Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Eddy-covariance carbon balance, photosynthetic capacity and vegetation indices in a harvested boreal… Hawthorne, Iain 2008

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

Item Metadata

Download

Media
24-ubc_2009_spring_hawthorne_iain.pdf [ 2.03MB ]
Metadata
JSON: 24-1.0066759.json
JSON-LD: 24-1.0066759-ld.json
RDF/XML (Pretty): 24-1.0066759-rdf.xml
RDF/JSON: 24-1.0066759-rdf.json
Turtle: 24-1.0066759-turtle.txt
N-Triples: 24-1.0066759-rdf-ntriples.txt
Original Record: 24-1.0066759-source.json
Full Text
24-1.0066759-fulltext.txt
Citation
24-1.0066759.ris

Full Text

Eddy-Covariance Carbon Balance, Photosynthetic Capacity and Vegetation Indices in a Harvested Boreal Jack Pine Stand  by IAIN HAWTHORNE B.Sc., with Honours, University of Edinburgh, 2004  A THESIS SUBMITTED IN PARTIAL FULLFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF Master of Science in THE FACULTY OF GRADUATE STUDIES (Soil Science)  THE UNIVERSITY OF BRITISH COLUMBIA (Vancouver) October 2008 © Iain Hawthorne 2008  ABSTRACT Eddy-covariance (EC) CO2 flux data were analysed and annual carbon (C) balances estimated for a four-year period (2004-2007) following clearcut harvesting of a boreal jack pine stand in northern Saskatchewan. The site was a source of C to the atmosphere for all years, with annual net ecosystem productivity (NEP) increasing from -153 g C m-2 yr-1 in 2004 to -63 g C m-2 yr-1 in 2007. This increase was mainly due to gross primary productivity (GPP) increasing significantly from 78 to 200 g C m-2 yr-1, while ecosystem respiration (R) increased only slightly from 231 to 263 g C m-2 yr-1 over the same period. In the 2006 growing season (GS), a field campaign was conducted to investigate the relationships between monthly destructive measurements of leaf area index (LAI) and daily measurements of the normalized difference vegetation index (NDVI) and photosynthetic capacity (Amax). The latter was derived from 5-day, 16-day, 30-day and annual Michaelis-Menten light response analyses using daytime measurements of NEP and incident photosynthetically active radiation. Digital-camera data were used to evaluate the potential of using the rectilinear-lens vegetation index (RLVI) as a surrogate for NDVI of a young forest stand. Results showed that LAI was linearly related to NDVI and RLVI, which was largely the result of changes in the deciduous vegetation component across the GS. These results indicate that RLVI could be used as a surrogate for NDVI up to a GS maximum LAI of 0.91 m2 m-2 observed in 2006. Measured mean (± 1 S.D.) GS LAI was 0.67 (± 0.24) m2 m-2 in 2006. LAI accounted for the majority of the variability in Amax at the 30-day time scale, while at shorter time scales air temperature was the dominant control. For 2004 to 2007, mean spring estimates of LAI were 0.25, 0.29, 0.38 (compared to 0.40 m2 m-2 from measurements) and 0.41 m2 m-2, respectively. Results suggest that a steady increase in the jack pine LAI component accounted for the annual increases in GPP and hence NEP over the four years.  ii  TABLE OF CONTENTS ABSTRACT ....................................................................................................................... ii TABLE OF CONTENTS .................................................................................................. iii LIST OF TABLES...............................................................................................................v LIST OF FIGURES .......................................................................................................... vii LIST OF SYMBOLS, ACRONYMS AND ABBREVIATIONS ......................................xi ACKNOWLEDGMENTS ................................................................................................xiv 1. INTRODUCTION ...........................................................................................................1 2. METHODS ......................................................................................................................5 2.1. Site description .........................................................................................................5 2.2. Atmosphere and soil climate ....................................................................................7 2.3. NEE measurement ....................................................................................................9 2.4. Net ecosystem productivity data screening ............................................................11 2.4.1. CO2 data screening stage one ...........................................................................11 2.4.2. NEP data screening stage two ..........................................................................11 2.4.3. Energy-balance closure.....................................................................................13 2.5. Gap-filling and partitioning NEP............................................................................15 2.5.1. Empirical non-linear R models .........................................................................15 2.5.2. Empirical non-linear GPP model......................................................................16 2.6. Michaelis-Menten light-response analysis using daytime NEP measurements .....17 2.7. Leaf area index measurements................................................................................17 2.7.1. Deciduous component ......................................................................................18 2.7.2. Jack pine component ........................................................................................19 2.8. Spectral vegetation indices .....................................................................................22 2.8.1 Normalized difference vegetation index............................................................22 2.8.2. Rectilinear-lens camera images ........................................................................23 2.8.3. Permanent stationary camera imaging setup ....................................................24 2.8.4. Camera imaging at destructive LAI sampling plots .........................................25 2.8.5. Image analysis ..................................................................................................25  iii  3. RESULTS AND DISCUSSION....................................................................................29 3.1. Site climate .............................................................................................................29 3.1.1. Site climate in 2006 .............................................................................................29 3.1.2. Site climate from 2004 to 2007 ...........................................................................31 3.2. Carbon balance .......................................................................................................34 3.2.1. Data screening ..................................................................................................34 3.2.2. Carbon dynamics for 2004 to 2007 ..................................................................36 3.3.2. Annual ecosystem respiration...........................................................................46 3.2.4. Annual gross primary productivity...................................................................47 3.2.5. Annual net ecosystem productivity ..................................................................47 3.3. Vegetation indices ..................................................................................................49 3.3.1. Leaf area index .................................................................................................49 3.3.2. Annual pattern of NDVI at HJP02 ...................................................................50 3.3.3. Growing season RLVI at HJP02 ......................................................................55 3.4. Relationships between Amax and vegetation indices ...............................................58 3.4.1. Relationship between Amax and LAI .................................................................58 3.4.2. Relationships of LAI to NDVI and RLVI ........................................................59 3.5.1. Relationship between Amax and NDVI ..............................................................65 4. CONCLUSIONS ...........................................................................................................67 REFERENCES ..................................................................................................................69 APPENDICES ...................................................................................................................74 Appendix 1: Site photographs from