on June 11 and continued for 80 days until August 30.  In 2003, CO2 enrichment was started at 15 h on May 6 and continued for 53 days until June 28.  This was when plants had reached the high wire and had at least 7 trusses with fruit. 3.2.5 Plant observations, growth and yield measurements In both years the yield of harvested fruit and plant growth were measured and leaves were observed for signs of injury caused by CO2 enrichment.  Fruit were harvested three times per week from the 40 plants in the centre row of each greenhouse  67 compartment.  The fruit were removed from the plant at the “breaker stage”, the stage when the fruit were just beginning to turn from green to red at the blossom end.  Each harvested fruit was identified with the greenhouse compartment and plant at picking and the mass was measured. Plant growth was measured in situ and non-destructively by measuring the stem elongation, stem diameter, leaf length and fruit diameter on a weekly basis.  In 2002, 10 plants per replicate (greenhouse compartment) where measured and in 2003, 7 plants were measured.  Stem elongation was estimated by measuring the change in growth after a week.  Every week the position of the shoot apex was marked on the supporting twine and the distance between that week and the previous week measured.  Stem diameter was measured in two dimensions (90° to each other) just above the truss with open flowers which was typically the third visible truss and about 0.15 m from the top of the plant. For leaf length measurements, the leaf closest to the truss with open flowers was identified as leaf one at the commencement of the measurement period (Figure 3.2). Every second leaf starting with leaf one was labelled (when the petiole was at least 30 mm long) and measured.  As the experiment progressed new leaves were added and measured weekly for 11 weeks in 2002.  In 2003 leaves were measured until they did not exhibit a further increase in length, usually after 6 weeks.  Twenty leaves per plant (odd numbered leaves 13 to 53) were measured in 2002 and in 2003 twelve leaves (1 to 25) were measured.  Leaf length in cm was measured with a ruler from the last pair of leaflets to the end of the terminal leaflet.  For the fruit measurements the trusses were numbered with truss 1 being the first truss produced by the plants.  On each truss the fruit labelled 1 was the most proximal with fruit 3 being the most distal and fruit 2 located between these  68  Figure 3.2 Schematic of a tomato plant (not to scale) with leaves, trusses with flowers and fruit indicated.  Leaf 1 is the leaf closest to the truss with open flowers, leaves above are identified as 0, -1, -2 etc. and leaves below are 2, 3, 4, 5, 6 etc.  Leaf 1 Support wire Rooting substrate Stem Support string  69 fruit.  The diameter of each fruit was measured in mm with an electronic calliper from the calyx to the blossom end (ie. top to bottom).  These measurements were carried out once per week. The measurement period for the 2002 experiment was about a month prior to the CO2 enrichment period.  The leaf length and stem height were first measured on April 23, stem diameter on May 1 and fruit diameter on May 8 starting at truss 3.  The first fruit harvest was on May 30.  In the 2003 experiment the measurement period closely coincided with the CO2 enrichment period.  The first fruit were harvested on April 14, stem height was measured on May 1 and leaf length, stem diameter and fruit diameter were first measured on May 8.  Fruit diameter was not measured until the eighth truss. 3.2.6 Calculation of fruit load and leaf area The fruit load was estimated by calculating the fruit volume from the measured fruit diameter.  The fruit volume was estimated by calculating the volume of a sphere:  Equation 3.1 3 3 4 rFV π= where FV is fruit volume in cm3 and r is fruit radius, in cm, measured from calyx to blossom end.  The volume was calculated for each fruit on a truss and then added together to give a total fruit volume per truss.  The fruit load of the plant was calculated by summing the volume of fruit from those trusses initiated after the onset of CO2 enrichment.  Therefore, every week an additional truss of fruit was added causing the fruit load to become progressively larger as the experiment progressed.  When the fruit were harvested the diameter was measured and it was dropped from the experiment.  70 Leaf area was estimated from the measurement of leaf length and a previously defined function relating these two measurements.  The function used was: Equation 3.2 ))(45.253.0( LLlneLA ∗+−= , R2=0.996. where LA is leaf area in m2 and LL is leaf length in meters, e is exponential and ln is natural logarithm.  The development of this function is presented in Appendix 1.1. 3.2.7 Data analysis All data were collected from repeatedly measuring the greenhouse compartments or plants over time.  Stem length, stem diameter, fruit load, fruit yield and the greenhouse environment variables were analyzed as repeated measures in a mixed model analysis of variance (ANOVA) using PROC MIXED, SAS version 9.1.3 (SAS Institute 2003).  The mixed model analysis can handle the fixed effects (CO2 treatment and time) and the random effects.  These random effects are the correlation pattern among the experimental units (greenhouse compartments) made over time (Littell et al. 1998).  In this experiment the experimental units were greenhouse compartments with plants being sub-replicates. To avoid the possibility of pseudo-replication mean stem length, stem diameter, harvest and fruit load were calculated per greenhouse and per date and then used in the analysis. The specifics of the data analysis were as follows.  It was necessary to pre-treat these data before the mixed model analysis.  The mean responses of stem length, stem diameter, fruit diameter and fruit fresh mass were calculated for each greenhouse compartment.  For the environmental variables (CO2 concentration, global radiation, temperature and relative humidity) daily means were calculated from the raw data for each greenhouse. During the mixed model analysis, residuals were examined to ensure  71 the homoscadesity of the variance and their normal distribution.  Stem length and mass per fruit were transformed using a natural logarithm and a square root transformation was used for the canopy fruit load data.  Following the procedure for mixed models outlined by Littell et al. (1998; 2000) and I. Bercovitz (pers. com. 2007) the autoregressive order 1 was chosen as the best fit variance/covariance matrices based on the Akaike information criterion.  In addition, for these type of data the autoregressive covariance structure has been recommended for repeated measurements made on experimental units over time (Littell et al. 1998; 2000).  This procedure was used to correct for inequality of variances among the sampling times.  The overall treatment mean for each measure was determined from the mixed model analysis using the least square means statement and treatment comparisons were made using the contrast statement.  Means were considered to be significantly different if the p value of the contrast was less than 0.05.  Treatment means were compared at the same date of data collection using a post hoc pair-wise comparison (least-squares mean with the pdiff option).  The two means were considered significantly different from each other if the p-value of the comparison was less than 0.05.  A Bonferroni adjustment was employed to decrease the probability of a type I error occurring when using a multiple comparison procedure (Littell et al. 1998; Dytham 2003; SAS Institute 2003) The repeated measures analysis was not used for leaf area and total harvested fruit per experiment.  The profile of leaf area expansion and leaf appearance over time was determined for both experiments.  For each identified leaf the mean leaf area was calculated per greenhouse compartment for each date of measurement.  The mean of the greenhouse compartments was then calculated and plotted versus days after the onset of  72 CO2 enrichment.  Twenty-one and 13 leaves were measured in 2002 and in 2003, respectively.  A leaf was assumed to have appeared when its area was 50% of the final leaf area of the first leaf measured.  In 2002 this was 0.030 m2 and in 2003 this was 0.023 m2.  The total amount of harvested fruit for each experiment and greenhouse compartment was calculated on a per plant and per square meter basis.  Least square means of each treatment were calculated and a pair-wise comparison was made using Proc GLM (SAS Institute 2003). 3.3 RESULTS The CO2 enriched plants in this study did not exhibit any visible injury to the leaves.  Necrosis of the margin of upper leaves was observed in commercial greenhouses during the time frame of these experiments (Appendix 3.1).  The application of CO2 to a daily mean of 800 ppm had a very modest but positive effect on the non-destructive indicators of plant growth measured here and fruit yield. 3.3.1 Greenhouse environment  Environmental data for the 2002 and 2003 experiments are presented in Figures 3.3 and 3.4 and summarized in Table 3.1.  The plots indicate the variable nature of global radiation, temperature, relative humidity and daytime CO2 concentration.  Carbon dioxide enrichment resulted in a mean atmospheric concentration of 846 ppm in 2002 and 804 ppm in 2003.  These are approximately twice as great as the mean ambient CO2 concentration of 420 ppm and 472 ppm in 2002 and 2003, respectively.  In the 2002 experiment, a small although significantly greater greenhouse temperature for the Enriched (22.5 C) versus Ambient (21.9 C) compartments occurred (Table 3.1).  Mean relative humidity was distinctly different between treatments with the Ambient being  73  Figure 3.3 The environment for the greenhouses receiving either ambient or enriched CO2 for the 2002 experiment at PARC-Agassiz, June 1 to August 30.  Panel A. mean daytime CO2, B. daily sum of global radiation, C. mean 24 hour temperature and D. mean 24 hour relative humidity. N=2. 200 300 400 500 600 700 800 900 1000 1100 1200 1300 1400 1500 -10 0 10 20 30 40 50 60 70 80 Days after onset of CO2 enrichment CO 2 ( pp m ) Ambient Enriched 0 5 10 15 20 25 30 35 -10 0 10 20 30 40 50 60 70 80 Days after onset of CO2 enrichment G lo ba l r ad ia tio n (M J m -2  d ay -1 ) All treatments 16.0 18.0 20.0 22.0 24.0 26.0 28.0 30.0 -10 0 10 20 30 40 50 60 70 80 Days after onset of carbon dioxide enrichment Te m pe ra tu re  ( C) Ambient Enriched 30.0 40.0 50.0 60.0 70.0 80.0 90.0 100.0 -10 0 10 20 30 40 50 60 70 80 Days after onset of carbon dioxide enrichment Re la tiv e H um id ity  ( % ) Ambient Enriched A B DC  74  Figure 3.4 The environment for the greenhouses receiving either ambient or enriched CO2 for the 2003 experiment at PARC-Agassiz, May 1 to June 28.  Panel A mean daytime CO2, B daily sum of global radiation, C. mean 24 hour temperature and D. mean 24 hour relative humidity. Gap in the data at 45 and 46 days was caused by a power failure.  N=2. 200 300 400 500 600 700 800 900 1000 1100 1200 1300 1400 -5 0 5 10 15 20 25 30 35 40 45 50 Days after onset of CO2 enrichment CO 2 ( pp m ) Ambient Enriched 0 5 10 15 20 25 30 -5 0 5 10 15 20 25 30 35 40 45 50 Days after onset  of CO2 enrichment G lo ba l r ad ia tio n (M J m -2  d ay -1 ) All treatments 10 20 30 40 50 60 70 80 90 100 -5 0 5 10 15 20 25 30 35 40 45 50 Days after onset  of carbon dioxide enrichment Re la tiv e hu m id ity  (% ) Ambient Enriched 16 18 20 22 24 26 28 30 -5 0 5 10 15 20 25 30 35 40 45 50 Days after onset of carbon dioxide enrichment Te m pe ra tu re  ( C) Ambient Enriched A B C D  75 Table 3.1 A summary of the daily mean greenhouse environment for the experiments at PARC-Agassiz. Year Treatment Daytime CO2 Daytime Temperature 24 hour Temperature Relative Humidity Global Radiation1   ppm C C % MJ m-2 Day-1 2002 Ambient 419.8 23.5 21.9 75.3  Enriched 846.1 24.1 22.5 68.4  p-value2 0.0004 0.021 0.017 <0.013 20.3 2003 Ambient 472.7 23.2 21.8 53.3  Enriched 804.4 23.2 21.7 54.7  p-value2 <0.0001 0.95 0.56 0.174 18.7 1. Global radiation was measured outside the greenhouse. 2. p-value from least square means pair wise comparison, n=2  76 75.3% and the Enriched 68.4% in 2002 (Table 3.1).  In 2003, the overall temperature and relative humidity were similar between the treatment greenhouses.  Average temperature for both was 21.8 C and relative humidity was 56.6% and 59.8% for Ambient and Enriched respectively.  In both experiments the day-to-day total global radiation was variable (Figure 3.3 and 3.4).  However, over the experimental period the mean daily radiation was similar being 20.3 and 18.3 MJ m-2, for 2002 and 2003 experiments, respectively. 3.3.2 Stem length and diameter Stem length and diameter were found not to be significantly affected by CO2 enrichment.  In the 2002 experiment, the mean stem length at the onset of CO2 enrichment was 2.40 m and after 80 days of exposure to CO2 enrichment increased to 4.80 m, 0.30 m more than for plants grown in ambient CO2 (4.50 m) (Figure 3.5A). Statistical analysis of these data indicated that the effect of enriched CO2 on length was not significant (p=0.378) and the treatment by date interaction was also non-significant (p=0.24).  In the 2003 experiment, which was 50 days in duration, plants in the Ambient treatment grew 1.24 m (3.13 to 4.37 m) while those in the Enriched treatment grew 1.32 m (3.13 to 4.45 m) (Figure 3.5B).  The statistical analysis indicated CO2 enrichment did not significantly increase length (p=0.44) and the time by treatment interaction was also insignificant (p=0.18) (Figure 3.5B).  For both years the daily growth rates were similar. In 2003, rates were 0.025 and 0.027 m day-1 for Ambient and Enriched treatments, respectively.  In 2002 the daily growth rates were 0.027 and 0.030 m day-1 for Ambient and Enriched treatments respectively.  77 Figure 3.5 Mean stem length for plants grown under CO2 enrichment and ambient CO2.  Panel A is the 2002 experiment, June 11 to August 30 and Panel B is the 2003 experiment, May 6 to June 28.  Each symbol is the mean response of plants from each greenhouse compartment, N=2.  There were no statistical differences between treatment means at any sampling date. 2.00 2.50 3.00 3.50 4.00 4.50 5.00 0 10 20 30 40 50 60 70 80 Days after onset of CO2 enrichment St em  le ng th  (m ) Ambient Enriched 2.00 2.50 3.00 3.50 4.00 4.50 5.00 5.50 6.00 0 10 20 30 40 50 60 70 80 Days after onset of carbon dioxide enrichment St em  le ng th  (m ) Ambient Enriched B A  78 Stem diameter was not significantly increased by CO2 enrichment in either experiment (2002, p=0.17 and 2003, p=0.05, Figure 3.6).  Overall, the mean stem diameter for both experiments was 8.8 mm for plants in the Ambient treatment and 9.4 mm and 9.6 mm for plants in the Enriched treatment in 2002 and 2003, respectively. Stem diameter was significantly affected by date of sampling (2002, p<0.0001 and 2003 p<0.0005) indicating changes with time.  Plots of the mean response indicate the dynamic nature of stem diameter (Figure 3.6).  The CO2 treatment by date of sampling interaction term was not significant, indicating that when CO2 enrichment affected stem diameter in 2003, it was consistent over sampling date.  On a weekly basis, when the Bonferroni adjustment was employed for the post hoc mean comparison there were no statistically significantly differences in either experiment.  However, for both studies there was a consistent trend of increased stem diameter. 3.3.3 The profile of leaf area expansion Rate of leaf appearance and expansion in leaf area for leaves were recorded during both experiments.  To help measure leaf appearance, the number of leaves which met or surpassed half of the final length of the first leaf was determined (Figures 3.7 and 3.8).  In 2002, 17 leaves in the Ambient treatment and 19 leaves in the Enriched treatment met or surpassed the 50% final area threshold of 300 cm2 (Figure 3.7).  In 2003, one more leaf met the 50% final area threshold of 230 cm2 in the Enriched compared to the Ambient treatment (Figure 3.8).  However, this distinction is minor as the Enriched treatment was just above the threshold (Figure 3.8).  In both years the maximum area per leaf was similar for both the treatments.  79  Figure 3.6 Mean stem diameter for plants grown under CO2 enrichment and ambient CO2 at PARC-Agassiz.  Panel A is the 2002 experiment, June 11 to August 30 and Panel B is the 2003 experiment, May 6 to June 28. Each symbol is the mean response of plants from each greenhouse compartment, N=2.  There were no statistical differences between treatment means at each sampling date. 6.0 6.5 7.0 7.5 8.0 8.5 9.0 9.5 10.0 10.5 11.0 11.5 12.0 0 10 20 30 40 50 60 70 80 Days after onset of CO2 enrichment St em  d ia m et er  (m m ) Ambient Enriched 6.0 6.5 7.0 7.5 8.0 8.5 9.0 9.5 10.0 10.5 11.0 11.5 12.0 0 10 20 30 40 50 60 70 80 Days after onset of carbon dioxide enrichment St em  d ia m et er  (m m ) Ambient Enriched B A  80  Figure 3.7 Mean area per leaf for plants grown under ambient CO2 (Panel A) and CO2 enrichment (Panel B) for the 2002 experiment at PARC-Agassiz.  Data were collected weekly from May 7 to August 30.  The horizontal line is 50% of the final leaf area of the first leaf measured.  Each symbol is the mean of 14 leaves. 0 50 100 150 200 250 300 350 400 450 500 550 600 650 -35 -25 -15 -5 5 15 25 35 45 55 65 75 85 Days after onset of CO2 enrichment Le af  a re a (c m2 ) 13 15 17 19 21 23 25 27 29 31 33 35 37 39 41 43 45 47 49 51 Leaf Id: A 0 50 100 150 200 250 300 350 400 450 500 550 600 650 -35 -25 -15 -5 5 15 25 35 45 55 65 75 85 Days after onset of carbon dioxide enrichment Le af  a re a (c m2 ) 13 15 17 19 21 23 25 27 29 31 33 35 37 39 41 43 45 47 49 51 53 Leaf id: B  81  Figure 3.8 Mean area per leaf for plants grown under ambient CO2 (Panel A) and CO2 enrichment (Panel B) for the 2003 experiment at PARC-Agassiz.  Data were collected weekly from May 6 to June 28.  The horizontal line is 50% of the final leaf area of the first leaf measured.  Each symbol is the mean of 20 leaves. 0 50 100 150 200 250 300 350 400 450 500 550 600 650 0 5 10 15 20 25 30 35 40 45 50 Days after onset of CO2 enrichment Le af  a re a (c m2 ) 1 3 5 7 9 11 13 15 17 19 21 23 25 0 50 100 150 200 250 300 350 400 450 500 550 600 650 0 5 10 15 20 25 30 35 40 45 50 Days after onset of carbon dioxide enrichment Le af  a re a (c m2 ) 1 3 5 7 9 11 13 15 17 19 21 23 25 Leaf id: B A  82 For both the 2002 and 2003 experiments, CO2 enrichment did not appear to cause a notable increase or decrease in the leaf area of the canopy throughout the experiment. Since leaf pruning was practised leaf area was fairly constant throughout the experiment. In 2002 the estimated mean and standard error for leaf area per plant was 1.147 ± 0.014 and 1.096 ± 0.015 m2 for the plants in ambient and enriched CO2, respectively.  In 2003, the estimated leaf area per plant was 0.886 ± 0.001 and 0.889 ± 0.011 m2 for the plants in the ambient and enriched CO2, respectively. 3.3.4 Canopy fruit load The effect of CO2 enrichment on fruit load was quantified by determining the increase in fruit volume.  The shape of the curves is mainly caused by the method of data collection, where only those fruit on trusses formed after the onset of CO2 enrichment were measured (Figure 3.9).  Therefore, with time a greater number of trusses were measured resulting in a greater fruit load (date effect was p<0.0001 for both years).  In 2002, CO2 enrichment did not appear to significantly affect the fruit load (p=0.15) and there was an non-significant treatment by date interaction (p=0.98) indicating a consistent response for both treatments with time.  In 2003, CO2 enrichment had a weak effect on fruit load (p=0.05) and a statistically significant treatment by date interaction was present (p=0.035).  Pair-wise comparisons at each date indicated that the fruit load was significantly greater at date 36 (p=0.0008), 43 (p=0.0002) and 50 (p=0.0005).   83 Figure 3.9 Fruit load increase for plants grown under CO2 enrichment and ambient CO2.  Panel A is the 2002 experiment, June 11 to August 30 and Panel B is the 2003 experiment, May 6 to June 28.  In each experiment plants initially had 7 trusses with 3 fruit but only trusses formed after the onset of CO2 enrichment were measured.  Therefore, every 7 days a newly formed truss was added and measured until the fruit were harvested or the experiment ended. * indicates treatment means were statistically different at p< 0.001. 0 200 400 600 800 1000 1200 1400 1600 0 10 20 30 40 50 60 70 80 Days after onset of CO2 enrichment Fr ui t l oa d pe r p la nt  (c m3 ) Ambient Enriched A 0 200 400 600 800 1000 1200 1400 1600 0 10 20 30 40 50 60 70 80 Days after onset of carbon dioxide enrichment Fr ui t l oa d pe r p la nt  (c m3 ) Ambient Enriched B * * *  84 3.3.5 Fruit harvest, mass and number Fruit harvest was quantified by determining individual fruit mass, total fruit mass and fruit number per plant and per square meter of greenhouse area.  For the experimental period of 80 days in 2002, the individual fruit mass, total fruit mass and number per plant were significantly greater for plants exposed to enriched CO2 (Table 3.2).  On a greenhouse area basis (per m2), plants in the Enriched greenhouses exhibited a slightly greater total fruit mass but the number of fruit was not increased.  In the 2003 experiment, after 50 days of CO2 enrichment harvested fruit measured on both a per plant and per square meter basis were similar between the treatments with no statistical differences present (Table 3.2). Total yield was further analyzed by examining the cumulative profile of yield (mass and number) over time on a per plant basis.  In 2002 both the mass and the fruit number per plant were statistically significantly increased by CO2 enrichment (mass p=0.0053 and number p=0.0069).  As well, for both measures there was a significant date by treatment effect (mass p<0.0001 and number p=0.0003) indicating that the treatments responded distinctly with time.  For harvested mass these differences began at 24 days and for fruit number at 38 days after the onset of CO2 enrichment (Figure 3.10).  Analysis of the fruit harvested in 2003 indicated that there were no differences between the treatments for fruit number or mass on a per plant basis (p=0.52 and p=0.97, respectively) and an insignificant treatment by date interaction (p=0.4) (Figures 3.11). 3.4 DISCUSSION This discussion will focus on several non-destructive measures of growth that were examined over-time and compared for plant grown in ambient and enriched CO2.  85 Table 3.2 A summary of the fresh tomato fruit harvest for the experiments at PARC – Agassiz. Year Treatment Fruit mass Total number  Total mass   g per plant-1 per m-2 kg plant-1 kg m-2 2002 Ambient 167.3 56.2 76.4 9.9 13.6  Enriched 177.7 58.5 81.8 10.9 15.3  p-value1 0.025 0.0067 0.17 0.005 0.04 2003 Ambient 200.3 39.6 50.7 8.21 10.4  Enriched 194.4 40.5 52.3 8.23 10.5  p-value1 0.1904 0.53 0.37 0.98 0.85 1. p-value from least square means pair wise comparison, n=2.  86 Figure 3.10 Cumulative harvested fresh fruit mass for plants grown under CO2 enrichment and ambient CO2 from the 2002 experiment, June 11 to August 30. Panel A is the cumulative fresh fruit mass harvested. Treatment means from 24 days were significantly different at p<0.05, means at 36 days and beyond were significantly different at p<0.0001. Panel B is the cumulative fruit number harvested.  Treatment means from 38 days were significantly different at p<0.05, means at 45 days and beyond were significantly different at p<0.0001. 0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0 10.0 11.0 12.0 0 10 20 30 40 50 60 70 80 Days after onset of CO2 Y ie ld   ( kg  p la nt- 1 ) Ambient Enriched A p<0.0001 p<0.05 0 5 10 15 20 25 30 35 40 45 50 55 60 65 0 10 20 30 40 50 60 70 80 Days after onset of carbon dioxide enrichment Y ie ld   ( fr ui t p la nt-1 ) Ambient Enriched B p<0.05 p<0.0001  87 Figure 3.11 Cumulative harvested fresh fruit mass for plants grown under CO2 enrichment and ambient CO2 from the 2003 experiment, May 6 to June 28.  Panel A is the cumulative fresh fruit mass harvested and Panel B is the cumulative fruit number harvested. Treatment means were not significantly different at the p<0.05 level for any date. 0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0 10.0 11.0 12.0 0 10 20 30 40 50 60 70 80 Days after onset of CO2 Y ie ld  (k g pl an t-1 ) Ambient Enriched 0 5 10 15 20 25 30 35 40 45 50 55 60 65 0 10 20 30 40 50 60 70 80 Days after onset of carbon dioxide enrichment Y ie ld  (f ru it pl an t-1 )  Ambient Enriched A B  88 The potential to use these results in plant-based CO2 management decisions will be deferred to the General Discussion (Chapter 6).  At the outset of this discussion it is noteworthy to mention that the application of CO2 to a daily mean of 800 ppm had, at best, a very modest effect on these measures.  These results are not unprecedented, as Wittwer and Honma (1979) reported no significant differences in the vegetative growth of tomato grown in 400, 800 or 1200 ppm of CO2. 3.4.1 The non-destructive measurement of plant growth under CO2 enrichment Stem length was not significantly affected by CO2 enrichment in the 2002 or 2003 experiments.  In the shorter 2003 experiment, plants in ambient CO2 and CO2 enrichment exhibited nearly identical stem length.  In the longer 2002 experiment stem lengths were slightly longer for plants receiving CO2 enrichment.  It is noteworthy to mention that the mean daytime temperature of the compartments receiving CO2 enrichment was 0.6 C greater than the compartments at ambient CO2 in 2002 (Table 3.1).  Over the duration of the experiment this increase in temperature would have contributed to increased growth. According to Heuvelink (2005) the rate of stem elongation is mainly controlled by temperature; although CO2 enrichment can cause longer plants because they are growing faster. Stem diameter is larger in plants that have received CO2 enrichment (Ainsworth & Long 2005; Heuvelink & Dorais 2005).  In both years of the present study, stem diameter was not significantly affected by CO2 enrichment.  For both experiments the mean stem diameter for the Ambient treatment was 8.8 mm.  The mean stem diameter for the Enriched treatments was 9.4 and 9.6 mm for 2002 and 2003 experiments, respectively.  Increased assimilate production caused by CO2 enrichment would help  89 maintain stem diameter as these extra assimilates could be directed to the stem once the requirements of the fruit have been met.  However, increased temperature as was the case in 2002 could reduce the effect of CO2 enrichment because higher temperatures tend to reduce stem diameter (Portree 1996). The maximum area of individual leaves was unaffected by CO2 enrichment (Figures 3.7 and 3.8).  In mature, fruiting crops the leaf area of plants grown under high CO2 can be smaller (Picken et al. 1986; Nederhoff 1994; Heuvelink 2005; Heuvelink & Dorais 2005) or be unchanged in area (Nederhoff 1994; Heuvelink & Dorais 2005; Hao et al. 2006) compared to plants in ambient CO2.  One would expect that leaf area would increase with elevated CO2 as the increased assimilates would be available for leaf growth.  A reduction in leaf area could be caused by water stress reducing the cell expansion when the leaf is developing.  In tomato, competition for assimilates with the fruit, can result in shorter leaves as the fruit are a stronger sink than the leaves (Ho & Hewitt 1986; Valantin et al. 1998).  Bertin and Gary (1998) contended that leaf area is independent of the assimilate supply (source activity) when the minimum structural leaf mass per unit area has been attained. CO2 at times had a statistically significant effect on fruit load.  For example, in 2003 there was a statistically significant increase in fruit load due to enrichment towards the end of the experiment.  This is not a surprising finding as towards the end of the experiment the plants have been exposed to CO2 enrichment longer and more fruit are being monitored.  In this work, the number of fruit was fixed at three per truss and the number of trusses was similar between the treatments.  Therefore, differences in fruit load were caused by the size of the fruit, which was determined by measuring the volume  90 of the fruit.  As well as available assimilates, final fruit size is also dependent on a number of other factors such as cell number in the ovary before anthesis (Ho 1995), seed number, fruit number, and temperature (Nederhoff 1994). 3.4.2 Effects of CO2 enrichment on fresh fruit yield The main goal of CO2 enrichment in tomato greenhouses is to increase the mass of fruit produced.  Enrichment of the greenhouse atmosphere to double the ambient CO2 resulted in a small increase in harvested yield in the 2002 experiment and no effect for the 2003 experiment.  In the 2002 study, which was 80 days (about 11 weeks), the number of fruit per plant and mean fruit mass were 4 and 6% higher, respectively for plants exposed to CO2 enrichment (Table 3.2).  This led to an overall 9% increase in the total harvested fruit mass per plant (Table 3.2).  Also notable that in this experiment these increases became statistically apparent as early as 24 days after exposure to CO2 enrichment (Figure 3.10).  According to Heuvelink and Dorais (2005), doubling of the atmospheric CO2 concentration is believed to increase fruit yield from 11 to 32% for greenhouse vegetable crops.  Therefore, my data were at the low range of the predicted yield increase expected for CO2 enrichment.  Comparisons to literature values can be problematic because yield depends not only on the concentration of CO2 but also on the duration of the experiment.  The following selected experiments have reported yield increases for tomato:  43% (106 days of CO2 enrichment to 1000 ppm), 18% (86 days of CO2 enrichment to 1000 ppm), 41% (76 days of CO2 enrichment to 1000 ppm), 58% (76 days of CO2 enrichment to 1000 ppm) (Wittwer & Honma 1979), 33% (53 days of CO2 enrichment at about 900 ppm) (Hicklenton & Jolliffe 1978), 31% (66 days of enrichment to 790 ppm) (van de Vooren et al. 1986), 22% (CO2 enrichment to 900 ppm for 70 days)  91 (Yelle et al. 1990), 16% (140 days, CO2 enrichment to 450 ppm) (Nederhoff 1994), 14% (112 days of CO2 enrichment to 1000 ppm) (Tripp et al. 1991) and -10% (62 days of CO2 enrichment to 800 ppm) (Hao et al. 2006). A fruit under greenhouse conditions takes 6 to 8 weeks (42 to 63 days) to develop from anthesis to harvest.  Therefore, it could take as long as 8 weeks (56 days) to clearly measure the benefits of CO2 enrichment on harvested fruit.  In the above cited studies, most were greater than 60 days and a couple were well over 100 days.  It is possible that for the environmental conditions in the 2003 experiment, 50 days was not a long enough time of exposure to CO2 enrichment for harvested fruit to be affected.  Along with duration of exposure to CO2 enrichment, crop management practises could also affect the plant response to CO2 enrichment.  In accordance with industry practises for beefsteak tomato production, the number of fruit per truss was restricted.  The rationale for fruit pruning is to promote fruit growth to a larger size which commands a higher return.  As well, fruit pruning ensures a steady production of future fruit because the plant maintains a favourable vegetative/generative balance.  Typically commercial greenhouses prune to leave four to five fruit per truss.  In the relatively short term experiments I carried out, where the fruit numbers were fixed, CO2 enrichment can only affect the mass of the fruit and prevent fruit or flower abortion.  Since these experiments were carried out in the summer enough assimilates could have been present to prevent fruit and flower abortion (Nederhoff 1994).  Heuvelink and Dorais (2005) indicate that sink strength rather than source strength determines assimilate partitioning.  Therefore, further increasing assimilate production through CO2 enrichment may not increase fruit growth if fruit requirements for assimilates have been met.  It is possible that summer light conditions in  92 2003 (no whitewashing of greenhouse compartments) and ambient CO2 have met the assimilate needs of the fruit.  In 2002 when light was reduced by the whitewash, plants in ambient CO2 were deficient in assimilates which was overcome by CO2 enrichment. 3.4.3 Comments on the experiments and non-destructive measurements The plant measures examined exhibited few statistically significant differences that were attributed to CO2 enrichment.  Differences in plant growth may have been quantifiable had these measures been made differently or had other measures been examined.  As well, an inability to detect significant differences between the treatments is also in part caused by the greenhouse environment, the variability of these measurements and the lack of replication.  In 2002 the greenhouse environment was modified by the application of a heavy coat of white wash.  The white wash reduced the light penetration into the greenhouse and the photosynthetic response to CO2 enrichment (Appendix 5.1). In 2003 no white wash was applied and photosynthetic rate (Appendix 5.1) and the total fruit mass harvested per day were higher compared to the 2002 study.  The duration of exposure to CO2 enrichment could also have been a factor, especially for the 2003 experiment.  In both experiments the crop was grown under ambient CO2 until the harvest of fruit was well under way.  When CO2 enrichment is delayed until the first or later trusses are harvested most of the fruit will have been formed before CO2 enrichment could affect them (Nederhoff et al. 1992). The variability of plant growth in this work was undoubtedly larger than would be found in a commercial greenhouse.  Research greenhouses being only a fraction of the size of a commercial greenhouse have less ability to buffer the ambient environment, resulting in hot and cold spots and the infrastructure used to hold the glass causes shading  93 to parts of the crop.  As well, in my observation, small greenhouses are more prone to uneven crop densities in tomato experiments.  A limited number of replicates results in low degrees of freedom which increases the criteria necessary for finding a statistically significant difference.  In this experiment the experimental unit was a greenhouse compartment and since there were only four available, the replication for each treatment was only two.  As Nederhoff (1994) has indicated, insufficient replication is a common problem in greenhouse research. The merits and disadvantages of these non-destructive measures are outlined as follows.  Stem length was very straightforward to measure and I believe it has a high accuracy and precision as it increases with time and in one direction.  Stem diameter is a commonly employed non-destructive measure of growth used by greenhouse tomato growers to assess the vegetative/generative balance of their plants.  Unlike stem length which increases with time the objective for the grower is to keep the plant stem diameter at 10 mm (G. Yakel pers com; Peet & Welles 2005).  The diameter is measured with a calliper close to the youngest truss with open flowers (Portree 1996).  Sometimes the stem is not round at this measurement point which would lower the accuracy and precision.  Leaf area was determined by measuring the length of the leaf blade (from leaf tip to basal pair of leaflets) and calculating the area from an established leaf length to area function (Appendix 1.1).  This method caused a minimum of disturbance to the leaf which unlike established scientific methods requires the leaf to be removed from the plant before measurement.  The leaf area of a crop is an important factor in crop productivity as the amount of intercepted light controls growth and yield.  Currently, greenhouse growers track only the length of leaves, and measuring the leaf area could be  94 beneficial for crop management.  Schwarz and Klaring (2001) also used allometery to estimate leaf area and concluded that leaf width explained 90% of the variation in leaf area and was superior to leaf length which only explained 75% of the variation.  I found leaf length to explain 99.6% of the variation in leaf area (Appendix 1.1).  However, Schwarz & Klaring (2001) admit that the leaf width can be difficult to measure in situ.  I also agree with their conclusion that a difficulty with this method is that the relationship would have to be determined for each cultivar (Schwarz & Klaring 2001). Fruit load was determined by the calculating the volume of fruit on a plant.  The volume of a fruit was determined by measuring the ‘polar’ fruit diameter (from calyx to blossom end) and the volume approximated by calculating the volume of sphere.  The polar diameter was an easy measurement to make in situ; however at no point in the development of a fruit is it perfectly spherical in shape.  Young fruit have a somewhat acorn-like appearance that gradually becomes more flattened at the blossom end as the future matures.  There could be considerable experimental error with these estimations and in hind-sight a destructive analysis should have been carried out using a water displacement method to check the volume and the fruit dry mass as determined by Gautier et al. (2001).  However, even though exposure to CO2 enrichment will increase the size of fruit, there is no indication in the literature that the shape of the fruit would also be affected. 3.5 CONCLUSIONS In this chapter I have compared the weekly growth and yield of plants under enriched and ambient CO2 using non-destructive measures.  95 1. The application of CO2 to a daily mean of 800 ppm, double the ambient CO2 had little or no effect on weekly measurements of stem length and diameter, leaf area and fruit volume. 2. In the 2002 experiment the weekly fresh fruit yield was significantly heavier after 24 days and more fruit was produced after 38 days for plants treated with CO2 enrichment. 3. After 80 days of exposure to CO2 enrichment overall fruit yield was about 9% greater (4% more fruit per plant and 6% increased fruit mass) in the 2002 experiment.  This modest increase compared to the published literature might have been caused by the application of whitewash to the greenhouse glass. 4. Exposure to CO2 enrichment for 50 days in 2003 did not increase fruit yield despite the absence of whitewash.  It is possible that the high light levels resulted in enough assimilate production in ambient CO2 to meet the needs of the fruit. However, during the experiment the fruit load was found to be significantly increased by CO2 enrichment.  Had the experiment been run longer significant differences in fruit fresh mass and number may have occurred. 5. Plant measurements were generally very straight forward to carry out but can suffer from inherent variability of the measurement point making the interpretation unclear.  96 3.6 LITERATURE CITED Ainsworth, E., Davey, P., Bernacchi, C., Dermody, O., Heaton, E., Moore, D., Morgan, P., Baidu, S., Ra, H., Zhu, Z., Curtis, P., and Long, S. 2002. A meta-analysis of elevated [CO2] effects on soybean (Glycine max) physiology, growth and yield. Global Change Biology 8: 695-709. Ainsworth, E.A. and Long, S.P. 2005. Tansley review.  What have we learned from 15 years of free-air CO2 enrichment (Face)?  A meta-analytic review of the responses of photosynthesis canopy properties and plant production to rising CO2.  New Phytologist 165:  351-372. Atherton, J.G. and Harris, G.P. 1986. Flowering.  In Chapman and Hall Ltd., Cambridge, Great Britain, pp 167-200. Bertin, N. and Gary, C. 1998. Short and long term fluctuations of the leaf mass per area of tomato plants - Implications for growth models. Annals of Botany 82: 71-81. Bowes, G. 1993. Facing the inevitable:  plants and increasing atmospheric CO2. Annual Review of Plant Physiology and Plant Molecular Biology 44: 609-632. Centritto, M., Lee, H.S.J., and Jarvis, P.G. 1999. Increased growth in elevated [CO2]:  an early, short-term response? Global Change Biology 5: 623-633. Cure, J.D. and Acock, B. 1986. Crop responses to carbon dioxide doubling:  a literature survey. Agricultural and Forest Meteorology 38: 127-145. Drake, B.G. and Gonzŕlez-Meler, M.A. 1997. More efficient plants:  a consequence of rising atmospheric CO2? Annual Review of Plant Physiology and Plant Molecular Biology 48: 609-39. Dytham, C. 2003. Choosing and Using Statistics.  A Biologist's Guide.  