the air and ground ................................................74 Appendix 2: Cumulative footprint contributions...........................................................78 Appendix 3: Volumetric displacement technique..........................................................79 A3.1 Equipment..........................................................................................................80 A3.2 Technique ..........................................................................................................80 A3.3 Total needle area of shoot..................................................................................81 Appendix 4: Median-absolute-deviation outlier removal technique .............................82 Appendix 5: Epson 3100z PhotoPC image capture details ...........................................83 Appendix 6:Growing season nocturnal NEP vs. σw ......................................................86  iv  LIST OF TABLES Table 1. Site information, values quoted regarding vegetation are the 2006 annual mean (± 1 S.D. where available). Climate values are the mean (± 1 S.D.) of all years data (20042007). Albedo represents GS values between 10:00 and 14:00 hrs (CST) and Bowen ratio is calculated from GS daytime values. Soil mineral and carbon content details were taken from Zha et al (2008). ...............................................................................6 Table 2. Measured atmospheric and soil climate variables. ................................................8 Table 3. 2006 16-day LAI sample period average NDVI values estimated using changing contributions of Si components..................................................................................23 Table 4. 5-day LAI sample period climate conditions in 2006 (S1, DOY 130 to 134; S2, DOY 163 to 167; S3, DOY 203 to 207; S4, DOY 227 to 231; S5, DOY 269 to 273). Albedo (So/Si) was calculated between 10:00 and 14:00 hrs (CST). Bowen ratio (H/LE) was estimated from daytime measurements. P is shown as the cumulative daily value measured using the Geonor precipitation gauge. Daily mean values θ are shown, while the remaining climate variables, Ta, Ts, Qi and D are daily-daytime-mean values....31 Table 5. Annual mean climate values for years 2004 to 2007. Albedo (So/Si) was calculated between 10:00 and 14:00 hrs (CST). GS Bowen ratio (H/LE) was estimated from daytime measurements. For 2004 and 2007 P was measured using a TBRG at HJP02 and OJP, respectively and is representative of GS P. For 2005 and 2006, annual total P is shown in brackets, as measured by the Geonor model R100 precipitation gauge. θ is the annual mean of all GS values and all other climate variables (Ta, Ts and D) are the annual mean of daytime values. ................................................................................34 Table 6. EC data coverage (%) during data screening (DS) stages one and two. For DS 2, (1) represents data left after outlier-removal, and (2) represents data left after σwth application for all four years......................................................................................35 Table 7. 2004 to 2007 annual estimates of coefficients from the REL, RLT, GPP and NEP Michaelis-Menten models using un-binned (no bin), and 50-point binned-averaging windows (bin). The model equations used are shown in brackets inside column one (Model). .....................................................................................................................38 Table 8. Michaelis-Menten Amax, α, Rd, r2, RMSE and MAE (for Eq. 10) for 5-day LAI sample periods (SP) at HJP02 during 2006. ..............................................................42  v  Table 9. Michaelis-Menten Amax, α, Rd, r2, RMSE and MAE (for Eq. 10) for GS months during 2006................................................................................................................42 Table 10. EPSON 3100z PhotoPC settings .......................................................................83  vi  LIST OF FIGURES Figure 1. Annual nighttime NEP vs. σw for 2004 and 2006. The points and associated vertical lines represent 100-point bin-averages with their standard errors. The dot-dashed and dashed lines indicate the conservative estimates of σwth, 0.12 and 0.15 m s-1 for 2004 and 2006, respectively, below which data were removed. An increasing number of large –ve values with large standard errors can be seen during 2006 at low σw. See Appendix 6 for GS nighttime NEP vs. σw for 2004 and 2006....................................................14 Figure 2. JP foliage dry weight vs. height. Square symbols represent the estimate of spring (May and June) JP average total foliage dry weight for small, medium and large JP. Upwards pointing triangles represent the estimate of summer (July and August) JP average total foliage dry weight for small, medium and large JP. Downwards pointing triangles represent the estimate of autumn (September) JP average total foliage dry weight for small, medium and large JP. Regression lines have been modelled using the range of heights measured at the site during this study. The coefficient of determination (r2), probability value (P) and linear regression parameters for JP foliage dry weight vs. height (Reg) are shown as well as the slope of the linear regression forced through zero (Zero reg)...................................................................................................................20 Figure 3. Image taken on July 22nd 2006 (DOY 203) showing permanent camera (C2) set up, with EC tower in the background, and the radiometric measurement boom clearly visible. The image also shows what the maximum LAI (0.91 m2 m-2 from measurements) looked like for 2006 at HJP02. The herb component stands out due to the evident senescence (yellowing) occurring...........................................................25 Figure 4. (a) Daily-daytime-mean soil (Ts, at 2 cm depth) and air (Ta) temperatures, and vapour pressure deficit (D). (b) Daily (24-hr) mean volumetric water content (θ1 and θ2 are the average of two probes (South and West of EC tower) at depths 0-15 cm and 1530 cm, respectively) and daily cumulative P at HJP02 in 2006. The upper and lower limits of the lower grey band represent the wilting point for θ1and θ2, respectively. The upper and lower limits of the upper grey band represent the field capacity for θ2, and θ1, respectively. The vertical dotted lines indicate the date LAI samples were taken and are labelled accordingly (S1 to S5). ................................................................................30 Figure 5. Monthly mean climate values for years 2004 to 2007. Ta, H, LE, and D were calculated from their daytime values. For 2004 and 2007 P was measured using a vii  TBRG at HJP02 and OJP, respectively and is representative of GS P. For 2005 and 2006 P was measured using a Geonor model R100 precipitation gauge. Albedo (So/Si) was calculated between 10:00 and 14:00 hrs (CST). Bowen ratio (H/LE) was estimated from daytime measurements. Monthly mean θ was calculated from all available data. ...................................................................................................................................33 Figure 6. Panel 1. 2006 GS daytime binned NEP vs. Qi with the Michaelis-Menten model line fit to half-hourly data. Panel 2. 2006 GS nocturnal binned NEP vs. Ts with REL and RLT model lines fit to the half-hourly data and panel 3. GS daytime binned GPP vs. Qi with Michaelis-Menten model line fit to half-hourly data. Each panel shows the measurements grouped into 50-point windows after sorting values in an ascending order of the independent variable. .............................................................................39 Figure 7. 2006 NEP and R nocturnal, daily(24-hr) and daytime values, and GPP values 40 Figure 8. r2 for Michaelis-Menten light response analysis for years 2004 to 2007 using 16day moving-windows of daytime NEP measurements..............................................45 Figure 9. Cumulative R (upper panel) and GPP (middle panel) and NEP (lower panel) for 2004 to 2007. Annual total values quoted beside the years were calculated using REL. The total RLT values were < 10 g C m-2 yr-1 different from the REL. .........................48 Figure 10. LAI components during the 2006 GS. Values of the JP component were derived from allometric relationships as described in Section 2.7.2. Also shown are the standard errors of the measurement of the deciduous component. Where LAIdc components are very similar in magnitude, values on the x-axis have been adjusted by ± 3 days to separate symbols for clarity. ......................................................................50 Figure 11. Daily mean RNIR and RVIS between 10:00 and 14:00 hrs (CST) during the GS.52 Figure 12. NDVI vs. RVIS from daytime estimates calculated as described in Section 2.8.1 for 2004 to 2007. All r2 are significant at the 5% level and the estimated linear models are shown. The dashed lines represent the mean GS values of NDVI and RVIS for the respective year. ..........................................................................................................53 Figure 13. GS daytime mean NDVI values calculated as specified in Section 2.8.1, with all days with P removed for 2004 to 2007. Values quoted inside the panels are the average of the spring (DOY 100 to 150), summer (DOY 160 to 230) and the GS mean (DOY 100 to 300) NDVI values calculated from the daytime mean values........................54 Figure 14. NDVI mean values for the spring, summer and GS of 2004 to 2007 with the standard errors shown as black horizontal lines within the respective markers. The 95% viii  confidence intervals (95% CI) are also shown with the linear regression line fit for spring NDVI values. ..................................................................................................55 Figure 15. Seasonal pattern of daily mean (all images taken between 10:00 and 14:00 hrs (CST)) values of RLVI at HJP02 in 2006. All days when P occurred were removed, excluding the RLVI quadrat samples taken in September (DOY 271). Daily mean and 5-day LAI sample period RLVI values for the permanent cameras (C2 and C3), and the mean RLVI quadrat values are shown. All five LAI sample periods were centred on the date of LAI sampling .................................................................................................57 Figure 16. Relationship between Amax and LAI at HJP02 for 2006. The r2 and lines of best fit for linear regressions between 5-day (dotted line), 16-day (dashed line), and 30-day (solid) LAI sample period Amax values vs. LAI are presented in the figure panel as well as the 95% CI (dot dashed line) of the 30-day Amax vs. LAI linear regression. .........60 Figure 17. Relationship of LAI to NDVI for 2006. The r2 and line of best fit for linear regressions between LAI and 5-day (dotted line), 16-day (dashed line), and 30-day (solid line) LAI sample periods NDVI mean values are shown. Also shown are the 95% CI for the LAI vs. 5-day NDVI linear regression......................................................61 Figure 18. Relationship of LAI to RLVI for 2006. The r2 and line of best fit drawn for linear regressions between 5-day (dotted line), 16-day (dashed line) LAI sample period RLVI and RLVI quadrat (solid line) mean values are shown. Also shown are the 95% CI for the LAI vs. 5-day LAI sample period RLVI linear regression. .................................63 Figure 19. r2 values and lines of best fit for linear regression analysis between NDVI and RLVI 5-day and 16-day LAI sample period values (dotted line and dashed line, respectively), and, 5-day LAI sample period NDVI vs. RLVI quadrat mean values (solid line). Also shown are the 95% CI for the NDVI 5-day LAI sample period vs. 5day LAI sample period RLVI linear regression. .......................................................64 Figure 20. r2 values and line of best fit shown for linear regressions between monthly Amax and NDVI, including and excluding data from 2007. The July Amax vs. NDVI values have been labelled by year. Also shown are the 95% CI for the Amax vs. NDVI excluding 2007 in the linear regression.....................................................................66 Figure 21. Overhead digital photograph taken before 2004 looking northeast over HJP02, showing stand boundaries, access routes and the windrows of CWD created by postharvest debris management. ......................................................................................74  ix  Figure 22. Overhead digital photograph taken in 2005looking from the east over the site, showing stand boundaries, access routes and windrows of CWD made by post harvest debris management. The site hut is located at the black square, EC tower at the red square and C2 and C3 at the green squares. ..............................................................75 Figure 23. Site overview May 2006. .................................................................................76 Figure 24. Site overview July 2006. ..................................................................................76 Figure 25. Site overview October 2006.............................................................................76 Figure 26. Greyscale digital elevation map, showing small changes in elevation at the site (< 7 m) with the EC tower location displayed (red square). ..........................................77 Figure 27. Annual cumulative daytime footprint. Values beside the lines inside the figure represent the % of the total flux that is measured within the bounds of that area. ....78 Figure 28. Annual cumulative night footprint. Values beside the lines inside the figure represent the % of the total flux that is measured within the bounds of that area. ....78 Figure 29. Example image from C2 image bundle (08/07/2006). Red markers indicate areas where 1 x 50 sample pixels have been taken for further analyses (See Figures 30 and 31)..............................................................................................................................84 Figure 30. 1 x 50 sample pixels taken from the locations of the red markers in Figure 29. The right-hand panel is the exposed soil, the centre panel is the green leaf and the right-hand panel is the shaded area. ............................................................................................84 Figure 31. DN and RLVI values from Figure 30 sample pixels, showing GIth values (0.58) above, which pixels would be classified G. Panel one shows that R has the strongest intensity reading from the soil and senesced vegetation surfaces. Panel three shows that some pixels are classified G as a result of colour casting green onto a light coloured shaded area, where there was no vegetation..............................................................85 Figure 32. GS nighttime NEP vs. σw for 2004 and 2006. The points and associated vertical lines represent 100-point bin-averages with their standard errors. The dot-dashed and dashed lines indicate the conservative estimates of σwth, 0.12 and 0.15 m s-1 for 2004 and 2006, respectively, below which data were removed. An increasing number of large –ve values with large standard errors can be seen during 2006 at low σw.................86  x  LIST OF SYMBOLS, ACRONYMS AND ABBREVIATIONS Amax Photosynthetic capacity estimated from µmol CO2 m-2 s-1 Michaelis-Menten light response function Blue DN B BRDF Bidirectional reflection distribution functions C g Carbon CCD Charge-coupled device CI Confidence intervals Carbon dioxide CO2 CSI Campbell Scientific Incorporated CST year/month/day/hour Central standard time CWD Coarse woody debris kPa Vapour pressure deficit D dc Deciduous vegetation components Shrubs dc1 dc2 Herbs and wildflower vegetation component Grass and sedge vegetation component dc3 Double difference time series di DN Digital number DOY Day of year DS Data screening Dwarf-shrub Bearberry mat E0 Approximately the activation energy divided by the gas constant EC Eddy-covariance EBC Energy-balance closure Buoyant force Fb EC CO2 flux Fc µmol CO2 m-2 s-1 FFP days Frost-free period fov Field of view W m-2; unitless Heat flux density into the surface; Green DN G Gravitational acceleration 9.8 m s-2 g Ground area sampled m2 Ga Greenness index GI GIth Greenness index threshold GPP g C m-2 d-1/g C m-2 Gross primary productivity yr-1 GS Growing season GSL days Growing season length Surface-atmosphere sensible heat flux W m-2 H IRGA Infra-red gas analyser JP Jack pine vegetation component mm Average length of the needles. L LAI m2 m-2 Total leaf area index LAIdc Total deciduous LAI 2 -2 m m Leaf area index of an individual deciduous LAIdc(i) component LAIJP JP LAI m2 m-2 Total deciduous LAI LAItdc xi  LE Ln m MADj MADth MAE Mdj MDW MODIS MS n NDVI NEE NEP NEPdm NEPnm OLS P* PAR Pg  W m-2 cm-2 g  g stems m-2 µmol CO2 m-2 s -1 g C m-2 d-1/g C m-2 yr-1 µmol CO2 m-2 sr-1 µmol CO2 m-2 s -1 mm; unitless W m-2  PLA PLdc(i)) Pt Q  m2 m2  Qi Qo R  µmol photons m-2 s-1 µmol photons m-2 s-1 g C m-2 d-1/g C m-2 yr-1; unitless  W m-2  r1,2,3 r2* Rd  µmol CO2 m-2 s-1  Rdm  µmol CO2 m-2 s -1  REL RH RLT RLVI RMSE Rn  %  W m-2  Surface-atmosphere latent heat flux Total needle area of the shoot Mass Median of the absolute deviation about the median difference Threshold for median absolute deviation Mean absolute error Median sum of differences Mean dry weight Moderate Resolution Imaging Spectroradiometer Number of stems per m2 Number of samples; Number of