Malden, MA, USA, Blackwell Publishing. Ehret, D.L., Lau, A., Bittman, S., Lin, W., and Shelford, T. 2001. Automated monitoring of greenhouse crops. Agronomie 21: 403-414. Gautier, H., Guichard, S., and Tchamitchian, M. 2001. Modulation of competition between fruits and leaves by flower pruning and water fogging and consequences on tomato leaf fruit growth. Annals of Botany 88: 645-652. Gruda, N. 2005. Impact of environmental factors on product quality of greenhouse vegetables for fresh consumption. Critical Reviews in Plant Sciences 24: 227-247. Hao, X., Wang, Q., and Khosla, S. 2006. Responses of a greenhouse tomato crop to summer CO2 enrichment. Canadian Journal of Plant Science 86: 1395-1400. Helmer, T., Ehret, D.L., and Bittman, S. 2005. CropAssist, an automated system for direct  97 measurement of greenhosue tomato growth and water use. Computers and Electronics in Agriculture 48: 198-215. Heuvelink, E. 1996. Re-interpretation of an experiment on the role of assimilate transport resistance in partitioning in tomato. Annals of Botany 78: 467-470. Heuvelink, E. 2005. Development processes. In Heuvelink, E., ed. Tomatoes. CABI Publishing, Cambridge, MA, pp 53-83. Heuvelink, E. and Dorais, M. 2005. Crop growth and yield. In Heuvelink, E., ed. Tomatoes. CABI Publishing, Cambridge, MA, pp 85-144. Hicklenton, P.R. and Jolliffe, P.A. 1978. Effects of greenhouse CO2 enrichment on the yield and photosynthetic physiology of tomato plants. Canadian Journal of Plant Science 58: 801-817. Ho, L.C. 1978. The regulation of carbon transport and the carbon balance of mature tomato leaves. Annals of Botany 42: 155-64. Ho, L.C. 1995. Carbon partitioning and metabolism in relation to plant growth and fruit production in tomato. Acta Horticulturae  412: 396-409. Ho, L.C. 1996. The mechanism of assimilate partitioning and carbohydrate compartmentation in fruit in relation to the quality and yield of tomato. Journal of Experimental Botany 47: 1239-1243. Ho, L.C. and Hewitt, J.D. 1986. Glasshouse crop production. In Chapman and Hall Ltd., Cambridge, Great Britain, pp 201-240. Ho, L.C. and Shaw, A.F. 1977. Carbon economy and translocation of  14C in leaflets of the seventh leaf of tomato during leaf expansion. Annals of Botany 41: 833-848. Ho, L.C., Shaw, A.F., Hammond, B.W., and Burton, K.S. 1983. Source-sink relationships and carbon metabolism in tomato leaves 1. 14C assimilate compartmentation. Annals of Botany 52: 365-372. Ho, L.C., Sjut, V., and Hoad, G.V. 1982. The effect of assimilate supply on fruit growth and hormone levels in tomato plants. Plant Growth Regulation 1: 155-171. Kimball, B.A. 1983. Carbon dioxide and agricultural yield:  an assemblage and analysis of 430 prior observations. Agronomy Journal 75: 779-788. Kramer, P.J. 1981. Carbon dioxide concentration, photosynthesis, and dry matter. BioScience 31: 29-33. Littell, R.C., J. Pendergrast, and R. Natarajan 2000. Modelling covariance structure in the analysis of repeated measures data. Statistics in Medicine 19: 1763-1819.  98 Littell, R.C., P.R. Henry, and C.B. Ammerman 1998. Statistical analysis of repeated measures data using SAS procedures. Journal of Animal Science 76: 1216-1231. Mortensen, L.M. 1987. Review:  CO2 enrichment in greenhouse. Crop Responses. Scientia Horticulturae 33: 1-25. Murray, D.R. 1995. Plant responses to carbon dioxide. American Journal of Botany 82: 690-697. Nederhoff, E.M., De Koning, A.N.M., and Rijsdijk, A.A. 1992. Leaf deformation and fruit production of glasshouse grown tomato (Lycopersicon esculentum Mill.) as affected by CO2, density and pruning. Journal of Horticultural Science 67: 411- 420. Nederhoff, E.M. and Vegter, J.G. 1994a. Photosynthesis of stands of tomato, cucumber and sweet pepper measured in greenhouses under various CO2-concentrations. Annals of Botany 73: 353-361. Nederhoff, E.M. and Vegter, J.G. 1994b. Canopy photosynthesis of tomato, cucumber and sweet pepper in greenhouses: measurements compared to models. Annals of Botany 73: 421-427. Nederhoff, E.M. 1994. Effects of CO2 concentration on photosynthesis, transpiration and production of greenhouse fruit vegetable crops. Wageningen. Peet, M.M. and Welles, G. 2005. Greenhouse tomato production. In Heuvelink, E., ed. Tomatoes. CABI Publishing, Cambridge, MA, pp 257-304. Picken, A.J.F., Stewart, K., and Klapwijk, D. 1986. Germination and vegetative development. In Chapman and Hall Ltd., Cambridge, Great Britain, pp 111-166. Poorter, H. and Navas, M.L. 2003. Plant growth and competition at elevated CO2:  on winners, losers and functional groups.  Tansley Review. New Phytologist 157: 175-198. Portree, J. 1996. Greenhouse Vegetable Production Guide. Victoria, BC, Extension Systems Branch, BCMAFF. Pritchard, S.G., Rogers, H.H., Prior, S.A., and Peterson, C.M. 1999. Elevated CO2 and plant structure: a review. Global Change Biology 5: 807-837. Rogers, H.H. and Dahlman, R.C. 1993. Crop responses to CO2 enrichment. Vegetatio 104/105: 117-131. SAS Institute . Statistical Analysis Software.  2003. Cary, NC, SAS Institute Inc. Schwarz, D. and Klaring, H.P. 2001. Allometry to estimate leaf area of tomato. Journal of Plant Nutrition 24: 1291-1309.  99 Slack, G. 1986. CO2 enrichment of tomato crops. In CRC Press, Boca Raton, Florida, pp 151-163. Tripp, K.E., Peet, M.M., Pharr, D.M., Willits, D.H., and Nelson, P.V. 1991. CO2- enhanced yield and foliar deformation among tomato genotypes in elevated CO2 environments. Plant Physiology 96: 713-719. Valantin, M., Gary, C., Vaissičre, B., Tchamitchian, M., and Bruneli, B. 1998. Changing sink demand affects the area but not the specific activity of assimilate sources in cantaloupe (Cucumis melo L.). Annals of Botany 82: 711-719. van de Vooren, J., Welles, G.W.H., and Hayman, G. 1986. Glasshouse crop production. In Chapman and Hall Ltd., Cambridge, Great Britain, pp 581-623. Wittwer, S.H. and Honma, S. 1979. Greenhouse tomatoes, lettuce and cucumbers.  225. 1979. East Lansing, Michigan State University Press. Yelle, S., Beeson Jr., R.C., Trudel, M.J., and Gosselin, A. 1990. Duration of CO2 enrichment influcences growth, yield, and gas exchange of two tomato species. Journal of American Society for Horticultural Science 115: 52-5  100 4 Canopy profiles of starch and leaf mass per area in greenhouse tomato and the relationship with leaf area and fruit load5,6,7 4.1 INTRODUCTION Effective CO2 dosing is difficult for greenhouse growers because they do not have meaningful and timely plant measures to evaluate.  Plant-based measures would be helpful because sensors that measure the quantity of CO2 in the greenhouse atmosphere do not convey information on plant assimilation and utilization of CO2.  Therefore, growers do not currently know when CO2 dosing is at an optimal level for the plant.  Enrichment with CO2 increases photosynthesis almost immediately, but as discussed in Chapter 3, the extra photosynthate often takes weeks to become measurable in plant parts remote from assimilation sites.  To be useful for CO2 decision-making, plant measures should be responsive to CO2 enrichment on an hourly or at least daily basis.  The measurement of photosynthesis is an obvious candidate as it is very sensitive to CO2 concentration. However, the measurement of whole-plant photosynthesis is not easily achievable for a large canopy like tomato, although there has been some progress made on modeling photosynthesis (Chapter 2).  The measurement of net carbon exchange of a leaf (leaf photosynthesis) is easier to perform but it requires expensive equipment and scaling up to the whole canopy from a single leaf is not easily done.  Potentially a more useful measure for the grower is the monitoring of major products of photosynthesis, particularly sucrose and starch.  Levels of these products depend on the activities of photosynthesis and carbon  5 Aversion of this chapter will be submitted for publication.  Edwards, D., Jolliffe, P. and Ehret, D.  Canopy profiles of starch and leaf mass per area in greenhouse tomato and the relationship with leaf area and fruit load.  6 A version of this chapter has been published.  Edwards, D., Jolliffe, P., Ehret, D. and Baylis, K.  2008. Towards a plant-based method of CO2 management.  Acta Horticulturae. 797:273-278.  7 A version of this chapter has been published.  Edwards, D.R., Jolliffe, P.A. and Ehret, D.L.. 2004. Carbon status of CO2 –enriched tomato plants under commercial greenhouse conditions.  Acta Horticulturae. 633:279- 286.  101 deposition into sinks and therefore they reflect the general carbon status of the plant.  Of these two measures starch seems to be a better candidate as its pools are more stable than sucrose (Körner et al. 1995; Bertin et al. 1999).  In fact, Gary (1989) has recommended that a carbon balance model would be a more sensible way to manage greenhouse climate control decisions. To be useful for growers, a plant-centered approach has to be feasible on site and rapidly applicable to a CO2 management decision.  In Appendix 6, I report that starch content can vary from 0% to 43% of leaf dry mass (Figure A6.9).  Accordingly, leaf starch can build up sufficiently to cause detectable increases in leaf mass.  In fact leaf mass per unit area (LMA) is a recognized plant response to growth under elevated CO2 (Körner et al. 1995; Heuvelink & Marcelis 1996; Bertin et al. 1999; Roumet et al. 1999).  Therefore, it is worth testing whether leaf starch and/or LMA could be useful in decision-making for CO2 dosing. This is a unique approach to CO2 management as no studies were found in the literature that used starch or LMA as indicators for CO2 dosing. 4.1.1 Leaf starch Carbon dioxide enters a leaf through the stomata, disperses through the intercellular spaces and diffuses into the chloroplast.  In the chloroplast exposed to adequate light energy, the CO2 is fixed into triose phosphate by the Calvin Cycle.  Triose phosphate is exported to the cytoplasm through the phosphate translocator to be made into sucrose, the chief export sugar of most plants.  When photosynthesis occurs at a fast rate, phosphate cannot be recycled fast enough, which results in some of the triose phosphate being unable to leave the chloroplast since a counter flow of phosphate is needed.  When this situation occurs, triose  102 phosphate is made into starch which is stored in the chloroplast until it can be used when photosynthesis declines. Although this is the basic pathway of starch formation, regulation is more complex, depending on the attributes of the source (assimilate producer), the sink (assimilate consumer) and the transport system between them (Farrar 1996; Minchin & Thorpe 1996). A more holistic way to look at starch accumulation has been expressed by Körner et al. (1995).  Starch accumulates when more assimilates are produced than the plant is able to translocate from the chloroplast to the cytoplasm and utilize at the source.  This may result from downstream insufficiencies in: transfer from the photosynthetic cell to the phloem, transport within the phloem, unloading from the phloem, and utilization by the sinks.  A bottleneck at any of these points could cause a backup and result in starch accumulating in source leaves.  Tomato plants with low sink strength are known to accumulate starch (Ammerlaan et al. 1986; Bertin & Gary 1998).  Also, tomato plants growing in commercial greenhouses have been regarded as sink limited (Nederhoff 1994) because some fruit are regularly removed to ensure uniform fruit growth.  If commercial greenhouses are chronically oversupplying CO2, resulting in assimilate production beyond what is needed for fruit growth, it would be expected that large amounts of starch will accumulate in the shoot canopy. 4.1.2 Leaf mass per unit leaf area Leaf mass per unit leaf area (LMA), sometimes known as specific leaf mass or specific leaf weight, is the ratio of leaf dry mass over leaf area.  It is less commonly used than specific leaf area (SLA) which is the reciprocal of LMA.  For my purposes, LMA is  103 preferred to SLA because in this work LMA will be compared to starch; LMA increases ought to relate directly to leaf starch accumulation, instead of inversely as with SLA. The influence of starch on leaf dry mass has only been quantified in a few studies (Ehret & Jolliffe 1985; Cao & Tibbitts 1997; Bertin & Gary 1998; Bertin et al. 1999).  This is a little surprising as SLA is central to the estimation of LAI, an important component of many growth models (Bertin & Gary 1998; Marcelis et al. 1998; Bertin et al. 1999; Heuvelink 1999; Lee & Heuvelink 2003; Ewert 2004).  In potato and bush bean, starch was found to have a very strong (R2 =0.97) influence on LMA (Ehret & Jolliffe 1985; Cao & Tibbitts 1997).  In tomato, Bertin and coworkers reported starch exhibited a strong influence on LMA (R2=0.93) for one study (Bertin & Gary 1998) and a weaker one in a later study (Bertin et al. 1999). Starch measurement requires laboratory procedures, and might be overly complicated for routine use by a commercial grower. However, in terms of ease of measurement, LMA would be a better candidate for use by a grower since if a fixed area of leaf tissue is sampled then only an oven and a balance are needed for LMA determination.  However, LMA is also affected by structural and soluble constituents and its dependence on variation in starch needs to be quantified. In order to employ either starch or LMA as CO2 management tools, we need to know how both of these measures vary throughout the canopy, how they may change during the growing season, and what, if any, are the effects of plant factors such as leaf area and fruit load.  This information is required to recommend where and when tissue should be collected for analysis.  Also, this information needs to be collected in commercial greenhouses because  104 commercial growing conditions and cultivation techniques are distinct from research venues, and commercial greenhouses are the places where the assessments will be applied. 4.1.3 Objectives of chapter The goal of this chapter was to investigate the levels of leaf starch and LMA in tomato shoot canopies under commercial growing conditions, and to determine the influence of season, leaf area and fruit load on leaf starch accumulation and LMA.  My specific objectives were: 1. to determine the relationship the between LMA and leaf starch, 2. to describe the vertical profiles of LMA and leaf starch in the shoot canopy during the production season, 3. to describe the interplay between fruit load and leaf starch in the canopy. 4.2 MATERIALS AND METHODS 4.2.1 Environment and crop culture in commercial greenhouses Three years of data were collected from beefsteak tomato plants growing in several commercial greenhouses located in the lower mainland of BC.  The data collection was consistent among sites and there was considerable uniformity in both plant environment and crop culture among the greenhouses. Greenhouse environment.  The three greenhouses were of Venlo design and cladded with glass.  All, or a sizable portion, of their areas were devoted to beefsteak tomato production.  The greenhouses were:  (1) South Alder Greenhouses Ltd (SA) located in Surrey, BC, 4 ha in area, where data were collected from 2000 to 2002, (2) CanAgro Produce Ltd (CA), located in Delta, BC, 5 ha in area, where data were collected in 2000 and 2001,  105 and (3) Gipaanda Greenhouses Ltd (GI), also located in Delta, BC, 7.3 ha in area where data were collected in 2002.  Greenhouse temperature, CO2 concentration, relative humidity (or humidity deficit) and global radiation (external to the greenhouse) were obtained from computer records maintained by the greenhouse managers.  South Alder and CA logged and stored these measures every 5 minutes throughout the year which enabled downloading of these data from their greenhouse environmental control system.  Only daily summaries were available from GI as their raw environmental data were regularly purged.  Monthly summaries of temperature, relative humidity (or humidity deficit), global radiation and CO2 conditions for each greenhouse are presented in Appendix 4.1. Plant culture.  All plants in the study were the indeterminate, beefsteak tomato (Lycopersicum esculentum Mill.) cultivar Rapsodie (Syngenta Seeds Inc, Boise, ID).  In 2001, Rapsodie plants in the CA greenhouse were grafted onto a Beaufort (DeRuiter Seeds Inc, Tecumseh, ON) rootstock in the nursery to increase plant vigour.  In 2002 both SA and GI greenhouses grafted Rapsodie onto the rootstook Maxifort (DeRuiter Seeds Inc, Tecumseh, ON) also to increase plant vigour. The tomato plants were raised in a high wire production system.  In this system, the plants were trained to plastic twine which was hung from an overhead hook, which in turn was attached to a wire suspended about 3 m above the ground.  Once the shoots grew to the level of the wire the plants tips were lowered (by un-spooling excess twine from the hook) and moved forward along the wire.  Shoot elongation was approximately 30 cm per week. As elongation progressed the greenhouse workers trained new growth to the twine, and lowered and moved the crop forward on the wire each week.  This method ensured that the top portion of the canopy received maximum sunlight, and it is thought to decrease stem and  106 fruit diseases (van de Vooren et al. 1986).  As well, the high wire system in combination with de-leafing (removal of leaves from bottom one to three trusses) improves fruit quality and ease of picking. The crops were maintained and fruit were harvested by the respective greenhouse staff according to conventional commercial practices for growing beefsteak tomatoes in the lower mainland of BC.  In BC, normal commercial production starts in December with the planting of the new crop and ends in November the following year when crop removal is complete.  The new crop was obtained from a propagator and set into the greenhouses at a density of 2.5 plants per m2.  Plant density was increased twice during the year by training a side shoot from the main stem (also trained to the high wire).  One shoot was allowed to develop soon after planting and the second shoot was trained in February.  By the end of March the shoot density had reached the maximum of 3.75 shoots (sometimes referred to as heads) per m2, and this density was maintained until late August when a side shoot was removed.  By mid-September the shoot density was further reduced to 2.5 heads per m2, and by early October the growing point was removed to promote development of the remaining fruit.  Fruit harvesting typically commenced in mid March and ceased by mid-October.  The crop was removed by early November and the greenhouse was then cleaned and prepared for the next production cycle. In 2002 the production schedule was shifted forward for the crop in SA because plants were not set into the greenhouse until February.  However, it should be noted that the high light levels in February that year increased the rate of crop development.  Consequently, to some degree, the SA crop was able to catch up to the earlier plantings established in other greenhouses.  107 4.2.2 Collection of leaf length and fruit diameter data from plant canopies Canopy profiles of leaf length and fruit load were measured in situ for several dates each year (see Appendix 4.2 for a list of the dates).  On each day of data collection a row was randomly selected from the middle of the greenhouse.  Within this row four plants (not together) were selected for measurements and from each of these plants, six leaves at different levels in the canopy were marked for future collection and analysis for starch and LMA.  On each plant that served as a data source, the lengths of all leaves from a single shoot were measured, and their positions on the shoot were recorded. Identification of leaf position in the tomato canopy can be complicated in tomato because of the indeterminate growth of the plant and the weekly removal of lower leaves.  I followed the industry practice of identifying leaf position in the canopy as follows: descending from the shoot apex, leaf position 1 is the leaf closest to the first truss with open flowers (Figure 3.2).  This leaf could be above or below the truss.  Leaves in this position have reached approximately 40% of their full expansion, and leaves progressively younger than (i.e. above) leaf 1 were identified as 0, -1, -2 etc., and leaves progressively older (i.e. below leaf 1) were 2, 3, 4 and so on (Figure 3.2).  Leaf length in centimetres was measured with a ruler from the base of the last pair of leaflets to the end of the terminal leaflet.  Fruit load was measured by recording the position of trusses in the canopy, and the diameters of all fruit in each truss.  Diameter of the fruit was measured in millimetres with an electronic calliper from the calyx to the blossom end (i.e. top to bottom).  Trusses were labelled consecutively downward, starting with truss 1 being the truss containing mostly flowers at anthesis.  Truss position in the canopy was also fixed relative to the closest leaf.  108 4.2.3 Calculation of leaf area and fruit load Leaf length and fruit diameter were converted to leaf area and fruit volume, respectively, by predetermined relationships.  Leaf area was determined from an allometeric function that related blade length to leaf area for the cultivar Rapsodie (Appendix 1.1). Equation 4.1 ))(45.253.0( LLlneLA ∗+−= where LA is leaf area in m2 and LL is leaf length in meters.  The fruit volume was calculated by assuming the fruit approximated a sphere.  Fruit diameter was converted to volume of a sphere using  Equation 4.2 3 3 4 rFV π= where FV is fruit volume in cm3 and r is fruit radius, calyx to blossom end, in cm.  Volume was calculated for each fruit on a truss and those values were added together to give a total volume of fruit for the truss. 4.2.4 Leaf collection, tissue analysis, and calculation of LMA and leaf starch After the in situ leaf and fruit measurements were made, tissue was collected for leaf mass per area (LMA) and starch analysis.  Between 12 and 13 h the six marked leaves per plant were collected, placed in a plastic bag, and stored with a cold pack in an insulated container.  In most cases the first leaf collected was the leaf in position 1 followed by leaf selection every three to four leaves descending through the canopy.  Occasionally a leaf above leaf 1 was collected if it was deemed large enough.  For each of the selected leaves the middle four leaflets were identified and ten disks were punched from each leaflet using a 10 mm diameter cork borer.  The ten leaf punches per leaflet were placed in an aluminum foil  109 packet and immersed in liquid nitrogen.  The total harvested area of these 10 punches was 0.000785 m2 (7.85 cm2) and consumed most of the leaflet.  These samples will be referred to as leaflet samples.  This procedure took approximately two hours and was carried out in the head house of the respective greenhouse.  The leaflet tissue samples were then transported to the University of British Columbia and transferred to a –70 C freezer where they were held until analysis. Sample preparation for LMA analysis entailed freeze drying the leaflet samples and storing them with anhydrous calcium sulphate in a desiccator or in a –70 C freezer.  Prior to the determination of their dry mass the leaflet tissue samples were thawed and transferred to a desiccator with anhydrous calcium sulphate (Drierite), to allow them to equilibrate to room temperature.  After equilibration the mass of each sample was determined in grams.  Leaf mass per area (g m-2) was calculated by dividing the dry mass of these leaflet samples (10 disks) by the leaf area for the 10 disks (0.000785 m2). The procedure for preparing and analyzing samples used for leaf starch is detailed in Appendix 6.  Therefore, only an overview will be provided here.  Two methods of leaf starch determination were employed:  amyloglucosidase-glucose oxidase-peroxidase (AGOP) (see Appendix 6, section A6.2.1) and near-infrared spectroscopy (NIRS) (see Appendix 6, section A6.2.6).  After the dry mass was obtained for the calculation of LMA (see above) the leaflet samples were transferred to small 1.5 ml Eppendorf-style centrifuge tubes and stored in a desiccator or in a –70 C freezer with calcium sulphate.  Samples destined for analysis by NIRS were thawed if necessary and coarsely broken up with a spatula prior to having their spectra determined.  The starch content of a leaflet sample was determined from a predictive equation that was developed between the NIRS spectra and the reference method (AGOP).  110 Samples destined for analysis by AGOP were thawed and ground to a fine powder by hand using a mortar and pestle.  After grinding, the leaflet samples were massed, the soluble sugars were extracted using hot ethanol (80%) and the starch was hydrolysed to glucose by the enzyme amyloglucosidase.  The liberated glucose was quantified using the enzyme glucose oxidase-peroxidase in the presence of a chromatophore.  The original starch content of the tissue was determined with the aid of wheat starch and glucose standards.  In both analyses the starch measured was on a per leaflet sample basis and thus was divided by 0.000785 m2 to give starch on a per unit leaf area (m2) basis, or divided by the leaf dry mass to give starch on a percentage basis (g g-1).  Starch on a per leaf basis was calculated by first determining the mean starch per area of the leaf using the four leaflet samples.  The starch per area for a leaf was then multiplied by the area of the leaf.  Starch per shoot was calculated by totalling the starch per leaf from the functions (described below) for each leaf position.  Starch per area of greenhouse space was determined by multiplying the starch per shoot by the 3.75, the number of shoots per m2 for the timeframe examined. 4.2.5 Data analysis Data pre-treatment.  Several types of regression analysis were used to quantify plant responses.  During the analysis it was sometimes necessary to transform or remove outlying data to ensure the data exhibited a normal distribution and the variance was homoscedastic. That the data were normally distributed was tested using the Shapiro-Wilk and Kolmogorov- Smirnov tests in the univariate procedure of SAS version 9.1.3 (SAS Institute 2003).  In the leaf starch - LMA analysis using Model II regression, 14 of the 1426 observations were considered outlying data based on the results of the stem and leaf, box and residual plots (in the univariate procedure of SAS version 9.1.3).  These 14 observations were removed.  111 Weighting.  Weighting of the regression equations was used to improve the estimation of the predicted line by giving more influence to data that had a higher weight. Prior to either linear or non-linear regression analysis the mean and variance of the dependent variable were calculated for each leaf position.  The regression was performed on the mean values with the inverse variance of the mean used as the weight.  This enabled those leaf positions having a smaller variance to be weighted more strongly and exert more influence on the subsequent predicted line.  The procedure was done for the LMA, leaf starch per area, leaf area and starch per leaf. Linear regression.  Leaf mass per unit leaf area, starch per unit leaf area and starch per leaf were all dependant variables regressed with the independent variable, leaf position, using linear regression.  Prior to analysis the means were plotted against leaf position to give an indication of what type of model should be fitted.  Depending on this pattern it was necessary to fit quadratic, cubic, quartic or quintic functions.  If a relationship between the dependant and independent variable was evident or suspected to be present, the analysis continued to the ANOVA, using the regression procedure of SAS version 9.3.  If the p-value of the ANOVA was 0.05 or less then there was evidence that the independent variable was a significant source of variation in the behaviour of the dependent variable.  A t-test was used to test the significance of the coefficients being zero and thus not influential in the regression.  If the null hypothesis (coefficient equal to zero) was rejected then the measure was included.  The measure was dropped if the null hypothesis was accepted.  Often a trial and error approach was necessary to determine the best fit for a particular dependent variable.  112 Model II regression.  Model II regression was used to determine the functional relationship between starch per area and LMA for each leaflet sample.  Model II regression was necessary because neither variable was fixed, instead being random and measured with error (Sokal & Rohlf 1995; Legendre 2001).  The Major Axis method of Model II regression was chosen since leaf starch per area and LMA are measured in the same units, g m-2 and their variances of error were deemed similar (Legendre 2001).  Data were analysed using the method and computer program of Legendre (2001). Non-linear regression.  Non-linear regression was used to model the pattern of leaf area increase in the canopy.  As leaf area exhibits a sigmoid response with canopy position (Appendix 4.6) several non-linear functions were investigated.  The Gompertz function was chosen over the logistic (Richards), Weibull and Beta functions because it was the only function that achieved convergence for the leaf area data.  Equation 4.3 ε+= −− d ec dL beY )1(  where ⎟⎠ ⎞⎜⎝ ⎛= b a cd ln , a is asymptotic leaf area, L is leaf position, b is leaf area when L=0, c is a growth rate coefficient, and d is the exponent (Jolliffe & Lin 1997).  The fit of the predicted line to the data were deemed to be acceptable by examining plots of the residues versus leaf position and visually assessing how well the predicted line fit the data.  It was not possible to calculate a pseudo-R2 as the Gompertz equation does not contain a constant and would not converge when a constant was added.  These functions were also used to calculate the leaf position at 95% of the maximum leaf area.  113 4.3 RESULTS These studies produced a large amount of descriptive data collected during three cropping years.  To improve the readability of this Chapter one or two exemplar plots have been shown and the remaining results are reported in Appendices 4.3 through 4.7. 4.3.1 The influence of starch on leaf mass per area Depending on the location of the leaf in the canopy and date of sample collection the percentage of starch ranged from 1 to 37% of the leaf dry mass (Figure 4.1).  Among the leaf positions, leaves in position 1 (near the top of the canopy) exhibited the highest concentration of starch at 21.3% of the dry mass (mean of dates and greenhouses).  Starch in these upper leaves was significantly greater than for leaves or for position 13 (lower canopy) at 2.7% starch (ANOVA, p<0.0001).  These data suggest that starch variation should have a significant effect on LMA with the potential effect being dependent on leaf canopy position and date of sampling. To resolve the influence of leaf position and seasonality on the relationship between leaf starch and LMA, more intensive sampling was carried out in 2001 and 2002.  In the March 2001 data set, there was a positive relationship between LMA and leaf starch per unit leaf area, with a fairly balanced distribution of data (Figure 4.2A).  This finding contrasts to the results of the 14 other data sets.  Using the September 2001 data set as an example (Figure 4.2B-D), a cluster of LMA data were present between 0 and 3 g m-2 of starch.  This cluster corresponded to a wide range of LMA values, 25 to 65 g m-2 (Figure 4.2B).  Closer inspection revealed that this cluster of data were from leaves in the lower region of the canopy. When this cluster of data were separated and re-plotted it was evident that two  114  Figure 4.1 Starch as a percentage of leaf dry mass for tomato leaf tissue collected at three positions in the canopy from two commercial greenhouses during the 2000 growing season.  SA is South Alder and CA is CanAgro.  Leaves were collected from position 1 (next to the truss with flowers), position 7 (mid-canopy) and from leaf position 13 (lower canopy).  Bars above each column indicate standard errors of the means, N=12. 0.0 5.0 10.0 15.0 20.0 25.0 30.0 35.0 40.0 45.0 May 26 Jun 29 Aug 11 Oct 6 May 31 Jun 26 Aug 17 Oct 5 Date St ar ch  (% ) 1 7 13 Greenhouse SA Greenhouse CA Leaf position:  115 distinct relationships were present in the data sets (Figure 4.2 C and D), one for the leaves in the top part of the canopy and another for leaves at the bottom part.   This pattern also occurred in the remaining data sets for 2001 and 2002 (Appendix 4.3). The relationship between starch and LMA was quantified using a Model II regression procedure (Figure 4.3 and Table 4.1).  The influence of starch on LMA, as indicated by the slope of the regression line, was up to 10 times larger for the lower canopy leaves than for the upper canopy leaves.  The slopes for the leaves from the upper canopy ranged from 1.5 to 2.8 and those from the lower canopy exhibited slopes that ranged from 7.0 to 80.0 (Table 4.1).  The greater slope for the lower canopy leaves indicates that starch has a larger effect on LMA there than for the upper canopy leaves.  However, it should be noted that the correlation coefficients (R) were often lower for the lower canopy than for the upper canopy leaves. A low R indicates that starch is only explaining a small amount of the variation in LMA.   The equation constants (y-intercepts) from the upper and lower canopy leaves were approximately in the same range; although the lower canopy leaves tended to have lower regression constant values.  The difference in y-intercepts (when starch is zero) indicates that that the non-starch mass of the leaf may also be distinct for the two strata of the canopy. 4.3.2 Canopy profiles of LMA and starch An overview of leaf starch and LMA data indicated seasonal and at times yearly differences.  The levels of leaf starch per unit leaf area in a shoot canopy ranged from 2 to 17 g m-2 while LMA was less variable being between 28 and 74 g m-2 (Figure 4.4B).  A more detailed analysis of the canopy was carried out to determine how these measures changed with leaf position.  These profiles were also used to clarify the two distinct relationships between leaf starch and LMA present in the canopies described in the previous section.  116 Figure 4.2 The starch content per unit leaf area and corresponding leaf mass per unit leaf area from samples of 0.000785 m2 of tomato leaflet tissue collected from South Alder (SA) and CanAgro (CA) greenhouses.  Panel A: data collected in March 2001.  Panel B: data collected in September 2001.  Panel C is redrawn from B with data from leaf position <=7 for SA and <=6 for CA.  Panel D is redrawn from B with data from leaf position >=6 for SA and data for leaf position >=7 for CA. 0.0 10.0 20.0 30.0 40.0 50.0 60.0 70.0 80.0 0 10 20 30 40 x y Series1 Series2 0.0 10.0 20.0 30.0 40.0 50.0 60.0 70.0 80.0 0 10 20 30 40 Leaf starch (g m-2) LM A  (g  m -2 ) Gnh 1 Gnh 2 0.0 10.0 20.0 30.0 40.0 50.0 60.0 70.0 80.0 0 10 20 30 40 Leaf starch (g m-2) y SA CA 0.0 10.0 20.0 30.0 40.0 50.0 60.0 70.0 80.0 0 10 20 30 40 Leaf starch (g m-2) LM A   ( g m -2 ) SA CA A B C D  117 Figure 4.3 The starch content and corresponding leaf mass area from samples of 0.000785 m2 of tomato leaflet tissue from South Alder (SA) and Gipaanda (GI) greenhouses.  Data were collected in June 2002. Panel A: upper canopy and Panel B, lower canopy.  Line equations for the Model II regression are in Table 4.1. 0.0 10.0 20.0 30.0 40.0 50.0 60.0 70.0 80.0 0.0 2.0 4.0 6.0 8.0 10.0 12.0 14.0 16.0 18.0 20.0 Leaf starch (g m-2) LM A  (g  m -2 ) SA GI 0.0 10.0 20.0 30.0 40.0 50.0 60.0 70.0 80.0 0.0 2.0 4.0 6.0 8.0 10.0 12.0 14.0 16.0 18.0 20.0 Leaf starch (g m-2) LM A  (g  m -2 )   SA    GI A B  118 Table 4.1 Model II regression equations for the influence of leaf starch (X) on leaf mass area (Y) for upper (A) and lower (B) regions of tomato plant canopies.  Data were collected from commercial tomato greenhouses during 2001 and 2002.  Table 4.1A Model II regression equations for the influence of leaf starch (X) on leaf mass area (Y) for upper regions of tomato plant canopies. 95% confidence interval Year Mn Gnh1 L2 Equation Intercept  Slope 3R P4 2001 Mar SA All3 Y=16.7+2.0X 15.1, 18.0 1.9, 2.2 0.91 0.001 2001 Mar CA All3 Y=19.2+1.6X 17.4, 20.7 1.5, 1.8 0.91 0.001 2002 May SA ≤ 11 Y=22.0+2.7X 16.2, 25.9 2.2, 3.3 0.76 0.001 2002 May GI ≤ 11 Y=29.0+1.7X 26.4, 31.2 1.5, 2.0 0.84 0.001 2001 Jun SA ≤ 11 Y=33.1+1.6X 29.9, 35.6 1.4, 2.0 0.81 0.001 2002 Jun SA ≤ 11 Y=22.7+2.6X 18.3, 25.8 2.2, 3.3 0.81 0.001 2002 Jun GI ≤ 11 Y=28.0+2.8X 19.3, 32.6 2.1, 4.2 0.61 0.001 2002 Jul SA ≤ 9 Y=28.8+2.0X 26.2, 30.8 1.7, 2.4 0.88 0.001 2002 Jul GI ≤ 9 Y=27.6+2.8X 4.6, 35.1 1.7, 6.1 0.46 0.001 2002 Aug SA ≤ 11 Y=31.5+1.9X 27.8, 34.1 1.6, 2.3 0.81 0.001 2002 Aug GI ≤ 10 Y=29.0+2.1X 22.1, 33.1 1.6, 2.9 0.68 0.001 2001 Sep SA ≤ 7 Y=27.5+1.4X 21.7, 31.7 1.2, 1.8 0.77 0.001 2001 Sep CA ≤ 6 Y=27.3+1.9X 22.7, 30.3 1.5, 2.5 0.71 0.001 2002 Sep SA ≤ 8 Y=25.8+1.5X 22.1, 28.4 1.2, 2.0 0.78 0.001 2002 Sep GI ≤ 8 Y=22.1+1.9X 13.9, 26.6 1.9, 3.8 0.65 0.001  Table 4.1B Model II regression equations for the influence of leaf starch (X) on leaf mass area (Y) for lower regions of tomato plant canopies. 95% confidence interval Year Mn Gnh1 L2 Equation Intercept  Slope 3R P4 2002 May SA ≥ 12 Y=9.3+15.0X 15.7, 147.9 7.1,156.6 0.36 0.052 2002 May GI ≥ 12 Y=15.8+11.0X 2.2, 20.8 7.0, 25.5 0.57 0.002 2001 Jun SA ≥ 12 Y=11.1+17.9X - - 0.41 0.028 2002 Jun SA ≥12 Y=16.0+18.8X 8.0, 19.7 13.7, 29.9 0.64 0.001 2002 Jun GI ≥ 12 Y=16.1+15.3X 3.9, 22.9 10.2, 30.4 0.57 0.002 2002 Jul SA ≥ 11 Y=21.3+11.3X 16.7, 24.0 9.0, 15.5 0.73 0.001 2002 Jul GI ≥11 Y=-1.0+31.0X - - 0.35 0.018 2002 Aug SA ≥ 12 Y=18.3+17.6X 5.2, 24.0 12.6, 29.0 0.64 0.001 2002 Aug GI ≥ 11 Y=26.0+ 7.0X 20.2, 29 5.4, 10.1 0.72 0.001 2001 Sep SA ≥ 8 Y=-87.3+80.0X - - 0.12 0.245 2001 Sep CA ≥ 7 Y= 9.7+20.5X - - 0.36 0.007 2002 Sep SA ≥ 9 Y= 8.7+14.0X -9.2, 15.1 9.5, 26.8 0.52 0.001 2002 Sep GI ≥ 9 Y=20.0+19.3X 7.0, 24.0 12.8, 42.0 0.50 0.001 1. Mn is month, Gnh is greenhouse, SA is South Alder, CA is CanAgro and GI is Gipaanda. 2. L is leaf position, where leaf 1 is the leaf closest is the first flowering truss. 3. R is the correlation coefficient. 4. One tailed probability for slope, - means confidence interval could not be calculated.  119   Figure 4.4 The seasonal range of leaf starch (A) or leaf mass per unit leaf area (LMA) (B) collected from tomato cv Rapsodie canopies.  Plants were from three commercial greenhouses during the 2000 to 2002 growing seasons. 0 2 4 6 8 10 12 14 16 18 1 2 3 4 5 6 7 8 9 10 11 12 Month Le af  s ta rc h (g  m -2 ) 2000 2001 2002 10 20 30 40 50 60 70 80 1 2 3 4 5 6 7 8 9 10 11 12 Month LM A  (g  m -2 ) 2000 2001 2002 A B  Jan     Feb   Mar    Apr   May    Jun     Jul     Aug    Sep    Oct    Nov    Dec  120 The vertical profile of LMA in the canopy exhibited one of four patterns.  In the majority of data sets the LMA declined steadily from the top to the bottom of the canopy (examples in Figure 4.5A, E G, Table 4.2 and Appendix 4.4).  In two of the data sets LMA exhibited modest curvature near the top or bottom of the canopy (examples in Figures 4.5A and E).  Leaf mass per area in another two data sets exhibited a parabolic like appearance (examples in Figure 4.5C and Appendix 4.4H), and in a further two data sets the LMA was highly variable and not possible to model simply (examples in Figures 4.5C and Appendix 4.4H).  Despite these different profiles, the fitted regression lines (where appropriate) were usually successful at modeling the decline in LMA, with R2 values ranging from the lowest at 0.21 to the highest at 0.93 (Table 4.2 and Appendix 4.5). The profile of starch concentration per unit leaf area also declined from the top to the bottom of the canopy, but exhibited patterns distinct from those found for LMA.  For all data sets the maximum leaf starch contents were between 10 and 20 g m-2 for leaves in position 4 or above (Figures 4.5D, F, H and Appendix 4.4).  In the majority of the data sets the very top leaves sampled exhibited slightly less starch than did the immediately lower leaves (Figure 4.5 B, D, F and H).  Maximum starch content also varied substantially with the greenhouse and/or date of sampling.  In all of the data sets the minimum amount of leaf starch was consistently 2 g m-2 or less.  The leaf position where the minimum starch content was first reached was variable among the canopies, reflecting differences in the decline profile for starch.  In general, the minimum in starch did not occur before leaf position 6 (Figures 4.5D, F and H and Appendix 4.4). The vertical profile of leaf starch per unit leaf area exhibited two general patterns of decline.  