needles Normalized difference vegetation index Net ecosystem exchange Net ecosystem productivity Daytime measured NEP Nighttime measured NEP Ordinary least squares Precipitation; Probability value Photosynthetically active radiation Total number of pixels in one image classed as green Projected leaf area Deciduous component projected leaf area Total number of pixels in one image Sum of net radiative fluxes from energy sinks or sources other than H Downwelling photosynthetically active radiation Outgoing photosynthetic photon flux density Ecosystem respiration; Red DN R(Ts) annual empirical constants Coefficient of determination Daytime mean R estimated from MichaelisMenten light response function Daytime modelled R from nocturnal empirical R(Ts) model Empirical logistic R model Relative humidity Lloyd and Taylor Arrhenius R model Rectilinear-lens vegetation index Root mean squared error Net radiation  xii  Near-infrared waveband reflectance Respiration at reference temperature Visible waveband reflectance Time varying parameter applied to annual models S.D. Standard deviation Mole mixing ratio sc µmol / mol dry air Half hourly CO2 storage Sc -2 Wm Downwelling solar radiation Si W m-2 Si in the near-infrared wavelengths 750-1000 nm SiNIR W m-2 Si in the visible wavelengths 350-750 nm SiVIS SLF Surface layer fluxes W m-2 Outgoing solar radiation flux density So W m-2 So in the near-infrared wavelengths 750-1000 nm SoNIR W m-2 So in the visible wavelengths 350-750 nm SoVIS 2 Specific leaf area SLA m The temperature when respiration is equal to T0 zero o C Air temperature Ta o C Soil temperature Ts Longitudinal wind velocity u m s-1 Frictional velocity u* Frictional velocity threshold below which NEP m s-1 u*th data is rejected Lateral wind velocity v m-3; cm-3 Volume; Displacement volume of needles V VI Vegetation index Vertical wind velocity m s-1 w Photosynthetic rate estimated from Michaelisα Menten light response function Dry air density ρa CO2 density ρc Half-hourly standard deviation of vertical wind m s-1 σw velocity m s-1 σw threshold below which NEP data are rejected σwth % Volumetric water content θ % Volumetric water content at the 0 to 15-cm depth θ1 % Volumetric water content at the15 to 30-cm θ2 depth * Symbols that are not italics when reported inside MATLAB figure windows RNIR Rref RVIS rw  xiii  ACKNOWLEDGMENTS Many thanks for the financial assistance I received from the Fluxnet Canada Research Network (FCRN), Canadian Carbon Program (CCP), the Boreal Ecosystems Research and Monitoring Sites (BERMS) program and Teaching Assistantships provided by the University of British Columbia (UBC) Faculty of Land and Food Systems. The technical and academic assistance received from the FCRN, CCP, and BERMS management and staff, and all members of the UBC Biometeorology Group was greatly appreciated. I would like to more personally acknowledge the support and assistance I received from Professor Andy Black, Zoran Nesic, Nicholas Grant, and Dominic Lessard, my family and close friends.  xiv  1. INTRODUCTION The boreal forest biome accounts for 17% of the world’s vegetated land surface and 29% (289 million hectares) of the land surface of Canada (979 million hectares) (WWF-Canada, 2008). The carbon (C) balance of this biome is strongly influenced by climate variability, ecosystem type, harvesting, insect damage and natural forest fires. With the largest anthropogenic disturbance mechanism to date being that of global climate change, there is an urgent need to understand potential effects this could have on the enormous C stocks of boreal Canada. The vast area covered by the boreal region of Canada makes use of satellite remote sensing technology an attractive option for helping quantify its C balance. A major problem with satellite-based data like those of MODIS1 is that they are temporally (16- and 8-day averages under cloud free conditions) and spatially (500-1000 m) coarse and not truly representative of the polygonal diversity of disturbed areas (Chen J, 1999; Turner et al., 2003). Airborne data can provide a finer spatial resolution, but cannot easily provide the temporal coverage. To better represent the large number of disturbed areas throughout boreal Canada and the world, methods of validating satellite data need to include established ground-based research sites (Cheng et al., 2006). Several different vegetation indices (VI) have been recognised as vital input parameters for process-based C models to characterize plant canopies and their seasonality, the most prevalent of which are leaf area index (LAI) and the normalized difference vegetation index (NDVI). LAI is defined as one-half of the total green leaf area per unit ground surface area (Chen and Black, 1992). It is very difficult to measure LAI accurately both directly and indirectly, with direct measurements being undesirably destructive and highly time consuming while indirect methods are not always possible for certain vegetation covers, in particular when  1  they are low lying and do not afford a below-canopy measurement of downwelling radiation (Jonckheere et al., 2004; Weiss et al., 2004). NDVI is estimated from normalized-spectralreflectances in the near infrared (750-1000 nm) and visible regions (380-750 nm) and is used to track the health and greenness of vegetation, i.e. it measures the chlorophyll of vegetation (Wilson and Meyers, 2007; Jenkins et al., 2007). Using the eddy-covariance (EC) technique, carbon dioxide (CO2) (Fc), water vapour (E), and sensible heat (H) fluxes can be measured at a single location over an area of several hectares, depending on the prevailing atmospheric turbulence conditions and measurement height (Baldocchi et al., 2003; Moncrieff et al, 1997;). Ground-based measurements of LAI and NDVI can be made around an EC tower to capture their site-specific relationships with flux densities. It is important that a spatial assessment is made accounting for LAI and NDVI heterogeneity (Chen, 1999; Cheng et al., 2006), and that there is some knowledge of the prevailing EC-flux footprint. EC measurements of Fc can be used to calculated the net ecosystem CO2 exchange (NEE) or net ecosystem productivity (NEP), where NEP = -NEE (see Section 2.3). A positive NEP value indicates C uptake by the ecosystem, i.e., it is an ecosystem C sink. The EC technique is widely used to measure NEP and in combination with different modelling techniques to partition NEP into its components of ecosystem respiration (R) and gross primary productivity (GPP), where NEP is GPP minus R (Falge et al., 2001; Reichstein, et al., 2005). The Michaelis-Menten light response function fitted to day-time NEP EC measurements (Falge et al., 2001; Suyker et al., 2001; Griffis et al., 2003) is a commonly used analysis tool for investigating relationships between the photosynthetic capacity (Amax) and VI (Barr et al., 2004; Humphreys et al., 2005; Humphreys et al., 2006; Zang et al., 2006; Richardson et al., 2007).  1  Moderate Resolution Imaging Spectroradiometer (MODIS) mounted on NASA’s terra (EOS AM) and Aqua (EOS PM) satellites (http://modis.gsfc.nasa.gov/about/, 04/05/08).  2  Since 2000, the Canadian Carbon Program (CCP) (previously Fluxnet-Canada) has used EC to measuring CO2, E and H exchanges over forest and peatland sites across Canada. Together with climate measurements, relationships can be obtained and controls on the C balance identified. For the purpose of inter-site comparisons of annual NEP a standard calculation technique was developed for Canadian sites in 2003 (see A1 in Barr et al., 2004). Two main strengths of the CCP output are that (1) flux data from a range of sites of different age classes are collected and studied; and (2) the data represent a continuous temporal record with which to investigate the effects of changing climate. Due to the complexity of the Earth’s ecosystems, there remains uncertainty in both C balance estimates from EC and ground based VI estimates. Three areas of concern common to both are (1) flux errors or uncertainty (2) data screening (3) scaling differences between ‘near’ e.g., tower-based measurements and ‘far’ e.g., satellite remotely-sensed estimates, and, one concern specific to EC C balance estimates is (4) model suitability for gap-filling and partitioning of NEP into R and GPP. One of the Boreal Ecosystem Research and Monitoring Sites (BERMS) research sites, called HJP02 (Harvested Jack Pine – harvested in 2000, scarified in 2002), in northern Saskatchewan provides an invaluable opportunity to investigate the ground-based relationships between C balance components and VI in a recently-harvested boreal jack pine (JP) site. HJP02, which is the youngest stand in a chronosequence of JP stands, represents a growth stage when annual NEP is expected to be negative and half-hourly C fluxes are small. It is, however, a highly dynamic stage important from an environmental, ecological, and forestry economic point of view.  3  The objectives of this thesis research are: 1. To quantify the changes in annual NEP, GPP and R over the first four years after harvesting at HJP02. 