For all but two of the data sets (Figure 4.5B) the profile of   121 Figure 4.5 Leaf mass per unit leaf area (LMA) and starch per unit leaf area from leaves in tomato plant canopies from two commercial greenhouse, South Alder (SA) and CanAgro (CA).  Leaf position 1 is the leaf closest to the first flowering truss, - leaves are younger and + leaves are older.  Panel A and B, data collected in March 2001, Panel C and D, data collected in September 2001.  Line equations are given in Table 4.2. 15.0 20.0 25.0 30.0 35.0 40.0 45.0 50.0 55.0 60.0 65.0 -3 0 3 6 9 12 15 18 21 24 Leaf position LM A  (g  m -2 )  SA  CA 0.0 2.0 4.0 6.0 8.0 10.0 12.0 14.0 16.0 18.0 20.0 -3 0 3 6 9 12 15 18 21 24 Leaf position Le af  st ar ch  (g  m -2 )  SA  CA 15.0 20.0 25.0 30.0 35.0 40.0 45.0 50.0 55.0 60.0 65.0 -3 0 3 6 9 12 15 18 21 24 Leaf position LM A  (g  m -2 )  SA  CA 0.0 2.0 4.0 6.0 8.0 10.0 12.0 14.0 16.0 18.0 20.0 -3 0 3 6 9 12 15 18 21 24 Leaf position Le af  st ar ch  (g  m -2 )  SA  CA A B C D  122 Figure 4.5 Leaf mass per unit leaf area (LMA) and starch per unit leaf area from leaves in tomato plant canopies from two commercial greenhouses, South Alder (SA) and CanAgro (CA).  Leaf position 1 is the leaf closest to the first flowering truss, - leaves are younger and + leaves are older.  Panel E and F, data collected in May 2002, Panel G and H, data collected in September 2002.  Line equations are given in Table 4.2. 15.0 20.0 25.0 30.0 35.0 40.0 45.0 50.0 55.0 60.0 65.0 -3 0 3 6 9 12 15 18 21 24 Leaf position LM A  (g  m -2 )  SA  GI 15.0 20.0 25.0 30.0 35.0 40.0 45.0 50.0 55.0 60.0 65.0 -3 0 3 6 9 12 15 18 21 24 Leaf position LM A  (g  m -2 )  SA  GI 0.0 2.0 4.0 6.0 8.0 10.0 12.0 14.0 16.0 18.0 20.0 -3 0 3 6 9 12 15 18 21 24 Leaf position Le af  st ar ch  (g  m -2 )  SA  GI 0.0 2.0 4.0 6.0 8.0 10.0 12.0 14.0 16.0 18.0 20.0 -3 0 3 6 9 12 15 18 21 24 Leaf position Le af  st ar ch  (g  m -2 )  SA  GI E F G H  123  Table 4.2 Linear regression equations for leaf mass or leaf starch per unit leaf and canopy leaf position for Figure 4.5. Gnh1 Leaf mass per unit leaf area (g m-2)2 Leaf starch per unit leaf area (g m-2)2 Year Month  Equation3 4R2  Equation3 4R2 2001 March  SA Y=34.2+0.3X-0.095X2 0.88  Y=9.7-0.6X-0.2X2+0.008X3 0.88 2001 March CA Y=45.4-1.5X 0.93  Y=14.5+0.07X-0.1X2+0.006X3 0.96 2001 September SA Y=47.5-2.4X-0.2X2 0.62  Y=14.2-3.3X+0.3X2-0.007X3 0.90 2001 September CA mean of 42.86 -  Y=7.4+1.6X-0.1X2-0.003X3 0.66 2002 May SA Y=56.1+0.19X-0.61X2+0.03X3 0.92  Y=11.4-0.6X-0.07X2+0.004X3 0.98 2002 May GI Y=50.7-1.5X 0.81  Y=13.5-1.7X+0.06X2 0.93 2002 September SA Y=40.7-0.9X 0.69  Y=13.8-2.1X+0.1X2-0.002X3 0.90 2002 September GI Y=48.4-1.1X 0.61  Y=23.4-5.7X+0.8X2-0.01X3 0.91 1. Gnh is greenhouse, SA is South Alder, CA is CanAgro and GI is Gipaanda. 2. All regressions were weighted by the inverse of the variance. 3. Y is leaf mass per area or leaf starch and X is leaf position. 4. R2 is the coefficient of determination.  124 decline in starch was reminiscent of the attenuation of light in the canopy (Figure 4.5D, F and G).  For these data sets the starch content declined rapidly in the top third of the canopy and then settled to a minimum value for the remainder of the leaves (Figure 4.5D, F and G and Appendix 4.4).  These data were modeled occasionally with a quadratic but more commonly with a cubic polynomial.  Two data sets, collected in March 2001, exhibited a distinct profile that did not seem to exhibit a light attenuation-type profile.  For those, the starch concentration was highest from leaves in position -3 through to about 8, below which it declined to a minimum of 1 g m-2 by the last sampled leaf (Figure 4.5B).  These data were best described using a cubic polynomial. The leaf position where the starch concentration first reached a minimum was determined visually from these figures using both the raw data and predicted responses. Examination of these data revealed a seasonal effect.  Profiles collected in March exhibited a steady reduction in starch content throughout the canopy, June to August profiles reached a minimum starch content by leaf position 11 or 12 and data collected in September exhibited profiles where the minimum starch was reached closer to the top, at leaf positions 8 or 9. 4.3.3 Profile of leaf area in the canopy The total leaf area of the shoot canopy varied among sampling dates and greenhouses. Leaf area per shoot ranged from 0.572 to 1.378 m2 in 2001 and 0.624 to 1.088 m2 in 2002 (Table 4.3).  There was a small reduction in total leaf area per shoot from June to August 2002.  The leaf area predicted from the Gompertz functions were used to determine the first leaf position exhibiting full leaf expansion.  Since the Gompertz function exhibits an upper asymptote which increases until infinity, the leaf position whose area was 95% of the area of the last sampled leaf was considered to be the first fully expanded leaf (Table 4.3).  The leaf  125 Table 4.3 Canopy and individual leaf areas from tomato shoot canopies measured from commercial greenhouses for several months in 2001 (A) and 2002 (B).  Table 4.3A 2001   Canopy leaf area per shoot (m2)2  95% maximum area3 Month Gnh1  No. of Leaves Meas. Model  Area of a leaf (m2) L March SA  22 1.190 1.098  0.0657 5 March CA  25 1.378 1.491  0.0906 9 May SA  19 0.628 0.648  0.0573 11 May CA  22 1.079 0.975  0.0626 10 June SA  20 0.712 0.644  0.0509 13 June CA  - - -  - - September SA  22 0.572 0.571  0.0337 5 September CA  24 0.629 0.701  0.0368 4 1. Gnh is greenhouse, SA is South Alder and CA is CanAgro. 2. Leaf area is the mean of 4 plants, only leaves greater than 1 cm  length were included, Meas is measure and Model is leaf area from the Gompertz model. 3. 95% max is 95% of maximum leaf area and the leaf position (L) where attained, calculated from Gompertz model. N=4.   Table 4.3B 2002   Canopy leaf area per shoot (m2)2  95% maximum area 3 Month Gnh1  No. of Leaves Meas. Model  Area of a leaf (m2) L May SA  24 1.088 1.075  0.0645 9 May GI  22 0.967 0.957  0.0650 10 June SA  29 1.010 0.944  0.0409 6 June GI  26 0.868 0.859  0.0473 8 July SA  29 0.969 0.961  0.0439 8 July GI  26 0.710 0.699  0.0399 7 August SA  27 0.624 0.608  0.0309 6 August GI  23 0.714 0.705  0.0422 7 September SA  27 0.844 0.819  0.0400 5 September GI  23 0.961 0.936  0.0570 5 1. Gnh is greenhouse, SA is South Alder and GI is Gipaanda. 2. Leaf area is the mean of 4 plants, only leaves greater than 1 cm length were included, Meas is measure and Model is leaf area from the Gompertz model. 3. 95% max is 95% of maximum leaf area and the leaf position (L) where attained, calculated from Gompertz model. N=4.  126 position at 95% maximum area was found to be variable, varying from 4 to 13 for the 2001 data and 5 to 10 for 2002 data (Table 4.3).  There appeared to be a seasonal effect, as the leaf position at full expansion moved closer to the top of the canopy as the season progressed and this was not related to the number of leaves in the canopy (Table 4.3). 4.3.4 Canopy profiles of starch content per leaf Starch content, on a per leaf basis, was obtained by combining the information available on the area of each leaf (m2) and the concentrations of starch per unit leaf area (g m-2).  In all cases it was possible to fit a function to the vertical profile of starch per leaf in the canopy.  However, a variety of polynomials (quadratic, cubic, quartic and quintic) were used to describe the response of starch per leaf to leaf position (Figure 4.6, Table 4.4 and Appendix 4.7).  In most of the sampled canopies, total starch per leaf was initially low at the top of the canopy, rising to a peak by leaf position 3 to 5 and declining to near zero for leaves at position 9 and higher (Figure 4.6).  While young leaves had high starch concentrations, they did not contribute greatly to canopy starch because their areas were small.  The major contributors to canopy starch were the leaves in position 3 to 5 as they were closer to full expansion, even if their starch concentrations were lower than the younger leaves. Examining the contribution of each leaf revealed that 90% of the total canopy starch occurs for leaves in position 9 and younger (Figure 4.6).  Further analysis with the profile of leaf area indicated that 40 to 91% of the total canopy starch occurred before leaves had reached 95% of full expansion (Table 4.5).  The data in Table 4.3B indicate that leaves become fully expanded by position 5 to 10, well beyond the peak starch concentration per leaf which occurred at leaf position 1 (Figure 4.6).  The total amount of starch held in a shoot varied from 1.12 g to 10.76 g (Table 4.5).  Canopies early in the season (March 2001 and May  127  Figure 4.6 Starch content per leaf and the cumulative contribution of each leaf to total starch for shoot canopies of plants in two commercial greenhouses.  Greenhouses were South Alder (SA) and Gipaanda (GI).  All data were collected in 2002, Panel A: May, B: June, C: July, D: August and E: September.  S is for starch and C is for cumulative.  Line equations for starch per leaf are given in Table 4.4. 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 -3 0 3 6 9 12 15 18 21 24 Leaf position St ar ch  p er  le af  (g ) 0 10 20 30 40 50 60 70 80 90 100 Cu m ul at iv e sta rc h (% ) AS GS AF GF 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 -3 0 3 6 9 12 15 18 21 24 Leaf position St ar ch  p er  le af  (g ) 0 10 20 30 40 50 60 70 80 90 100 Cu m ul at iv e sta rc h (% ) SA GA SA GA 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 -3 0 3 6 9 12 15 18 21 24 Leaf position St ar ch  p er  le af  (g ) 0 10 20 30 40 50 60 70 80 90 100 Cu m ul at iv e sta rc h (% ) SA GA SA GA 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 -3 0 3 6 9 12 15 18 21 24 Leaf position St ar ch  p er  le af  (g ) 0 10 20 30 40 50 60 70 80 90 100 Cu m ul at iv e sta rc h (% ) S_SA S_GI C_SA C_GI 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 -3 0 3 6 9 12 15 18 21 24 Leaf position St ar ch  p er  le af  (g ) 0 10 20 30 40 50 60 70 80 90 100 Cu m ul at iv e sta rc h  (% ) S_SA S_GI C_SA C_GI A B C D E  128 Table 4.4 The equations for the vertical profile of starch per leaf in the tomato canopies from two commercial greenhouses from Figure 4.6. Month Gnh 1 Equation2 R2 May SA Y = 0.34 + 0.071X + 0.0026X2 - 0.0034X3 + 0.00031X4 - 0.0000077X5 0.99 May GI Y = -0.23 + 0.73X - 0.18X2 + 0.019X3 - 0.00088X4 + 0.000016X5 1.0 June SA Y = 0.17 + 0.051X - 0.012X2 + 0.00075X3 - 0.000015X4 0.74 June GI Y = 0.047 + 0.17X - 0.033X2 + 0.0021X3 - 0.000044X4 0.92 July SA Y = 0.23 - 0.021X + 0.00054X2 0.59 July GI Y = 0.12 + 0.11X - 0.026X2 + 0.0020X3 – 0.000048X4 0.90 August SA Y = 0.21 - 0.019X + 0.00052X2 0.69 August GI Y = -0.14 + 0.291X - 0.056X2 + 0.0038X3 - 0.000086X4 0.94 September SA Y = 0.12 + 0.048X - 0.0096X2 + 0.00056X3 - 0.000011X4 0.62 September GI Y = -0.12 + 0.43X - 0.099X2 + 0.0078X3 - 0.00020X4 0.89 1. Gnh is greenhouse, SA is South Alder and GI is Gipaanda. 2. Equations determined by linear regression, Y is starch per leaf (g) and X is leaf position.  All regressions were weighted by the inverse of the variance.  129 Table 4.5 The mean and total leaf starch per shoot for tomato shoot canopies, and starch accumulation before measurable fruit load and at full leaf expansion from commercial tomato greenhouses. Year Month Gnh1 Cum. LS to FLE2 Cum. LS before MFL3,4 Total starch per shoot5 Total starch per land area    (%) (%) (g) (g m-2) 2001 March SA 55 - 5.55 20.8  March CA 69 - 10.76 40.4  June SA 92 - 2.45 9.2  September SA 58 - 1.12 4.5  September CA 59 - 1.17 4.4 2002 May SA 89 79 5.17 19.4  May GI 91 80 5.36 20.1  June SA 62 70 2.07 7.8  June GI 73 66 2.92 11.0  July SA 66 60 1.77 6.6  July GI 71 71 2.04 7.7  August SA 57 68 1.55 5.8  August GI 67 73 2.68 9.8  September SA 40 55 2.21 8.3  September GI 67 87 2.87 10.8 1. Gnh is greenhouse, A is South Alder, C is CanAgro and G is Gipaanda. 2. Cumulative sum of leaf starch as a percentage of total starch in the canopy at the point where leaves are 95% of full expansion. 3. Cumulative sum of leaf starch as a percentage of total canopy starch to 3 leaves below the oldest truss before the fruit load becomes measurable. 4. MFL is measurable fruit load having a mean fruit volume ≥2.5 cm3. 5. Starch per leaf summed for all leaves in the canopy. .  130 2002) appeared to have more starch than did later season canopies (Table 4.5). On a per square meter of greenhouse space the amount of starch was a mean of 12.4 g, varying from 4.4 to 40 g per m2 of greenhouse floor space (Table 4.5). 4.3.5 Relationships between fruit load and starch per leaf in the canopy The strengths and distributions of sinks were determined by quantifying the fruit load per truss in the canopy (Figure 4.7).  This was calculated from the total fruit volume per truss (Figure 4.7) and the cumulative volume for all trusses in the canopy (Table 4.6).  The fruit load for the canopies ranged from 1355 to 2046 cm3, corresponding to means of 24.3 to 29.5 fruit distributed on a total of 9 to 11 trusses (Table 4.6).  By themselves the number of trusses with fruit and total number of fruit were not necessarily good indicators of total fruit load. This was because the size of the fruit also contributes to fruit load.  Fruit load became measurable usually by truss 4, which was between leaf position 7 and 8 for the majority of the canopies (Table 4.6).  Trusses younger than the fourth consisted of flower buds, flowers, and small fruit with a mean diameter of less than 17 mm. Starch per leaf and fruit load were plotted together to illustrate their interaction in the canopy (Figure 4.7).  The region of the canopy where substantial starch accumulated per leaf occurred either before the fruit load was measurable or where the fruit load was very low (Figure 4.7).  That is 55 to 87% of the total canopy starch had accumulated at trusses above positions where fruit load became measurable (Table 4.6).  Finally, a sizable portion of the fruit load occurs in the lower region of the canopy, where there were no leaves because of leaf pruning (Figure 4.7).  131  Figure 4.7 The modelled starch content per leaf and total fruit volume per truss from the shoot canopies of plants in two commercial greenhouses.  Greenhouses were South Alder (SA) and Gipaanda (GI).  All data were collected in 2002, Panel A: May, B: June, C: July, D: August and E: September.  S is for starch and F is for fruit.  Line equations for starch per leaf are given in Table 4.4. 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 -3 0 3 6 9 12 15 18 21 24 27 30 Leaf position St ar ch  p er  le af  (g ) 0 100 200 300 400 500 600 Fr ui t v ol um e pe r t ru ss  (c m 3 ) AS GS AF GF 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 -3 0 3 6 9 12 15 18 21 24 27 30 Leaf position St ar ch  p er  le af  (g ) 0 100 200 300 400 500 600 Fr ui t v ol um e pe r t ru ss  (c m 3 ) SA GA SA GA 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 -3 0 3 6 9 12 15 18 21 24 27 30 Leaf position St ar ch  p er  le af  (g ) 0 100 200 300 400 500 600 Fr ui t v ol um e pe r t ru ss  (c m 3 ) SA GA SA GA 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 -3 0 3 6 9 12 15 18 21 24 27 30 Leaf position St ar ch  p er  le af  (g ) 0 100 200 300 400 500 600 Fr ui t v ol um e pe r t ru ss  (c m 3 ) S_SA S_GI F_SA F_GI 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 -3 0 3 6 9 12 15 18 21 24 27 30 Leaf position St ar ch  p er  le af  (g ) 0 100 200 300 400 500 600 Fr ui t v ol um e pe r t ru ss  (c m 3 ) S_SA S_GI F_SA F_GI A B C D E  132  Table 4.6 The size and distribution of fruit load in tomato shoot canopies for several months in 2002 from two commercial greenhouses. Month Gnh1 Trusses  Fruit   Total Fruit2 >2.5 cm3 L3 Total No. Total volume cm3 May SA  11 8 6-7  29.5 2046.3 May GI  10 7 8-9  27.3 1393.9 June SA  10 8 7-8  31.7 1686.2 June GI  9 6 7-8  32.3 1713.3 July SA  11 7 7-8  29.3 1850.7 July GI  10 6 7-8  26.4 1850.7 August SA  10 7 7-8  26.5 1490.0 August GI  12 8 8-9  35.3 2029.3 September SA  9 6 7-8  24.3 1355.2 September GI  9 6 8-9  28 1880.9 1. Gnh is greenhouse, SA is South Alder and GI is Gipaanda. 2. 2.5 cm3 corresponds to a 17 mm diameter fruit, mean of all fruit on the truss. 3. L is leaf position, where fruit load was first measurable from the top of the plant.   133 4.4 DISCUSSION These studies have added to previous knowledge on the accumulation and distribution of starch and leaf mass in the shoot canopy of tomato plants.  In earlier research there have been numerous studies where a leaf, or a few leaves, have been measured for starch.  Also, LMA (or its inverse, specific leaf weight) has usually been determined for the entire shoot. The present studies have detailed the vertical profiles of starch and LMA and their relationships to fruit load and growing season.  An important feature of the present work is that it was carried out with plants growing under commercial conditions, rather than in controlled environment chambers where maximum PAR levels are relatively low. 4.4.1 Leaf starch and LMA values Plants exposed to elevated CO2 have been reported to accumulate more starch than plants in ambient CO2, and this results in increased LMA (Ehret & Jolliffe 1985; Körner et al. 1995; Heuvelink & Marcelis 1996; Bertin et al. 1999; Roumet et al. 1999).  On a whole canopy basis the starch contents reported here were usually higher than in most published studies.  Direct comparison of literature values to my work is complicated because previous studies have sometimes not described the ages of the sampled leaves or their positions in the canopy.  A further difficulty is the diversity of units that have been used to document starch content.  When sufficient information was available these units (mg starch per g fresh mass, mg glucose per g fresh mass, µmol glucose per g fresh mass, mmol glucose per g dry mass, percentage of dry mass (g g-1), mg starch per g dry mass, mg of glucose per dm-2 of leaf area) were converted to g m-2.  If that were not possible then percentage on a dry mass basis was used (starch per dry mass (g g-1)).  134 The mean starch concentration I observed for leaves in a canopy under CO2 enrichment ranged from 2 to 17 g m-2, which was roughly equivalent to 6 to 18 % starch per leaf dry mass for a canopy (Figure 4.1).  Studies in tomato where at least 3 leaves per canopy were measured and CO2 enrichment was used, reported starch content as follows:  5.83 to 8.15 g m-2 (Ammerlaan et al. 1986), 1 to 5.21 g m-2, (Bertin et al. 1999), 2 g m-2 (Bertin & Gary 1998) and 11.5 to 14.7% starch (Gent 1984).  Other studies where only a portion of the canopy or a single leaf of tomato were measured reported starch contents of 0.9 to 14.6 g m-2 for the lower third of the canopy (Tripp et al. 1991), 0.31 to 3.97 g m-2 for the seventh leaf from the base of de-topped plants (probably mid-canopy) (Ho & Shaw 1977), 2.83 to 3.84% starch at the fifteenth leaf from the top, (mid-canopy) (Nederhoff et al. 1992), 5 to 30% in the third leaf from bottom of 31 day old plants (De Groot et al. 2001) and 16 to 24% starch in fully expanded mature leaves (Gao et al. 1998).  All of these studies used enriched CO2 except for De Groot et al. (2001).  Therefore, my findings overlap with previously reported values but also included higher maximum values and wider ranges of starch for plant canopies than in other published experiments.  This can be attributed to the scope of my work because my data were collected throughout the growing season and over several years.  As well, my data were collected from plants grown under commercial conditions where high PAR and CO2 levels are available to support rapid carbon intake by leaf photosynthesis. In contrast to the starch data, most previous studies have measured the leaf mass per unit leaf area for the entire plant canopy and have seldom separated the canopy into multiple regions.  In the literature, LMA is less commonly used than SLA, which is inversely related to LMA and leaf starch levels.  To allow comparisons between this work and what is in the literature, reported SLA values (mostly in cm2 g-1, sometimes reported as dm2 kg-1, m2 kg-1)  135 have been converted to g m-2.  The range I observed for LMA of tomato canopies was 18 to 60 g m-2 for plants in commercial greenhouses (Figure 4.4).  These values are similar to what others have reported for greenhouse-raised tomato:  16.7 to 83.3 g m-2 (Cooman & Schrevens 2006), 37 to 100 g m-2 (de Koning 1993), 20 to 50 g m-2 (Bertin & Gary 1998) and 30 to 55 g m-2 (Bertin et al. 1999).  Heuvelink (1995b) attributed the very high LMA that de Koning (1993) reported to the use of CO2 enrichment.  However, pruning pre-treatments may also have contributed to the high LMA values that de Koning (1993) reported.  Compared to other crops, the above reports are somewhat lower in LMA, especially in comparison to trees.  The range of LMA reported for other species (where more than three regions in the canopy were measured) were:  31 to 56 g m-2 in cotton (Reddy et al. 1989), 25 to 110 g m-2 beech and birch (Uemura et al. 2006) and 52 to 94 g m-2 in poplar (Gielen et al. 2003).  It is not surprising that these studies would have higher LMA as their plants were not grown under cover.  Greenhouse cladding can reduce incident light as much as 45%, although 30% is more typical of a glass Venlo commercial greenhouse (Nederhoff & Vegter 1994a; Bailey 2002).  Also, trees (if they are not bearing fruit) can have limited sinks for exported assimilates compared to a fruiting tomato plant.  Both of these circumstances will affect the LMA. 4.4.2 Profile of leaf starch in the shoot canopy. Previous researchers  have noted that  the profile of leaf starch roughly follows the pattern of light attenuation in the canopy (Ammerlaan et al. 1986; Bertin et al. 1999).  That is, starch concentrations are highest for the unshaded leaves at the top of the canopy and decline steeply with descent through the canopy.  According to Acock et al. (1978), light attenuates exponentially through the canopy following the Beer-Lambert Law, and the top  136 25% of the canopy is responsible for over 60% of the photo-assimilate production.  In my data (Figure 4.6) and that of Ammerlaan et al. (1986), Pressman et al. (1997) and Bertin et al. (1999) starch values were initially very high and fell to their minimum by the midpoint of the canopy and stayed low for the remaining leaves in the canopy. In some of my datasets, however, leaf starch concentration was lower for the first measured leaf at the very top of the canopy than for the subsequent measured leaf.  This gave the profile an initial increase in starch until leaf position 3, followed by a rapid decline to the midpoint of the canopy.  This tendency was not well characterized by the linear regressions which were caused by insufficient samples collected at the top of the canopy.  In their starch profiles Ammerlaan et al. (1986), Pressman et al. (1997) and Bertin et al. (1999) did not observe this initial increase then decrease possibly because they sampled too few upper canopy leaves.  The dip in starch content for the very youngest leaves may have more than one cause.  It may indicate that despite their position at the top of the shoot the uppermost leaves may have had lower photosynthesis than the next tier of leaves.  This might be caused by incomplete development of the photosynthetic apparatus, and/or possibly it was caused by a more vertical leaf orientation that results in less PAR to be intercepted than by the leaves immediately below them.  Also, the young leaves tend to be curled, further reducing their exposure to incoming radiation (Appendix 3.1, Figure A3.1C).  Tightly curled upper leaves are a sign that growers use as an indication that the plants have a surplus of assimilates, although this has not been verified in the literature. Another possibility is that the youngest leaves had lower starch accumulation due to higher carbohydrate usage to meet their own demands for synthesis and growth.  137 Below the maximum starch levels near the top of the canopy there were two distinct profiles of starch decline in the canopy.  Early in the production season the starch content of leaves was similar for leaves in position –3 through to +3, and then starch declined with increasing leaf position (Figure 4.6).  This profile was very similar to that described by Ammerlaan et al. (1986), where starch was uniformly high for the top 25% of the canopy and then declined to a minimum by the mid-point of the canopy.  The second profile can be described as a rapid decline in starch content from the top to the mid-point of the canopy (Figure 4.6).  This profile was the most common found in this work, and it was also found by Bertin et al. (1999) and Pressman et al. (1997).  Ammerlaan et al. (1986), Pressman et al. (1997) and Bertin et al. (1999) used different leaf counting nomenclature than that used in my study.  Translating their method to the one I used allows the following comparisons. Ammerlaan et al. (1986) reported that starch values corresponding to leaf position -3 through to 3 was 15.0 g m-2, minimum starch content first occurred at leaf position 13 and remained at about 4 g m-2 until position 24, the last leaf they measured.  Bertin et al. (1999) reported initial starch was 8 g m-2 at leaf position 4, the minimum was first reach by leaf position 13 and it remained close to 1 g m-2 until the last leaf was reached, at position 22.  Pressman et al. (1997) reported a slightly different profile for leaves measured at 14 h.  At leaf position 1, starch was 15 g m-2 after which starch decreased linearly to 5 g m-2 by leaf position 12 (Pressman et al. 1997).  By leaf position 12 and greater, the rate of decline was markedly lower with starch reaching 2.5 g m-2 by leaf position 24, the last leaf they measured (Pressman et al. 1997).  In these experiments, five to seven leaves per canopy were sampled, and Bertin et al. (1999) and Ammerlaan et al. (1986) used CO2 enrichment while Pressman et al. (1997) did not.  As well, their data were only collected for a single day during the  138 growing season; although Pressman et al. (1997) generated their profiles for several times during the day.  In my study the maximum starch content varied from 10 to 20 g m-2 depending on the month of sampling and the greenhouse (Figure 4.6).  The minimum was about 1.0 g m-2 regardless of date of sampling or greenhouse, although the leaf position where this minimum was reached did vary with these factors. There may be several reasons for the differing profiles in my work, compared to other studies.  The profile in the early March data (Figure 4.6) and that presented by Ammerlaan et al. (1986) might indicate that at that time there was greater penetration of light into the canopy contributing to higher starch production in mid-canopy leaves.  However, the leaf area of these canopies were amongst the highest recorded which should cause an opposite effect on mid-canopy starch due to increasing shading (Table 4.3).  Another reason could be the different environment and/or environmental histories.  Generally light levels are lower early in the season but CO2 is dosed more at that time because of increased need for heating and the low requirement for ventilation (Appendix 4.1).  The higher CO2 supply could cause greater carbohydrate production and increased starch.  A third reason could be fruit load.  In the early season the crop would have a lower fruit load, because of lower fruit number per truss rather than fewer trusses.  Ammerlaan et al. (1986) did not quantify their fruit load beyond saying there were more than 3 fruit per truss and canopy fruit load was not measured in their March data sets.  In a later section the relationship between leaf area and fruit load and starch pools will be discussed in detail. 4.4.3 The profile of LMA in the shoot canopy The approximately linear decline in LMA, from about 60 g m-2 near the top of the canopy to about 25 g m-2 by the bottom (Figures 4.5 and Appendix 4.4), has not always been  139 observed in previous studies.  My results are in contrast with the approximately parabolic profile that Bertin & Gary (1998) and Bertin et al. (1999) reported for tomato canopies.  In their cases, LMA was lowest from sympod 4 to 6 (mid-canopy, equivalent to leaf position 13 to 19) relative to higher LMA values measured above and below this region (Bertin & Gary 1998; Bertin et al. 1999).  In only two instances was that pattern of response evident in my data: greenhouse CA in September 2001 and greenhouse SA in August 2002 (Figures 4.5C and Appendix 4.4G).  I am uncertain as to the cause of these parabolic responses in LMA (and those reported by Bertin et al. 1999).  They do not seem to be in strict parallel with changes in starch and/or soluble sugar accumulation, although they did occur near the time of shoot thinning, which may have disturbed the usual pattern of carbon balance in the shoots. Other studies, however, have found linear decreases in LMA, similar to my results. de Koning (1993) measured LMA at several canopy positions monthly from February until November in the Netherlands.  Leaf mass per area was found to decline linearly with depth in the canopy for the months of February to May (de Koning 1993).  For June and July the LMA was fairly consistent at any leaf position.  From August to November LMA initially declined and then increased with canopy depth, giving a parabolic-like response (de Koning 1993).  The late season data in de Koning (1993) are surprising, especially since the highest LMA (100 g m-2) was for the oldest leaves.  Those leaves were sampled from August to November, but they would have been upper leaves in June and July when light conditions were especially high.  It is possible that their structural mass became larger during their formation and early development, which carried on until the time they were eventually observed.  The same explanation could account for the parabolic-like profile I observed for  140 my September and August data sets, although it was not uniform among the greenhouses sampled here. Other crops also exhibit interesting canopy profiles in LMA.  In cotton, Reddy et al. (1989) collected and modeled seasonal LMA for several nodes in the canopy.  They reported a convex parabolic response of LMA with days after emergence for the lower 20 nodes while the upper 5 nodes exhibited a linear increase in LMA with time (Reddy et al. 1989).  When their data are re-expressed per node position, as in Figure 4.5, LMA linearly decreases with depth in the canopy for most dates.  Interestingly as well, LMA was lowest for day 90, which was when the maximum boll load occurred or greatest sink strength was attained (Reddy et al. 1989). Other canopy profile work has been carried out in broadleaf and evergreen trees.  In two studies canopy profiles were measured at 1 m intervals for trees above 6 m in height (Gielen et al. 2003; Uemura et al. 2006).  In Populus grown in a high density plantation, the LMA was uniformly high (100 g m-2) for the top third (2 to 3 m) of the canopy and then declined steadily to a minimum of 60 g m-2 at the bottom of the canopy (8 m) for P. alba and P. nigra  (Gielen et al. 2003).  Interestingly, CO2 treatment did not affect the LMA of these poplars (Gielen et al. 2003).  This gradual decline in LMA for poplar contrasts to what has been reported for Fagus and Betula canopies (Uemura et al. 2006).  Both of those species exhibited a rapid, linear decline in LMA from the top to first meter of canopy depth.  Using Fagus crenata as an example, the LMA declined from 100 to 50 g m-2 after 1 m, and by the fourth meter the LMA was at a minimum of 30 g m-2 (Uemura et al. 2006).  This decline is similar to the data in Figure 4.5, where 50% of the LMA was lost in 1.5 m, roughly the distance between the top and the bottom of a tomato plant canopy.  If LMA was highly  141 dependent on starch and soluble sugars, this is the type of decline that would be expected - one that roughly follows the attenuation of light in the canopy. Körner et al. (1995) stated that canopy levels of soluble carbohydrate usually change much less than starch.  Soluble sugars were not measured in this study but Bertin et al. (1999) reported that soluble carbohydrates did not exhibit a set pattern in the canopy, always being fairly constant between 3.5 and 4.5 g m-2 from the top of the plant to the bottom and was not responsible for the fluctuations they saw in LMA. Aside from plant and environmental factors, measurement and sampling can also affect the observed values of LMA.  LMA is the ratio of the leaf dry mass divided by the leaf area and consequently is susceptible to changes in both these measures.  CO2 dosing promotes the accumulation of leaf mass and as results in Chapter 3 indicate it does not appear to affect leaf area in tomato. Since LMA is a ratio, a decrease in leaf area caused by other factors such as by water stress could obscure the true increase in dry mass.  Also, the measurement of leaf area is prone to underestimation, because leaves can fold or overlap during measurement and dehydration of detached leaves can cause shrinkage (Nageswara Rao et al. 2001).  In the present work the leaves were sub-sampled by taking a fixed amount of area (0.000785 m2 per leaflet, equal to 0.00314 m2 per leaf) from each leaf.  Fixing the leaf area reduces the potentially confounding changes in area of a leaf that could occur. 4.4.4 The influence of leaf starch on LMA Regardless of the greenhouse sampled or the date of sampling, I found two distinct relationships between starch and LMA (Table 4.1, Figure 4.2) occurring in two different regions of the shoot canopy. The two regions, delineated by the leaf position above and below where starch first reaches a minimum, will be referred to as the upper and lower  142 canopy regions, respectively.  Examination of each canopy region separately identified a linear relationship between leaf starch and LMA with various degrees of precision (Table 4.1).  The slopes of the linear portion of the upper canopy equation for the majority of data sets were distinct from the slopes of the lower canopy regressions (Table 4.1).  Starch was found to have a greater influence on LMA (correlation coefficient of approximately 0.50) for the lower canopy than the upper canopy (R≈ 0.76) (Table 4.1) based on slopes of the relationships.  The lower R for the lower canopy equations was likely the result of the small amounts of starch (less then 7 g m-2) occurring there.  The equation constant (y-intercept) is the LMA when starch is zero, and differences indicate that non-starch leaf mass per unit area may also be lower in the lower canopy. Only a few previous studies have quantified the influence of leaf starch on LMA.  In bush bean, Ehret & Jolliffe (1985) reported a strong positive relationship between starch and LMA (LMA = 1.02 (starch g m-2) + 32.4, R2=0.97).  In potato, Cao & Tibbitts (1997) concluded that LMA was a good indicator of leaf starch accumulation in potato, and they defined the relationship as LMA = 30.5 + 1.89 (starch g·m-2), R2=0.97.  It should be noted that both of these studies used linear regression which, as will be discussed below, is not the ideal statistical methodology to use.  In tomato, Bertin & Gary (1998) defined a relationship between leaf starch with LMA as LMA = 34.5 + 6.66 (starch g m-2), R2=0.93, also using linear regression.  In their later work Bertin et al. (1999), also discovered a distinct relationship for upper and lower leaves and showed these relationships separately with what seems to be linear regression lines.  Although they did not supply the equations for the regressions I was able to reconstruct them from their graphs.  For leaves in the upper canopy LMA = 36.89 + 2.09 (starch g m-2), R2=0.70 for the lower leaves LMA = 35.5 + 8.5 (starch g  143 m-2), R2=0.44 (Bertin et al. 1999).  However, it should be mentioned that in the studies by Ehret & Jolliffe (1985), Cao & Tibbitts (1997) and Bertin & Gary (1998) the data were grouped at the low and high ends with few intervening measurements which may overestimate the precision. At the outset of my research, it was not anticipated how difficult it would be to define the relationship between LMA and leaf starch.  Not only are the temporal and spatial patterns complex, there are statistical issues as well.  Analysis by linear regression should not be attempted as both X and Y are random variables and using them in a linear regression is a clear violation of the assumptions for this analysis (Sokal & Rohlf 1995).  Model II regression makes fewer assumptions about the data and is able to handle two random variables (Dytham 2003).  However, the techniques are not yet well developed and often are not included in statistical programs (Dytham 2003).  Judging by the R values I obtained, my results appear not always to be as precise as Ehret & Jolliffe (1985), Cao & Tibbitts (1997) or (Bertin & Gary 1998).  A possible cause for this difference may be the fact that the other studies were carried out in controlled environment chambers, allowing greater uniformity and stability in the environment than in the commercial greenhouses I used. 4.4.5 Source and sink relations in the shoot canopy The distribution of leaf starch, leaf area and fruit load in the shoot canopy were used to examine the source and sink relations in the shoot canopy.  Starch content per leaf was plotted to provide an improved understanding of the influence of leaf area on the distribution of starch in the canopy.  Expressing starch for an entire leaf is not commonly reported in the literature for tomato.  In my work when starch was summed for all the leaves in the shoot, the amount of starch ranged from 1.12 to 10.76 g per shoot (Table 4.6).  On a per square meter  144 basis of greenhouse space this amount of starch is 4.5 to 40.4 g m-2.  Typically, the starch per shoot I measured was much smaller than the 14 to 28.8 g per shoot reported by Ammerlaan et al. (1986).  In their work, Ammerlaan et al. (1986) found starch for the middle and upper leaves to be 0.47 and 0.40 g per leaf, respectively, which is comparable to what I found (Figure 4.6).  The lower canopy leaves in my work exhibited close to 0 g of starch while Ammerlaan et al. (1986) reported starch to be 0.33 g.  As well, the lower leaves in the Ammerlaan et al. (1986) study also exhibited greater leaf area and starch per area than what I measured.  Therefore, the differences in the starch per shoot between my work and Ammerlaan et al. (1986) were the result of a larger leaf area.  Ammerlaan et al. (1986) reported canopy leaf areas of 2.40 to 4.70 m2 while leaf areas in my work were 0.57 to 1.38 m2 for a canopy (single stem) (Table 4.3).  These differences in leaf area can be ascribed to raising plants under commercial and experimental conditions, where in industry high planting densities and leaf removal are practised.  Ammerlaan et al. (1986) did not report their plant density which precludes calculating the amount of starch per square meter of greenhouse space. Canopy profiles of starch per leaf and fruit load were plotted in tandem to examine the interplay of source and sink relations in the shoot. It is well known from the literature, especially from fruit pruning experiments, that low sink strength can cause starch to accumulate in tomato leaves (Ammerlaan et al. 1986; Bertin & Gary 1998).  In tomato the main sinks are the apex, young leaves, flowers and fruits.  The fruit are considered to be the most significant sinks because the apex and young leaves are able to support much of their carbohydrate needs with their own photosynthesis (Tanaka & Fujita 1974).  My work supports the view that fruit load is a prominent factor in controlling leaf starch accumulation  145 because 32 to 60% of the starch in the canopy accumulated in leaves above the trusses at which fruit load became sizable (Figure 4.7). Fruit growth, and rate of growth from anthesis to maturity, have been examined in detail for tomato by Ho et al. (1987).  Tomatoes under commercial greenhouse management produce a new truss (and three leaves) approximately every week.  From anthesis, (in my work the truss near the leaf at position 1, Figure 4.7 and identified as truss 1) a fruit takes about eight weeks to mature at 20 C (Heuvelink 1996).  A full canopy has seven to eight trusses of fruit, plus one truss at anthesis and one to two immature trusses possessing flower buds.  From anthesis (truss 1), each subsequent truss has fruit which are a week more developed on average.  This pattern of fruit development allows an approximate chronological age to be assigned to the fruit on each truss and allows comparison to literature values where fruit growth was measured over time. Ho et al. (1987) reported that both dry and fresh mass accumulation rates (g fruit per day-1) exhibited parabolic patterns of change over time.  The measures were near zero for week 1 (second truss), reached a maximum (0.2 and 0.3 g fruit per day-1, dry and fresh mass respectively) by week 3 (fourth truss) and declined again to near zero by week 7 (eighth truss).  As well, dry matter concentration in the phloem sap was found to be a maximum two weeks after anthesis, corresponding to the third truss.  Ho et al. (1987) indicated that maximum fruit growth and importation rates occur between week 2 and 4 corresponding to the third to fifth trusses.  From their data I concluded that growth, and thus sink strength, was very low for fruit at the second truss and younger. Further evidence for this is also provided in the studies by Ho et al. (1982) and Monselise et al. (1978).  Dry mass gain per fruit was very low for fruit on the first to third  146 truss (1 g per fruit) compared to fruit on the fourth truss (3 g per fruit) and fruit on the eighth truss (8 g per fruit) (Ho et al. 1982).  Increases in fruit volume were reported to be very low for fruit until the third to fourth truss with the maximum growth rate on a volume basis occurring from trusses 3 to 6 (Monselise et al. 1978).  These studies indicate that potential for fruit growth at the second truss and younger is low.  The accumulation of starch observed in Figures 4.7 indicates these fruit are unable to use all of the assimilates produced by the surrounding leaves.  Truss 2 is located close to leaf position 4 which is in the top quarter of the canopy and would have high photosynthesis rates (Acock et al. 1978).  These studies on fruit growth, in combination with the data in Figure 4.7, indicate that low sink strength in combination with high rates of photosynthesis causes starch to accumulate in leaves at the top of the canopy. The low levels of starch in the lower leaves indicates that starch initially accumulated at the top will eventually be used, presumably largely for fruit growth.  The mean amount of starch contained in the sampled canopies (from the top to 3 leaves below truss 2) ranged from 1.1 to 4.3 g with a mean of 2.1 g (Table 4.5).  Assuming that all the starch was converted to sucrose and the sucrose was directly responsible for fruit dry matter gain with no starch replenishment and no losses caused by respiration, 2.1 g of starch in the canopy would convert to 2.3 g of potential dry matter gain (sucrose).  Using the fruit growth data of Ho et al. (1987), and a dry matter gain of 2.3 g supplied by starch, fruit growth could roughly be sustained on truss 2 (4 fruit at 0.1 g fruit-1 day-1) for 5 days or on truss 3 (4 fruit at 0.68 g fruit-1 day-1) for about three-quarters of a day.  Hence, the levels of starch accumulated in the tomato canopy under CO2-enriched commercial conditions seem to be in excess of the short- term (minutes to hours) requirements needed to sustain fruit growth.  147 Little starch was found to accumulate below the fourth truss (Figure 4.7).  That truss is located between leaves 9 to 12, at least half way down the canopy depending on the total number of leaves retained before pruning.  Near that truss leaf photosynthesis would be lower than above as a result of shading, and growing fruit are present as large sinks.  My starch observations suggest that fruit at and below this truss may be source-limited. It seems puzzling that the upper leaves do not have sufficient ability to export their surplus assimilates to meet some of the demand from fruit growing in the lower trusses.  In work with radioactive tracers Tanaka & Fujita (1974) concluded that the three leaves below a truss were mainly responsible for feeding it, although they did indicate that excess assimilates could be translocated to other sinks.  Ho & Hewitt (1986) also concluded that, in a multi-truss plant, leaves supply local trusses with assimilates but some overlapping in supply does occur.  Heuvelink (1995a) questioned this assumption and has shown that assimilates can move great distances, from one stem to another on a multi-stem plant. Although Heuvelink (1995a) could not prove his hypothesis of one common assimilate pool for tomato, he and others did indicate that assimilate partitioning was flexible (Heuvelink & Buiskool 1995) and concluded that proximity to the source is less important than sink strength and the growth potential of the sink (Heuvelink 1996). It appears that the reproductive structures of tomato plants are sink limited at the top and source limited at the bottom.  This is somewhat contrary to the findings of Hocking & Steer (1994) and Gautier et al. (2001) who suggested that source activity exceeds sink activity.  It is possible that the leaves at truss 2 and above are immature and either not exporting assimilates or do not have a vascular system mature enough to fully export their assimilates.  However, the third leaf below truss 2, in leaf position 7 was 55 to 87% of full  148 expansion.  According to Ho & Shaw (1977), a leaf becomes a net exporting organ at 15 to 25% of full expansion.  In addition, in all my data sets the maximum starch per leaf was for leaves at position 3.  These leaves were 71% of full expansion and should be unhindered in their potential for export. Given the evidence for starch accumulation at the top of the shoot, and the indications of source limitation for the growth of fruit lower in the canopy, this discussion suggests that greenhouse tomato plants are not immediately able to exploit all of the carbon they assimilate during CO2 dosing. There may be some impediment to the remobilization of carbon via the transport system from the young leaves to the developing fruit.  This would not seem to be due to a high hydraulic resistance in the phloem, which is thought to be negligible (Heuvelink 1996).  Water deficits in the upper leaves could slow transport since sucrose is exported in solution and 90% of water import into the fruit comes from the phloem (Ho et al. 1987).  However, this is not borne out by studies of tomato growth and dry matter accumulation under saline conditions (Gao et al. 1998).  Another possibility is that a local deficiency in phosphorus may be affecting the phosphate translocator on the chloroplast membrane, which would prevent the triose phosphate from leaving the chloroplast and cause a build up of starch.  Whatever the cause, this issue should be taken up in future research, because it would seem that excess starch accumulation in leaves of the upper shoot is preventing commercial tomato plants from benefiting fully from the additional carbon inputs afforded by CO2 enrichment. 4.4.7 Assessment of LMA and starch as a management tool for CO2 dosing These studies have provided considerable detail of the variations in leaf starch and mass distribution in shoots of tomato plants growing in commercial conditions under CO2  149 enrichment.  Detailed discussion of the potential for such measures to be used in management decisions on CO2 dosing will be postponed until Chapter 6, when all of the plant-based approaches will be considered together. 4.5 CONCLUSIONS 1. Profiles of leaf starch per unit area and LMA were somewhat similar, but not identical, in the sampled tomato plant canopies.  The vertical profile in starch seemed to be controlled by PAR availability and fruit load, leading to starch accumulation in the upper canopy, where PAR was high and sink availability was low, and negligible starch low in the canopy, where the reverse was true. 2. The profile of LMA usually, but not always, exhibited a linear decline with leaf position.  LMA often continued to decline in the canopy past the point where starch levels had reached their minimum.  Therefore leaf starch on a per area basis was only modestly able to explain the variation in LMA. 3. The relationship between leaf starch and LMA was stronger for leaves in the upper part of the canopy compared to the leaves in the lower canopy.  Comparing starch and LMA, the difference between the two regions of the canopy can at least be partly attributed to the more rapid decline in starch concentration in the upper canopy, and the continued decline in LMA in the lower canopy.  It seems for the lower canopy that LMA is not a good predictor of starch. 4. The vertical profile of leaf area in the canopies was sigmoidal and very similar across sampling dates.  The position where leaves first became fully expanded moved closer to the top of the canopy as the season progressed (from leaf position 9 to 5).  150 5. Maximum content of starch per leaf occurred in the top third of the canopy, at leaf positions 3 to 5, where leaves were close to full expansion.  That region of the canopy was well above the strata at which fruit load becomes significant.  This indicates that starch levels in leaves above truss 3 are sink limited, or are limited in the capacity of the transport system to convey assimilates to the sinks. 6. The low levels of leaf starch below truss 4, (the truss at leaf position 7 or 8) occur where sink capacity is high, providing evidence that in that portion of the canopy fruit growth is source-limited. 7. Under commercial growing conditions, CO2 enriched tomato plants accumulate levels of starch in their upper canopy that are in excess of their immediate ability to exploit that carbon for fruit growth.  151 4.6 LITERATURE CITED  Acock, B., Charles-Edwards, D.A., Fitter, D.J., Hand, D.W., Ludwig, L.J., Warren Wilson, J., and Withers, A.C. 1978. The contribution of leaves from different levels within a tomato crop to canopy net photosynthesis:  An experimental examination of two canopy models. Journal of Experimental Botany 29: 815-827. Ammerlaan, A.W.S., Joosten, M.H.A.J., and Grange, R.I. 1986. The starch content of tomato leaves grown under glass. Scientia Horticulturae 28: 1-9. Bailey, B.J. 2002. Optimal control of carbon dioxide enrichment in tomato greenhouses. Acta Horticulturae  578: 63-69. Bertin, N. and Gary, C. 1998. Short and long term fluctuations of the leaf mass per area of tomato plants - Implications for growth models. Annals of Botany 82: 71-81. Bertin, N., Tchamitchian, M., Baldet, P., Devaux, C., Brunel, B., and Gary, C. 1999. Contributions of carbohydrate pools to the variations in leaf mass per area within a tomato plant. New Phytologist 143: 53-61. Cao, W.X. and Tibbitts, T.W. 1997. Starch concentration and impact on specific leaf weight and element concentrations in potato leaves under varied carbon dioxide and temperature. Journal of Plant Nutrition  20: 871-881. Cooman, A. and Schrevens, E. 2006. A monte carlo approach for estimating the uncertainty of predictions with the tomato plant growth model, Tomgro. Biosystems Engineering 94: 517-524. De Groot, C.C., Marcelis, L.F.M., van den Boogaard, R., and Lambers, H. 2001. Growth and dry-mass partitioning in tomato as affected by phosphorus nutrition and light. Plant, Cell and Environment 24: 1309-1317. de Koning, A.N.M. 1993. Growth of a tomato crop.  Measurements for model validation. Acta Horticulturae 328: 141-146. Dytham, C. Choosing and Using Statistics.  A Biologist's Guide.  2003. Malden, MA, USA, Blackwell Publishing. Ehret, D.L. and Jolliffe, P.A. 1985. Leaf injury to bean plants grown in carbon dioxide enriched atmospheres. Canadian Journal of Botany 63: 2015-2020. Ewert, R. 2004. Modelling plant responses to elevated CO2:  How important is leaf area index. Annals of Botany 93: 619-627. Farrar, J.F. 1996. Sinks-integral parts of a whole plant. Journal of Experimental Botany 47: 1273-1279.  152 Gao, Z., Sagi, M., and Lips, S.H. 1998. Carbohydrate metabolism in leaves and assimilate partitioning in fruits of tomato (Lycoperscion esculentum L.) as affected by salinity. Plant Science 135: 149-159. Gary, C. 1989. Interest of a carbon balance model for on-line growth control:  the example of a daylight dependent night temperature control. Acta Horticulturae 248: 265-268. Gautier, H., Guichard, S., and Tchamitchian, M. 2001. Modulation of competition between fruits and leaves by flower pruning and water fogging and consequences on tomato leaf fruit growth. Annals of Botany 88: 645-652. Gent, M.P.N. 1984. Carbohydrate level and growth of tomato plants.  I.  The effect of carbon dioxide enrichment and diurnally fluctuating temperatures. Plant Physiology 76: 694- 699. Gielen, B., Liberloo, M., Bogaert, J., Calfapietra, C., De Angelis, P., Miglietta, F., Scarascia- Mugnozza, G., and Ceulemans, R. 2003. Three years of free-air CO2 enrichment (POPFACE) only slightly affect profiles of light and leaf characteristics in closed canopies of Populus. Global Change Biology 9: 1022-1037. Heuvelink, E. 1995a. Dry matter partitioning in a tomato plant:  one common assimilate pool? Journal of Experimental Botany  46: 1025-2033. Heuvelink, E. 1995b. Growth, development and yield of a tomato crop:  periodic destructive measurements in a greenhouse. Scientia Horticulturae 61: 77-99. Heuvelink, E. 1996. Re-interpretation of an experiment on the role of assimilate transport resistance in partitioning in tomato. Annals of Botany 78: 467-470. Heuvelink, E. 1999. Evaluation of a dynamic simulation model for tomato crop growth and development. Annals of Botany 83: 413-422. Heuvelink, E. and Buiskool, R.P.M. 1995. Influence of sink-source interaction on dry matter production in tomato. Annals of Botany 75: 381-389. Heuvelink, E. and Marcelis, L.F.M. 1996. Influence of assimilate supply on leaf formation in sweet pepper and tomato. Journal of Horticultural Science 71: 405-414. Ho, L.C., Grange, R.I., and Picken, A.J. 1987. An analysis of the accumulation of water and dry matter in tomato fruit. Plant, Cell and Environment 10: 157-162. Ho, L.C. and Hewitt, J.D. 1986. Glasshouse crop production. In Chapman and Hall Ltd., Cambridge, Great Britain, pp 201-240. Ho, L.C. and Shaw, A.F. 1977. Carbon economy and translocation of  14C in leaflets of the seventh leaf of tomato during leaf expansion. Annals of Botany 41: 833-848. Ho, L.C., Sjut, V., and Hoad, G.V. 1982. The effect of assimilate supply on fruit growth and  153 hormone levels in tomato plants. Plant Growth Regulation 1: 155-171. Hocking, P.J. and Steer, B.T. 1994. The distribution and identity of assimilates in tomato with special reference to stem reserves. Annals of Botany 73: 315-325. Jolliffe, P.A. and Lin, W.C. 1997. Predictors of shelf life in long English cucumbers. Journal of American Society for Horticultural Science 122: 686-690. Körner, Ch., Oelaez-Riedl, S., and van Bel, A.J.E. 1995. CO2 responsiveness of plants:  a possible link to phloem loading. Plant, Cell and Environment 18: 959-600. Lee, J.H. and Heuvelink, E. 2003. Simulation of leaf area development based on dry matter partitioning and specific leaf area for cut chrysanthemum. Annals of Botany 91. Legendre, P.  2001. Model II regression - User's guide.  Departement de sciences biologiques, Universite de Montreal. Marcelis, L.F.M., Heuvelink, E., and Gourdriaan, J. 1998. Modelling biomass productuion and yield of horticultural crops:   a review. Scientia Horticulturae 74: 83-111. Minchin, P.E.H. and Thorpe, M.R. 1996. What determines carbon partitioning between competing sinks? Journal of Experimental Botany 47: 1293-1296. Monselise, S.P., Varga, A., and Bruinsma, J. 1978. Growth analysis of the tomato fruit, Lycopersicon esculentum Mill. Annals of Botany 42: 1245-1247. Nageswara Rao, R.C., Talwar, H.S., and Wright, G.C. 2001. Rapid assessment of specific leaf area and leaf nitrogen in peanut (Arachis hypogaea L.) using a chlorophyll meter. Journal of Agronomy and Crop Science 186: 175-182. Nederhoff, E.M., De Koning, A.N.M., and Rijsdijk, A.A. 1992. Leaf deformation and fruit production of glasshouse grown tomato (Lycopersicon esculentum Mill.) as affected by CO2, density and pruning. Journal of Horticultural Science 67: 411-420. Nederhoff, E.M. and Vegter, J.G. 1994a. Photosynthesis of stands of tomato, cucumber and sweet pepper measured in greenhouses under various CO2-concentrations. Annals of Botany 73: 353-361. Nederhoff, E.M. 1994. Effects of CO2 concentration on photosynthesis, transpiration and production of greenhouse fruit vegetable crops. Wageningen. Pressman, E., Bar-Tal, A., Shaked, R., and Rosenfeld, K. 1997. The development of tomato root system in relation to the carbohydrate status of the whole plant. Annals of Botany 80: 533-538. Reddy, V.R., Acock, B., Baker, D.N., and Acock, M. 1989. Seasonal leaf area-leaf weight relationships in the cotton canopy. Agronomy Journal 81: 1-4.  154 Roumet, C., Laurent, G., and Roy, J. 1999. Leaf structure and chemical composition as affected by elevated CO2:genotypic responses of two perennial grasses. New Phytologist 143: 73-81. SAS Institute. 2003. Statistical Analysis Software. Cary, NC, SAS Institute Inc. Sokal, R.R. and Rohlf, F.J. 1995. Biometry The Princples and Practice of Statistics in Biological Research. New York, W. H. Freeman and Company. Tanaka, A. and Fujita, K. 1974. Nutrio-physiological studies on the tomato plant.  IV. Source-sink relationship and structure of the source-sink unit. Soil Science and Plant Nutrition 20: 305-315. Tripp, K.E., Peet, M.M., Willits, D.H., and Pharr, D.M. 1991. CO2-enhanced foliar deformation of tomato:  relationship to foliar starch concentration. Journal of American Society for Horticultural Science 116: 876-880. Uemura, A., Harayama, H., Koike, N., and Ishida, A. 2006. Coordination of crown strucutre, leaf plasticity and carbon gain within the crowns of three winter-deciduous mature trees. Tree Physiology 26: 633-641. van de Vooren, J., Welles, G.W.H., and Hayman, G. 1986. Glasshouse crop production. In Chapman and Hall Ltd., Cambridge, Great Britain, pp 581-623.  155 5 Temporal variations of leaf starch and mass8,9 5.1 INTRODUCTION 5.1.1 The diurnal variability of leaf starch and LMA For leaf starch and/or leaf mass per area (LMA) to be useful components in a plant based approach to greenhouse CO2 management, their temporal dynamics need to be understood, particularly in circumstances relevant to commercial plant production.  On a daily basis, the amount of starch formed in the leaf is affected by light, CO2, and the sink requirements for carbon.  Since starch is an overflow product of photosynthesis one would expect that its diurnal profile would resemble the diurnal photosynthesis rate but with a time lag.  Accordingly, leaf starch builds up during the day when carbon fixation exceeds carbon use or export by the leaf.  At night, or when carbon fixation becomes lower than carbon use, starch is broken down to maintain carbon export to the sinks.  Exposure to CO2 enrichment is known to increase the amount of starch and the LMA of some leaves, however, little is known about how continuous exposure to CO2 enrichment may affect the diurnal profile. Greenhouse tomato production in BC generally occurs exclusively under natural light, where there are substantial variations in light intensity, diurnally and seasonally. Results from growth chamber studies, where plants are exposed to relatively low and constant daytime light levels (Galtier et al. 1995), may have limited relevance to starch dynamics in plants growing in commercial greenhouses (Ho 1977; Hammond et al. 1984; Hrubec et al. 1985; Stutte et al. 1996; Li et al. 2001).  The research literature is unclear about diurnal patterns of starch for greenhouse-grown crops.  Starch accumulation has been found  8 A version of this chapter will be submitted for publication.  Edwards, D., Jolliffe, P. and Ehret, D.  Temporal variations of leaf starch and mass.  9 A version of this chapter has been published.  Edwards, D., Jolliffe, P. and Ehret, D. 2004. Carbon status of CO2- enriched tomato plants under commercial greenhouse conditions.  Acta Horticulturae. 633-279-286.  156 to:  follow a parabolic-like response with day length (Madsen 1968; Stutte et al. 1996), steadily increasing from sunrise to sunset and declining at night (Ammerlaan et al. 1986; Pressman et al. 1997; Bertin et al. 1999; Ayari et al. 2000b), or to rise to an asymptote during the latter part of the day (Galtier et al. 1995; Ayari et al. 2000a).  In my exploration of the literature, no studies were found that monitored tomato plants growing under steady-state fruit production such as would be the case in a commercial greenhouse.  Fruiting pattern would seem to be important in this context; in Chapter 4 it was seen that sink capacity seems to be a factor that helps to determine the vertical profiles of leaf starch and LMA. Leaf starch accumulation may be a useful indicator of plant carbon status, but with current methodology starch would be difficult for a greenhouse grower to measure.  A possible surrogate for leaf starch, investigated in Chapter 4, was leaf mass per unit leaf area (LMA).  Compared to starch, there is even less information from the literature about the diurnal profile of LMA.  Presumably it would tend to follow the trend in starch, and like starch, LMA is known to increase under CO2 enrichment (Bertin et al. 1999).  However, the findings of Chapter 4 indicate that leaf starch and LMA are not strictly coupled for some leaves, particularly those in the lower canopy. 5.1.2 The dynamics of leaf starch and LMA at the onset of CO2 enrichment In addition to diurnal variations, it is also of interest to explore plant carbon status and the dynamics of starch and LMA in response to the onset of CO2 enrichment.  Increasing the source activity by CO2 enrichment ought to quickly result in a build up of starch in the leaf, presuming that carbon mobilization from leaves and sink activity may not be able to respond as rapidly.  It is not usual for commercial greenhouses in British Columbia to grow their crop under ambient CO2 for an extended time, but there are periods of time where low light levels  157 cause photosynthetic carbon input to be low.  By knowing the timing of leaf starch build up when source strength is boosted growers can be better informed on how long to apply CO2 enrichment after periods of low photosynthetic activity.  Knowing the timing of starch increase would also help prevent the plants from being oversupplied with CO2 when adequate reserves of starch are already present. 5.1.3 Objectives of chapter The goal of this chapter is to provide temporal information on variation in leaf starch and LMA in tomato canopies.  Such information may clarify the patterns that apply to mature tomato plants growing in greenhouse houses, and it may help to identify when samples could be collected for use in a plant-based method of CO2 management.  The research involved experiments in two venues:  two commercial greenhouses that practised CO2 enrichment, and research greenhouses at PARC – Agassiz where CO2 dosing could be controlled and an experimental design employed.  These studies had the following specific objectives: 1. to establish the diurnal variability of leaf starch and LMA of tomato plants growing in CO2 enriched commercial greenhouses, 2. to determine the diurnal variability leaf starch and LMA of plants exposed to CO2 enrichment or ambient CO2, and 3. to establish the response of leaf starch and LMA to the onset of CO2 enrichment. 5.2 MATERIALS AND METHODS Experiments were carried out in two commercial greenhouses in the lower mainland of British Columbia and in research greenhouses at the Pacific Agri-Food Research Centre in Agassiz (PARC-Agassiz), British Columbia.  Some parts of the methodology for this Chapter  158 have already been presented.  In order to avoid needless repetition, a brief description is given here with the details referenced to the relevant chapter and section. 5.2.1 Greenhouse venues Commercial greenhouses.  Data were collected in 2001 from CanAgro Produce Ltd (CA) and South Alder Greenhouse Ltd (SA).  Most of the specifics of these greenhouses including the environment, crop culture and production systems have been explained in Chapter 4, section 4.2.1.  Briefly, both greenhouses used the high wire production system to grow the beefsteak tomato cultivar Rapsodie (Syngenta Seeds Inc, Boise, ID) and used supplemental CO2.  Supplemental CO2 was mainly generated from the combustion of natural gas for heat production.  At times during the summer liquid CO2 was also used when heating was not required.  Greenhouse temperature, CO2 concentration, relative humidity or humidity deficit, and global radiation (external to the greenhouse) were obtained from the climate records of the respective greenhouses. Research greenhouses.  Experiments were also conducted in the research greenhouses at PARC –Agassiz during the summers of 2002 and 2003.  In Chapter 3 the greenhouse setup and environment (section 3.2.1), experimental layout and design (section 3.2.4 and Figure 3.1) and plant culture (section 3.2.2) were fully described.  Briefly, four greenhouse compartments of 37.5 m2 with independent environmental controls were used.  Two of the compartments were dosed with CO2 to a set point of 1000 pm between 6 h and 21 h and the other two were exposed to ambient CO2.  In the centre of each compartment, 40 tomato cultivar Rapsodie plants were trained to a high wire production system.  A guard row of plants was planted around the perimeter of each compartment.  Experiments commenced in 2002 at 6 h on June 11 and in 2003 at 15 h on May 6.  159 5.2.2 Data collection at the onset of CO2 enrichment At PARC – Agassiz the changes in leaf starch and LMA were monitored for the first week of exposure to CO2 enrichment.  Plants were established in four greenhouse compartments and grown under ambient CO2 for 122 days in 2002 and 138 days in 2003. Then, CO2 enrichment was applied in two of the four greenhouse compartments (replicates). In both years, data were collected just prior to the application of CO2 enrichment from all replicates.  In 2002, CO2 enrichment was started at dawn of June 11 (123 days from grafting), with leaf data collected on June 10 (day 0), June 11 (day 1), June 15 (day 5) and June 17 (day 7).  At each sampling date, leaf data were collected at 8 h and 14 h from three randomly selected plants per greenhouse compartment.  For the morning collection, leaves were shaded with tin foil, which had been installed at dusk the night before and removed just prior to sample collection.  Three leaves were collected per plant canopy, one each from leaf positions 1, 7 and 13.  These leaves were located at the trusses with: open flowers (truss 1), the second truss with fruit (truss 3) and the fourth truss with fruit (truss 5), respectively.  Leaf samples were processed as explained in section 5.2.3. The experiment in 2003 was a repeat of the 2002 work with a few changes. In 2003, CO2 enrichment was started at 15 h on May 6 (day 0), after the collection of the pre- treatment leaves at 14 h (plants were 138 days from seeding).  Leaves were collected on the following dates at 14 h:  May 8 (day 2), May 10 (day 4) and May 13 (day 7).  Three leaves were collected per canopy from the same positions as in 2002, from four randomly selected plants per greenhouse compartment per sampling date.  Leaf samples were processed as explained in section 5.2.3.  160 5.2.3 Data collection for diurnal profiles Commercial greenhouses.  Leaves were collected between dawn of one day and dawn of the next day for the assessment of the diurnal profiles of leaf starch and LMA.  For that purpose, commercial greenhouses were visited on six occasions during the 2001 growing season.  Data were collected from SA on June 4-5, August 22-23 and September 19-20.  Data were collected from CA on June 12-13, August 20-21 and September 6-7.  Samples were collected at dawn (6 h to 7 h), mid morning (10 h to 11 h), early to mid afternoon (13:30 h to 15:30 h), late afternoon to evening (16:30 h to 20:00 h) and at dawn the next day (6 h to 7 h). The time of collection varied depending on the times when access to the respective greenhouse was possible.  At each sampling time, four rows were randomly selected from the centre of the greenhouse, and one plant was selected from the middle of the row.  From that plant the leaf closest to the truss with open flowers (the leaf in position 1) was selected and excised.  A full description of sample handling and analysis for LMA and leaf starch is presented in Chapter 4, section 4.2.4. Research greenhouses.  In 2002, leaves were collected during two day periods from plants in both CO2 treatments (Enriched and Ambient), and results from those samples were used to generate two consecutive diurnal profiles for leaf starch and LMA.  Leaves were collected at dawn (5:30 h), 10 h, 14 h and dusk (20 h) on August 24 and 25 (197 days from grafting and 74 days after the start of CO2 enrichment).  Three leaves were collected from the canopies of four randomly selected plants per greenhouse compartment.  Collected leaves were at position 1 (at the truss with open flowers), position 7 (two trusses below the truss with open flowers) and position 13 (four trusses below the truss with open flower).  A full description of the collection and analysis of leaf tissue for LMA and leaf starch is presented in Chapter 4, section 4.2.4.  161 5.2.4 Data analysis Commercial greenhouses.  In this venue it was not possible to apply a CO2 enrichment treatment, or to randomize and replicate the experimental units.  For each sampling time, the mean and 95% confidence intervals were calculated for leaf starch or LMA using SAS version 9.1.3 software (SAS Institute 2003).  For each greenhouse sampled, the raw environmental data (temperature, CO2 level, global radiation and humidity deficit or relative humidity) were plotted for the 24 hour period of the data collection and for the complete day before sampling. Research greenhouses.  The experimental design employed at PARC-Agassiz allowed an analysis of variance to be performed.  A mixed model ANOVA was used for the analysis of leaf starch and LMA data because both the diurnal and onset experiments contained fixed (the CO2 treatment and time) and random (experimental units) effects.  A full description of the mixed model methodology can be found in Chapter 3, section 3.2.7.  At each time or day of sampling, treatment means were compared using a priori contrasts.  In experiments where more than one leaf position per canopy was observed, leaf position was analyzed separately because of the differing light environment caused by intra-canopy shading on leaf starch and LMA.  Examination of the plots of the residuals for LMA and leaf starch indicated that the variance was heteroscedastic for the 2003 data set.  To improve the homoscedasticy, the starch and LMA values were transformed by taking their natural logarithms and then the results were analyzed using a mixed model ANOVA.  After the analysis the means were back transformed for presentation.  162 5.3 RESULTS The results concerning diurnal variations will be considered first, taking the findings from commercial and research greenhouses in order.  Then, the changes following the onset of CO2 enrichment will be considered, taking the 2002 and 2003 experiments in order. 5.3.1 The diurnal variability of leaf starch and LMA Commercial greenhouses.  Leaf starch and LMA tended to follow the diurnal profile of light but with several hours lag (Figures 5.1 and 5.2).  In the commercial greenhouses, leaf starch was lowest in the morning and reached a maximum between 14 and 16 h (Figure 5.1). This was 7 to 9 hours after sunrise.  Over this time starch increased from 46 to 114% of its base value (minimum to maximum).  For most of the profiles, starch was similar between dawn and about 11 h (Figure 5.1).  As well, for consecutive dawn measurements, five of the six profiles exhibited similar starch contents from one day to the next (Figure 5.1). Midday leaf starch concentrations were marginally higher at SA than at CA for two of the three periods sampled (Figure 5.1A and C).  Differences in starch accumulation may be related to environmental changes during the growing season, and between the two greenhouses on the dates of sampling (Figures 5.3 to 5.5).  In June, the higher starch concentrations at SA may have been caused by higher light levels and CO2 concentrations that occurred then (Figure 5.3A and B).  In September light levels were similar for the two greenhouses, but CO2 levels were higher at SA than at CA (Figure 5.5B).  In late August, however, leaf starch levels were only about one quarter as high at SA than at CA (Figure 5.1B).  That difference may reflect the markedly lower radiation levels received by SA on the August sampling dates (Figure 5.4B).  163  Figure 5.1 Diurnal and seasonal patterns of leaf starch concentration at leaf position 1 from tomato plants in two commercial greenhouses, South Alder (SA) and CanAgro (CA).  Data were collected dawn to dawn for selected dates in 2001, Panel A. SA June 4, CA June 12, Panel B. SA August 22, CA August 20, Panel C. SA September 19, CA September 6. Bars are 95 % confidence intervals of the mean. The dark horizontal bars indicate the night. 0 5 10 15 20 25 4 8 12 16 20 24 28 Portion of Day Le af  s ta rc h (g  m -2 ) SA CA 0 5 10 15 20 25 4 8 12 16 20 24 28 Portion of Day Le af  s ta rc h (g  m -2 ) SA CA 0 5 10 15 20 25 4 8 12 16 20 24 28 Hour Le af  s ta rc h (g  m -2 ) SA CA  4                8               12              16             20              24              4 C B A  164 Figure 5.2 Diurnal and season patterns of leaf mass per unit area (LMA) at leaf position 1 from tomato plants in two commercial greenhouses, South Alder (SA) and CanAgro (CA).  Data were collected dawn to dawn for selected dates in 2001, Panel A. SA June 4, CA June 12, Panel B. SA August 22, CA August 20, Panel C. SA September 19, CA September 6. Bars are 95 % confidence intervals of the mean. The dark horizontal bars indicate the night. 30 35 40 45 50 55 60 65 4 8 12 16 20 24 28 Portion of Day LM A  (g  m -2 ) SA CA 30 35 40 45 50 55 60 65 4 8 12 16 20 24 28 Portion of Day LM A  (g  m -2 ) SA CA B A 30 35 40 45 50 55 60 65 4 8 12 16 20 24 28 Hour LM A  (g  m -2 ) SA CA                                                                                         4 C  165  Figure 5.3 The diurnal environment from two commercial greenhouses, CanAgro (CA) and South Alder (SA) in 2001.  Panel A. is global radiation -exterior to greenhouse, Panel B is hourly CO2 concentration, Panel C is hourly temperature and Panel D is hourly humidity deficit (SA) or relative humidity (CA).  Data were collected June 3 to 5 for SA and June 11 to 13 for CA. 200 400 600 800 1000 1200 1400 1600 1800 2000 0 12 24 36 48 60 Days after onset of experiment C O 2  ( pp m ) SA CA 0 1 2 3 4 5 6 7 0 12 24 36 48 60 Hour H um id ity  d ef ic it (g  m -3 ) 0 10 20 30 40 50 60 70 80 90 100 R el at iv e hu m id ity   ( % ) CA SA 0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0 12 24 36 48 60 Days after onset of experiment G lo ba l r ad ia tio n (M J m  -2 h -1 ) SA CA 10 12 14 16 18 20 22 24 26 28 30 0 12 24 36 48 60 Hour Te m pe ra tu re  (C ) SA CA A B C D                               12          24          12 0                           12         24         12 0         12      24       12       24       12 0           12         24          12          24          12  166  Figure 5.4 The diurnal environment from two commercial greenhouses, CanAgro (CA) and South Alder (SA) in 2001.  Panel A is global radiation, measured exterior to the greenhouse, Panel B is hourly CO2 concentration, Panel C is hourly temperature and Panel D is hourly humidity deficit (SA) and relative humidity (CA).  Data were collected August 21 to 23 for SA and August 19 to 21 for CA. 0 0.05 0.1 0.15 0.2 0.25 0.3 0 12 24 36 48 60 Days after onset of experiment G lo ba l r ad ia tio n (M J m  -2 h -1 ) SA CA 10 12 14 16 18 20 22 24 26 28 0 12 24 36 48 60 Hour Te m pe ra tu re  (C ) SA CA 200 400 600 800 1000 1200 1400 1600 1800 2000 0 12 24 36 48 60 Days after onset of experiment C O 2  ( pp m ) SA CA 0 2 4 6 8 10 12 0 12 24 36 48 60 Hour H um id ity  d ef ic it (g  m -3 ) 0 10 20 30 40 50 60 70 80 90 100 Re la tiv e hu m id ity  (% ) CA SA A B C D 0            12         24          12          24           12 0         12         24       12        24        12 0        12      24       12      24       12 0                     24          12          24           12  167  Figure 5.5 The diurnal environment from two commercial greenhouses, CanAgro (CA) and South Alder (SA) in 2001.  Panel A is global radiation exterior to greenhouse, Panel B is hourly CO2 concentration, Panel C is hourly temperature and Panel D is hourly humidity deficit (SA) and relative humidity (CA).  Data were collected September 9 to 11 for SA and September 5 to 7 for CA. 0 0.05 0.1 0.15 0.2 0.25 0.3 0 12 24 36 48 60 Days after onset of experiment G lo ba l r ad ia tio n (M J m  -2 h- 1 ) SA CA 10 12 14 16 18 20 22 24 26 28 0 12 24 36 48 60 Hour Te m pe ra tu re  (C ) SA CA 200 400 600 800 1000 1200 1400 1600 1800 2000 0 12 24 36 48 60 Days after onset of experiment C O 2  (p pm ) SA CA 0 1 2 3 4 5 6 7 8 9 10 0 12 24 36 48 60 Hour H um id ity  d ef ic it (g  m  -3 ) 0 10 20 30 40 50 60 70 80 90 Re la tiv e hu m id ity  (% ) CA SA A B C D                              12         24         12 0                       12       24       12 0          12         24        12         24        120          12         24          12         24         12  168 The diurnal profile of LMA was largely similar to that of leaf starch.  However, at times LMA exhibited greater diurnal variability than did leaf starch.  For example, on August 20 LMA increased from 43.6 g m-2 at dawn (first day) to 55.2 at 17 h, a difference of 11.6 g m-2 (Figure 5.2B).  Over the same period starch rose from 9.8 g m-2 to 16.5 g m-2 corresponding to an increase of 6.7 g m-2 (Figure 5.1B).  Therefore, the rise in LMA exceeded the starch contribution by 4.9 g m-2. Research greenhouses.  In the research greenhouses at PARC-Agassiz, leaves from both CO2 treatments and canopy positions exhibited significant diurnal variation in leaf starch concentration (p<0.007).  Maximum leaf starch was consistently observed at about 14 h (Figure 5.6).  The minimum was sometimes recorded at dawn but sometimes starch concentration was equally low at the 10 h observation (Figure 5.6).  For leaves at position 1 at day one, enrichment with CO2 resulted in greater leaf starch at 5:30 h (p=0.0066), 14 h (p=0.048) and 20 h (p=0.049) than in ambient CO2 (Figure 5.6A).  For leaves at position 7 on day one, CO2 enrichment resulted in significantly more starch at 10 h (p=0.0097) and 14 h (p=0.0032) than in ambient CO2 (Figure 5.6B).  The lower canopy leaves (position 13) in either CO2 treatment exhibited the lowest amounts of starch and had the lowest diurnal variability (Figure 5.6C). The diurnal plots indicate that the upper leaves of plants exposed to CO2 enrichment also carried over larger amounts of starch to the next morning than did plants raised in ambient CO2 (Figures 5.6A and B).  At all leaf positions, plants in ambient CO2 started the days with relatively little leaf starch (Figure 5.6).  CO2 enrichment only increased the leaf starch of the top and middle leaves during the first day of monitoring, with no further  169 Figure 5.6 Diurnal profiles of leaf starch in plants exposed to ambient or enriched CO2 at PARC-Agassiz.  Data were collected on August 24 and 25, 2002, 74 days after the onset of CO2 enrichment.  Panel A: leaves from position 1 (top of canopy), Panel B: leaves from position 7 (middle) and Panel C: leaves from position 13 (lower).  The horizontal dark bars indicate night.  Vertical bars are 95% confidence intervals of the mean, N=2.  Means statistically significantly different at p<0.05 and p<0.01 are indicated by * and **, respectively. 0 1 2 3 4 5 6 7 0 4 8 12 16 20 24 28 32 36 40 44 48 Portion of day Le af  st ar ch  (g  m -2 ) A E 0 1 2 3 4 5 6 7 0 4 8 12 16 20 24 28 32 36 40 44 48 Portion of day Le af  st ar ch  (g  m -2 ) A E 0 1 2 3 4 5 6 7 0 4 8 12 16 20 24 28 32 36 40 44 48 Hour Le af  st ar ch  (g  m -2 ) Ambient Enriched A B C * ** * ** **                                  16      20             4        8        12      16      20       24  170 increase in peak starch level on the second day (Figure 5.6A and B).  In fact, during the second day, peak starch levels were similar in the ambient and enriched treatments.  Because the ambient treatments started with a lower minimum concentration at the beginning of the second day, this result required somewhat more rapid starch accumulation in ambient CO2 up to 14 h of that day (Figure 5.6A and B). In this experiment no statistical differences in LMA were found with time of sampling or CO2 enrichment treatment (p>0.05).  The lack of significance can be attributed to the high variability within a collection time (Figure 5.7). The diurnal environmental data for one day prior to, plus the two days of data collection, are summarized in Figure 5.8.  