2. To determine the relationship between Amax and LAI at HJP02. 3. To determine whether NDVI can be used to estimate LAI at HJP02. 4. To test whether digital cameras can measure vegetation greenness at HJP02 and how this relates to NDVI.  4  2. METHODS 2.1. Site description HJP02 is located approximately 135 km NE of Prince Albert, Saskatchewan near Narrow Hills Provincial Park, Lat: 53° 56’ 38.3” or 53.944737 °N, Long: 104° 38’ 59.1” or 104.649340 °W (tower location). Prior to clearcutting the site was similar to the BERMS OJP stand about 4 km to the southwest (for OJP site description see Baldocchi et al., 1997; Griffis et al., 2003). It was clearcut in August 2000 and then scarified in May 2002. After scarification the remaining coarse woody debris (stems and branches) (CWD) was moved into predominantly north-south windrows ~ 1-m-high. The stand has naturally regenerated to its current status (Table1). In 2006, the windrows had decreased in height to < 50 cm while JP seedling (> 30-cm-tall) and saplings (> 30-cm-tall but under 10 cm diameter at breast height) (hereafter referred to as JP for simplicity) and other low lying and short stature vegetation had colonized the site. The soil is an upland Brunisol originating on a glacial outwash plain. The soil texture is sand, excessively drained, and low in nutrients, with small amounts of organically available C. The surrounding area is covered with different aged stands (5-17 m) of JP, predominantly middle aged to mature directly surrounding the site, thus creating a decrease in the vegetation height where the site is situated (Table 1, mean (± 1 S.D.) JP height 0.57 (± 0.32) m). The clearcut covers an area of ~ 30 ha. The elevation at the tower is ~ 495 m above sea level, though over the clearcut it ranges from 489-498 m (pers. comm. Laura Chasmer (2006) A1).  5  Table 1. Site information, values quoted regarding vegetation are the 2006 annual mean (± 1 S.D. where available). Climate values are the mean (± 1 S.D.) of all years data (2004-2007). Albedo represents GS values between 10:00 and 14:00 hrs (CST) and Bowen ratio is calculated from GS daytime values. Soil mineral and carbon content details were taken from Zha et al (2008). Dominant tree species  Jack pine (Pinus Banksiana)  Jack pine height (m)  0.57 (± 0.32)  JP stand density (this study) (stems ha-1) GS mean JP LAI (this study) (m2 m-2)  16,416-28,000  GS mean Deciduous LAI (this study) (m2 m-2)  0.29 (± 0.02) Bearberry (Arctostaphylos rubra (Rehd. & Wilson) Fern) Various species, commonly asters and primula Various grasses, sedges and reindeer lichen (Cladina mitis (Sandst.) Hustich), green alder (Alnus crispa (Ait.) Pursh) 0.43 (± 0.19)  Dead above ground (t C ha-1)*  8.8 (± 1.3)  Dominant ground cover Herbaceous plants Other vegetation  -1 *  Fine woody debris (t C ha ) LFH  0.14 (± 0.04)  C(t C ha-1) *  0.96 (± 0.7)  -1 *  N(t C ha )  0.01 *  Sand/silt/clay in BC horizon (%)  92/6/2  Mineral soil C to 50 cm (t C ha-1) *  17.5 (± 3.3)  -1 *  Mineral soil N to 50 cm (t N ha ) o  Air temperature ( C) o  *  0.81 0.6 (± 13.9)  Soil temperature ( C)  4.4 (± 10.0)  Total precipitation (mm)  548  Growing season length (days)  118 (± 4)  Growing season albedo  0.16 (± 0.03)  Growing season Bowen ratio  1.5 (± 2.6)  Values taken from Zha et al. (2008).  6  2.2. Atmosphere and soil climate Air temperature (Ta), incoming and outgoing shortwave and longwave radiation, and the photosynthetic photon flux density (PAR) were all measured from the EC tower location (A1) (Figure 3) at the height of 2 m. Soil temperature (Ts) and soil volumetric content (θ) were measured radially around the tower at different depths.  Table 2 lists the climate variables measured in this study at HJP02, the instruments and their location. Half-hourly averages were recorded using a CR23X data logger (Campbell Scientific Inc (CSI), Lincoln, NE) and uploaded to a central site computer every 30 minutes.  7  Table 2. Measured atmospheric and soil climate variables. Variable Air temperature  Precipitation  Instrument Model and Manufacturer Copper-constantan Thermocouple, UBC Biometeorology Group, Vancouver, BC T-200, Geonor Inc., Norway  Instrument Location  Units  On scaffold tower, in aspirated radiation shield at 2 m  °C  50 m south of tower location  mm  Relative humidity  HMP45C, Vaisala Ojy., Helsinki, Finland  On scaffold tower at 2 m  %  Incoming global shortwave radiation Outgoing shortwave radiation  CM 3 in CNR1, Kipp & Zonen B.V., Delft, Netherlands CM 3 in CNR1, Kipp & Zonen B.V., Delft, Netherlands CG 3 in CNR1, Kipp & Zonen B.V., Delft, Netherlands CG 3 in CNR1, Kipp & Zonen B.V., Delft, Netherlands LI-190SA, LI-COR Inc., Lincoln, NE  On southward extended boom attached to scaffold tower, at 2 m On southward extended boom attached to scaffold tower, at 2 m On southward extended boom attached to scaffold tower at 2 m On southward extended boom attached to scaffold tower at 2 m Attached to top of scaffold tower, on south facing railing at 2.6 m On south eastward extended boom at 2 m  W m-2  Incoming longwave radiation Outgoing longwave radiation Incoming PAR Outgoing PAR  LICOR LI-190SA, LICOR Inc., Lincoln, NE  Soil heat flux  HFP01, Hukseflux B.V., Delft, Netherlands  Soil temperature  Copper-constantan Thermocouple, UBC Biometeorology Group, Vancouver, BC Soil volumetric water CS615, Campbell content Scientific Inc (CSI), Logan, UT Soil volumetric water CS615, CSI, Logan, UT content  9 and 10 cm below the ground surface south and north of tower respectively 2 cm below the ground surface south of tower  W m-2 W m-2 W m-2 µmol quanta m-2 s-1 µmol quanta m-2 s-1 W m-2 °C  At a 45 deg angle in pit south % of tower 0-15 cm and 15-30 cm below the soil surface At a 45 deg angle in pit west of % tower 0-15 cm and 15-30 cm below the soil surface 8  2.3. NEE measurement NEE was obtained using: NEE = Fc + S c  (1)  where Sc is rate of change in storage of CO2 in the air column below the EC-flux measurement height (2 m). Fc , the EC CO2 flux, was calculated from the half-hourly block-averaged covariance of the CO2 mole mixing ratio (sc) and vertical velocity (w): Fc = ρ a w' s c '  (2)  where ρa is molar density of dry air and the overbars denote mean values and primes denote fluctuations from the mean (Baldocchi, 2003). This was followed by a 3-axis coordinate rotation, bringing the mean vertical ( w ) and lateral ( v ) velocity components, and the covariances between these components ( w'v' ) to zero (Tanner and Thurtell, 1969). Sc was calculated as follows:  S c = z m ρ a ∆sc / ∆t  (3)  where ∆sc was calculated as the difference between sc of the following and previous half-hour ( ∆t ) at the measurement height (zm). Positive values of NEE indicate CO2 transfer to the atmosphere, while negative values indicate the opposite (Stull, 1998). Kidston (2006) found that the loss of low-frequency components of the covariance when using half-hour averaging at HJP02 was negligible. The high-frequency measurements (20 Hz) of sc were made using an LI-7000 (LI-COR Inc.) closed-path infrared gas analyser (IRGA). The IRGA was enclosed by a temperaturecontrolled housing (TCH) maintained at 37 ± 0.5 °C, located at the top of a scaffold tower (made 9  of 5-cm diameter tubing) near the air sampling inlet. Air was drawn down a short (4.5 m) 4-mm inner-diameter heated Synflex 1300 (Saint-Gobain, Performance Plastics, Wayne, NJ) insulated sample-tube at a rate of 12 L min-1, thus ensuring turbulent air flow within the tube to the IRGA. The inlet of the sample-tube was located 2 m above the ground surface on the northwest side of the tower, directly beside a 3-axis sonic anemometer-thermometer (CSAT3, CSI). The CSAT3 measured the longitudinal (u), v, and w components of the wind vector and Ta at 20 Hz. All highfrequency data from the CSAT3 and LI-7000 were saved on the site computer. Good data quality was ensured by employing proper field calibrations and maintenance, which included:  1. Daily calibrations of CO2 (zero and span) and H2O (zero) and using the values to apply software corrections to the measurements 2. Periodic resetting of the IRGA's zero and span values (every 3-6 months). This improved the software corrections in 1. 3. Regular sample-tube filter changes (1-3 months, depending on the time of the year and air quality) 4. Annual sample-tube replacements to reduce the high-frequency losses due to dirt accumulation on the sample-tube inner surface (Kidston, 2006)  Fluxes were calculated on the site computer and along with climate data were transferred daily, via satellite link, to UBC for initial inspection.  