The temperature was similar among the treatments and relative humidity was at times about 4% higher for the greenhouses at ambient CO2 (Figure 5.8).  The environment for both treatment compartments on day two of sample collection was 3 C lower in temperature, and the daily radiation was also lower at midday compared to day one (Figure 5.8). 5.3.2 Leaf starch and LMA after the onset of CO2 enrichment In the 2002 experiment, starch levels were relatively low compared to the data obtained in the diurnal study (Table 5.1A).  Starch concentrations and LMA were not affected by leaf position, time of sampling or exposure to CO2 enrichment (Table 5.1). Environmental monitoring during this experiment (Figure 5.9) indicated that temperature was similar among the greenhouse compartments, but relative humidity was slightly higher for the compartments with ambient CO2 (Figure 5.9C and D).  The  171  Figure 5.7 Diurnal profiles of leaf mass per unit area (LMA) in plants exposed to ambient or enriched CO2 at PARC-Agassiz.  Data were collected on August 24 and 25, 2002, 74 days after the onset of CO2 enrichment.  Panel A: leaves from position 1 (top of canopy), Panel B: leaves from position 7 (middle), and Panel C: leaves from position 13 (lower).  The horizontal dark bars indicate night.  Vertical bars are 95% confidence intervals of the mean, N=2. 20 22 24 26 28 30 32 34 36 38 0 4 8 12 16 20 24 28 32 36 40 44 48 Portion of day LM A  (g  m -2 ) A E 20 22 24 26 28 30 32 34 36 38 0 4 8 12 16 20 24 28 32 36 40 44 48 Portion of day LM A  (g  m -2 ) A E B A 20 22 24 26 28 30 32 34 36 38 0 4 8 12 16 20 24 28 32 36 40 44 48 Hour LM A  (g  m -2 ) Ambient Enriched   0                                                 4        8        12      16      20       24 C  172  Figure 5.8 The diurnal environment of the two treatments in the 2002 PARC-Agassiz experiments.  Ambient did not receive supplemental CO2 and Enriched was given supplemental CO2.  Panel A is global radiation -exterior to greenhouse.  Panel B is mean hourly CO2 concentration.  Panel C. mean hourly temperature, and Panel D is mean hourly relative humidity.  Data were collected between August 23 and 25, 2002. Except for global radiation, N=2. 0 0.5 1 1.5 2 2.5 3 0 6 12 18 0 6 12 18 0 6 12 18 0 Days after onset of experiment G lo ba l r ad ia tio n (M J m  -2  h r-1 ) Series1 200 400 600 800 1000 1200 1400 1600 0 6 12 18 0 6 12 18 0 6 12 18 0 Days after onset of experiment CO 2 (p pm ) Ambient Enriched 10 15 20 25 30 35 0 6 12 18 0 6 12 18 0 6 12 18 0 Hour Te m pe ra tu re   ( C) Ambient Enriched 0 20 40 60 80 100 120 0 6 12 18 0 6 12 18 0 6 12 18 0 Hour Re la tiv e hu m di ty   ( % ) Ambient Enriched DC BA  173 Table 5.1 Leaf starch and leaf mass per unit area from leaves collected prior to and during the first seven days of exposure to CO2 enrichment at PARC-Agassiz, June 10 to 17, 2002.  Table 5.1A Leaf starch per unit leaf area.  The means and 95% confidence intervals are presented for three leaf positions and at two times during the day. Leaf1 Time Trt2 Day 0 Day 1 Day 5 Day 7 Pos. (hr)  Starch (g m-2) Ambient 0.63 ±0.52 1.37 ±0.69 2.62 ±1.7 0.98 ±0.45 9 Enriched 0.78 ±0.29 0.77 ±0.40 1.35 ±0.92 1.53 ±0.80 Ambient 1.45 ±0.83 1.88 ±0.71 2.63 ±0.59 1.34 ±0.47 1 14 Enriched 1.78 ±0.58 1.19 ±0.47 2.81 ±1.23 1.41 ±0.40 Ambient 1.03 ±0.46 1.89 ±0.82 1.26 ±0.39 1.26 ±0.76 9 Enriched 0.53 ±0.25 0.66 ±0.26 1.97 ±0.88 1.14 ±0.72 Ambient 2.47 ±0.96 1.88 ±0.70 2.47 ±0.71 1.32 ±0.53 7 14 Enriched 1.57 ±0.67 1.98 ±0.47 4.15 ±1.47 2.34 ±0.71 Ambient 0.68 ±0.26 0.50 ±0.25 0.96 ±0.22 1.35 ±0.46 9 Enriched 0.66 ±0.19 1.27 ±0.69 1.24 ±0.51 1.02 ±0.40 Ambient 1.53 ±0.39 2.22 ±1.14 1.14 ±0.40 1.35 ±0.44 13  14 Enriched 1.75 ±0.67 2.57 ±01.06 2.85 ±0.98 2.21 ±1.07 1. Leaf position, leaf 1 is the leaf closest to the truss with open flowers. 2. Treatment, Ambient CO2 averaged 419.7 ppm and Enriched CO2 averaged 824.5 ppm.  Table 5.1B Leaf mass per unit area.  The means and 95% confidence intervals are presented for three leaf positions and at two times during the day. Leaf1 Time Trt2 Day 0 Day 1 Day 5 Day 7 Pos. (hr)  Leaf mass per unit area (g m-2) Ambient 27.07 ±3.87 24.94 ±3.55 25.20 ±1.63 24.98 ±2.22 9 Enriched 25.27 ±2.04 20.77 ±2.16 22.70 ±2.78 21.74 ±2.09 Ambient 25.28 ±2.64 28.38 ±2.81 28.9 ±2.89 28.45 ±2.19 1 14 Enriched 23.20 ±2.34 20.68 ±2.47 26.49±3.98 24.48 ±1.68 Ambient 28.17 ±2.96 25.76 ±4.43 22.79 ±2.74 22.99 ±2.49 9 Enriched 21.01 ±2.17 21.29 ±2.07 25.81 ±3.64 21.35 ±1.94 Ambient 22.45 ±1.39 25.50 ±3.01 27.01 ±1.87. 26.02 ±0.2.39 7  14 Enriched 27.55 ±3.86 26.20 ±2.10 29.97±3.44 25.49 ±1.72 Ambient 25.03 ±2.15 22.08 ±1.94 23.02 ±1.44 23.47 ±2.59 9 Enriched 24.87 ±3.88 23.27 ±2.25 24.71 ±1.52 21.94 ±1.77 Ambient 25.37 ±3.27 27.92 ±2.46 20.87 ±1.41 24.84 ±2.01 13  14 Enriched 25.65 ±2.13 29.62 ±3.03 28.94 ±2.60 25.78 ±2.72 1. Leaf position; leaf 1 is the leaf closest to the truss with open flowers. 2. Treatment; Ambient CO2 averaged 419.7 ppm and Enriched CO2 averaged 824.5 ppm.  174  Figure 5.9 Summary of the environmental data for greenhouse compartments treated with either Ambient or Enriched CO2 at PARC-Agassiz during the CO2 onset experiment in 2002.  Panel A is total global radiation per day measured external to the greenhouse.  Panel B is mean daily atmospheric CO2. Panel C is mean daily temperature and Panel D is mean daily relative humidity.  Data were collected between June 10 to 17 2002. N=2 except for global radiation and bars are 95% confidence intervals of the mean. 0 5 10 15 20 25 30 35 0 1 2 3 4 5 6 7 Days after onset of experiment G lo ba l ra di at io n (M J m  -2 d-1 ) Ambient 200 400 600 800 1000 1200 1400 1600 0 1 2 3 4 5 6 7 Days after onset of experiment CO 2 ( pp m ) Ambient Enriched 10 12 14 16 18 20 22 24 26 28 30 0 1 2 3 4 5 6 7 Days after onset of experiment Te m pe ra tu re  (C ) Ambient Enriched 20 30 40 50 60 70 80 90 100 0 1 2 3 4 5 6 7 Days after onset of experiment Re la tiv e hu m id ity  (% ) Ambient Enriched DC BA  175 greenhouse compartments receiving CO2 enrichment exhibited 600 ppm by day 1 which increased to a high of 1400 ppm by day 6 (Figure 5.9B).  The ambient CO2 compartments were about 400 ppm (Figure 5.9B).  The global radiation was variable during the latter part of this experiment.  For the first five days of the experiment (day 0 to 4) the global radiation was about 28 MJ m-2 d-1; afterwards it dropped to a low of 5 MJ m-2 d-1 on day 6 (Figure 5.9A). In the 2003 experiment, treatment with CO2 enrichment increased the leaf starch and LMA of the upper leaves.  For leaves in position 1, CO2 enrichment resulted in a significant increase in leaf starch at day 4 (p=0.0068) and day 7 (p=0.0035).  For the middle leaves (position 7) leaf starch was significantly increased by CO2 enrichment at day 2 (p=0.048) and day 7 (p=0.014) (Figure 5.10).  Treatment with CO2 enrichment did not significantly increase the starch of the leaves in position 13 (p=0.47) at any day sampled (Figure 5.10). Leaf mass per area exhibited an increase under CO2 enrichment but the effect depended on leaf position and duration of exposure.  Leaves in position 13, did not exhibit any significant differences in LMA attributable to CO2 enrichment during the study (p=0.23) (Figure 5.11).  Enrichment with CO2 significantly increased LMA at day 4 (p=0.014) and day 7 (p=0.0065) for leaves in position 1 and day 4 (p=0.009) for leaves in position 7 (Figure 5.11).  It should be noted that at day 4 (leaf 7), there was a dip in the LMA in the ambient treatment but the LMA in the enriched treatment did not increase. In the 2003 onset experiment, the global radiation exhibited limited variability, and the temperature, relative humidity and CO2 concentration of the greenhouse compartments were relatively stable (Figure 5.12).  Global radiation ranged between 14 and 25 MJ m-2 d-1 and the temperature was 20 to 22 C for the greenhouse compartments (Figure 5.12A and C).  176  Figure 5.10 Leaf starch per unit leaf area prior to and during the first seven days of exposure to Enriched or Ambient CO2 at PARC-Agassiz, May 6 to 13 2003.  Data were collected from three leaf positions at 14 h.  Panel A is for leaf position 1 (top of canopy), Panel B is for leaf position 7 (middle) and Panel C is for leaf position 13 (lower).  Symbols are the means (N=2) and bars are 95% confidence intervals of the mean.  Means significant different at p<0.05 and p<0.01 are indicated by * and **, respectively. 0.0 4.0 8.0 12.0 16.0 0 1 2 3 4 5 6 7 Days after treatment Le af  st ar ch  (g  m -2 ) A E 0.0 4.0 8.0 12.0 16.0 0 1 2 3 4 5 6 7 Days after treatment Le af  st ar ch  (g  m -2 ) A E 0.0 4.0 8.0 12.0 16.0 0 1 2 3 4 5 6 7 Days after treatment Le af  st ar ch  (g  m -2 ) Ambient Enriched * ** ** * B A C  177  Figure 5.11 Leaf mass per unit area (LMA) prior to and during the first seven days of exposure to Enriched or Ambient CO2 at PARC-Agassiz, May 6 to 13 2003.  Data were collected from three leaf positions at 14 h.  Panel A is from leaf position 1 (top of canopy).  Panel B is from leaf position 7 (middle) and Panel C is from leaf position 13 (lower). Symbols are the means (N=2) and bars are 95% confidence intervals of the mean.  Means significant different at p<0.05 and p<0.01 are indicated by * and **, respectively. 20 25 30 35 40 45 50 55 0 1 2 3 4 5 6 7 Days after treatment LM A  (g  m -2 ) A E 20 25 30 35 40 45 50 55 0 1 2 3 4 5 6 7 Days after treatment LM A  (g  m -2 ) AC AP EC EP 20 25 30 35 40 45 50 55 0 1 2 3 4 5 6 7 Days after treatment LM A  (g  m -2 ) Ambient Enriched ** * ** B A C  178 Figure 5.12 Summary of the environmental data for greenhouse compartments treated with either Ambient or Enriched CO2 at PARC-Agassiz during the CO2 onset experiment in 2003.  Panel A is total global radiation per day measured external to the greenhouse.  Panel B is mean daily atmospheric CO2. Panel C is mean daily temperature and Panel D is mean daily relative humidity.  Data were collected between May 5 to 13 2003. N=2 except for global radiation and bars are 95% confidence intervals of the mean.  CO2 enrichment was started at 15 h on day 0; therefore day -1 data (two days before the experiment commenced) were shown to confirm no pre-treatment differences in CO2 concentration. 200 400 600 800 1000 1200 1400 1600 -1 0 1 2 3 4 5 6 7 Days after onset of experiment CO 2 ( pp m ) Ambient Enriched 0 5 10 15 20 25 30 35 -1 0 1 2 3 4 5 6 7 Days after onset of experiment G lo ba l ra di at io n (M J m  -2 d-1 ) Ambient 20 30 40 50 60 70 80 90 100 -1 0 1 2 3 4 5 6 7 Days after onset of experiment Re la tiv e hu m id ity  (% ) Ambient Enriched 10 12 14 16 18 20 22 24 26 28 30 -1 0 1 2 3 4 5 6 7 Days after onset of experiment Te m pe ra tu re  (C ) Ambient Enriched DC BA  179 The greenhouse compartments receiving CO2 enrichment were for the most part about 400 ppm greater than those at ambient CO2 (Figure 5.12B).  Carbon dioxide was elevated at day 0 because CO2 enrichment commenced at 15 h, after the leaf samples were harvested for starch and LMA analysis.  Relative humidity for day -1, 1, 3 and 6 was significantly greater for the compartments receiving CO2 enrichment (Figure 5.12). 5.4 DISCUSSION This discussion will focus on the diurnal variations in leaf starch concentration (section 5.4.1) and LMA (section 5.4.2) and the changes in those measures after the onset of CO2 enrichment (section 5.4.3).  The potential to use these results in plant-based CO2 management decisions will be deferred to the General Discussion (Chapter 6). It was advantageous to carry out some of this research in commercial greenhouses because the results are directly relevant to commercial practice.  However, the use of commercial greenhouses also had shortcomings in that there was limited replication, no randomization and no control treatment at ambient CO2.  The PARC-Agassiz experiments were distinct from the commercial greenhouses, and differences in their external climate may account for some of the differences between the commercial greenhouses and PARC- Agassiz.  Also, in proportion to their size, the climate variation within the small research greenhouse compartments may have been more extreme than in the larger commercial greenhouses.  Plants at PARC-Agassiz exhibited higher variability about the mean and lower responsiveness for both leaf starch and LMA than did those in the commercial greenhouses. Large standard errors and confidence intervals have also been reported by Bertin et al. (1999) for leaf starch and LMA data in a similar diurnal experiment.  180 5.4.1 Diurnal profiles of leaf starch In accordance with Ayari et al. (2000a) and Erhioui et al. (2002) my studies indicated that starch concentration in the upper leaves appeared to be influenced by the diurnal profile of light.  Also in agreement with Ayari et al. (2000a) and Erhioui et al. (2002), days when light levels were low resulted in leaves having low levels of starch and little diurnal variation. On a sunny day, leaf starch usually started out being the lowest during the morning (not necessarily at dawn), then built to a maximum 2 to 3 hours past the solar maximum and then either stabilized, or declined in the afternoon/evening.  The values measured at PARC- Agassiz were lower than in the commercial greenhouses but were generally similar in shape. The lower starch at PARC-Agassiz was probably caused by lower light penetration into the greenhouse, especially in 2002. Leaves near the top of the canopy and under CO2 enrichment consistently exhibited significant amounts of starch at dawn, with values being 10 to 15 g m-2 for the commercially raised plants and 2 to 3 g m-2 for the CO2 enriched plants at PARC-Agassiz.  At PARC- Agassiz the control plants in ambient CO2 exhibited close to zero starch at dawn for both days sampled (Figure 5.4).  The amount of starch at dawn in the CO2 enriched plants was sizable, as it was often two thirds of the maximum starch measured later in the day. Appreciable amounts of starch at dawn in upper canopy leaves have been reported by Ayari et al. (2000b) who measured 17 g m-2 at 6 h, and Pressman et al. (1997) who measured 12 g m-2 also at 6 h (also two thirds of the maximum starch measured for that day).  These high levels indicate that leaves were unable to use or export carbon from all their starch reserves during the night.  Gary et al. (2003) indicates that it can take a significant amount of time for starch to be used during darkness.  Gary et al. (2003) reported that at the onset of darkness the upper leaves (in position 1 to 4) exhibited 19% (roughly equivalent to 8 g m-2 in my  181 work) starch per leaf dry mass.  After 10 hours of darkness leaf starch was reduced to 10% of leaf dry mass (approximately 4.0 g m-2).  Thereafter, the rate of decline in starch was even slower and starch did not entirely disappear until 60 hours of darkness (Gary et al. 2003). This pattern suggests that the immediate requirements of fruit sinks are easily met by a moderate leaf starch reserve, or perhaps the breakdown of starch is under circadian control (Li et al. 1992).  My diurnal profiles and the work of Gary et al. (2003) seem to indicate that CO2 enriched greenhouse tomato plants can form more starch during the day than they can use at night.  The high amounts of upper leaf starch at dawn could be the result of several days of surplus starch, and might reside in those leaves until nearby sink strengths become sufficient to draw down the reserve carbon. For the upper leaves of CO2 enriched plants, in both commercial greenhouses and at PARC-Agassiz, leaf starch levels were similar between dawn and 10 h, or they were even slightly lower at 10 h than at dawn.  This mid-morning plateau or dip in leaf starch was also measured in tomato by Madsen (1968), King et al. (1988) (at 2200 ppm CO2), and by Klages et al. (2001) in apple.  I would have expected leaf starch to build during this period as light availability and photosynthesis increase.  However, if during this period the fruit have a higher demand, or priority, for current assimilates than for carbon stored in starch, leaf starch reserves might remain stable. Peak starch levels were observed in the early afternoon, after which time the levels of leaf starch may diminish.  The trend in leaf starch in plants from commercial greenhouses in the afternoon is unclear, partly resulting from the fact that leaf samples were not collected at sunset in the commercial greenhouses.  In the PARC-Agassiz data, however, samples collected at sunset usually had lower starch than those collected at 14 h.  In published diurnal  182 studies, leaf starch has been found to exhibit one of three patterns in the later part of the day: (1) a maximum at 17 h with a decline thereafter (Madsen 1968), (2) no change later in the day (Delucia et al. 1985; Galtier et al. 1995; Hou 1997; Ayari et al. 2000a; Erhioui et al. 2002) or (3) an increase as the day progressed (Delucia et al. 1985; Ammerlaan et al. 1986; Pressman et al. 1997; Bertin et al. 1999; Ayari et al. 2000b).  Continued starch build-up in the late afternoon might occur, despite diminishing light for photosynthesis if sucrose transfer to sinks declines by more than what is produced by photosynthesis.  Translocation of assimilates to the sinks may be slowed by lower greenhouse temperatures at the end of the day (Heuvelink & Dorais 2005). Upper leaves from plants in commercial greenhouses sometimes exceeded 20 g m-2 of starch.  Carrying such a large quantity of starch may reduce future photosynthesis and may also be related to CO2 adaptation of these leaves (Stitt 1991).  Plants in ambient CO2 at PARC-Agassiz usually exhibited less starch in their top leaves than the enriched plants at dawn.  However, the daily gain in starch (dawn to 14 h) was larger than for the plants in enriched CO2.  This is surprising as the CO2 concentration was at least 300 ppm higher in the CO2 enriched compartments compared to ambient compartments.  These results may indicate that high starch has an inhibitory effect on photosynthesis and subsequent starch accumulation, and there is some evidence that prolonged exposure to high CO2 can diminish photosynthetic capacity (Stitt 1991; Besford 1993). On the other hand, carrying surplus leaf starch would be beneficial to maintain fruit growth when a plant enters a prolonged period of low photosynthesis, such as several days of cloudy weather.  For the CO2-enriched plants, the nightly draw-down of leaf starch was about one-third of the peak value.  This suggests that even leaves heavily laden with starch only  183 contain about three days of reserves if photosynthetic replenishment is insufficient.  A similar conclusion is reached from the observation that leaf starch can become low after only two days of low light (Figure 5.1B and 5.4A).  Moreover, for the individual leaves high starch levels do not persist as canopy development progresses.  Leaves become shaded by new growth, lowering the capability for assimilation, and nearby sinks become larger with higher capacity for assimilates. In agreement with the results presented in Chapter 4, leaves lower than position 9 contained only small amounts of starch.  The amplitude of the diurnal variation of leaf starch decreased with depth in the canopy.  Leaves in position 13, consistently maintained less than 3 g m-2 of starch throughout the day, and enrichment with CO2 only modestly increased the starch concentration but not LMA (Figures 5.4 and 5.5).  These results are similar to the findings of Ammerlaan et al. (1986), Pressman et al. (1997), and Bertin et al. (1999) where leaves in position 12 and lower exhibit little starch accumulation and have little diurnal variation. 5.4.2 Diurnal profiles of LMA There have only been a few studies of the diurnal variability of LMA or its reciprocal, specific leaf area.  The results in Chapter 4 established that leaf starch influences LMA, especially in the upper leaves.  In this chapter, all of the leaves from the commercially-raised plants were collected from the upper canopy, so it is not surprising that the diurnal profiles of LMA were similar to the profiles for leaf starch.  As with starch, LMA was similar between the dawn and mid-morning measurements.  After mid-morning, LMA usually exhibited an increase, reaching a maximum at 14 to 16 h.  After late afternoon, LMA exhibited no change or there was a decline at sunset which continued overnight.  These patterns are different from  184 those reported by Bertin et al. (1999) for tomato raised in greenhouses, and by Hrubec et al. (1985) for soybean in growth chambers.  Bertin et al. (1999) reported that the LMA of terminal leaflets exhibited a steady increase from 5 h to 13 h to 18h (sunrise to sunset).  Had Bertin and co-workers (1999) made more frequent measurements, it is possible that their profile for LMA could have been similar to what I found.  Hrubec et al. (1985) measured LMA of soybean (Glycine max) plants in enriched or ambient CO2 four times during the daylight period.  In either CO2 treatment LMA increased from the beginning of the light period to the end.  Since Hrubec’s study was carried out in growth chambers, with no diurnal variation in light through the daytime period, it may not be surprising that LMA would continue to build under these constant light conditions. At PARC-Agassiz, plants exposed to CO2 enrichment exhibited no increases in LMA compared to plants raised in ambient CO2, and the variation in the LMA values made it difficult to discern relationships with leaf starch.  For example, for leaves in position 1 on day one at 5:30 h, CO2 enriched plants exhibited a 3 g m-2 increase in starch over the ambient plants, while the corresponding difference in LMA was only 1 g m-2.  However, differences of 1 to 3 g m-2 are small against the background variability of starch and LMA.  In soybean, Hrubec et al. (1985) found CO2 enrichment resulted in a 5 g m-2 increase in LMA over plants in ambient CO2 at dawn.  Later in the day (12 hours) LMA increased to a 10 g·m-2 difference between treatments (Hrubec et al. 1985). The PARC-Agassiz experiments indicated LMA was similar among the three leaf positions in the canopy.  Leaves in position 1 exhibited more starch than did leaves in position 13, but LMA was similar between these leaves.  An explanation could be that the low levels of starch in the PARC-Agassiz experiments were inadequate to alter LMA.  Bertin  185 et al. (1999) observed that LMA was similar among leaves of different canopy positions, but like the PARC-Agassiz work those studies involved low leaf starch levels (0 to 4 g m-2). There was a substantial difference in LMA between the commercial greenhouses and the research greenhouses at PARC-Agassiz.  In the commercial greenhouses LMA ranged from 33 to 58 g m-2 and at PARC-Agassiz the range was lower, being 24 to 33 g m-2 (Figures 5.2 and 5.7).  The lower LMA measures at PARC-Agassiz were partly caused by lower leaf starch.  But, it is also likely that differences in soluble sugars and structural components contributed to this, in part caused by the lower  global radiation at PARC-Agassiz in accordance with the findings of Sims et al. (1998) and Poorter et al. (2006). 5.4.3 Leaf starch and LMA at the onset of CO2 enrichment In the 2002 experiment, CO2 enrichment did not cause a significant increase in either leaf starch or LMA (Table 5.1).  These findings contrast to the 2003 data where both leaf starch and LMA of the leaves in position 1 (top of canopy) were significantly increased by day two to four (Figures 5.10 and 5.11).  Comparison of the environment data between 2002 and 2003 revealed that CO2 concentrations were similar and global radiation measured outside the greenhouse was in fact substantially higher for the first three days of the 2002 experiment than the 2003 experiment (Figures 5.9 and 5.12).  Photosynthesis measurements carried out in both years, revealed that PAR for leaves in position 1 was 2.5 times higher in 2003 than in 2002 and photosynthesis was increased by CO2 enrichment in 2003 but not in 2002 (Appendix 5.1).  I believe these findings were caused by the application of whitewash in 2002 which was used to help control greenhouse temperature but substantially reduced the light reaching the crop.  The rest of this discussion will focus on the results from the 2003 experiment.  186 Many of the earlier studies where leaf starch was monitored at the onset of CO2 enrichment have focused on young tomato plants with few fruit.  As well, plant responses were examined in growth chambers and experiments were for longer or shorter timeframes then I used.  Ho (1978) measured starch in an upper leaf of young tomato plants (48 days old) one, two, five and seven days after transfer to an environment of 1000 ppm CO2.  By the fifth day, starch had increased 75% over the measurement made on the first day (Ho 1978). Direct comparisons to my data are difficult because Ho (1978) measured starch on a fresh mass basis; however, his increases were relatively lower and slower compared to the data in Figure 5.10A.  In the upper leaves, I found starch exhibited an 80% increase (compared to time zero) after two days that eventually increased to 89% after seven days of exposure to enriched CO2.  Ho (1978) has reported substantial increases in leaf starch which is surprising given the very low light levels (40 umol m-2 s-1) his plants were exposed to.  However, the increased precision of CO2 enrichment made possible in a growth chamber and the sink limitation (only 1 truss of fruit) of his plants may compensate for the lower light.  Yelle et al. (1989) examined the weekly response of leaf starch starting with one month old tomato plants exposed to CO2 enrichment for 10 weeks.  Plants under CO2 enrichment (900 ppm) exhibited a 38% increase in leaf starch for leaves in position 1 (referred to by Yelle as leaf 5) and 47% increase for leaves in position 9 (leaf 9 by Yelle) relative to plants in ambient CO2 (330 ppm) after one week (Yelle et al. 1989).  These results are contrary to what I have found for leaf position, where upper leaves exhibited greater increases in starch than did lower leaves (Figure 5.10).  I am unsure how to explain Yelle and co-workers findings for leaf position?  A possible explanation could be that the plants in Yelle’s study had not reached a steady state of fruit production and thus their source sink relations were altered.  Yelle et al.  187 (1989) started with 32 day old plants that were 102 days old by the end of the experiment.  In my work the plants were at a steady state of fruit production and in 2003 were 138 days old before CO2 enrichment was introduced. No relevant studies were found that examined the LMA of tomato at the onset of CO2 enrichment.  In bean (Liu, 1990) and soybean (Cure et al., 1987) LMA exhibited an increase after only 2 and 3 days, respectively, of exposure to enriched CO2.  These increases are slightly faster than I had measured for tomato (Figure 5.11).  Liu (1990) measured an increase of 35 to 41 g m-2 after 2 days of exposure to 1400 ppm CO2 which is similar to the increase I measured after 2 days (Figure 5.11A). In this study, CO2 enrichment often increased LMA beyond what was attributable to increases in starch (Figures 5.10A and 5.11A).  For leaf 1, the increase in LMA between time zero and day 7 was 18.1 g m-2 while the increase in starch was only 12.4 g m-2.  Since only a week separated these measurements it is likely the difference was caused by increased soluble sugars.  Bertin (1999) has measured soluble sugars in the range of 1 to 6 g m-2 in tomato leaves. 5.5 CONCLUSIONS Conclusions pertaining to plant-based indicators of CO2 dosing will be addressed in Chapter 6.  In addition: 1. Leaf starch and LMA tended to follow the diurnal profile of light but with several hours of lag.  The highest starch contents were measured between 14 h and 16 h with the lowest in the morning between sunrise and 11 h.  In most cases plants in commercial greenhouses and those under CO2 enrichment at PARC-Agassiz carried over leaf starch from one day to the next.  188 2. LMA generally followed the same profile as leaf starch.  Highest measurements were between 14 h and 16 h and the lowest in the mornings.  Increases in LMA from morning to afternoon were in excess of those found for starch. 3. Leaf position in the canopy influenced the extent of leaf starch.  Leaves at positions 1 and 7 exhibited diurnal variations in leaf starch while leaves in position 13 did not exhibit significant changes in starch.  CO2 enrichment significantly increased starch formation during the day for leaves in positions 1 and 7 but not for leaves in position 13.  LMA was not significantly increased by CO2 enrichment for any time during the day or because of leaf position. 4. CO2 enrichment, light and greenhouse growing conditions influenced leaf starch and to some extent LMA.  Plants in commercial greenhouses exhibited substantially higher starch and LMA, and on some days more diurnal variation, than did those in the research greenhouses at PARC-Agassiz. 5. The response of leaf starch and LMA measured at the onset of CO2 enrichment and for the following days depended more on the quantity of light reaching the leaves then CO2 enrichment.  In the first experiment (2002), plants in CO2 enrichment did not exhibit an increase in leaf starch or LMA relative to the plants in ambient CO2. This was attributed to poor light penetration into the greenhouse.  In the second experiment (2003), both leaf starch and LMA increased within four days of exposure to CO2 enrichment for leaves in positions 1 and 7.  These measures continued to increase to the end of the monitoring period (7 days).  Leaves in position 13 did not exhibit any increases in leaf starch or LMA during this experiment.   189 5.6 LITERATURE CITED Ammerlaan, A.W.S., Joosten, M.H.A.J., and Grange, R.I. 1986. The starch content of tomato leaves grown under glass. Scientia Horticulturae 28: 1-9. Ayari, O., Dorais, M., and Gosselin, A. 2000a. Daily variations of photosynthetic efficiency of greenhouse tomato plants during winter and spring. Journal of American Society for Horticultural Science 125: 235-241. Ayari, O., Samson, G., Dorais, M., Boulanger, R., and Gosselin, A. 2000b. Stomatal limitation of photosynthesis in winter production of greenhouse tomato plants. Physiologia Plantarium  110: 558-564. Bertin, N., Tchamitchian, M., Baldet, P., Devaux, C., Brunel, B., and Gary, C. 1999. Contributions of carbohydrate pools to the variations in leaf mass per area within a tomato plant. New Phytologist 143: 53-61. Besford, R.T. 1993. Photosynthetic acclimation in tomato plants grown in high CO2. Vegetatio 104/105: 441-448. Cure, J.D., Rufty, T.W., and Israel, D.W. 1987. Assimilate utilization in the leaf canopy and whole-plant growth of soybean during acclimation to Elevated CO2.  Botanical Gazette 148:  67-72. Delucia, E.H., Sasek, T.W., and Strain, B.R. 1985. Photosynthetic inhibition after long-term exposure to elevated levels of atmospheric carbon dioxide. Photosynthesis Research 7: 175-184. Erhioui, B.M., Gosselin, A., Hao, X., Papadopoulos, A.P., and Dorais, M. 2002. Greenhouse covering materials and supplemental lighting affect growth, yield, photosynthesis and leaf carbohydrate synthesis of tomato plants. Journal of American Society for Horticultural Science 127: 819-824. Galtier, N., Foyer, C.H., Murchie, E., Alred, R., Quick, P., Voelker, T.A., Thepenier, C., Lasceve, G., and Betsche, T. 1995. Effects of light and atmospheric carbon dioxide enrichment on photosynthesis and carbon partitioning in the leaves of tomato (Lycopersicon esculentium L.) plants over-expressing sucrose phosphate synthase. Journal of Experimental Botany  46: 1335-1344. Gary, C., Baldet, P., Bertin, N., Devaux, C., Tchamitchian, M., and Raymond, P. 2003. Time-course of tomato whole-plant respiration and fruit and stem growth during prolonged darkness in relation to carbohydrate reserves. Annals of Botany 91: 429- 438. Hammond, J.B.W., Burton, K.S., Shaw, A.F., and Ho L.C. 1984. Source-sink relationships and carbon metabolism in tomato leaves 2. carbohydrate pools and catabolic enzymes. Annals of Botany 53: 307-314.  190 Heuvelink, E. and Dorais, M. 2005. Crop growth and yield. In Heuvelink, E., ed. Tomatoes. CABI Publishing, Cambridge, MA, pp 85-144. Ho, L.C. 1977. Effects of CO2 enrichment on the rates of photosynthesis and translocation of tomato leaves. Annals of Applied Biology 87: 191-200. Ho, L.C. 1978. The regulation of carbon transport and the carbon balance of mature tomato leaves. Annals of Botany 42: 155-64. Hou, G. 1997. Effects of light, CO2 and temperature on carbohydrate metabolism in marigold.  (Tagetes Patula). University of Kentucky. Hrubec, T.C., Robinson, J.M., and Donaldson, R.P. 1985. Effects of CO2 enrichment and carbohydrate content on the dark respiration of soybeans. Plant Physiology 79: 684- 689. King, A.I., Joyce, D.C., and Reid, M.S. 1988. Role of carbohydrates in diurnal chilling sensitivity of tomato seedlings. Physiol Plant 86: 764-768. Klages, K., Donnison, H., Wunsche, J., and Boldingh, H. 2001. Diurnal changes in non- structural carbohydrates in leaves, phloem exudate and fruit in 'Braeburn' apple. Australian Journal of Plant Physiology 28: 131-139. Li, B., Geiger, D.R., and Shieh, W.J. 1992. Evidence for circadian regulation of starch and sucrose synthesis in sugar beet leaves. Plant Physiology 99: 1393-1399. Li, C.R., Sun, W.Q., and Hew, S.H. 2001. Up-regulation of sucrose metabolizing enzymes in Oncidium goldiana grown under elevated carbon dioxide. Physiologia Plantarium 113: 15-22. Liu, H.-T. 1990. Physiological limitations to the growth response of bean plants (Phaseolus vulgaris L.)  to carbon dioxide enrichment. The University of British Columbia. Madsen, E. 1968. Effect of CO2 - concentration on the accumulation of starch and sugar in tomato leaves. Physiologia Plantarum 21: 168-175. Poorter, H., Pepin, S., Rijkers, T., de Jong, Y., Evans, J.R., and Korner, C. 2006. Construction costs, chemical composition and payback time of high- and low- irradiance leaves. Journal of Experimental Botany 57: 355-371. Pressman, E., Bar-Tal, A., Shaked, R., and Rosenfeld, K. 1997. The development of tomato root system in relation to the carbohydrate status of the whole plant. Annals of Botany 80: 533-538. SAS Institute. 2003. Statistical Analysis Software.  Cary, NC, SAS Institute Inc. Sims, D.A., Seemann, J.R., and Luo, Y. 1998. Elevated CO2 concentration has independent effects on expansion rates and thickness of soybean leaves across light and nitrogen  191 gradients. Journal of Experimental Botany 49: 583-591. Stitt, M. 1991. Rising CO2 levels and their potential significance for carbon flow in photosynthetic cells. Plant, Cell and Environment 14: 741-762. Stutte, G.W., Yorio, N.C., and Wheeler, R.M. 1996. Interacting effects of photoperiod and photosynthetic photon flux on net carbon assimilation and starch accumulation in potato leaves. Journal of American Society for Horticultural Science 121: 264-268. Yelle, S., Beeson Jr., R.C., Trudel, M.J., and Gosselin, A. 1989. Acclimation of two tomato species to high atmospheric CO2 I. Sugar and starch concentration. Plant Physiology 90: 1465-1472.  192 6 General discussion This thesis has explored several plant-based approaches that may have potential for use in the management of CO2 in commercial greenhouse tomato production.  This has involved evaluations at several levels of organization:  crop and greenhouse simulation modeling, the non-destructive analysis of plant growth, and the dynamics of leaf starch.  It was not the intent of this work to attempt to replace existing management approaches, involving environmental monitoring.  Rather, the objective was that this work would lead toward the use of plant-based information as part of the decision-making practices on CO2 dosing in commercial tomato greenhouses. 6.1 RESEARCH VENUES The venues where this research was carried out contributed to some of the strengths and the weaknesses of the work.  Collecting data in commercial beefsteak tomato greenhouses was necessary in order to produce meaningful results for the greenhouse industry.  The research facilities at PARC-Agassiz, and the commercial greenhouses (South Alder, Can Agro and Gipaanda) were dissimilar in size and climate, and the crops performed differently between the PARC-Agassiz and commercial sites.  The commercial greenhouses were large, 40 000 to 50 000 m2 in size and were in a moderate climate being located close to the coast of Georgia Strait.  The PARC-Agassiz research greenhouses were 37.5 m2 each and were located near the eastern end of the lower Fraser Valley.  Being distant from the coast resulted in PARC-Agassiz having lower light levels and higher summer temperatures than the commercial houses.  This climate difference, coupled with the smaller greenhouses, caused a less uniform and more stressful growing environment for the tomato crop at PARC- Agassiz.  I suggest that this may have been responsible for some of the differences between  193 crops in Agassiz and the commercial sites.  As well as having more commercially relevant growing conditions, commercial greenhouses can also be rich sources of data because the growers maintain extensive records of crop growth and the greenhouse environment.  Those data were invaluable to the simulation modelling chapter (Chapter 2).  A complaint concerning greenhouse model development has been that it has focussed on research venues, rather than commercial greenhouses (Challa 2002).  In this work I went directly to a commercial greenhouse to use the data that a grower would have access to for the development and application of the models. There were also disadvantages to carrying out research in the commercial greenhouses.  Although the greenhouse industry was very accommodating for this research, their production practises limited the experimental designs that were possible.  In the commercial greenhouses it was not possible to have an ambient CO2 control because growing crops exclusively at ambient CO2 (without CO2 enrichment) is not practised by any commercial tomato greenhouses in the lower mainland of BC.  As well, replication of the CO2 treatments could not be carried out because the greenhouses could not be partitioned into separate parts having independent CO2 controls.  Although the commercial greenhouses log many types of data, I found they lacked key information for the development of the greenhouse models (Chapter 2).  The models could have been improved if it were known, on an hourly basis, how much CO2 was available from heating, how much of the CO2 was dosed into the greenhouse and what portion was expelled from the greenhouse through the stack. Also, the choice of tomato cultivars can change quickly in commercial greenhouses.  The cultivar Rapsodie, which was the most popular beefsteak cultivar at the inception of this  194 work, is now no longer used.  However, the growing and CO2 dosing methods are still the same, and information obtained in this thesis can likely be applied to newer cultivars. Although different from the commercial greenhouses, the PARC-Agassiz site was invaluable for allowing an experimental design involving replication of the CO2 treatments plus ambient CO2 controls.  At PARC-Agassiz, the replication was limited to two experimental units (greenhouses).  This limitation decreased the degrees of freedom available for the statistical tests and may have been partially responsible for not detecting statistical differences in plant growth or yield caused by CO2 enrichment (Chapter 3).  Limited replication is a commonly acknowledged problem in carrying out of greenhouse research (Nederhoff 1994). 6.2 EVALUATION OF THE PLANT-BASED INDICATORS FOR GUIDING CO2 DOSING This research has led to a number of insights into the carbon relations of greenhouse tomato plants, which have been discussed in previous chapters.  Here the main focus of discussion will be on the potential for plant-based indicators to be used as an aid in guiding CO2 dosing decisions.  In evaluating that prospect, the main lines of my research were: 1. Investigation of the costs and benefits of CO2 enrichment using plant and greenhouse simulation models (Chapter 2). 2. Determining whether some non-destructive measures of plant growth can be useful for guiding CO2 dosing (Chapter 3). 3. Characterizing the distribution of starch and LMA by the development of their profile in the shoot canopy and the determination of the influence of fruit load and leaf area (Chapter 4).  