10  2.4. Net ecosystem productivity data screening Two different data quality control stages are described below. CO2 data screening stage one is used to capture instrument errors, while NEP data screening stage two is trying to find a theoretically sound data set for investigating ecosystem C fluxes. 2.4.1. CO2 data screening stage one Incoming data was screened daily to monitor instrument behaviour. Automated cleaning of the data involved removing any half-hours where the sonic anemometer recorded less than 20,000 points and if Fc < –100 or > 100 µmol m-2 s-1. The data were then manually cleaned and points removed where there was evidence of instrument malfunctions, e.g., using response time plots from calibrations. High-frequency data traces of sc, Ta, u, v, and w data were plotted and manually checked when the half-hourly standard deviation of sc, water, H and LE, measured using the IRGA seemed high, additionally correlations between all scalars and the wind vector components at such times were checked. 2.4.2. NEP data screening stage two Investigation of NEP outliers (> 5 or < -4 µmol m-2 s-1) left after data screening stage one and examination of their high-frequency data traces revealed interesting characteristics in the behaviour of the surface-layer motions at the site. Stable atmospheric conditions, resulted in periodic intermittent turbulence and wave events during the evening, night (after dark) and morning daylight hours. Gravity waves are an atmospheric wave motion that do not require a mean wind or shear to form, they are caused by rising or falling air parcels disrupting a statically stable atmosphere, analogous to surface-waves one sees when a pebble is tossed into a pond (Stull, 1998). Examination of high-frequency data obtained at the site since 2004 indicated that wave events began to occur more frequently as the stand aged. They are similar to those reported  11  to occur over a 21-m-tall boreal aspen forest (OA) 70 km southwest of HJP02, which were shown to be generated locally, rather than propagating down from higher altitudes (Lee, et al., 1997). A similar phenomenon was described for wave-like wind fluctuations observed in the stable surface-layer over a 1.5-m-tall sorghum field (Maitani, a) 1984, b) 1989). Further investigation of this phenomenon at HJP02 was beyond the scope of this thesis. NEP outlier-values apparently caused by the occurrence of wave-like wind fluctuations and intermittence were removed using the two following screening steps:  1. An outlier-removal technique similar to the Papale et al (2006) approach, using the median-average-deviation (MAD) of NEP over 13-day periods and a threshold (MADth) below which data were rejected was applied (A4). The outlier-detection algorithm is based on the magnitude of half-hourly values relative to those before and after, within the 13-day data blocks.  2.  u* (friction velocity) and σw (S.D. of vertical wind velocity component) -thresholds (u*th and σwth, respectively), below which data were rejected were compared, with the σwth being preferred as it maintained a greater number of data points while maintaining a similar frequency distribution histogram as the more traditional u*th (see Section 3.2.1).  Data screening stage two techniques are commonly only applied to nocturnal data (potential radiation ≤ 0.01 W m-2). The tendency for the initiation and termination of stable atmospheric conditions to be in the late and early daylight hours, respectively, can result in the NEP outlier events being excluded from the nighttime data but included during the daytime. The data screening stage two techniques were therefore applied to NEP collected during the nighttime and daytime. The chosen MADth was conservatively set at 4. Papale et al. (2006) show that as one 12  increases the chosen MADth the number of points considered bad is decreased as their deviation from the median average difference (MADj) is allowed to be greater (See A4). The σwth value below which all data points were removed was determined for each year as the point at which the variance of the100-point bin-averaged σw vs. NEP begins to decrease (Figure 1). Using this technique, the σwth values determined were 0.12, 0.12, 0.15, and 0.15 m s-1 for 2004 to 2007, respectively. 2.4.3. Energy-balance closure The degree of energy balance closure (EBC) is commonly used to assess the extent to which the magnitude of surface-layer fluxes (SLF) i.e., H and LE, are over/under estimated (Twine et al., 2000, Massman and Lee, 2002; Wilson et al., 2002; Barr et al., 2006). The surface energy balance can be expressed as:  Rn = H + LE + G + Q  (4)  where Rn is the net radiation flux density, G the soil heat flux density, and Q is the sum of fluxes associated with any other energy sinks or sources, or sensible and latent heat storage changes in the air-layer between the EC and Rn measurement height and soil surface, which was assumed to be negligible.  13  1 2004 2006  NEP (µmol m−2 s−1)  0  −1  −2  0  0.1  0.2  0.3 0.4 −1 σ (m s )  0.5  0.6  0.7  w  Figure 1. Annual nighttime NEP vs. σw for 2004 and 2006. The points and associated vertical lines represent 100-point bin-averages with their standard errors. The dot-dashed and dashed lines indicate the conservative estimates of σwth, 0.12 and 0.15 m s-1 for 2004 and 2006, respectively, below which data were removed. An increasing number of large –ve values with large standard errors can be seen during 2006 at low σw. See Appendix 6 for GS nighttime NEP vs. σw for 2004 and 2006. The closure fraction is defined as (Wilson et al., 2002):  EBC =  H + LE Rn − G  (5)  Kidston (2006) demonstrated that SLF values at the site were spatially well behaved and that the lack of EBC was likely due to differences in the footprints of the measured available energy flux (A) (A = Rn – G) and the total surface-layer heat flux (T) (T = H + LE). Consequently, he recommended that no energy balance closure correction should be applied to the SLF at HJP02, so none was done in this study. 14  2.5. Gap-filling and partitioning NEP Post-processing of NEP was performed using modified versions of the FCRN protocol. 2.5.1. Empirical non-linear R models To calculate R, a modified FCRN protocol used the Lloyd and Taylor (1994) Equation 11 for describing the relationship between nocturnal NEP (R) and shallow soil temperature (Ts), rather than the standard FCRN empirical logistic equation (REL) as follows (Barr et al., 2004): R EL = rw  r1 1 + exp r2 (r3 −Ts )  (6)  where rw is a time varying parameter while r1, r2 and r3 are annual empirical constants and Ts is the soil temperature at 2-cm depth in °C. The Lloyd and Taylor (1994) (RLT) Equation 11 was compared allowing all free parameters to be data-set dependent taking the form:  RLT = rw Rref exp  ⎛ 1 1 − E0 ⎜⎜ ⎝ 283.15−T0 Ts −T0  ⎞ ⎟ ⎟ ⎠  (7)  where rw is a time varying parameter, Rref is the mean R at the reference temperature (in this case 283.15 K), E0 is approximately the activation energy divided by the gas constant (8.315 J K-1 mol-1), T0 is the temperature (in Kelvins) at which the predicted R reaches zero and Ts is the soil temperature (in Kelvins) at the 2-cm depth (Lloyd and Taylor, 1994). The Lloyd and Taylor model is more sensitive to temperature increase at higher Ts. In addition to the two previously described R empirical models, exponential and logarithmic empirical models were also compared (Richardson & Hollinger, 2005). Due to the high sensitivity of the latter two models at high Ts (20-40 °C) and the low number of data points available at those temperatures, these functions resulted in unreasonably high half-hourly R values at higher temperatures so were not used further. The annual R(Ts) non-linear regression models were fit using the available daytime and nighttime data from the non-growing season (NGS) NEP and growing season (GS) NEP 15  nighttime measurements (NEPnm). The GS was defined as all times when Ta (2-m height) and Ts (2-cm depth) are both > 0 oC. The annual relationship was used to model R, while the rw parameter was calculated inside a 100-point moving-window with a 20-point step. These windows selected modelled R and corresponding NEPnm and a linear regression, forced through zero, was performed using the model vs. measured values. The window used a fixed number of acceptable data points rather than a fixed period of time. The additional time-varying parameter (rw) was the slope of the regression and used to multiply the modelled values calculated using the annual-model fit for each 20-point-window. Measured values are left untouched and using the adjusted (multiplied by rw) model output, gaps in R during the night and day are filled. 