195 4. Examining the temporal dynamics of leaf starch and LMA by investigating their diurnal variability and their responsiveness to the onset of CO2 enrichment (Chapter 5). Several criteria were used to evaluate the suitability of these plant-based indicators for the greenhouse growers.  These criteria are:  (1) be known from the scientific literature to be relevant to CO2 enrichment, (2) have acceptable precision and accuracy, (3) be measurable on site by the grower, (4) be convenient to measure and interpret, preferably in real time, (5) not be unduly damaging to the plants (6) be acceptable to the grower’s philosophy of growing plants and understanding of effects of CO2 enrichment on the crop, (7) have moderate technological requirements with low maintenance, and (8) be inexpensive. 6.2.1 Plant-based modeling Much of the recent research on CO2 management has focused on the development of climate control models (Alscher et al. 2001a; 2001b; 2001c; Chalabi et al. 2002a; Chalabi et al. 2002b; Dieleman et al. 2005; Aaslyng et al. 2006).  The tomato production system is well-suited to modeling efforts because of the high degree of environmental control and the strict management of the crop.  However, to my knowledge, no tomato greenhouse growers in BC use a model that includes a plant component or helps in making CO2 dosing decisions. The work described in Chapter 2 combined plant and greenhouse climate models, and used them to predict fruit yields and to explore the costs and benefits of several CO2 enrichment scenarios.  This approach was retrospective, where grower-collected fruit yield and greenhouse environment data were used in plant and greenhouse CO2 models.  Fruit yield determined by the Acock et al. (1978) and Aikman (1996) models underestimated the seasonal yield collected by a grower by 10% and did not accurately simulate the short term  196 (week to week) variability in yield.  In this work, the lack of direct and detailed information on CO2 injection into the greenhouse necessitated the use of empirical relationships between CO2 generation and greenhouse CO2 concentrations.  These models are specific to each greenhouse.  However, using the approaches adopted here, other growers ought to be able to develop and apply their own empirical relationships, using data from their own greenhouses. Under current production techniques where heat and CO2 generation are coupled, the modeling results indicated that about 14% of CO2 produced by the greenhouse was used by the crop during a production season.  During the summer, despite higher vent opening, CO2 usage by the crop increased to an average of 25%, caused by increased photosynthesis. Simulation modelling allowed the evaluation of the costs of CO2 enrichment in relation to the increased revenue from fruit production at high, intermediate and low CO2 rates of dosing. With a CO2 cost of $0.10 per kg, high CO2 usage was the most profitable.  However, if the cost of CO2 increased to about $0.50 per kg then the low and high CO2 dosing was similar if fruit price was held constant. The models used in Chapter 2 linked CO2 in the greenhouse atmosphere to canopy photosynthesis and fruit yield.  Using a photosynthesis sub-model is a convenient approach to connecting plant performance to atmospheric CO2.  However, I found that the connection from photosynthesis to yield was imprecise.  As well, this approach still does not assess the carbon needs of the plant. As a plant-based indicator for CO2 management the modeling approach possesses many attractive features for a greenhouse grower.  The models I used do not require plant tissue and only minimal input measurements; therefore, no damage to the crop is involved. The growers already have most of the required technology (computer and sensors) to collect  197 the required input data, and they are experienced in using this technology.  I found the models to have reasonable relevance to CO2 enrichment through their use of a photosynthesis sub-model, although a drawback was lower precision and accuracy for fruit yield than desired.  The modeling approach was also hampered by the limited information available on CO2 injection into the greenhouse, which necessitated the use of greenhouse-specific relationships between CO2 generation and greenhouse CO2 concentrations.  Finally, the approach I took was retrospective which is useful for the broader aspects of decision-making. A real-time approach, however, would be more helpful for the hourly decision-making that growers desire. 6.2.2 Non-destructive growth measures At the plant level, Chapter 3 investigated several measures that growers record to track crop growth that may also be useful indicators for CO2 management decisions.  Tomato growth under CO2 enrichment has been extensively documented (Kimball 1983; Slack 1986; Atherton & Harris 1986; Picken et al. 1986; Mortensen 1987; Nederhoff 1994; Gruda 2005; Heuvelink & Dorais 2005; Peet & Welles 2005) and researchers have usually assessed growth by a destructive analysis of the plant or shoot.  In a commercial greenhouse, however, growth measurements need to be non-destructive and cause a minimum of plant disturbance and interference with production activities.  For that reason, my studies focussed on the collection and interpretation of non-destructive growth measures. Non-destructive analysis was carried out in research greenhouses at PARC-Agassiz for tomato plants grown in enriched or ambient CO2 (Chapter 3).  Application of CO2 to a daily mean of 800 ppm (double the ambient CO2) did not cause injury to the plants and resulted in non-significant or at best modest, increases in growth and yield.  Stem length,  198 stem diameter and leaf area exhibited no changes that were attributed to CO2 enrichment. Yield and fruit load were significantly increased by CO2 enrichment, but this response was inconsistent (Chapter 3). The lack of a strong response of the plants to CO2 enrichment makes it difficult to assess the usefulness of the non-destructive measures.  Two possible conclusions can be drawn from my data:  CO2 enrichment caused increased plant growth but the destructive measures and/or experimental design were unable to resolve it, or CO2 enrichment did not increase plant growth and the non-destructive measures confirmed this.  In either case the inability of these measures to track growth limits their role in a plant-based method of CO2 management.  It is noteworthy to mention that differences in growth may have been quantifiable had these measures been made differently or had other non-destructive measures been examined. Had the non-destructive measures exhibited a larger response to CO2 enrichment, they may have had a role to play in decision-making for CO2 dosing.  Non-destructive measures possess a number of desirable attributes for a plant-based indicator: they are measurable on site without damaging the crop (although repeated touching in a short timeframe may affect growth), are easy to measure, are routinely monitored by growers, involve low technology with minimal maintenance, and the equipment they require is inexpensive.  In my work, the variability of some of these measurements was high.  In a commercial greenhouse these measures would exhibit greater precision than I found because of greater consistencies in the greenhouse environment and plant spacing.  Had any of these measures shown a timely response to CO2 enrichment, future work could focus on using  199 more sophisticated technology such as imaging cameras, linear variable displacement transducers or suspended load cells (Ehret et al. 2001; Helmer et al. 2005). 6.2.3 Starch and LMA To assess if starch or LMA could play a role in a plant-based method for managing CO2 dosing, basic information on the distribution of starch in the shoot and its temporal dynamics were investigated.  As well, since it is inconvenient for a grower to measure starch, LMA was evaluated as a possible surrogate measure for starch. The overall canopy profiles of starch and LMA were similar for upper leaves but were often distinct for the lower leaves (Chapter 4).  For upper leaves, starch seemed to be influenced by the high PAR availability and low fruit load, leading to starch accumulation in this region.  Starch was negligible for leaves low in the canopy, where PAR was low and fruit load was high.  The profile of LMA usually, but not always, exhibited a linear decline with leaf position.  LMA often continued to decline in the canopy past the point where starch levels had reached their minimum.  Therefore leaf starch on a per leaf area basis was able to explain the variation in LMA only above the mid-canopy levels. The diurnal profiles of starch and LMA in the upper leaves followed that of light, but with several hours lag.  For upper leaves (positions 1 and 7) peak leaf starch and LMA was measured between 14 h and 16 h and the lowest values were observed in the morning.  It was also observed that leaf starch and LMA did build until after 10 h, which may indicate a source limitation between dawn and at least mid-morning.  For the upper leaves it was apparent that a sizable amount of starch was carried over from the end of one day to the beginning to the next.  Lower leaves (position 13) exhibited near zero starch and did not exhibit significant diurnal variation in starch or LMA.  Increases in LMA from morning to  200 afternoon were in excess of those found for starch, likely caused by increases in soluble sugars, although this pool was not measured. During a transition from ambient CO2 to CO2 enrichment, leaf starch increased for upper leaves within two days and continued to increase to the end of the monitoring period (seven days).  LMA did not exhibit a significant increase until the fourth day.  For lower leaves, leaf starch and LMA were not significantly affected by the onset of CO2 enrichment. The canopy and diurnal profiles of starch and LMA may have the necessary responsiveness to be used in a plant-based method of CO2 management.  Using the canopy profiles of starch per leaf and fruit load, leaves in position 7 to 9 may best reflect the carbon status of the canopy.  Leaves in this region are high enough in the canopy to accumulate starch and are adjacent to a truss which has significantly-sized fruit and thus a requirement of carbon (Chapter 4).  Leaves in this region also exhibit significant diurnal variation in leaf starch and appear to carry over starch from one day to the next.  A possible application would be to examine leaf 7 at the end of the day and the following dawn.  If starch or LMA were similar in these two measurement periods, CO2 enrichment could be delayed until some of these reserves have been used.  However, these reserves should not be zero, as starch is needed for vegetative and fruit growth in the morning before photosynthesis can meet all the needs of the plant.  Also notable is the observation that the upper leaves, especially those in position 1, carried over several days of starch reserves from one day to the next.  As the canopy profiles have shown, these reserves will eventually be used for fruit growth in the following weeks.  However, in the meantime these upper leaves may have a surplus of carbohydrate which could depress their current photosynthesis.  Under these conditions, CO2 enrichment may be an inefficient method of increasing photosynthesis.  201 Potentially, starch and LMA are very suitable as plant-based indicators for managing CO2 enrichment.  The scientific literature is rich with studies where starch and LMA have been identified as responsive to CO2 enrichment (Madsen 1968; Gent 1984; Ehret & Jolliffe 1985;Gent 1986; Tripp et al. 1991; Holbrook et al. 1993; Körner et al. 1995; Heuvelink & Marcelis 1996; Bertin et al. 1999; Roumet et al. 1999; Ayari et al. 2000).  In my work, I was able to identify a region of leaves in the canopy and suggest a time of day to monitor these leaves.  At this point, a grower is better able to measure LMA than starch.  However, technical hurdles for the measurement starch, especially in real time, could be overcome with improvements in near-infrared spectroscopy or the development of quick starch measurement kits.  Although some sample collection is necessary for these approaches damage could be minimized by using single leaflets. 6.3 ADDITIONAL IMPLICATIONS OF THIS RESEARCH Outcomes of this research will be of interest to developers of policies on CO2 emissions, greenhouse growers and researchers.  Of interest to CO2 policy makers is the amount of CO2 generated and consumed over a season of commercial tomato production. During a typical production season, a tomato greenhouse generated 125.5 kg per m2 (heating and dosing) of CO2 of which 17.4 kg per m2 was utilized by the crop.  For the entire production season, 14% of the CO2 generated by the greenhouse was utilized by the crop. This figure increases to 25% during the summer, when higher light levels increased CO2 fixation and less CO2 was generated because heating requirements were lower.  It was also apparent from my study that CO2 generation was largely out of sync with the CO2 needs of the crop.  More CO2 is produced than needed by the crop because of heating requirements in the late fall, winter and early spring.  If this excess CO2 could be captured and stored until  202 summer then the growers would not have to purchase liquid CO2 thereby reducing their carbon footprint. Greenhouse growers will also be interested in modelling yield and plant growth and further efforts to apply the Acock-Aikman models, or models similar to these, for determining yield.  Yield forecasting is becoming important for the selling of tomatoes, especially to chain stores who need a steady supply of fruit.  I also believe the growers will be interested in my method for non-destructively estimating the leaf area of a canopy, which would be useful in monitoring the balance between vegetative and generative growth of the plant. Researchers may also find the detailed starch and LMA profiles useful.  Some researchers use sub-samples of SLA (inverse of LMA) and canopy dry mass to estimate canopy leaf area.  I found LMA to exhibit significant variation through the canopy and over the course of the day.  Therefore researchers who employ that method of estimating leaf area may find the need to standardize their sampling techniques or to take multiple samples to reduce their measurement error. 6.4 CONCLUSIONS ON PLANT-BASED APPROACHES At the outset of this research it was well known that supplemental CO2 can promote growth and yield of tomato plants in commercial greenhouse cultivation.  It was also known that tomato plants under CO2 enrichment can accumulate more starch in their leaves than those grown in ambient CO2 (Körner et al. 1995; Bertin et al. 1999).  To my knowledge, plant growth, starch or LMA have not been tested as methods to help manage CO2 dosing in greenhouses.  This research has investigated those possibilities in commercial and research tomato greenhouses, leading to the following main conclusions:  203 1. Crop photosynthesis and yield models can help with decisions on CO2 use in relation to costs and benefits, but are ineffective in describing short-term patterns.  This limits their ability to be used in day-to day fine-tuning of CO2 dosing. 2. Non-destructive measures of growth were difficult to evaluate because they did not respond strongly to CO2 enrichment in this work.  The measures evaluated here seem to be too sluggish in their responses to be used to fine tune CO2 dosing. 3. Upper canopy leaf starch shows promise as a guide for managing CO2 dosing.  Based on canopy profiles of starch, leaf area and fruit load, leaves in positions 7 to 9 were identified as those upon which to focus.  The diurnal profiles revealed that upper canopy leaves carryover substantial amounts of starch from one day to the next. Monitoring starch at its peak time of accumulation (14 h to 16 h), and at sunset and sunrise will indicate how much the peak starch reserves are used and exported overnight.  If there is little change in starch between peak, or sunset, levels and sunrise the following day, then a grower could conclude that the plants are in a carbon-surplus state.  This could be used to justify a postponement in further CO2 enrichment to later in the day or the next day. 4. For the entire canopy, starch was only partly able to explain the variation in LMA. However, for upper canopy leaves starch was found to contribute positively to LMA. This relationship was evident from both the canopy and diurnal profiles.  Therefore LMA is promising as a substitute for starch for those leaves.  However, leaves in position 7 to 9 are at the lower region of the upper canopy and thus have less starch than leaves higher up.  At leaf positions 7 to 9, therefore, LMA might not be as good a predictor for starch as for the uppermost leaves.  It remains unclear if diurnal plots  204 of LMA at positions 7 to 9 are useful indicators of plant carbon status as they exhibited larger variation than did starch. 5. At this time I do not think that these approaches are ready for use by the grower.  In order for the simulation modelling to move toward implementation for CO2 dosing management the total amount of CO2 generated per hour and the percentage of that injected into the greenhouse need to be known.  These data would enable the models to work in real time.  This approach could also be improved if a plant growth model or canopy carbon status model could be used in addition to the canopy photosynthesis model.  In order to implement the starch or LMA plant-based method for managing CO2 enrichment several things need to be known.  The LMA - starch relationship should be investigated in more detail for leaves in positions 7 to 9.  Leaves in this region are at the lower end of the upper canopy and it would be useful to intensively study this region, especially with regard to diurnal dynamics.  I believe starch is the better indicator of plant carbon status, and a real time method of starch analysis needs to be developed.  Ultimately, using starch as an indicator for CO2 dosing needs to be tested against traditional methods of dosing.  The possibility exists that knowledge of plant carbon status will allow growers to fine tune their CO2 applications, maintaining high yields but using CO2 more efficiently.  Consequently, there will be lower costs for the grower and lower environmental impacts from excess CO2 emissions.  205 6.5 LITERATURE CITED Aaslyng, J.M., Andreassen, A., Korner, O., Lund, J., Jakobsen, L., Pedersen, J., Ottosen, C., and Rosenqvist, E. 2006. Integrated optimization of temperature, CO2 screen use and artificial lighting in greenhouse crops using a photosynthesis model. Acta Horticulturae 711: 79-88. Acock, B., Charles-Edwards, D.A., Fitter, D.J., Hand, D.W., Ludwig, L.J., Warren Wilson, J., and Withers, A.C. 1978. The contribution of leaves from different levels within a tomato crop to canopy net photosynthesis:  An experimental examination of two canopy models. Journal of Experimental Botany 29: 815-827. Aikman, D.P. 1996. A procedure for optimizing carbon dioxide enrichment of a glasshouse. Journal of Agricultural Engineering and Research 63: 171-184. Alscher, G., Krug, H., and Liebig, H. 2001a. Optimisation of CO2 and temperature control in greenhouse crops by means of growth models at different abstraction levels.  I. Control strategies, growth models and input data. Gartenbauwissenschaft 66: 105- 114. Alscher, G., Krug, H., and Liebig, H. 2001b. Optimisation of CO2 and temperature control in greenhouse crops by means of growth models at different abstraction levels.  II. Growth models and parameter generation for lettuce crops. Gartenbauwissenschaft 66: 153-163. Alscher, G., Krug, H., and Liebig, H. 2001c. Optimisation of CO2 and temperature control in greenhouse crops by means of growth models at different abstraction levels. IIISimulation and optimisation with the combined models. Gartenbauwissenschaft 66: 213-218. Atherton, J.G. and Harris, G.P. 1986. Flowering.  In Chapman and Hall Ltd., Cambridge, Great Britain, pp 167-200. Bertin, N., Tchamitchian, M., Baldet, P., Devaux, C., Brunel, B., and Gary, C. 1999. Contributions of carbohydrate pools to the variations in leaf mass per area within a tomato plant. New Phytologist 143: 53-61. Chalabi, Z.S., Biro, A., Bailey, B.J., Aikman, D.P., and Cockshull, K.E. 2002a. Optimal control strategies for carbon dioxide enrichment in greenhouse tomato crops - Part II. Using the exhaust gases of natural gas fired boilers. Biosystems Engineering 81: 323- 332. Chalabi, Z.S., Biro, A., Bailey, B.J., Aikman, D.P., and Cockshull, K.E. 2002b. Optimal control strategies for carbon dioxide enrichment in greenhouse tomato crops - Part 1: Using pure carbon dioxide. Biosystems Engineering 81: 421-431. Challa, H. 2002. Crop models for greenhouse production systems. Acta Horticulturae 593 : 47-53.  206 Dieleman, J.A., Meinen, E., Marcelis, L.F.M., de Zwart, H.F., and van Henten, E.J. 2005. Optimisation of  CO2 and temperature in terms of crop growth and energy use. Acta Horticulturae 691:  149-154. Ehret, D.L., Lau, A., Bittman, S., Lin, W., and Shelford, T. 2001. Automated monitoring of greenhouse crops. Agronomie 21: 403-414. Gruda, N. 2005. Impact of environmental factors on product quality of greenhouse vegetables for fresh consumption. Critical Reviews in Plant Sciences 24: 227-247. Helmer, T., Ehret, D.L., and Bittman, S. 2005. CropAssist, an automated system for direct measurement of greenhosue tomato growth and water use. Computers and Electronics in Agriculture 48: 198-215. Heuvelink, E. and Dorais, M. 2005. Crop growth and yield. In Heuvelink, E., ed. Tomatoes. CABI Publishing, Cambridge, MA, pp 85-144. Kimball, B.A. 1983. Carbon dioxide and agricultural yield:  an assemblage and analysis of 430 prior observations. Agronomy Journal 75: 779-788. Körner, Ch., Oelaez-Riedl, S., and van Bel, A.J.E. 1995. CO2 responsiveness of plants:  a possible link to phloem loading. Plant, Cell and Environment 18: 959-600. Mortensen, L.M. 1987. Review:  CO2 enrichment in greenhouse. Crop Responses. Scientia Horticulturae 33: 1-25. Nederhoff, E.M. 1994. Effects of CO2 concentration on photosynthesis, transpiration and production of greenhouse fruit vegetable crops. Wageningen. Peet, M.M. and Welles, G. 2005. Greenhouse tomato production. In Heuvelink, E., ed. Tomatoes. CABI Publishing, Cambridge, MA, pp 257-304. Picken, A.J.F., Stewart, K., and Klapwijk, D. 1986. Germination and vegetative development. In Chapman and Hall Ltd., Cambridge, Great Britain, pp 111-166. Slack, G. 1986. CO2 enrichment of tomato crops. In CRC Press, Boca Raton, Florida, pp 151-163.  207 Appendix 1.1 Leaf length verses leaf area for tomato cv Rapsodie.  Panel A is the raw response and Panel B is the mean response.  The equation of the line is Y = e(-0.53+2.45(ln(X)), R2=0.996.  The equation was weighted with the inverse of the variance and fitted to the mean response.  Those data with a variance of 0 were not plotted in Panel B.  0.00 0.02 0.04 0.06 0.08 0.10 0.12 0.14 0.00 0.10 0.20 0.30 0.40 0.50 0.60 Leaf length (m) Le af  a re a (m 2 ) Series1 A 0.00 0.02 0.04 0.06 0.08 0.10 0.12 0.14 0.00 0.10 0.20 0.30 0.40 0.50 0.60 Leaf length (m) Le af  a re a (m 2 ) Response Equation B  208 Appendix 2.1 The CO2 levels (ppm) used for low, intermediate and high CO2 enrichment scenarios at several light regimes used for modelling in Chapter 2.   Appendix 2.1A The CO2 (ppm) used for the “CO2 Intermediate” scenario at several light ranges. Light Month (W m-2) Jan Feb Mar Apr May Jun Jul Aug Sep Oct  Nov Dec >750 X X X 800 500 500 500 400 800 X X X 750-550 X 900 900 800 500 500 500 400 800 900 X X 550-350 X 900 900 800 500 600 500 400 800 900 1000 X 350-150 1000 1000 1000 800 700 700 700 600 800 1000 1000 1000 <150 1000 1000 1000 800 700 700 700 600 800 1000 1000 1000 X means no events were in these ranges.   Appendix 2.1B  The CO2 (ppm) used for the “Low CO2” scenario at several light ranges. Light Month (W m-2) Jan Feb Mar Apr May Jun Jul Aug Sep Oct  Nov Dec >750 X X X 400 400 400 400 400 500 X X X 750-550 X 600 600 400 400 400 400 400 500 600 X X 550-350 X 600 600 400 400 400 400 400 500 600 700 X 350-150 700 700 700 600 500 500 500 500 500 700 700 700 <150 700 700 700 600 500 500 500 500 500 700 700 700 X means no events were in these ranges.   Appendix 2.1C  The CO2 (ppm) used for the “High CO2” scenario at several light ranges. Light Month (W m-2) Jan Feb Mar Apr May Jun Jul Aug Sep Oct  Nov Dec >750 X X X 1000 700 700 700 600 1000 X X X 750-550 X 1100 1100 1000 700 700 700 600 1000 1100 X X 550-350 X 1100 1100 1000 700 800 700 600 1000 1100 1200 X 350-150 1200 1200 1200 1000 900 900 900 800 1000 1200 1200 1200 <150 1200 1200 1200 1000 900 900 900 800 1000 1200 1200 1200 X means no events were in these ranges.  209 Appendix 3.1 Leaf injury and CO2 enrichment In commercial greenhouses during the summer, BC tomato growers have observed injury to the leaves which they associate with the high use of CO2.  This injury is chlorosis and necrosis of the margins of the terminal leaflet for leaves located in the upper canopy (usually above the truss with open flowers) (Figure A3.1).  This injury to the young leaves has also been noted by van Berkel (1984); however, in the research literature the injury thought to be attributed to high CO2 has been more often reported for older leaves.  van Berkel (1984), Tripp et al. (1991b), Tripp et al. (1991a), Nederhoff et al. (1992) and Hao et al. (2006) have described chlorosis, necrosis, crispness, stiffness and curling or rolling of the leaf blade in older leaves.  When the stiffness and curling of the leaves becomes severe it has been called the ‘short leaf syndrome’ and can result in a significant decrease in the leaf area and eventually fruit production (Nederhoff et al. 1992).  This injury is also evident in experiments where pure CO2 sources have been used (van Berkel 1984).  Therefore it is believe not to be caused by contaminants in the CO2 or caused by noxious products (nitrogen oxides, carbon monoxide and ethylene) from the combustion of natural gas.  The injury observed by the BC growers (Figure A3.1) seems more subtle compared to what the above researchers have reported.  However, it is possible that this injury is impairing the crop beyond what is suggested from these moderate symptoms.  More information surrounding the development of this injury would be desirable, so that it could be avoided or better monitored. Plants grown at PARC-Agassiz were exposed to either CO2 enrichment or ambient CO2 and were observed for signs of leaf injury.  A complete description of the methodology  210  Figure A3.1 Examples of leaf injury observed from tomato crops under CO2 enrichment. Plate A. an uninjured leaf, B. a terminal leaflet with bubbly/patchy regions between the veins, C. twisting of the rachis of a young leaf, D. young terminal leaflet with chlorosis at leaf margin, also called leaf scorch, E. necrosis of leaf margin and interveinal chlorosis and F. rolling of leaflets from mature leaf. A F E D C B  211 and materials can be found in Chapter 3.  In the upper canopy, leaves were observe for the presence of:  chlorosis or necrosis on the margins of the leaflets (Figure A3.1D and E), curling or twisting of the rachis (Figure A3.1C) and leaflets exhibiting a “bubbly” or “patchy” look (Figure A3.1B).  The lower leaves were monitored for lesions on the blade, chlorosis and a stiff and curled appearance (Figure A3.1F). In this study none of the typical injuries associated with CO2 dosing were observed in any region of the canopy.  In this timeframe (June to August), commercial growers observed occasional chlorosis and necrosis of the margins of the terminal leaflet of leaves located in the upper canopy (above the truss with open flowers) but not in the lower canopy (Figure A3.1).  It is unclear to me why the plants in my work did not exhibit any symptoms as CO2 was dosed at a very high rate.  Carbon dioxide was applied from 6 h to 21 h to a set point of 1000 ppm.  However, because of the need to ventilate to cool the greenhouses, the mean daily measured CO2 concentration was lower, 846 and 804 ppm, respectively in 2002 and 2003 (Table 3.1, Figures 3.4 and 3.5).  I think this should be sufficient CO2 to induce injury as commercial greenhouse growers enrich to a lower level, 500 to 700 ppm and frequently note leaf damage in the upper canopy.  As well, van Berkel (1984), reported a five-fold increase in the number of leaflets with chlorotic margins when CO2 was increased from 430 to 790 ppm and Nederhoff et al. (1992) reported symptoms at 550 ppm of CO2 dosed from sunrise to sunset.  Both Moe (1984) and Mortensen (1987) have noted that leaf damage under high CO2 seems to be variable and not always observed.  As well, van Berkel (1984) and Mortensen (1987) noted that more injury seems to occur when CO2 enrichment is used and global radiation is high.  In my study the mean daily radiation measured outside of the greenhouse structure was 20.3 and 18.3 MJ m-2 per day for 2002 and 2003, respectively  212 (Table 3.1, Figures 3.4 and 3.5).  These light levels are similar to those studies where leaf injury occurred (16.9 to 22.7 MJ m-2 per day (Nederhoff et al. 1992) and 19.3 to 23.3 MJ m-2 per day (Hao et al. 2006)) but are slightly lower than the seasonal daily mean of 22.4 MJ m-2 recorded by a commercial greenhouse.  It is possible that for the 2002 experiment no symptoms were observed because the greenhouse was whitewashed which reduced the light reaching the crop.  I estimate that light penetration was reduced 60% by the white wash.  In 2003, no white wash was applied therefore I am unsure why no signs of injury were present. The cultivar Rapsodie is not immune to these symptoms since I have observed them in commercial greenhouse and Hao et al. (2006) also observed symptoms in his experiments at Agriculture and Agrifood Canada in Harrow, Ontario. It is conceivable that even without white wash the light reaching the crop was too low.  The research greenhouses at Agassiz being small, have more infrastructure to support the glass and as a result let in less light then the larger commercial greenhouses.  Less light would lower the source activity of the plants and management practises that lower source activity (such as increased plant density) are believed to prevent CO2 injury and prevent the formation of the ‘short leaf syndrome’ (Heuvelink 2005).  It is noteworthy to mention that in an unreported experiment at PARC-Agassiz, leaf symptoms similar to those described above were observed when plants were pruned to one fruit per truss (Edwards, unpublished data). Therefore, a more accurate description might be the injury is the result of the assimilate production not balanced with assimilate consumption. Low source activity or high sink activity prevents the large accumulation of assimilates and a rapid growth rate of the leaf apex.  According to a theory by Nederhoff (1992), the ‘short leaf syndrome’ occurs when insufficient calcium reaches the apex.  213 Therefore, the lower light would reduce the source activity, preventing a large accumulation of assimilates and result in slower growth rate of the apex and which would allow the xylem to keep up with the calcium requirements of the apex.  I do not have the data here to prove or disprove Nederhoff’s (1992) theory, but if high assimilates (e.g. starch) are present in these asymptomatic plants then her theory may be questionable. Literature Cited Hao, X., Wang, Q., and Khosla, S. 2006. Responses of a greenhouse tomato crop to summer CO2 enrichment. Canadian Journal of Plant Science 86: 1395-1400. Heuvelink, E. 2005. Development processes. In Heuvelink, E., ed. Tomatoes. CABI Publishing, Cambridge, MA, pp 53-83. Moe, R. 1984. CO2 enrichment in Scandinavia . Acta Horticulturae 162: 217-225. Mortensen, L.M. 1987. Review:  CO2 enrichment in greenhouse. Crop Responses. Scientia Horticulturae 33: 1-25. Nederhoff, E.M., De Koning, A.N.M., and Rijsdijk, A.A.  1992. Leaf deformation and fruit production of glasshouse grown tomato (Lycopersicon esculentum Mill.) as affected by CO2, density and pruning. Journal of Horticultural Science 67: 411-420. Tripp, K.E., Peet, M.M., Pharr, D.M., Willits, D.H., and Nelson, P.V. 1991a. CO2-enhanced yield and foliar deformation among tomato genotypes in elevated CO2 environments. Plant Physiology 96: 713-719. Tripp, K.E., Peet, M.M., Willits, D.H., and Pharr, D.M. 1991b. CO2-enhanced foliar deformation of tomato:  relationship to foliar starch concentration. Journal of Amercian Society for Horticultural Science 116: 876-880. van Berkel, N. 1984. Injurious effects of high CO2 concentrations on cucumber, tomato, chrysanthemum and gerbera. Acta Horticulturae 162: 101-112.    214  Appendix 4.1 A summary of the environment from experiments conducted in commercial greenhouses in 2001 and 2002.  Appendix 4.1A The environment data for the summer months measured at South Alder commercial greenhouse in 2001. 3CO2 (ppm) Temperature4 (C) Relative Humidity4 (%) Global Radiation3  (MJ m-2 day-1) Month Mean Min1 Max2 Mean Min Max Mean Min Max Mean Min Max March 1430 362 3438 18.1 14.3 26.1 80.4 57.0 92.0 9.86 1.10 18.56 April 1339 350 3440 18.5 13.9 26.8 91.0 53.0 100 15.29 2.39 24.18 May 1103 478 2528 19.2 14.6 29.6 70.2 41.0 91.0 20.78 4.91 30.09 June 770 363 2222 19.3 13.8 28.8 76.2 54.0 93.0 20.58 5.30 30.23 July 721 334 3106 20.4 13.6 28.8 73.9 50.0 100 22.78 5.77 30.24 August 552 345 2023 20.7 14.8 31.5 69.4 31.60 88.0 17.49 2.72 26.42 September 699 395 2409 19.4 14.4 27.9 71.2 43.0 90.0 14.65 2.95 20.95   Appendix 4.1B The environment data for the summer months measured at CanAgro commercial greenhouse in 2001. 3CO2 (ppm) Temperature4 ( C) Humidity Deficit4 (g m-3) Global Radiation3  (MJ m-2 day-1) Month Mean Min1 Max2 Mean Min Max Mean Min Max Mean Min Max March 1117 263 2150 18.7 14.4 27.3 3.09 0 7.70 10.07 0.87 18.98 April 974 293 1954 19.5 14.3 29.0 3.79 1.10 11.20 15.72 2.51 24.99 May 659 268 1648 19.3 14.6 27.9 4.80 0 12.70 21.28 4.46 29.89 June 495 208 1354 19.8 14.3 29.3 5.05 0 13.10 21.63 8.07 29.60 July 404 211 1202 19.6 14.2 29.6 5.84 1.10 16.60 23.57 4.26 30.31 August 426 184 1304 20.1 14.3 31.4 5.51 1.10 17.70 18.03 1.96 27.09 September 486 226 1262 19.2 14.1 26.6 6.15 1.70 12.00 18.24 9.46 20.79 1. Minimum. 2.  Maximum. 3.  CO2 and global radiation were based on daytime measurements and global radiation was measured outside the greenhouse. 4. Temperature and relative humidity were based on data recorded for 24 hours.  215 Appendix 4.1C The environment data for the summer months measured at South Alder commercial greenhouse in 2002. 3CO2 (ppm) Temperature4 ( C) Relative Humidity4 (%) Global Radiation3  (MJ m-2 day-1) Month Mean Min1 Max2 Mean Min Max Mean Min Max Mean Min Max March 937 337 1502 19.0 15.2 25.8 78.1 63.0 91.0 9.18 3.53 14.19 April 884 365 1924 18.5 12.5 27.8 74.7 35.0 100.0 15.0 2.92 25.89 May 734 323 3440 18.5 13.0 28.0 74.7 44.0 92.0 19.45 4.30 29.65 June 639 398 1395 19.7 13.1 31.2 74.1 43.0 92.0 23.41 7.46 30.62 July 634 418 1486 20.3 12.7 32.1 75.3 35.0 94.0 22.65 5.08 29.85 August 662 416 1406 19.9 11.3 29.8 84.5 53.0 100.0 20.78 9.35 25.80 September 727 451 1401 18.8 13.0 27.2 80.8 49.0 100.0 14.90 5.72 21.41   Appendix 4.1D The environment data for the summer months measured at Gipaanda commercial greenhouse in 2002. 3CO2 (ppm) Temperature4 ( C) Relative humidity4 (%) Global Radiation3  (MJ m-2 day-1) Month Mean Min1 Max2 Mean Min Max Mean Min Max Mean Min Max March 1271 906 1626 18.7 16.8 19.5 78.1 72.0 83.0 9.53 3.14 17.82 April 1003 730 1221 18.6 17.2 20.1 82.8 73.0 100.0 14.95 3.46 25.82 May 937 693 1261 19.5 18.2 21.0 83.5 76.0 91.0 19.47 3.83 28.99 June 696 641 779 19.5 17.8 22.9 79.9 73.0 83.0 23.92 6.30 30.68 July 667 584 763 20.1 17.7 23.6 78.4 72.0 88.0 23.34 6.47 30.13 August 657 590 710 19.8 17.2 23.5 65.4 19.3 82.0 21.06 11.84 25.78 September 685 593 824 18.5 17.2 19.8 77.1 19.4 88.0 15.06 4.31 20.94 1. Minimum. 2. Maximum. 3. CO2 and global radiation were based on daytime measurements and global radiation was measured outside the greenhouse 4.  Temperature and relative humidity were based on data recorded for 24 hours.  .  216 Appendix 4.2 Dates of data collection in commercial greenhouses 2000 to 2002.  Year Greenhouse 2000  2001 2002 South Alder May 26 March 6 May 16  June 29 1May 4 June 24  August 11 June 5 July 21  October 6 September 10 August 21    September 27 CanAgro May 31 March 7 -  June 26 1May 7 -  August 17 September 6 -  October 5 - - Gipaanda - - May 20  - - June 21  - - July 19  - - August 20  - - September 25 1. only leaf length data was collected.   217 Appendix 4.3 The starch content and corresponding leaf mass per area (LMA) of tomato leaf tissue from South Alder (SA) and Gipaanda (GI) commercial greenhouses.  Panels A-C data collected in May 2002, D-F data collected in September 2002.  Data from all leaf positions are in panels A and D, panel B has data from leaf positions <=11, C has data from leaf positions >=12, E has data from leaf positions <=8 and F has data from leaf positions >=9.  Each sample was 0.000785 m2 of leaf tissue. 15 25 35 45 55 65 75 85 0 10 20 30 40 Leaf starch (g m-2) Le af  m as s a re a (g  m -2 ) Series1 Series2 15 25 35 45 55 65 75 85 0 10 20 30 40 Leaf starch (g m-2) y Series1 Series2 15 25 35 45 55 65 75 85 0 10 20 30 40 STMA LM A  (g  m -2 ) Gnh 1 Gnh 2 15 25 35 45 55 65 75 85 0 10 20 30 40 x LM A  (g  m -2 ) Series1 Series2 15 25 35 45 55 65 75 85 0 10 20 30 40 Leaf starch (g m-2) LM A  (g  m -2 ) SA GI 15 25 35 45 55 65 75 85 0 10 20 30 40 Leaf starch (g m-2) SA GI A B C D E F  218  Appendix 4.3 The starch content and corresponding leaf mass per area of tomato leaf tissue from South Alder (SA) and Gipaanda (GI) commercial greenhouses. Panels G-I data collected in June 2001, J-L data was collected in June 2002. Data from all leaf positions are in panels G and J, H has data from leaf positions <=11, I has data from leaf positions >=12, panel K has data from leaf positions <=11, panel L has data from leaf positions >=12.  Each sample was 0.000785 m2 of leaf tissue. 15 25 35 45 55 65 75 85 0 10 20 30 40 x y Series1 Series2 15 25 35 45 55 65 75 85 0 10 20 30 40 x y Series1 Series2 15 25 35 45 55 65 75 85 0 10 20 30 40 Leaf starch (g m-2) y SA GI 15 25 35 45 55 65 75 85 0 10 20 30 40 x LM A  (g  m -2 ) Gnh 1 15 25 35 45 55 65 75 85 0 10 20 30 40 x LM A  (g  m -2 ) Gnh 1 15 25 35 45 55 65 75 85 0 10 20 30 40 Leaf starch (g m-2) LM A  (g  m -2 ) SA G H I J K L  219  Appendix 4.3 The starch content and corresponding leaf mass per area from tomato leaf tissue from South Alder (SA) and Gipaanda (GI) commercial greenhouses. Panels M-O data collected in July 2002, P-R data was collected in August 2002. Data from all leaf positions are in panels M and P, Panel N has data from leaf positions <=9, O has data from leaf positions >=10, Q has data from leaf positions <=11 for SA and <=10 for GI and R has data from leaf positions >=12 for  SA and >=11 for GI.  Each sample was 0.000785 m2 of leaf tissue. 15 25 35 45 55 65 75 85 0 10 20 30 40 x y Series1 Series2 15 25 35 45 55 65 75 85 0 10 20 30 40 x y Series1 Series2 15 25 35 45 55 65 75 85 0 10 20 30 40 Leaf starch (g m-2) y SA GI 15 25 35 45 55 65 75 85 0 10 20 30 40 x LM A  (g  m -2 ) Series1 Series2 15 25 35 45 55 65 75 85 0 10 20 30 40 x LM A  (g  m -2 ) Series1 Series2 15 25 35 45 55 65 75 85 0 10 20 30 40 Leaf starch (g m-2) LM A  (g  m -2 ) SA GI M N O P Q R  220 Appendix 4.4 Leaf mass per area of leaves in tomato plant canopies from South Alder (SA) and Gipaanda (GI) commercial greenhouses.  Leaf position 1 is the leaf closest to the first flowering truss, - leaves are younger and + leaves are older. Panel A and B, data collected in June 2001, Panel C and D, data collected June 2002, line equations are in Appendix 4.5 15.0 20.0 25.0 30.0 35.0 40.0 45.0 50.0 55.0 60.0 -3 0 3 6 9 12 15 18 21 24 Leaf position LM A  (g  m -2 ) Gnh 1 0.0 2.0 4.0 6.0 8.0 10.0 12.0 14.0 16.0 18.0 -3 0 3 6 9 12 15 18 21 24 Leaf position Le af  st ar ch  (g  m -2 ) Gnh 1 15.0 20.0 25.0 30.0 35.0 40.0 45.0 50.0 55.0 60.0 -3 0 3 6 9 12 15 18 21 24 Leaf position LM A  (g  m -2 ) SA GI 0.0 2.0 4.0 6.0 8.0 10.0 12.0 14.0 16.0 18.0 -3 0 3 6 9 12 15 18 21 24 Leaf position Le af  st ar ch  (g  m -2 ) SA GI A B C D  221  Appendix 4.4 Starch from leaves in tomato plant canopies from South Alder (SA) and Gipaanda (GI) commercial greenhouses.  Leaf position 1 is the leaf closest to the first flowering truss, - leaves are younger and + leaves are older.  Panel E and F, data collected in July 2002, Panel G and H, data collected in August 2002, line equations are in Appendix 4.5. 15.0 20.0 25.0 30.0 35.0 40.0 45.0 50.0 55.0 60.0 -3 0 3 6 9 12 15 18 21 24 Leaf position LM A  (g  m -2 )  SA GI 0.0 2.0 4.0 6.0 8.0 10.0 12.0 14.0 16.0 18.0 -3 0 3 6 9 12 15 18 21 24 Leaf position Le af  st ar ch  (g  m -2 ) SA GI 15.0 20.0 25.0 30.0 35.0 40.0 45.0 50.0 55.0 60.0 -3 0 3 6 9 12 15 18 21 24 Leaf position LM A  (g  m -2 ) SA GI 0.0 2.0 4.0 6.0 8.0 10.0 12.0 14.0 16.0 18.0 -3 0 3 6 9 12 15 18 21 24 Leaf position Le af  st ar ch  (g  m -2 ) SA GI E F G H  222  Appendix 4.5 Linear regression equations for leaf mass per area or leaf starch and canopy leaf position for Appendix 4.4. Gnh1 Leaf mass per area (g m-2)2 Leaf starch (g m-2)2 Year Month  Equation3 4R2 Equation3 4R2 2001 June SA Y = 55.4 - 1.60.3X 0.92  Y = 14.0 - 2.0X + 0.08X2 0.93 2002 June SA mean of 30.3   Y = 13.4 - 2.2X + 0.1X2 - 0.002X3 0.73 2002 June GI Y = 52.1 - 0.9X 0.67  Y = 12.8 - 1.5X + 0.05X2 0.92 2002 July SA Y = 47.7 - 1.0X 0.52  Y = 7.4 - 0.8X + 0.02X2 0.63 2002 July GI Y = 49.2 - 0.7 0.23  Y = 13.2 - 2.3X + 0.2X2 - 0.003X3 0.96 2002 August SA Y = 51.1 - 0.8X 0.64  Y = 20.0 - 3.8X + 0.2X2 - 0.005X3 0.96 2002 August GI Y = 44.4 + 1.8X - 0.4X2 + 0.02X3 0.76  Y = 15.5 - 1.8X + 0.06X2 0.84 1. Gnh is greenhouse, SA is South Alder, CA is CanAgro and GI is Gipaanda. 2.  All regressions were weighted by the inverse of the variance. 3. Y is leaf mass per area or leaf starch and X is leaf position. 4. R2 is the coefficient of determination.  223  Appendix 4.6 Leaf area profile of tomato plant canopies from South Alder (SA), CanAgro (CA) and Gipaanda (GI) commercial greenhouses.  Leaf position 1 is the leaf closest to the first flowering truss, - leaves are younger and + leaves are older.  Panel A, data collected in March 2001, equations for lines are:  SA:  Y = 0.032e[(0.43(1-e(-0.56X))/0.56],  CA: Y = 0.033e[(0.36(1-e(-0.34X))/0.34], Panel B, data collected in September 2001, equations for lines are:  SA:  Y = 0.0094e[(0.93(1-e(-0.70x))/0.70], CA:  Y = 0.021e[(0.43(1-e(-0.73x))/0.73].  Panel C, data collected in May 2002, equations for lines SA: Y = 0.026e[(0.32(1-e(-0.34x))/0.34]  and GI:  Y = 0.019e[(0.42(1-e(-0.33x))/0.33], Panel D, data collected September 2002, equations for lines are SA:.Y = 0.0045e[(2.0(1-e(-0.9x))/0.9] and GI: Y = 0.007e[(1.82(1-e(-0.85x))/0.85]. All regressions are weighted by the inverse of the variance. 0.00 0.02 0.04 0.06 0.08 0.10 0.12 -9 -6 -3 0 3 6 9 12 15 18 21 24 Leaf position Le af  a re a (m 2) SA CA 0 0.02 0.04 0.06 0.08 0.1 0.12 -9 -6 -3 0 3 6 9 12 15 18 21 24 Leaf position Le af  a re a (m 2) SA GI 0.00 0.02 0.04 0.06 0.08 0.10 0.12 -9 -6 -3 0 3 6 9 12 15 18 21 24 Leaf posit ion Le af  a re a (m 2 ) SA CA 0 0.02 0.04 0.06 0.08 0.1 0.12 -9 -6 -3 0 3 6 9 12 15 18 21 24 Leaf posit ion Le af  a re a (m 2 ) SA GI A B C D  224  Appendix 4.6 Leaf area profile of tomato plant canopies from South Alder (SA), CanAgro (CA) and Gipaanda (GI) commercial greenhouses.  Leaf position 1 is the leaf closest to the first flowering truss, - leaves are younger and + leaves are older. Y = be[(c(1-e(-dx))/d].  Panel E, data collected in May and June 2001, line equations are: SA. Y = 0.01e[(0.50(1-e(-0.273X))/0.273],  CA. Y = 0.017e[(0.42(1-e(-0.31X))/0.31] and SA Jn. Y = 0.0095e[(0.39(1-e(-0.22X))/0.22] Panel F, data collected in June 2002, line equations are:  SA. Y = 0.012e[(0.70(1-e(-0.54x))/0.54], CA. Y = 0.012e[(0.60(1-e(-0.42x))/0.42]. Panel G, data collected in July 2002, line equations are SA.Y = 0.007e[(0.84(1-e(- 0.45x))/0.45], GI. Y = 0.011e[(0.63(1-e(-0.17x))/0.17], Panel H, data collected August 2002, line equations are SA.Y = 0.0064e[(0.92(1-e(-0.56x))/0.56], GI.Y = 0.0043e[(1.27(1-e(- 0.54x))/0.54]. All regressions are weighted by the inverse of the variance. 0.00 0.02 0.04 0.06 0.08 0.10 0.12 -9 -6 -3 0 3 6 9 12 15 18 21 24 Le af  a re a (m 2) SA CA 0.00 0.02 0.04 0.06 0.08 0.10 0.12 -9 -6 -3 0 3 6 9 12 15 18 21 24 Leaf position Le af  a re a (m 2) SA GI 0.00 0.02 0.04 0.06 0.08 0.10 0.12 -9 -6 -3 0 3 6 9 12 15 18 21 24 Le af  a re a (m 2 ) SA CA SA Jn 0.00 0.02 0.04 0.06 0.08 0.10 0.12 -9 -6 -3 0 3 6 9 12 15 18 21 24 Leaf position Le af  a re a (m 2 ) SA GI E F G H  225  Appendix 4.7 Starch content per leaf for tomato plant canopies from South Alder (SA) and CanAgro (CA) commercial greenhouses..  Leaf position 1 is the leaf closest to the first flowering truss, - leaves are younger and + leaves are older.  Panel A, data collected in March 2001, line equations are:  SA. Y = 0.34 + 0.11X - 0.19X2 + 0.0071X3, R2=0.80, CA. Y = 0.53 + 0.093X - 0.013X2 + 0.0004X3, R2=0.80. Panel B, Data collect in June 2001, line equation SA. Y = 0.19 + 0.022X - 0.0058X2 + 0.00025X3, R2=076.  Panel C, data collected in September 2001, line equations are:  SA. Y = 0.084 + 0.014X - 0.0029X2 - 0.00012X3, R2=0.38, CA. Y = 0.14 - 0.019X -0.00075X2, R2=0.51. Note different scales for Y-axis. 0.00 0.20 0.40 0.60 0.80 1.00 1.20 1.40 -3 0 3 6 9 12 15 18 21 24 Leaf position St ar ch  p er  le af  (g ) SA 0.00 0.20 0.40 0.60 0.80 1.00 1.20 1.40 -3 0 3 6 9 12 15 18 21 24 Leaf posit ion St ar ch  p er  le af  (g ) Gnh 1 Gnh 2 0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 -3 0 3 6 9 12 15 18 21 24 Leaf position St ar ch  p er  le af  (g ) SA CA A B C  226 Appendix 5.1 The effect of CO2 enrichment on photosynthetic rates for tomato cv Rapsodie canopies at PARC-Agassiz in June 2002 and 2003. Year1 Treatment Leaf Photosynthesis CO2 PAR3  CO2 Position2 umol m-2 s-1 ppm umol m-2 s-1 2002 Ambient 1 10.87 A 484 B 387.5 A  Enriched 1 10.95 A 717 A 390.7 A  Ambient 7 to 9 3.92 C 482 B 225.4 B  Enriched 7 to 9 5.70 B 750 A 179.1 BC  Ambient 13 to 16 2.43 D 427 B 148.6 C  Enriched 13 to 16 2.30 D 698 A 113 C  p- value4 - 0.0001 0.0001 0.0001 2003 Ambient 1 12.55 B 339 C 980.2 A  Enriched 1 16.27 A 596 B 834.4AB  Ambient 7 to 9 9.41C 343 C 686.8 B  Enriched 7 to 9 10.29 C 604 B 489.9 C  Ambient 13 to 16 4.27D 360 C 421.9 CD  Enriched 13 to 16 5.05 D 672 A 289.4 D  p- value3 - 0.0001 0.0001 0.0001 1. In 2002 data were collected on June 10, 11, 12, 14 and 17 between 13 h and 15 h and in 2003 June 4, 6, 9, 17 and 25 between 11 h and 15 h. 2. Leaf position, leaf 1 is the leaf closest to the truss with flowers. 3. PAR is photosynthetically active radiation. 4.  p-value from the ANOVA, pair wise were made using the pdiff option of SAS, identical letters. following the treatment mean indicate no statistical significant differences at the α=0.05 level. In 2002 was N=534 and in 2003 was N=270.  227 0 200 400 600 800 1000 1200 1400 1600 1800 0 10 20 30 40 50 60 70 80 Days after onset of carbon dixoide enrichment Fr ui t l oa d (c m3 ) Ambient Enriched  Appendix 5.2 The weekly profile of mean fruit load for tomato plants grown at PARC- Agassiz under CO2 enrichment and ambient CO2 from June 11 to August 30, 2002.  Plants had 7 trusses with 3 fruit.  Each symbol is the mean of 2.  228  Appendix 5.3 Canopy profiles of starch content and leaf mass per unit area (LMA) of leaves from tomato plant canopies treated by enriched or ambient CO2 at PARC-Agassiz.  Data was collected July 4, 2003, 58 days after CO2 enrichment commenced.  Leaf position 1 is the leaf closest to the first flowering truss, - leaves are younger and + leaves are older.  Panel A. leaf starch, line equations are Ambient: Y = 5.57 - 1.67X + 0.23X2 - 0.013X3 + 0.00024X4, R2=0.65 and Enriched Y = 6.69 - 1.21X + 0.083X2 - 0.0019X3, R2=0.62.  Panel B. LMA, line equations are Ambient:  Y = 40.22 - 0.48X, R2=0.40 and Enriched:  Y = 46.25 - 0.44X, R2=0.51. All regressions are weighted by the inverse of the variance. 0 1 2 3 4 5 6 7 0 3 6 9 12 15 18 21 24 27 x Le af  st ar ch  (g  m -2 ) Ambient Enriched 25 30 35 40 45 50 0 3 6 9 12 15 18 21 24 27 Leaf position LM A  (g  m -2 ) Ambient Enriched A B  229 Appendix 6.1 Quantification of leaf starch using enzymatic and near-infrared spectroscopic methods The aim of this work is to assess the accuracy and precision of two methods of quantifying starch in tomato leaf tissue.  These methods are enzyme degradation and near infrared spectroscopy. A6.1 INTRODUCTION According to Burrell (2003), worldwide annual starch production from photosynthesis is 2850 million tonnes.  The importance of starch as a storage carbohydrate is well known; starch in roots, tubers and seed endosperms provides 35 to 80% of a person’s daily caloric intake depending on where they live in the world (Burrell 2003).  As well as a food stuff, starch is also an important raw material for many industrial products such as: biodegradable plastics, seed coatings for fertilizers, dusting powders for tablets and additives for paper, yarns, concrete, gypsum board and others (Burrell 2003). The importance of starch to the plant is reflected in that half of the assimilated carbon by plant photosynthesis is stored as starch (Zeeman et al. 2002).  In higher plants starch is stored in plastids with its function depending on the type of plastid and tissue (Tetlow et al. 2004).  Thus starch storage can be long-term (weeks to years) and short-term (hours to days). In leaves starch storage is short term; starch is synthesized and degraded on a diurnal basis within the chloroplast.  Its role is as a carbon buffer, supplying sucrose when photosynthesis is low or not occurring.  The exact mechanisms involved in leaf starch turnover (synthesis and degradation are now believed to be more complex than initially believed and Tetlow et al. (2004), Smith et al. (2003) and Zeeman et al. (2004) have discussed them.  230 A6.1.1 The composition and structure of starch A starch granule is a complex, water insoluble polymer of glucose residues that are in two arrangements, amylose and amylopectin.  Amylose constitutes 20 to 30% of a starch molecule; it is a mostly straight chain of glucose units attached together, end-to-end, by α1-4 glycosidic linkages (Figure A6.1A).  Amylopectin constitutes 70 to 80% of a starch molecule, and consists of amylose chains with one in twenty or thirty glucose units having an α1-6 glycosidic linkage, which forms a branched pattern (Figure A6.1B).  The distribution of these branches allows adjacent chains of amylose to line up together and twist into double helices (Figure A6.1B).  When packed together the helices form a lattice that gives part of the starch granule a crystalline structure (Smith et al. 1997).  The overall semi-crystalline structure of the granule is caused by interspersion of amorphous regions, containing amylose and amylopectin in a less ordered state (Jobling 2004) (Figure A6.1C).  A cross section of a starch granule viewed by light microscopy has what appears to be “growth rings” (Figure A6.1C).  These growth rings are concentric regions of alternating amorphous and crystalline structure.  The granule grows outward, by adding on branch points and straight chain.  How the plant regulates the arrangement of branching points is not yet fully understood (Tetlow et al. 2004; Smith et al. 2003; Zeeman et al. 2004). A6.1.2 The enzymatic measurement of starch The enzymatic measurement of starch is by the hydrolysis of the α1-4 and α1-6 linkages and measurement of the librated glucose.  In practise the analysis of starch in plant tissue is complicated because starch is embedded in a heterogeneous matrix of saccharides and the granule has a tight, hydrophobic structure making hydrolysis of these bonds difficult (Batey 1982).  Rose et al. (1991) noted that the quantification of leaf starch can be problematic because most plant physiologists are not experienced in analytical chemistry and  231 O CH OH OH OH O 2 O CH OH OH OH O 2 O CH OH OH OH OH O 2CH OH OH OH O 2 O α1-6 bond α1-4 bond CH OH OH OH O 2 HO CH OH OH OH O 2 OHO Crystalline lamella Amorphous lamella Semi -Crystalline Structure Amorphous background Semi -Crystalline Structure Growth Ring A. B. C.   Figure A6.1 The components and granule structure of a starch molecule.  A. is the two types of glycosidic linkages between glucose molecules.  B. is the arrangement of amylose and amylopectin molecules to give the semi-crystalline structure.  C. is a slice through a granule, where the zones of amorphous and semi-crystalline structure are present.  Redrawn from Smith et al. (1997) and Jobling (2004).  232 there are a myriad of techniques in the literature employed to analyse starch.  He also notes that these techniques seldom provide the complete recipe-type instructions needed to use them (Rose et al. 1991). Method overview.  The key steps for starch analysis are:  tissue preparation, the extraction of soluble sugars, gelatinization, hydrolysis and measurement of glucose.  After collection, tissues should be frozen quickly in liquid nitrogen, freeze-dried to prevent any enzymatic conversion of starch to sugar and finely ground to a particle size of 500 um. Soluble sugars must be removed from the tissue prior to starch analysis as they too can be hydrolyzed to glucose, causing an overestimation of starch.  The two most common methods employed for removing soluble sugars are heating in 80% ethanol or in a mixture of methanol:chloroform:water (Rose et al. 1991; Anonymous 2000; Anonymous 2005; Haissig & Dickson 1979; Hendrix 1993).  After washing and then centrifugation, the soluble sugars will be in the supernatant and the starch in the pellet.  The supernatant can be discarded or kept for the analysis of soluble sugars.  Gelatinization causes the starch granules to hydrate and swell, thereby increasing the accessibility of the starch polymer to the hydrolyzing agents (Rose et al. 1991).  To gelatinize, the pellet is suspended in water and autoclaved or boiled (Batey 1982; Anonymous 2005).  Hydrolysis of the glycosidic linkages can be achieved by incubation with either acid or hydrolytic enzymes such as amyloglucosidase or α-amylase.  The liberated glucose is usually assayed colourimetrically after reaction with glucose oxidase/peroxidase and a chromatogen such as o-dianisidine (Haissig & Dickson 1979; Rose et al. 1991; Anonymous 2005) or 4-aminoantipurine (Anonymous 2000; Karkalas 1985; Kunst et al. 1983).  The colour change in the solution can be determined spectrophotometerically and is directly proportional to the amount of glucose.  The amount  233 of starch in a sample can be calculated by the glucose reading multiplied by 0.9 (conversion of glucose to starch mass) (Denison et al. 1990) or by including pure starch standards with the analysis. A6.1.3 Hydrolytic enzymes for starch hydrolyses The consensus of the literature is that enzymatic methods are preferable to acid hydrolysis for complex plant matrices.  Appropriate enzymes exhibit very good specificity to the target (α1-4 and α1-6) linkages, with little hydrolysis of other polysaccharides such as cellulose and hemi-cellulose, and use modest amounts of hazardous chemical (Mitchell 1990; Beutler 1983; Haissig & Dickson 1979; Rose et al. 1991; Karkalas 1985; Hendrix 1993). The most popular enzyme used is amyloglucosidase, an enzyme derived from the fungus Aspergillus niger that has a specificity for hydrolyzing only starch (Beutler 1983).  The excellence of amyloglucosidase for starch hydrolysis is not unanimously agreed upon. Passos et al. (1999), Denison et al. (1990), Beutler (1983) and Haissig & Dickson (1979) express concern that amyloglucosidase can hydrolyse the β1-6 linkages of cellulose and the enzyme purity may be low.  In any event it seems prudent to test the enzyme activity on pure starch and cellulose before employing it. A6.1.4 Near-infrared analysis of leaf starch Near-infrared analysis has been described as one of the most “practical and exciting techniques to hit the agricultural and food industries since the Kjeldahl test for nitrogen was implemented” (Deaville & Flinn 2000).  Near-infrared analysis offers several advantages to the analyst:  it greatly increases sample throughput, requires a minimum of sample preparation, does not require reagents (Deaville & Flinn 2000; Foley et al. 1998) and has the potential to be used in situ and thus in remote sensing (Matson et al. 1994; Foley et al. 1998). The main drawback associated with this technique are the costs of the instrumentation and its  234 heavy reliance on chemometrics, which has led some to conclude that it is a black box (Deaville & Flinn 2000). Near infrared spectroscopy (NIRS) is considered a secondary method of measuring the chemical constituents of tissues.  Like other types of spectroscopy, NIRS depends on the Beers-Lambert law, where energy absorption is proportional to the concentration of the absorbing material.  In the case of NIRS this energy is in the near infrared region (750 to 2500 nm) of the electromagnetic spectrum, an area of the spectrum where organic compounds have excellent transmittance and reflectance properties.  Absorptions in the near infrared region are caused by the bending and stretching of C-H, N-H and O-H bonds of organic compounds in plant or animal tissues (Shenk & Westerhaus 1995).  The nature and number of these bonds reflects the chemical composition of the tissue and determine the wavelengths and amount of energy absorbed/reflected.  The spectrum of infrared energy that is reflected from the samples therefore contains detail on the chemical composition of that material (Foley et al. 1998).  An absorbance spectrum is developed by measuring the reflected light from the substance for each wavelength in the near infrared region. Absorbance is related to reflectance by: Equation A6.1:  Absorbance=log10[1/Reflectance] A spectrum from a sample has an appearance of a smoothly rolling line with a few shoulders (Deaville & Flinn 2000).  This is caused by overlapping bands because the reflectance of the sample is the summation of all its chemical components such as starch, sugars, cellulose, water, protein etc. (Deaville & Flinn 2000).  The shape of the spectral line or the rate of change of the slope with wavelength conveys the chemical information (Deaville & Flinn  235 2000).  At this point NIRS relies on chemometrics to relate the spectrum to the reference chemistry and this is the reason NIRS is considered a secondary method of measurement. The analyst has to develop a statistical model to test the intensity of the relationship between a particular absorbance and the reference method (Foley et al. 1998).  This relationship is empirical and has to be repeatedly tested as it can change with species, cultivar or growing condition.  Chemometrics for NIRS consist of mathematical pre- treatment of the spectral data (if necessary), calibration of the spectra with the reference method, development of a predictive equation and validation of this equation.  In this work the reference method for NIRS is the amyloglucosidase-glucose oxidase-peroxidase (AGOP) method of starch analysis.  This method will be described fully in section A6.2.1.  The chemometrics used for NIRS will be elaborated on in section A6.2.6. A6.1.5 Objectives The objective of this chapter is to investigate the suitability, accuracy and precision of two laboratory methods for measuring starch in tomato leaf tissue.  These methods are the direct measurement of starch by enzyme hydrolysis and the indirect measurement by near- infrared spectroscopy. A6.2 MATERIALS AND METHODS The analysis of leaf starch was carried out on 5748 samples using either wet chemical techniques or near-infrared reflectance spectroscopy (NIRS) or both techniques.  These samples were from the following data sets:  Industry 2000, Industry 2001, Industry 2002, Agassiz 2002 and Agassiz 2003 and Industry 2005.  Industry data sets were collected from plants in commercial greenhouses located in and around Delta, BC while Agassiz data sets were collected from plants in the research greenhouses at the Pacific Agri-Food Research  236 Centre of Agriculture and Agri-Food Canada in Agassiz, BC (PARC-Agassiz).  See Chapters 2 to 5 for the greenhouse specifications, CO2 practises and plant growing techniques.   Wet chemical techniques for analyzing starch were performed on all or some samples in all the data sets, while near-infrared reflectance spectroscopy (NIRS) was performed on Industry 2002, Agassiz 2002, Agassiz 2003 and Industry 2005 data sets.  The majority of the wet chemistry analysis was carried out at the University of British Columbia with a small amount analyzed at PARC-Agassiz.  Acquisition of NIRS spectra was carried out exclusively at PARC-Agassiz.  The development of the wet chemistry or reference method, based on amyloglucosidase-glucose oxidase-peroxidase, will be described followed by a description of the NIRS method. A6.2.1 Amyloglucosidase-glucose oxidase-peroxidase method of starch analysis Starch was quantified using an enzymatic method containing amyloglucosidase to break the starch into glucose units and glucose oxidase-peroxidase to measure the resulting glucose product (Anonymous 2000; Anonymous 2005; Hendrix 1993; Rose et al. 1991; Karkalas 1985; Beutler 1983; Kunst et al. 1983).  The development of this method was an iterative process; therefore, the final method will be described first and the development and troubleshooting of the method in the succeeding sections.  The amyloglucosidase-glucose oxidase-peroxidase (AGOP) method can be divided into five steps:  tissue preparation, extraction of soluble sugars, gelatinization, solublization and hydrolysis of starch to glucose and measurement of glucose product.  These steps are outlined below: 1. Leaf tissue preparation.   Prior to leaf sample collection aluminum foil envelopes were hand-made and labeled with two identification numbers.  One identification number was for the leaf sample collection and a second random identifying number was for the dry mass and starch analysis.  This system ensured that the analyst did not know the contents of  237 the envelope and thus could not introduce bias into the analysis.  Leaf samples were collected from plants growing in several greenhouses in the lower mainland of British Columbia.  The middle four leaflets (excluding the terminal leaflet) were sampled. Within 2 hours of leaf collection, 10 punches per leaflet were made with a 0.01 m (1 cm) cork borer and immediately transferred to an aluminum foil envelope and plunged into liquid nitrogen.  The total harvested leaf area was 0.000785 m2 (7.85 cm2).  The leaf tissue was transported in liquid nitrogen to the University of British Columbia or Agriculture and Agri-Food Canada at Agassiz (AAFC) and then freeze-dried under vacuum.  After drying, the samples were weighed and ground by hand using a mortar and pestle to fine particles (approximately 500 um).  Samples were then transferred to a 1.5 ml eppendorf-style centrifuge tube and stored in a –70 C freezer or in a desiccator with drierite.  Immediately prior to the analysis, samples were thawed, re-weighed and transferred to test tubes (16 x 100 mm) –one sample per tube. 2. Removal of soluble sugars.  Soluble sugars such as sucrose, fructose and glucose were removed from samples, to prevent the overestimation of starch.  Using a method adapted from a commercial starch analysis kit (STA20 Sigma-Aldrich Canada LTD, Oakville, ON) and Rose et al. (1991), 5 ml of 80% ethanol was added to each test tube, capped with a glass marble and incubated at 80-85 C for 5 minutes.  The tubes were removed from the water bath, mixed, another 5 ml of 80% ethanol was added and the samples centrifuged for 10 minutes at 1000g.  The supernatant was discarded and the pellet resuspended in 10 ml of ethanol, again mixed and centrifuged for 10 minutes.  The supernatant was decanted and discarded.  238 3. Gelatinization of starch.  The starch was gelatinized by re-suspending the pellet in 3 ml of distilled water and incubation in a boiling water bath for 20 minutes (Rose et al. 1991; STA20 Starch kit, Sigma-Aldrich). 4. Hydrolysis of starch to glucose.  One-millilitre aliquots of 0.1 M sodium actetate buffer (pH=4.3) and amyloglucosidase enzyme (EC 3.2.1.3, 1244.4 U per ml, from Aspergillus niger, Sigma-Aldrich A3514, later replaced by A1602) diluted 1:100 was added to a 1 ml aliquot of the starch extract.  The mixture was gently mixed and the solution was incubated for 1 hour at 60 C in a shaking water bath.  To stop any potential hydrolysis of cellulose the amyloglucosidase was inactivated by immersion into a boiling water bath for 3 minutes. 5. Measurement of glucose product.  Glucose was measured using the glucose oxidase peroxidase (GOP) assay (510A Glucose kit, Sigma-Aldrich) (Rose et al. 1991).  The samples were diluted either 5, 10 or 20 times, a 0.5 ml of the diluted extract was transferred to a test tube and 5 ml of glucose oxidase peroxidase solution with the colour reagent o-dianisidine were added and tubes were incubated at 37 C for 30 minutes.  The absorbance of the resulting coloured product was measured using a spectrophotometer at 450 nm within 30 minutes. In 2003, and later the glucose analysis kits from Sigma-Aldrich were unavailable. The GOP assay was then made following the directions of Anonymous (2000) and Kunst et al. (1983).  Glucose oxidase from Aspergillus niger (G7141, Sigma-Aldrich), peroxidase from horseradish (P8125, Sigma-Aldrich), 4-aminoantipurine, phenol, sodium phosphate dibasic dodecahydrate and potassium phosphate were mixed according to Anonymous (2000).  The resulting coloured product was measured using a spectrophotometer at 505 nm.  239 6. Calibration.  In order to determine the starch concentration of the leaf tissue, known starch standards were analysed with each set of samples starting at step 3. Six samples of wheat starch in the 0 to 10 mg range and one sample of approximately 22 mg were analysed. Using Excel software (Microsoft Corp. 1985), linear regressions were performed between starch concentration and absorbance for each dilution factor (5, 10 and 20) used.  The starch content of the leaf samples was determined by the following equation: Equation A6.2: ⎟⎠ ⎞⎜⎝ ⎛= standard of Abs. (mg) standard ofion concentratstarch  sample of Abs.(mg)content Starch where Abs is absorbance.  The determined amount of glucose liberated from starch standard samples was used as a check for the hydrolysis performance of the enzymes.  At step 5, 4 glucose standards between 0 and 0.16 mg were prepared and their absorbance determined. The glucose content from the starch standards was read from the curve. A6.2.2 Refinement and troubleshooting of the AGOP method of starch analysis Investigations were carried out to fine-tune the AGOP method of starch analysis.  It was unclear from the literature how long samples should be incubated with the amyloglucosidase enzyme.  An experiment was designed to examine the amount of glucose released after various incubation times of wheat starch with amyloglucosidase; these results are in Figure A6.3.  Thirty-five, 6 mg samples of wheat starch were weighed out and transferred to separate test tubes.  The AGOP protocol for starch analysis was followed from steps 3 to 5 with the following deviations.  At step 4 (hydrolysis of starch to glucose) the samples were incubated for either: 0, 7, 15, 30, 45, 60 or 120 minutes after which the reaction was halted and the formation of the colour product measured at 450 nm.  240 One of the cautions of using amyloglucosidase is that it can be unspecific, hydrolysing cellulose as well as starch –resulting in an overestimation of starch.  An experiment was also devised to determine if amyloglucosidase significantly hydrolyses cellulose.  Approximately 30 mg of pure cellulose was weighed, exposed to the AGOP method and the resulting absorbance determined at 450 nm. A6.2.3 Accuracy and precision of AGOP method for pure starch samples The calibration procedure (section A6.2.1 step 6) provided information that was useful for determining the accuracy of the AGOP method.  Starch can be calculated from the amount of glucose liberated.  The calculation is: Equation A6.3: Calculated starch (mg) = amount of glucose in sample (mg) x 3 (step 2) x 3 (step 3) x dilution factor in step 5 (5, 10 or 20) x 0.9 (glucose to starch conversion factor) This formula can be shortened to: Equation A6.4: Calculated starch (mg) = amount of glucose (mg) x 8.1 x dilution step 5 To assess the degree of accuracy and precision of this method a linear regression was conducted between the starch weighed out prior to analysis and amount of starch recovered from the analysis.  These data were analyzed using the regression procedure of SAS (version 8.2, Cary, North Carolina) the results of which are in Figure A6.4.  Analysis of the residuals produced from the regression, using the univariate procedure, indicated that the residuals did not exhibit a normal distribution and the variance was heteroscedastic.  Plots of the residuals revealed a large degree of scatter primarily at the highest starch concentration, (approximately 22 mg).  Transformation of the data using the natural logarithm and square  241 root were unsuccessful in reducing the heteroscedasticity of the variance.  A plot of the raw data indicated that recovered starch was often underestimated at the highest measured starch concentration.  This was re-enforced by the observation that it was often difficult during the analysis to get the starch to fully dissolve into solution at the highest concentration.  Using the aforementioned rationale and the results of the stem and leaf plot and box plot, of the univariate analysis, seven data points identified as outliers were removed from the data set. After outliers were removed, the residuals exhibited a normal distribution.  However, the variance remained heteroscedastic even after transformations where applied. A6.2.4 Accuracy and precision of the AGOP method for starch in tomato leaf tissue The accuracy of the AGOP method was tested using leaf samples spiked with wheat starch (Table A6.2B).  Leaves were collected, their petioles were re-cut, placed in a water- filled vase and put in a darkened room to deplete their endogenous starch.  After 3 days the leaves were removed and processed for starch following steps 1 to 5 of the AGOP method with the following modification.  Samples were processed as per step 1, after grinding, the samples were bulked together, thoroughly mixed and 10 samples of 30 mg of leaf tissue were transferred to separate test tubes.  After the extraction of soluble sugars (step 2), 3 mg of wheat starch was added to 5 of the 10 samples.  Mean and standard deviation were calculated using Excel software. The accuracy and precision of the AGOP method was compared to a commercially available starch analysis kit (STA20 Sigma-Aldrich Canada, Oakville, ON) using tomato leaf tissue (Table A6.2A).  Leaf tissue was collected as per step 1 of the AGOP method (section A6.2.1) with the following changes.  After grinding, samples were pooled, thoroughly mixed and 14 aliquots of 30 mg of ground leaf tissue were transferred to separate test tubes for the  242 Table A6.1 The dry mass (DM) composition of treatments used for determining precision of the Amyloglucosidase-Glucose Oxidase-Peroxidase and near infrared spectroscopy methods of starch quantification. Treatment1 HS DM2 LS DM2 T DM R DM % / % mg mg mg mg 0/100 0.0 ±0 24.3 ±0.2 24.3 ±0.2 18.9 ±1.2 10/90 4.6 ±0.2 21.8 ±0.2 26.4 ±0.2 21.4 ±0.7 20/80 9.4 ±0.2 19.5 ±0.2 28.9 ±0.2 23.2 ±1.6 30/70 14.2 ±0.2 16.9 ±0.2 31.1 ±0.2 26.3 ±0.6 40/60 18.5 ±0.3 14.6 ±0.1 33.1 ±0.3 28.4 ±0.7 50/50 23.6 ±0.2 12.1 ±0.1 35.8 ±0.2 31.2 ±1.5 60/40 28.4 ±0.2 9.7 ±0.1 38.1 ±0.2 33.2 ±1.7 70/30 33.0 ±0.2 7.5 ±0.2 40.5 ±0.3 34.9 ±1.2 80/20 37.6 ±0.1 4.6 ±0.2 42.2 ±0.2 37.4 ±0.7 90/10 42.4 ±0.3 2.5 ±0.2 45.0 ±0.4 38.9 ±3.1 100/0 47.1 ±0.2 0.0 ±0 47.1 ±0.2 39.6 ±2.3 1. The proportion of high starch/low starch in treatment. 2. Dry mass is from a leaf area 0.000785 m2 from leaves confirmed to have high and low starch content.  HS is leaf tissue with 20 mg of starch.  LS is tissue with 1 mg of starch.  T is resultant mass of blending HS and LS as per the treatment proportion.  R is the dry mass recovered after grinding the sample.  Presented means are followed by standard deviation, n=8. .  243 analysis of starch.  Seven of the leaf samples were analysed using the AGOP method and seven were analysed using the Sigma starch kit.  Mean and standard errors were calculated using Excel software.  The null hypothesis that the two methods were detecting similar amounts of starch was tested using the t-test function (2 tail), also in Excel software. A6.2.5 Precision of the AGOP method over a range of starch concentration in tomato leaf tissue Evaluation of the precision of the AGOP method was carried out over a range of leaf starch contents using the Industry 2005 data set.  This data set was also used for quantifying the precision of the near-infrared spectroscopy method (section A6.2.8).  One hundred leaf samples were collected and processed as per step 1 in section A6.2.1 from the upper (high starch) and lower (low starch) canopies of plants in a commercial greenhouse.  To confirm that starch contents were distinct and related to leaf dry mass, a small number of each type of “high” and “low” samples were analyzed.  In order to prepare samples that would exhibit a range of starch contents, samples belonging to the high starch group were combined and thoroughly mixed.  The same procedure was carried out for the low starch samples.  Using the average sample dry mass as a guide, high and low starch samples were weighed out and combined together in varying proportions to give 11 sequential dilutions in starch content (Table A6.1).  That is 100, 90, 80, 70, 60, 50, 40, 30, 20, 10 and 0% of the high starch samples were mixed with 0, 10, 20, 30, 40, 50, 60, 70, 80, 90 and 100 of the low starch samples (Table A6.1).  Each of the 11 treatments was replicated 8 times for a total of 88 samples. After the near-infrared spectra were obtained, the samples were analyzed for starch using the AGOP method outlined in section A6.2.1.  A linear function was fitted to the leaf starch content (mg) and sample dry mass (mg), using the regression procedure of SAS. Residuals were generated and analyzed using the univariate procedure.  Results of the  244 Table A6.2A A comparison of the Amyloglucosidase-Glucose-Oxidase-Peroxidase (AGOP) method and Sigma Starch kit for detecting starch in homogenous tomato leaf tissue samples.  Tissue Starch  mg mg AGOP1 30.3±0.2 1.4±0.8 Sigma Starch Kit2 30.3 ±0.2 1.5±0.9 T-test NS p=0.87 (NS) N=7, T-test at p=0.05 level.  Means are followed by the standard deviation.  Calibration equations were: 1. absorbance(@450 nm)=0.029(starchmg), R2=0.98. 2. absorbance(@450 nm)=0.028(starchmg) R2=0.99. NS is not significant   Table A6.2B The accuracy of Amyloglucosidase-Glucose-Oxidase-Peroxidase (AGOP) method to detected wheat starch added to tomato leaf tissue samples.  Tissue Added Starch Recovered Starch2  mg mg mg Sample1 30.4 ±0.3 0 0.3 ±0.05 Sample + Starch1 30.3 ±0.2 3.3±0.1 3.5 ±0.3 Recovery   98% 1. N=5, mean followed by standard deviation. 2. Calibration equation was absorbance(@450 nm) = 0.034(starchmg), R2=0.99 and absorbance(@450 nm)=4.48(glucosemg), R2=1  245 univariate analysis indicated that the residuals were not normally distributed.  To correct this, three data points were deleted, being deemed outliers as indicated by the stem and leaf plot and box plot.  After these deletions, the residuals appeared normally distributed and tests for normality were passed but the variance remained heteroscedastic.  Transformation of the leaf starch content by the natural logarithm did improve homoscedasity of the variance. A6.2.6 Near-infrared spectroscopic method of leaf starch analysis Near infrared spectroscopy (NIRS) was used to determine the starch contents of leaf samples from Industry 2002, Agassiz 2002, Agassiz 2003 and Industry 2005 data sets.  Initial calibration equations were formed from the Industry 2002 samples and modified for the Agassiz 2002 and Agassiz 2003 data sets.  Separate calibrations were used for the Industry 2005 samples as the particle size and plant cultivar were distinct from the majority of samples.  The general methodology for NIRS will be outlined below followed by the specific analysis for each data set.  The NIRS methodology outline consists of: tissue preparation, NIRS instrumentation and equipment, obtaining the NIR spectra, pre-treatment of data, sample selection for reference chemistry, calibration and validation. Tissue preparation.  Leaf tissue was collected as per the wet chemistry method (AGOP).  After freeze drying, the leaf samples were weighed, then transferred to 1.5 ml eppendorf-style centrifuge tubes and stored in a –70 C freezer or in a desiccator with drierite. Prior to obtaining the near-infrared spectra, samples were thawed and broken into smaller pieces to facilitate mixing (Figure A6.2B). NIRS instrumentation and equipment.  The NIRS system (Figure A6.2A) consisted of a fibre-optic probe attached to a NIRSystems 6500 scanning monochromator (Foss North American, Silver Spring, MD) containing a tungsten light source.  The NIRS system was  246 Figure A6.2 The near infrared spectroscopy system used for analyzing leaf starch.  In A is a view of the system, B is a close up of a leaf sample, C and D are a views of the fiber optic probe scanning the reference and a sample, respectively. D C B A  247  interfaced to a personal computer to store and process the data using the WinIsi (Version 1.5, Infrasoft International LLC, Port Matilda, PA) scanning and chemometerics software. To ensure output from the light source was uniform, the NIRSystem was warmed for 1 hour prior to sample scanning. Obtaining the near-infrared spectra.  Using a lab jack the reference standard or sample was lifted to the tip of the fibre optic probe until they were almost touching (Figures A6.2C and A6.2D).  The reference standard was a ceramic plate and is necessary to correct against background noise and instrument drift (Figure A6.2C) (Norris 1989).  For sample scanning, an eppendorf tube containing the sample was inverted and the sample gently transferred to the lid (Figure A6.2B).  The sample was then positioned in a sample holder of clay, placed on a lab jack and raised to just under the tip of the probe (Figure A6.2D).  After scanning the reference, the near-infrared spectrum for each sample was generated by measuring the reflectance at 699 wavelengths, which were every 2 nm between 1100 and 2498 nm.  This raw reflectance data was converted to absorbance by taking the reciprocal log10 of the reflectance data (Absorbance=log [1/R]).  This function is reported to linearly relate sample composition to absorbance in the near infrared region (Swift 2003). Pre-treatment of data.  After the spectra are obtained, a mathematical pretreatment of the spectral data was used to improve calibration accuracy.  Since it is often necessary to correct for particle and background scatter, a number of treatments: none, standard normal variate, and standard normal variate detrending were examined (Swift 2003).  Standard normal variate scales each spectrum to a standard deviation of 1.0, and standard normal variate detrending removes the linear and quadratic curvature of each spectrum (Swift 2003; Deaville & Flinn 2000).  A second useful transformation used was derivatization.  Either first  248 or second derivatives were used to reduce the quantity of data, remove baseline shifts and to help discern overlapping absorption bands.  A combination of the scatter correction technique and derivatization were investigated to determine which yielded the best model (Swift et al. 2007).  Further data reduction was accomplished by using principle component analysis, where the spectral data were compressed into a small number of independent components or scores.  The first principle component is the linear combinations of wavelengths that vary the most from sample to sample (Westerhaus 1989).  The second principle component is independent of the first one and has the next most variation from sample to sample and so on, until all the variability is accounted for (Westerhaus 1989). Sample selection for reference chemistry.  Using the chemometrics software, samples for calibration were selected based on their spectral information.  The Winisi software has two algorithms, CENTER and SELECT that use principal component analysis and Mahalanobis distances (H) for calibration development.  The CENTER algorithm (Shenk & Westerhaus 1991a; Shenk & Westerhaus 1991b) uses a Mahalanobis distance calculated as the difference between a sample spectra and the mean spectra of the population (Swift 2003). This specific measure is denoted as GH (Global Mahalanobis) in the NIRS literature and is used to define the spectra of the population, eliminate spectra that are extreme and identify those spectra that should be analyzed by the reference method.  Spectral outliers are those that have large GH values, which Shenk & Westerhaus (1991b) suggest should have 3.0 as an upper limit.  The SELECT algorithm (Shenk & Westerhaus 1991a; Shenk & Westerhaus 1991b) was used to find spectrally unique samples and identify redundant samples.  