2.5.2. Empirical non-linear GPP model Using the daytime measurements of NEP (NEPdm) and daytime modelled R (Rdm) values from the nocturnal R(Ts) relationship, GPP was calculated during the day throughout the GS as: GPP = NEPdm + Rdm  (8)  From the derived GPP values the annual GS light-response model, a rectangular hyperbolic function (Michaelis-Menten), was fit using incoming PAR (Qi), measured in µmol photons m-2 s1  , as  the forcing variable:  GPP = rw  αAmax Qi αQi + Amax  (9)  where Amax is the photosynthetic capacity (µmol CO2 m-2 s-1) and α is the quantum yield. The same linear regression moving-window approach used when modelling R was used to calculate the time varying parameter rw. Measured values are left untouched and using the adjusted (multiplied by rw) model output, gaps in GPP are filled.  16  All gaps in C balance components were filled using their modelled values. This procedure was carried out for each year of available data (2004 through 2007). Providing FCRN adjusted estimates (REL and RLT) of annual NEP, R and GPP, as well as annual model coefficients. 2.6. Michaelis-Menten light-response analysis using daytime NEP measurements Following Suyker et al (2001) and Griffis et al (2003) Amax was calculated using the MichaelisMenten rectangular hyperbolic function applied to NEPdm (i.e., potential downwelling radiation ≤ 0 W m-2). NEP =  αQi Amax − Rd αQi + Amax  (10)  where Rd, the intercept of the plot of NEPdm vs Qi (See Section 3.2.2), is an estimate of mean daytime R. For all models regression coefficients were fitted using the ordinary-least-squares (OLS) technique, which requires that outliers have been removed leaving normally distributed NEPdm and NEPnm values. The relationships between Amax with climate and with site derived VI were investigated using model fits created from daytime NEP data within 5-day, 16-day and 30-day windows centred on LAI sample dates (i.e., LAI sample period) where possible, and using 5- and 16-day moving-windows in their respective time period steps across the GS. 2.7. Leaf area index measurements Due to the differences in plant structure at HJP02, which can be broadly separated into deciduous (dc) and JP components, two different destructive approaches were used to estimate LAI separately for the dc and JP. It was assumed that all of the dc components had flat leaves with far greater width than thickness allowing for the projected surface area (PSA) to be  17  measured and represent half of the total leaf area, a common assumption for this purpose (Campbell and Norman, 1998). 2.7.1. Deciduous component On May 12th (DOY 132) 2006 a destructive LAI measurement field campaign began. Two destructive samples were taken from each of four 100-m-long transects located in the area surrounding the EC tower (Driese & Reiners, 1997), and considered to be representative of the vegetation within the 90% cumulative-daily-flux footprint (A1). To ensure random sampling along each transect, two random numbers (between 1 and 100) were generated in MATLAB, giving the sampling distance (m) from one end of the transect. A coin was then tossed to decide which side of the transect to sample from, the sampler then moved 2 m perpendicular to the transect, and collected all the vegetation, excluding JP, within a 0.25-m2 quadrat. The samples were packaged in moisture-controlled cool boxes (using two damp sponges inside a 1’x 1’ x 3’ Coleman cooler and four gel ice packs) and sent immediately via courier to the UBC Soil Physics and Biometeorology Laboratory, Vancouver for processing and analysis. In the laboratory the samples were split into four deciduous components, shrubs (dc1), sub-shrubs (dc2), herbaceous plants (herbs) (dc3) and grasses (dc4). Although reindeer lichen was an abundant ground cover at the site, its contribution to the productivity at the site was likely very low (Pratt, 1976) so its LAI contribution was not sampled, but visually assumed to be < 0.1 m2 m-2. Total sub-sample fresh weight from one sample location was regularly < 10 g; however, where dc(i) was large (> 20 g) measurements were made on 10 g sub-samples. The total PSA for each component (PLtdc(i)) in m2 was measured using a planimeter (model LI-3100c, LI-COR, Inc). All samples and sub-samples were then oven dried at 80 °C and where necessary PLtdc(i) was estimated from the dry weight allometric relationship between total dc(i) and the sub-sample of dc(i). LAI for each deciduous component (LAIdc(i)) was estimated using: 18  LAIdc(i) =  PL tdc(i) Ga  (11)  where Ga is the total ground area sampled (2 m2). The total deciduous LAI (LAItdc) was calculated as the sum of all LAIdc(i). No less than 8 samples (2 m2) were taken, as described above, on one day of each month from May through September (DOY 132, 165, 204, 229 and 271, respectively). 2.7.2. Jack pine component On July 23rd (DOY 204) 2006 the height of 78 JP within the EC footprint were measured. They were sorted by height and divided into three evenly numbered groups (small, medium and large height) a mean height value was calculated for each height class (0.24, 0.53 and 0.96 m, respectively). Three representative JP outside the EC footprint were sampled. On August 17th (DOY 229) 2006 four representative JP (one small, two medium, one large) were also sampled. All samples were couriered to the UBC Soil Physics and Biometeorology Laboratory for processing and analysis in the same way as with the monthly dc. Average shoot specific leaf area (SLA) of 16-shoots (three from the upper third, four from the middle third and nine from the bottom third) from one August JP representative of the site mean height (0.57 m) was estimated using the volumetric displacement technique developed by Burdette (1979). The geometric equation used represented a hemi-cylindrical needle shape, like that of the species (Chen et al., 1997) (A3). All JP were defoliated and their foliage/needles oven dried at 80 °C and their dry weight recorded. Linear regression was used to obtain relationships (with and without regression line forced through zero) between foliage dry weight and height (Figure 2).  19  150  Foliage Dry Weight (g)  125  Measured Reg r2 = 0.89; P = 0.00; y = 1.24x -28.87 Zero reg slope = 0.83 Spring Summer Autumn  100  75  50  25  0 0  25  50  75 JP Height (cm)  100  125  150  Figure 2. JP foliage dry weight vs. height. Square symbols represent the estimate of spring (May and June) JP average total foliage dry weight for small, medium and large JP. Upwards pointing triangles represent the estimate of summer (July and August) JP average total foliage dry weight for small, medium and large JP. Downwards pointing triangles represent the estimate of autumn (September) JP average total foliage dry weight for small, medium and large JP. Regression lines have been modelled using the range of heights measured at the site during this study. The coefficient of determination (r2), probability value (P) and linear regression parameters for JP foliage dry weight vs. height (Reg) are shown as well as the slope of the linear regression forced through zero (Zero reg). Using JP density (stems m-2) and height (m) measurements the average number of stems per m2 was made for each height class (S). The total JP LAI (LAIjp) for each deciduous destructive sample date was estimated using the mean dry weight (MDW) in g obtained as described above (Figure 2), S and the specific leaf area (SLA) (total shoot leaf area/dry mass of shoot needles) (See A3.3) in m2 per g (0.019 ± 0.005) for small, medium and large height classes as follows:  20  LAI jps =  S s MDWs SLA 2  (12)  LAI jpm =  S m MDWm SLA 2  (13)  LAI jpl =  S l MDWl SLA 2  (14)  where subscript s, m, and l denote small medium and large height classes, respectively and Ss, Sm, and Sl were 0.5, 0.43 and 0.07 stem m-2. From calculations the total JP LAI contribution for each month was estimated in accordance with Figure 2 using  LAI jp = LAI jps + LAI jpm + LAI jpl  (15)  LAI for each sample date was estimated as the sum of LAItdc and LAIjp. The biggest advantage for this study in using an allometric technique to estimate LAI is that it removed the need for including a clumping factor in LAI estimation (Chen, J., et al 1997), which was clearly very complex between the deciduous and JP components. Furthermore, the low-lying stature of the vegetation, particularly the dwarf-shrub (bearberry) makes LAI estimates from indirect approaches, requiring below canopy radiometric measurements, unreliable; however, in using this direct approach potential errors could arise during the estimation of the foliage area:mass ratio used to convert foliage mass of a tree to area, from the variance of the leaf area and the variance of the stem density in the stand, these measurement errors accumulate and it is difficult to keep the total error under 25% (Chen et al., 1997). The LAI value calculated for each of the five sample dates was used to represent the central mean of 5-day, 16-day and 30-day time periods.  