In this case the Mahalanobis distance, referred to as NH, was calculated to measure the difference between a sample’s spectra and its nearest neighbors (Narra et al. 2005; Swift 2003).  Shenk  249 & Westerhaus (1991a) and Swift (2003) point out that an NH value less then 0.6 indicates the spectra of two samples are not spectrally unique and that one sample can represent neighboring samples.  In practical terms this means that the reference chemistry can be carried out on the one representative sample as opposed to all of them.  Once the calibration equation has been developed, only samples identified as spectrally unique need to have their starch contents determined by wet chemistry. Equation calibration and validation.  In NIRS, calibration relies on chemometrics, where regression modeling is used to relate the spectra generated by the instrument with the values obtained from the reference method.  Partial least-squares (PLS) regression is preferred as it uses the full spectrum to develop the relationship between the spectral absorbance and the reference method (Deaville & Flinn 2000; Narra et al. 2005).  In PLS regression the residuals at each wavelength are standardized (divided by the standard deviations of the residuals at that wavelength) before calculating the next factor (Shenk & Westerhaus 1995).  Equation validation is conducted to assess the predictive ability of the selected calibration equation.  In this procedure the sample set was split into several groups and the calibration preformed on one group and used to predict the outcome of another group until every sample was used for both calibration and validation (Foley et al. 1998; Deaville & Flinn 2000).  Consequently, calibration and validation were simultaneously.  The best-fit equation was selected as the equation with the highest (close to 1) coefficient of determination (R2) and one minus the variance ratio (1-VR).  Variance ratio was the unexplained variance divided by the total variance.  If the unexplained variance was only a small proportion of the total variance the ratio will be close to 1 (Narra et al. 2005).  Since PLS regression uses cross-validation to prevent over-fitting of the equations (Shenk &  250 Westerhaus 1995), the best equation will also have a low standard error of calibration and low standard error of cross-validation (Narra et al. 2005).  In practice the NIRS prediction equation is a work in progress, the equation can always be improved with the addition of new samples (Stuth et al. 2003). A6.2.7 Calibration development for near-infrared spectroscopy Industry 2002 and Agassiz data sets.  Development of a calibration equation was initiated using the industry 2002 data set, containing 950 leaf samples.  A preliminary calibration equation was first developed from a subset of 50 samples.  The samples were selected and their spectra obtained followed by determination of sample starch using the wet chemistry method - AGOP.  These selected samples were thought to represent a range of starch contents and were selected on the assumption that dry mass is directly related to starch content.  Once the starch was determined by wet chemistry, a calibration equation was developed using the total amount of starch in a sample.  The spectra were then obtained for the remaining 900 samples.  Using the CENTER and SELECT algorithms, 635 of the 900 samples were selected as spectrally unique (samples with GH>3.0) and required their starch contents to be determined by wet chemistry.  Once the wet chemistry was complete these data were added to the calibration model and used for further calibration development. The spectra of 2751 leaf samples from the Agassiz data sets were obtained.  Using the CENTER and SELECT algorithms and a sample criterion of GH>5.0, 253 samples were selected for wet chemistry.  Wet chemistry was also performed on a further 81 samples, suspected of having very high starch contents based on their dry mass.  Once the wet chemistry was performed on these 334 samples, these data were also added to the calibration model for further model development.  251 To produce the most robust calibration equation, data from the industry 2002 and Agassiz data sets were combined and reanalyzed.  Five hundred and thirty-five samples that were spectrally different (using CENTER and SELECT algorithms) and had wet chemistry data were used to build the calibration equation.  To further improve the calibration, combinations of scatter correction, derivatization, detrend and standard normal variate were tested to determine which produced the most robust partial least squares model.  Model fit was evaluated by determining the lowest standard error of cross validation (SECV) and highest coefficient of determination (Table A6.3). Comparison to reference method.  Using linear regression, the starch content determined by near infrared spectroscopy was compared to the starch content measured by the AGOP method (reference method and independent variable) (Figure A6.5).  The regression and univariate procedure of SAS indicated that 14 of the 535 (2.6%) were outlying data; caused by one or all of, improper scanning, poor wet chemistry or sample mix- up.  It was not possible to re-analyse these samples after the wet chemical analysis for starch. Removing these data greatly improved the distribution of residuals, shifting them towards a normal distribution.  However, the distribution of residuals still failed tests for normality, having high kurtosis and long tails, a symptom of further outlying data.  Since results of the above analysis indicated curvature at the higher concentrations of starch, further data manipulation was carried out.   In order to ascertain if the NIRS was adequate to predict starch at lower concentrations and to determine what the concentration cut off should be, starch data measured by the reference method (AGOP) was sequentially removed from highest to lowest concentration.  When the quadratic term was no longer significantly different from zero  (p>0.05), the starch data above that concentration was not included in  252 Table A6.3 Calibration statistics of cross validation of prediction equation used to estimate starch for tomato leaf tissue for relevant data sets.  Data were per 0.000785 m2 of leaf tissue. Experiment No. Samples Mean SD SEC SECV R2 1-VR   mg mg mg mg Samples 535 2.65 3.85 1.73 0.64 0.80 0.95 Industry 2005 88 10.36 6.30 0.94 1.052 0.98 0.97 SD is standard deviation, SEC is standard error of calibration, SECV is standard error of cross validation, R2 is coefficient of determination for calibration and 1-VR is variance ratio.  253 the analysis (Figure A6.6) and a simple linear regression was performed. Industry 2005 data set.  Precision of the NIRS method over a range of leaf starches was investigated.  The experimental design and tissue preparation is as explain in section A6.2.4.  In brief, 11 treatments of varying leaf starch concentrations were prepared and replicated eight times (Table A6.1).  This method differs from the above data sets in that samples were finely and not coarsely ground to facilitate sample blending.  After the near- infrared spectra were obtained and the wet chemistry completed, a calibration equation was developed (Table A6.3).  The starch in leaf tissue from the NIRS method was compared to that determined from the wet chemistry using regression analysis. A6.3 RESULTS Experiments were carried out to clarify methodological uncertainties of the enzymatic and near infrared spectrophotometry for starch detection in tomato leaf tissue and to ascertain the accuracy and precision of both methods. A6.3.1 Exposure to amyloglucosidase for starch analysis The optimal exposure time for amyloglucosidase to hydrolyse the α1-4 and α1-6 linkages between glucose units of wheat starch was determined by a time-course experiment. After 15 minutes of incubation with amyloglucosidase the formation of the coloured product was almost maximal as indicated by the absorbance of the solution (Figure A6.3).  A small increase in absorbance (10%) occurred between 15 and 60 minutes with no sensible increase between 60 and 120 minutes (Figure A6.3).  These results indicate 60 minutes of exposure to amyloglucosidase is sufficient to hydrolysed the glycosidic linkages in pure starch samples. The ability of amyloglucosidase to hydrolyse β1-4 linkages of cellulose was a concern and consequently investigated.  Amyloglucosidase activity on pure cellulose was very low after  254 0 0.05 0.1 0.15 0.2 0.25 0.3 0 10 20 30 40 50 60 70 80 90 100 110 120 130 Time (Minutes) A bs or ba nc e at  4 50  n m   Figure A6.3 The coloured product produced with increasing time of incubation of starch to amyloglucosidase.  The line is drawn through the mean values.  255 60 minutes of incubation being 0.01 at 450 nm, indicating very little hydrolysis of the β1-4 linkages of cellulose. A6.3.2 Accuracy and precision AGOP method of starch analysis The accuracy and precision of the AGOP method was examined for both pure starch standards and leaf tissue samples.  The accuracy of the AGOP analysis for recovering starch from known pure starch standards is illustrated in Figure A6.4.  Recovered starch is calculated from the amount of glucose librated from the starch standards using equation A6.4.  Results of the ANOVA indicate the recovered starch is almost exclusively dependent on the known initial amount of starch, as indicated by the high coefficient of determination (R2) of 0.99 and p<0.0001 for the regression equation (Figure A6.4).  The relationship (slope, β1) between the starch standard and the recovered starch was 0.99 (p<0.0001), slightly below (1%) the desired outcome of 1.0, indicating a perfect relationship.  The constant or β0 value of the regression equation was 0.06 and was not significantly distinct from 0 at the 5% probability level (p=0.69).  The precision of the starch recovery from standards was satisfactory as indicated by 7.7% coefficient of variation, ±0.006 mg standard error and the 95% confidence limit of ±1.06 mg of the coefficient of the regression line. In collected samples of tomato leaves it is impossible to know the true starch content of the tissue and thus the degree of accuracy of AGOP method for leaf tissue.  The approach taken to ascertain accuracy was to compare the AGOP method to a second method, a commercially available starch analysis kit.  Both methods detected similar amounts of starch (1.4 to 1.5 mg) from tomato leaf samples (Table A6.2A).  The means of each method were tested for statistical differences using a T-test and were similar (p=0.8) (Table A6.2A).  A  256 0.0 3.0 6.0 9.0 12.0 15.0 18.0 21.0 24.0 27.0 0.0 3.0 6.0 9.0 12.0 15.0 18.0 21.0 24.0 27.0 Actual Starch (mg) R ec ov er ed  S ta rc h (m g) Regression line 95% CL   Figure A6.4 The calculated recovered amount of starch from the AGOP method of starch detection.  Included is the regression line, Y = 0.99X, R2=0.99 with 95% confidence limits of the predicted line.  N=241.  257 second evaluation of accuracy was determined by the recovery of starch from leaf samples spiked with pure starch.  In this experiment 98% of the added starch was recovered from the leaf tissue using the AGOP method (Table A6.2B). A6.3.3 Accuracy and precision of NIRS method of starch analysis NIRS is a secondary or indirect method of analysis that relies on chemometrics to develop a relationship between a sample’s spectra and the reference method.  Thus the accuracy and precision of NIRS is dependent on that of the reference chemistry plus the NIRS instrument error; so it is unlikely that it will be improved over that of the reference method (Foley et al. 1998).  Accuracy and precision can best be ascertained for NIRS from the calibration statistics of the prediction equation (Table A6.3).  The mean and standard deviation can be misleading because they reflect the inherent variability of the samples; an improved measure are the standard errors of calibration (SEC), cross validation (SECV) and the equation fit statistics (Table A6.3).  The Industry 2002 and Agassiz data sets exhibited a coefficient of determination (R2 =0.80) and a variance ratio of 0.95 indicating a good fit of the NIRS to the reference method. A6.3.4 Comparison between starch predicted by NIRS analysis and the reference method for leaf samples Starch determined by NIRS was plotted against starch determined by the reference method (AGOP) to ascertain the suitability of the NIRS prediction for leaf starch from the Industry 2002 and Agassiz samples (Figure A6.5A).  The plots and regression analysis indicated that a small number of outliers were present (Figure A6.5A).  After the outliers were removed (Figure A6.5B), the relationship between NIRS starch and AGOP starch was strongly quadratic (p<0.0001), R2=0.96.  Both the linear and quadratic terms, but not the  258  Figure A6.5 Leaf starch from samples measured by the reference method (AGOP) and predicted from near infrared spectrophotometery (NIRS). In panel A is the raw data, N=535.  In panel B outliers have been removed, N=521.  The regression equation is Y=-0.024+1.17X-0.031X2, R2 = 0.96, also included is Y=1X line. Each sample is of leaf area 0.000785 m2. 0.0 3.0 6.0 9.0 12.0 15.0 18.0 21.0 24.0 27.0 0.0 3.0 6.0 9.0 12.0 15.0 18.0 21.0 24.0 27.0 N IR S St ar ch  (m g) 0.0 3.0 6.0 9.0 12.0 15.0 18.0 21.0 24.0 27.0 0.0 3.0 6.0 9.0 12.0 15.0 18.0 21.0 24.0 27.0 Reference Method Starch (mg) N IR S St ar ch  (m g) Sample Predicted 1:1 line A B  259 intercept were significantly distinct from zero (p<0.0001).  The quadratic response indicates that the NIRS under-predicted high concentrations of leaf starch (Figure A6.5B). The results in Figure A6.5B also indicate that the quadratic function matches the unity line for approximately the first third of the starch concentration.  This was confirmed by regression analysis, where the quadratic term was not significantly distinct from zero for reference starch concentrations between 0 and 8.4 mg.  Therefore, 63 of the 521 samples with starch contents greater than 8.4 mg were excluded from the analysis (Figure A6.6).  The linear regression on this subset of data was significant (p<0.0001) and the model fit was strong, as indicated by the coefficient of determination (R2) of 0.92.  Both the intercept (0.10) and coefficient (0.98) were significantly distinct from zero (p<0.001).  The coefficient when tested was not significantly different from 1 indicating a strong relationship between the reference method and the NIRS (Figure A6.6). A6.3.5 Precision of starch quantification over a range of leaf starch concentrations The precision of starch detection in leaf samples was determined from specially blended leaf samples containing from 1 to 23 mg of starch (Table A6.1 and Figure A6.7). These data have also been referred to as Industry 2005. AGOP method.  In Figure A6.7A there is a positive linear relationship (R2=0.99 and ANOVA p<0.0001) between the amount of starch measured in a sample using the AGOP method and the sample dry mass.  The coefficients and the intercept were also both significant at p< 0.0001 (Figure A6.7A).  The intercept is expected to be significant as a small amount of starch is present in the low starch samples.  The regression coefficient and its standard error were 0.9±0.02 mg.  Precision of the analysis was satisfactory as indicated by the coefficient of variation of 9.9%.  260 0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0 10.0 0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0 10.0 Reference Method Starch (mg) N IR S St ar ch  (m g) Sample Predicted 95% CL Pred   Figure A6.6 Leaf starch from samples measured by the reference method (AGOP) method and predicted from near infrared spectrophotometery (NIRS).  Outliers and reference method data greater than 8.4 mg have been removed.  Regression line and 95% prediction limits for the regression line, N=458.  The regression equation is Y = 0.10 + 0.98X , R2 = 0.92.  Each sample is of leaf area 0.000785 m2.  261  Figure A6.7 The starch recovered from samples blend with increasing amounts of a high starch sample. Panel A is the AGOP method, the regression line is Y=-20.4+0.9X, R2=0.98 and N=85. Panel B is the near infrared spectroscopy (NIRS) method, the regression line is Y= -20.2 + 0.9X, R2=0.98, N=86.  95% CL Pred is 95% prediction limits for the regression line.  See Table A6.1 for the corresponding high and low starch mass for each treatment. 0.0 3.0 6.0 9.0 12.0 15.0 18.0 21.0 24.0 21.0 24.0 27.0 30.0 33.0 36.0 39.0 42.0 45.0 48.0 51.0 Sample dry weight (mg) St ar ch  (m g) Sample Predicted 95% CL Pred 0.0 3.0 6.0 9.0 12.0 15.0 18.0 21.0 24.0 21.0 24.0 27.0 30.0 33.0 36.0 39.0 42.0 45.0 48.0 51.0 Sample dry weight (mg) St ar ch  (m g) Sample Predicted 95% CL Pred A B  262 NIRS method.  Prior to analysis by the AGOP method the near infrared spectra were determined for the samples.  The starch content as determined by the AGOP method (reference method for NIRS) was used for the calibration of the NIRS method.  As already mentioned a separate NIRS calibration was preformed on these data because of differences in particle size and plant cultivar compared to the main data set (Table A6.3).  The calibration statistics for these data were adequate; the R2 was 0.98 and a variance ratio of 0.97 (Table A6.3). The starch content determined by NIRS when regressed on leaf dry mass exhibited a significant linear response (Figure A6.7B).  The coefficient and intercept were both significant at p<0.0001 (Figure A6.7B).  The regression coefficient and its standard error was 0.9 ±0.01 mg, respectively.  Precision of the analysis was satisfactory as indicated by the coefficient of variation of 8.9%.  The equations that describe starch content determined by AGOP or NIRS to leaf dry mass were virtually identical for both methods, although the scatter around the regression line was distinct for each method (Figure A6.7A and B). NIRS prediction of starch.  As for the other samples, the starch content determined by NIRS was regressed against leaf starch determined by the reference method (Figure A6.8). The NIRS was very effective in predicting the leaf starch content as determined by the reference method.  The coefficient was 0.99 and upon testing was found not to be significantly different from 1.  The intercept was 0.16, found to be not significantly different from 0.  It should be noted that the NIRS did not underestimated the starch beyond 8.4 mg as indicated by the linear relationship (no curvature) (Figure A6.8).   263 0.0 3.0 6.0 9.0 12.0 15.0 18.0 21.0 24.0 0.0 3.0 6.0 9.0 12.0 15.0 18.0 21.0 24.0 Reference Method (mg) N IR S (m g) Sample Predicted 95% CL Pred   Figure A6.8 Starch measured by the reference method and predicted from near infrared spectrophotometery from the leaf samples blended with increasing amounts of a high starch sample.  The equation of the regression line is Y=0.16 + 0.99X, R2=0.99, N=84.  95% CL Pred is 95% prediction limits for the regression line.  264 A6.3.6 Starch content of collected samples The distribution of sample starch contents for all but the Industry 2005 data set are illustrated in a histogram (Figure A6.9).  The majority (38.1%) of the 5665 samples exhibited between 0 and 1.2 mg of starch (Figure A6.9).  Adding up the histogram categories reveals that 89% of the samples contained between 0 and 8.4 mg of starch (Figure A6.9).  Eight percent of samples exhibited greater than 10.2 mg of starch with only 1% of samples containing 18 mg or more of starch (Figure A6.9). A6.4 DISCUSSION The goal of this chapter was to clarify methodological uncertainties of the enzymatic and near infrared spectrophotometry for starch detection in tomato leaf tissue and to ascertain the accuracy and precision of both methods. A6.4.1 Methodological considerations for AGOP A methodological consideration in this work is that full hydrolysis of starch is reached, thus ensuring that the method is reasonably accurate and precise.  The hydrolysis of glycosidic linkages depends on the concentration of the enzyme, the temperature and the amount of time for incubation.  For the amyloglucosidase enzyme (1244.4 U per ml, Sigma- Aldrich A3514/A1602) used here the manufacturers information is 1 unit will liberate 1.0 mg of glucose from starch in 3 minutes at 55 C (Anonymous 2005).  In the AGOP method, 12.4 mg of glucose (11.2 mg of starch) would be liberated in 3 minutes given that 12.4 U of amyloglucosidase was added to each sample.  In fact, each sample of wheat starch was dissolved into 3 ml of water and a 1 ml aliquot was used for analysis.  For the highest sample here, approximately 24 mg of starch, the actual amount of starch was 24 mg divided by 3 ml, yielding 8 mg (8.8 mg of glucose) per test tube.  Thus, more than enough enzyme was present to hydrolyse the starch  265 0 5 10 15 20 25 30 35 40 0.6 3 5.4 7.8 10.2 12.6 15 17.4 19.8 22.2 24.6 27 29.4 Sample Starch Content (mg) Pe rc en ta ge   Figure A6.9 A histogram illustrating the distribution of sample starch concentration for all experiments.  The leaf area of each sample was 0.000785 m2.  The number of samples was 5665.  266  Rose et al. (1991) has reviewed some of the difficulties with using the literature for developing a starch methodology.  In the literature the incubation time varied anywhere between 20 minutes (Anonymous 2005) to 48 hours (Haissig & Dickson 1979).  Results of the time course experiment indicated that 60 minutes was an adequate time to hydrolyse the starch (Figure A6.3).  A minimum time of exposure for amyloglucosidase is desirable to increase sample throughput and to reduce the possibility of any hydrolysis of the β1-4 linkages of cellulose.  After 60 minutes of incubation at 60 C almost no celluloytic activity was detected with the amyloglucosidase enzyme used here. A6.4.2 Accuracy and precision of the AGOP method The amount of starch recovered from a pure wheat starch standard was 99%, 1% lower than the actual amount of starch weighed out (Figure A6.4).  Other reports of wheat starch recovery using amyloglucosidase were (mean and standard deviation):  92.4 ±1.1% (Denison et al. 1990) and (with α-amylase) 98.7 ±0.5% (Karkalas 1985).  It should be noted that the above studies were carried out on single concentrations of starch whereas the work here (Figure A6.4) is for a range of starch concentrations (1.5 to 24.0 mg).  The comparison of variability between statistics that measure central tendency and predictive statistics such as regression are not straightforward.  Variability appeared larger in my work (coefficient of variation of 7.7%); however, the scope of the experiment was large with the presentation of data from the starch standards collected over several years from about 40 experiments where many different batches of enzymes were used.  This does illustrate the robustness of the analysis. For the quantification of leaf starch the AGOP method uses two standard curves.  A standard curve derived from wheat starch is used to determine the unknown starch content of  267 the leaf tissue and a glucose standard curve is used to ensure the full hydrolysis of the wheat starch.  Wheat starch is a storage starch and it is possible that it is not representative of starches that are formed and broken down diurnally in the leaf.  Zeeman et al. (2002) for Arabidopsis and Santacruz et al. (2004) for potato, indicate that leaf starch granules are smaller and contain less amylose but are otherwise similar in structural organization to that of starch from non-photosynthetic tissue. In leaf tissue there is no direct way to check that complete enzymatic hydrolysis of starch has occurred (Haissig & Dickson 1979).  The approach here was to compare the AGOP method to a commercially available kit and to analyze samples spiked with wheat starch.  Results of the AGOP method on leaf starch compared favourably with a commercially available starch kit from Sigma Aldrich (Table A6.2A).  The Sigma kit exhibited a slightly lower standard deviation but the high cost of the kit ($4 per sample verses $0.54 for AGOP) makes these kits unfeasible to use.  The AGOP method recovered 98% of the added starch (spiked) to leaf samples (Table A6.2B).  These results indicate that the leaf tissue did not possess any interfering substances for the detection of starch.  The precision of the AGOP method was determined for leaf samples manipulated to have increasing amounts of starch (Figure A6.7).  In regression the precision of the fitted line can be determined by the standard error of the slope and the coefficient of determination (R2). The R2 was 0.98 – very high - the slope was 0.9 mg mg-1 with a standard error of ±0.02 mg mg-1.  Few published studies that examine starch in leaves present a measure of accuracy, precision or even variance.  Hendrix (1993) presents error bars for his data that are in the same range as reported here.  As well as a measurement method for leaf starch, the AGOP method had a further role as a reference method for NIRS.  268 A6.4.3 Near-infrared spectroscopy for the analysis of leaf starch The choice of an analytical method depends on the degree of accuracy and precision desired, the cost, the equipment available and comfort of the analyst with a given technique. The enzymatic analysis of starch is a time consuming analysis.  I estimate from the first step (sample grinding) to the last step (lab cleanup) the analysis took me 16 hours to measure 50 samples.  Sample throughput can be improved by automation (Denison et al. 1990; Hendrix 1993) and the use of commercial kits (Campbell et al. 1999); however, even with these aides the analyst usually has to restrict the number of samples processed.  This analytical bottleneck can compromise the experimental design by inadequate sampling of experimental units and/or the need to pool samples.  The expense and time-consuming nature of the analysis of plant constituents has led to the adoption of near infrared technology by many agricultural and manufacturing industries.  As well as speed (3 hours to prepare and scan 50 samples) NIRS allows the simultaneous measurement of multiple constituents, is non- destructive, amenable to small sample sizes, requires a minimum of sample preparation and does not use reagents (Deaville & Flinn 2000; Foley et al. 1998).  However NIRS is a secondary method of analysis, relying heavily on chemometrics that has according to Deaville & Flinn (2000) unfairly given it the reputation of being a “black box”. A reality of being a secondary method is the accuracy and precision of NIRS is dependent on that of the reference chemistry as well as affected by the NIRS instrument error.  For this reason, it is unlikely that NIRS can improve upon the analytical performance of the reference method (Foley et al. 1998).  Paulsen et al. (2003) states that the best achievable standard error of cross validation (SECV) for NIRS is about 1.3 times the standard error of the reference method.  Comparisons of the standard errors of calibration and cross validation with other reported values are difficult because of their dependence on  269 the units of measurement.  Most authors, if they show calibration statistics, express the starch as a percentage (Hattey et al. 1994; Ball et al. 1998; Paulsen et al. 2003; Kim & Williams 1990) or on a dry mass basis (Hou 1997; Stamm Katovich & Becker 1998; Xiccato et al. 2003).  In this work, the data were expressed per leaf sample, where a sample was a fixed leaf area of 0.000785 m2.  No NIRS studies were found that reported starch on a leaf area basis so it seemed a moot point to express the starch per square centimetre or metre.  I was reluctant to express the data as a percentage or on a dry mass basis for a number of reasons. Since part of the focus of this chapter was on the accuracy and precision of starch measurement I wanted the reader to easily read the tables and figures without having to convert between pure starch and leaf starch on a percentage or dry mass basis. My findings indicate NIRS to be a very good predictor of leaf starch content, as it compared favourably to the reference method (Figure A6.5).  The most striking result was the saturation-type response of the NIRS for some tissue starch contents above 8.4 mg.  This response for starch is evident in the literature but has not been expounded upon.  Significant deviation from a linear response between measured starch and starch predicted by NIRS at higher concentrations was found in work by Hou (1997), Xiccato et al. (2003) and Card et al. (1988).  The Industry 2005 data contained samples with up to 21.0 mg of starch with no curvature apparent (Figure A6.7 and Figure A6.8).  I believe these data indicate that the NIRS technique is adequate for the detection of relatively high amounts of starch.  The lack of starch detection in the Industry 2002/Agassiz data sets is likely caused by a problem with the scanning.  The relatively large particle size caused a loose stacking with gaps and overlapping of the samples occurring in the sample holder.  This may have caused the NIRS beam not to thoroughly penetrate the sample and thus not detect all the starch.  The smaller  270 particle size of Industry 2005 would cause the samples to pack more closely together with less overlapping and fewer gaps resulting in a more complete scan.  Samples with low amounts of starch appear not affected because the leaf samples are thinner which would result in less stacking. A6.4.4. Starch content of samples The vast majority of the samples were within the working ranges of the NIRS and AGOP calibration.  Approximately 90% of the samples in this work, exhibited leaf starch contents of 8.4 mg or less (Figure A6.9).  The importance of this finding is that these samples fell with in the linear range for the NIRS calibration.  For the AGOP calibration all but 6 of the 5665 (0.1%) samples were within the range of the AGOP calibration. A6.5 CONCLUSIONS The objective of this work was to investigate the suitability, accuracy and precision of two laboratory methods for measuring starch in tomato leaf tissue.  These methods are the direct measurement of starch by enzyme hydrolysis and the indirect measurement by near- infrared spectroscopy.  The findings are: AGOP method: 1. 1 hour of exposure to 12.4 U of amyloglucosidase incubated at 60 C will completely hydrolyse the glycosidic linkages of wheat starch and will not hydrolyse the cellulose in leaf tissue. 2. Accuracy of AGOP method is 99%, based on examining the amount of wheat starch recovered from measured standards.  271 3. The precision of the method is high (R2=0.99), having a low standard error of estimates of ±0.006 mg and a 95% confidence interval of ±1.02 mg. 4. Ninety-nine point nine percent (99.9%) of analyzed leaf samples fell within the calibration range of the AGOP method. 5. Accuracy and precision indicates that this method is adequate as a reference method for NIRS. NIRS method: 1. The accuracy of the NIRS method depends on the accuracy of the AGOP method.  The precision was good with R2 =0.98 and a variance ratio of 0.97. 2. For coarsely ground leaf samples, NIRS was adequate for measuring the starch in samples with less than 8.4 mg of starch.  For finely ground samples the NIRS was as good as the AGOP method, being able to quantify starch up to 21.0 mg (the upper range of starch in samples). 3. Leaf particle size, packing and sample presentation will affect the viability of NIRS; samples should be ground to 500 um. 4. Ninety percent (90%) of the 5665 leaf samples analyzed had less than 8.4 mg of starch and thus most of the data from NIRS was viable. 5. Accuracy and precision of the NIRS method will be less than the reference method caused by compounded experimental errors.  However, this must be balanced with the increased speed of sample throughput, multiple analyses of constituents and non- destructive nature of the NIRS analysis.  272 A starch granule is comprised solely of glucose assembled in a complex arrangement that gives it good stability.  The glycosidic linkages can be hydrolysed with the enzyme amyloglucosidase and the liberated glucose measured.  This was the mode of action of the Amyloglucosidse-Glucose-Oxidase-Peroxidase (AGOP) method of starch analysis.  The AGOP method was the reference method for the near-infrared spectroscopic method (NIRS) of starch analysis.  Near-infrared technology measures the reflectance from a sample and uses mathematics and statistics to quantify the amount of starch.  NIRS has many advantages, for this work the greatly increased sample throughput compared to the AGOP method was most notable.  The drawback of NIRS is its reliance on statistical models to link near-infrared spectra to the reference chemistry (known as chemometics).  Future work for NIRS should test its feasibility for in situ sampling of leaf tissue in greenhouses.  This technique could be a useful tool for measuring the starch content of intact leaves and thus the starch content of plant canopies. A6.6 LITERATURE CITED Anonymous. 2005. Starch Hydrolysis. http://www.bio-link.org/pdf/starch.pdf. 2000.  Bio- Link . February 9. Anonymous. 2005. Starch Assay Kit STA-20. http://www.sigmaaldrich.com/sigma/bulletin/sta20bul.pdf. 2005.  Sigma. February 10. Anonymous. 2005. A1602  Amyloglucosidase from Aspergillus niger.  October 25. Ball, R.A., Sabbe, W.E., and DeLong, R.E. 1998. Starch and nitrogen status in soybean during shading and nutrient deficiency. Journal of Plant Nutrition 21: 665-685. Batey, I.L. 1982. Starch analysis using thermostable alpha-amylases. Starch/Starke 34: 125- 128. Beutler, H.O. 1983. Starch. In Bergmeyer, H.U., ed. Methods of Enzymatic Analysis. Academic Press, New York, pp 2-10. Burrell, M.M. 2003. Starch:  the need for improved quality or quantity - an overview. Journal  273 of Experimental Botany 54: 451-456. Campbell, J.A., Hansen, R.W., and Wilson, J.R. 1999. Cost-effective colorimetric microtitre plate enzymatic assays for sucrose, glucose and fructose in sugarcane tissue extracts. Journal of Science and Food Agriculture 79: 232-236. Card, D.H., Peterson, D.L., Matson, P.A., and Aber, J.D. 1988. Prediction of leaf chemistry by the use of visible and near infrared reflectance spectroscopy. Remote Sensing Of Environment 26: 123-147. Deaville, E.R. and Flinn, P.C. 2000. Near-infrared (NIR) spectroscopy:  an alternative approach for the estimation of forage quality and voluntary intake. In Givens, D.I., Owen, E., Axford, R.F.E., and Omed, H.M., eds. Forage Evaluation in Ruminant Nutrition. CAB International, New York, NY, pp 301-320. Denison, R.F., Fedders, J.M., and Tong, C.B.S. 1990. Amyloglucodidase hydrolysis can overestimate starch concentration of plants. Agronomy Journal 82: 361-364. Foley, W., McIlwee, A., Lawler, I., Aragones, L., Woolnough, A., and Berding, N. 1998. Ecological applications of near infrared reflectance spectroscopy - a tool for rapid, cost-effective prediction of the composition of plant and animal tissues and aspects of animal performance. Oecologia 116: 293-305. Haissig, B.E. and Dickson, R.E. 1979. Starch measurement in plant tissue using enzymatic hydrolysis. Physiologia Plantarium 47: 141-157. Hattey, J.A., Sabbe, W.E., Baten, G.D., and Blakeney, A.B. 1994. Nitrogen and starch analysis of cotton leaves using near infrared reflectance spectroscopy (NIRS). Commun. Soil Sci. Plant Anal. 25: 1855-1863. Hendrix, D.L. 1993. Rapid extraction and analysis of nonstructural carbohydrates in plant tissues. Crop Science 33: 1306-1311. Hou, G. 1997. Effects of light, CO2 and temperature on carbohydrate metabolism in marigold.  (Tagetes Patula). University of Kentucky. Jobling, S. 2004. Improving starch for food and industrial applications. Current Opinion in Plant Biology 7: 210-218. Karkalas, J. 1985. An improved enzymatic method for the determination of native and modified starch. Journal of Science and Food Agriculture 36: 1019-1027. Kim, H.O. and Williams, P.C. 1990. Determination of starch and energy in feed grains by near-infrared reflectance spectroscopy. Journal of Agricultural and Food Chemistry 38: 682-688. Kunst, A., Draeger, B., and Ziegenhorn, J. 1983. Colorimetric methods with glucose oxidase and peroxidase. In Bergmeyer, H.U., ed. Methods of Enzymatic Analysis. Academic  274 Press, New York, pp 178-185. Matson, P., Johnson, L., Billow, C., Miller, J., and Pu, R.  1994. Seasonal patterns and remote spectral estimation of canopy chemistry across the Oregon transect. Ecological Applications 4: 280-298. Mitchell, G.A. 1990. Methods of Starch Analysis. Starch/Starke 42: 131-134. Narra, S., Fermanian, T., and Swiader, J. 2005. Analysis of mono- and polysaccharides in creeping bentgrass turf using near infrared reflectance spectroscopy. Crop Science 45: 266-273. Norris, K. 1989. NIRS Instrumentation. In Marten, G.C., Shenk, J.S., and Barton II, F.E., eds. Near Infrared Reflectance Spectroscopy (NIRS):  Analysis of Forage Quality. USDA, pp 12-17. Passos, L.P., Vidigal, M.C., da Mota, A.F., and Farias, A.P.S.F. 1999. Evaluation of techniques and suitable sample number of soluble carbohydrate extraction in stem bases of elephantgrass. Communication in Soil Science and Plant Analysis. 30: 1153- 1164. Paulsen, M.R., Pordesimo, L.O., Singh, M., Mbuvi, S.W., and Ye, B. 2003. Maize starch yield calibrations with near infrared reflectance. Biosystems Engineering 85: 455- 460. Rose, R., Rose, C.L., Omi, S.K., Forry, K.R., Durall, D.M., and Bigg, W.L. 1991. Starch determination by perchloric acid vs enzymes:  evaluating the accuracy and precision of six colorimetric methods. Journal of Agricultural and Food Chemistry 39: 2-11. Santacruz, S., Koch, K., Amdersson, R., and Aman, P. 2004. Characterization of potato leaf starch. Journal of Agricultural and Food Chemistry 52: 1985-1989. Shenk, J.S. and Westerhaus, M.O. 1991a. Population definition, sample selection, and calibration procedures for near infrared reflectance spectroscopy. Crop Science 31: 469-474. Shenk, J.S. and Westerhaus, M.O. 1991b. Population structuring of near infrared spectra and modified partial least squares regression.  Crop Science 31: 1548-1555. Shenk, J.S. and Westerhaus, M.O. 1995. Forage analysis by near infrared spectroscopy. In Barnes, R., Miller, D., and Nelson, C.J., eds. Forages Volume II.  The Science of grassland agriculture. Iowa State University Press, pp 111-120. Smith, A.M., Kenyer, K., and Martin, C. 1997. The synthesis of the starch granule. Annual Review of Plant Physiology and Plant Molecular Biology 48: 67-87. Smith, A.M., Zeeman, S.C., Thorneycroft, D., and Smith, S.M. 2003. Starch mobilization in leaves. Journal of Experimental Botany 54: 577-583.  275 Stamm Katovich, E.J. and Becker, R.J. 1998. Seasonal fluctuations of carbohydrate levels in roots and crowns of purple loosestrife (Lythrum salicaria). Weed Science 46: 540- 544. Stuth, J., Jama, A., and Tolleson, D. 2003. Direct and indirect means of predicting forage quality through near infrared reflectance spectroscopy. Field Crops Research 84: 45- 56. Swift, M.L. 2003. Prediction of dry matter, crude protein degradability, and amino acid composition of corn silage and grass silage by near infrared reflectance spectroscopy (NIRS). The University of British Columbia. Swift, M. L., Edwards, D. R., Ehret, D., and Jolliffe, P. 2007. Prediction of starch in tomato leaves using NIRS and fibre optic probe technology. Proceedings of the 12th International Conference on Near Infrared Spectroscopy, pp 683-685. Tetlow, I.J., Morell, M.K., and Emes, M.J. 2004. Recent developments in understanding the regulation of starch metabolism in higher plants. Journal of Experimental Botany 55: 2131-2145. Westerhaus, M.O. 1989. Equation development. In Marten, G.C., Shenk, J.S., and Barton II, F.E., eds. Near Infrared Reflectance Spectroscopy (NIRS):  Analysis of Forage Quality. USDA. Xiccato, G., Trocino, A., De Boever, J.L., Maertens, L., Carabano, R., Pascual, J.J., Perez, J.M., Gidenne, T., and Falcao-E-Cunha, L. 2003. Prediction of chemical composition of European compound feeds for rabbits by near infrared reflectance spectroscopy (NIRS). Animal Feed Science and Technology 104: 153-168. Zeeman, S.C., Smith, S.M., and Smith, A.M. 2004. Tansley review.  The breakdown of starch in leaves. New Phytologist  163: 247-261. Zeeman, S.C., Tiessen, A., Pilling, E., Kato, K.L., Donald, A.M., and Smith, A.M. 2002. Starch synthesis in Arabidopsis. Granule synthesis, composition, and structure. Plant Physiol 129: 516-29. 

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                        
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            data-media="{[{embed.selectedMedia}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
https://iiif.library.ubc.ca/presentation/dsp.24.1-0066866/manifest

Comment

Related Items