21  2.8. Spectral vegetation indices 2.8.1 Normalized difference vegetation index The normalized difference vegetation index is a measure of chlorophyll and canopy cover. Site scale NDVI estimates were calculated as follows (Wilson & Meyers, 2007): NDVI =  R NIR − RVIS R NIR + RVIS  (16)  where RVIS and RNIR are the reflectivities for the visible (380-750) and near infrared (750-1000 nm) radiation, respectively. RVIS was calculated using: RVIS =  Qo Qi  (17)  where Qo is outgoing PAR (measured by downwards facing LI-190SA). It is expected that as LAI increases a greater proportion of Qi will be absorbed by the plants, decreasing RVIS. To calculate RNIR, the reflectance of near-infrared radiation, the incident solar radiation (Si) measured by the CNR1 was partitioned into its visible or shortwave (SiVIS) and near-infrared (SiNIR) components. SiVIS was assumed to be 45% of Si (Weiss, 1985), i.e., S iVIS = 0.45S i  (18)  S oVIS = RVIS S iVIS  (19)  With increasing LAI expected to reduce RVIS, the effect here would be a reduction in SoVIS. SiNIR was assumed to be 55 % of Si (Weiss, 1985), i.e., S iNIR = 0.55S i  (20)  Then SoNIR can be estimated from S oNIR = S o − S oVIS  (21) 22  and RNIR R NIR =  S oNIR S iNIR  (22)  It is possible that the proportion of the SiNIR and SiVIS components in the measured Si were not constant due to changing atmospheric conditions. The effect on NDVI of changing proportions of the SiNIR and SiVIS components in the measured Si is shown in Table 3. If the SiNIR portion of Si is larger than 45% then the calculated NDVI would be greater. To maintain consistency with Wilson and Meyers (2007), this effect was assumed not to occur and results were left unchanged. Table 3. 2006 16-day LAI sample period average NDVI values estimated using changing contributions of Si components. Portion of Si  1  Sample Period NDVI 2 3 4  5  SiNIR 0.40 SiVIS 0.60  0.40  0.44  0.53  0.48  0.42  SiNIR 0.45 SiVIS 0.55  0.42  0.46  0.56  0.51  0.44  SiNIR 0.50 SiVIS 0.50  0.45  0.49  0.58  0.53  0.46  There was no correction applied to NDVI for the soil, lichen and CWD background (i.e., non zero NDVI for zero vegetation cover). 2.8.2. Rectilinear-lens camera images Three digital cameras (model EPSON 3100z PhotoPC, Epson Ltd, Toronto, Ca) (C1, C2, C3) were deployed on May 12th (DOY 132) to capture interval-images for measuring seasonal greenness fractional ground cover, named the rectilinear-lens vegetation index (RLVI) in this study. Images were taken throughout the GS and within the EC footprint (A2). Another identical camera was used to capture RLVI images of the destructive LAI quadrat samples (RLVI  23  quadrat), for comparison with LAI and NDVI estimates, and to provide more information on spatial variability of site greenness fractional ground cover. 2.8.3. Permanent stationary camera imaging setup The three stationary cameras were mounted horizontally looking downwards inside a modified plastic camera case (model APP-1200, Pelican Inc, Ca) (dimensions 24 x 19 x 10.5-cm)2. The Pelican cases were attached to 3-m-tall steel A-frames, giving the cameras a field of view (fov) of ~ 7.5 m2 at a 2.75 m lens height (Figure 3). Each case had a custom-built white radiation shield to prevent over heating during extremely warm days. The three frames were located over typical site vegetation at the edge of the daytime flux footprint (A1). C2 had six JP in its fov, ranging from 10 to 67 cm in height and 5 to 60 cm in crown diameter, one green alder shrub (62cm-tall with a crown diameter of 80 cm) (in July) and a sparse distribution of herbs and grasses with no bearberry mat. C3 had one JP, 90-cm-tall with a crown diameter of 50 cm, and a significant bearberry mat cover (~ 20%) and a sparse distribution of herbs and grasses. C2 had more exposed soil, lichen and CWD than C3 in its fov. Both cameras were set to capture at least two images between 10:00 and 14:00 hrs (CST) and their memory cards were downloaded at least once a month.  2  Built by Neumann, N. (2002), Climate and Earth Observations Division, Meteorological Service of Canada, 11 Innovation Boulevard, Saskatoon, SK S7N 3H5.  24  Figure 3. Image taken on July 22nd 2006 (DOY 203) showing permanent camera (C2) set up, with EC tower in the background, and the radiometric measurement boom clearly visible. The image also shows what the maximum LAI (0.91 m2 m-2 from measurements) looked like for 2006 at HJP02. The herb component stands out due to the evident senescence (yellowing) occurring. 2.8.4. Camera imaging at destructive LAI sampling plots Photographs of the destructive LAI quadrat sampling plots began on June 11th (DOY 164) and were taken at breast height with the operator’s body facing the sun holding the camera at arm’s length. The small fov of the camera resulted in only the quadrat and some excess being in the image (~ 25% of total image area). All images were taken using the same camera set up (A4). 2.8.5. Image analysis Images bundles were organised by camera, into folders using FastStone Image Viewer 3.5 Freeware (http://www.faststone.org/, 28/07/08). C1 showed excessive camera movement (image shift) within the image capture sequence, due to wind moving the frame and accidental movement during maintenance, so it was removed from the study. C2 and C3 suffered a two25  week power failure in September, some cases of image shift and poor image quality resulting from a foggy lens. On such occasions images were removed leaving 70% (128 days) of all taken by cameras C2 and C3 between 10:00 and 14:00 hrs (CST). Image analysis was carried out using MATLAB 6.5.1. The program was designed to load individual images, assigning them to the concurrent EC flux half-hour. Using a loop command, every pixel in an image had its red, green, and blue (R, G and B, respectively) digital number (DN) light intensities (each ranging in intensity from 0 – 256) extracted with which a simple calculation was performed: GI =  G R+B  (23)  where GI is the greenness index for that pixel, which lies somewhere between 0 and an infinitely large number. During instrument installation on May 12th (DOY 132), a field experiment was conducted where one camera was set to collect an image every half-hour over the course of a full day. All images captured during daylight hours were analysed as described above and individual pixels visually classed as green (i.e., only containing green photosynthetic material) were inspected for their GI. From this a threshold (GIth) above which pixels could be classified as green was determined to be 0.58. A program was written to compare the GI of each pixel with GIth and classify it as green or not green. The program also calculated the total RLVI for each image using: RLVI =  Pg Pt  (24)  where Pg is the total number of pixels in the image classified as green and Pt is the total number of pixels. In this study, RLVI is representative of the area of the ground at two locations (~ 16 m2 total) and semi-random locations throughout the GS (> 4 m-2 per month).  26  Gamon et al. (2006) noted the abiotic factors affecting estimates of NDVI and likely the calculated RLVI. These factors included:  “temporary snow cover, sky conditions (cloudy vs. clear), time of day (due to sun angle and canopy/stand structure), and season (partly due to sun angle), rain, drought, fire, secondary succession (shrub regrowth and annual vegetation growth) and associated changes in species composition and stand structure.”  This study did not aim to solve these particular problems but strived to capitalise on a simple approach for investigating daily NDVI estimates from instruments commonly associated with EC towers. Comments are, however, made with regards to the potential impacts such factors may have had on the estimated spectral VI at HJP02, and where possible these effects have been minimised. Bidirectional reflection distribution functions (BRDF), considered but not corrected for in this study, are used to define how incident light is reflected by opaque surfaces. Changes in BRDF are a function of incident Si angle, the nadir view, and the wavelengths of interest relative to the surface being observed. Disney et al. (2004) and Wilson & Meyers (2007) suggest solar zenith angles should be < 30°, when using radiometric measurements to estimate daily mean NDVI. At HJP02 solar zenith angles varied between 30° and 50° across the GS. Hilker et al. (2008) use semi-empirical kernel driven models for standardising the BRDF of radiometric signals detected by a multi-angular spectral radiometer above a mature coastal Douglas-fir stand on Vancouver Island, BC. Some studies investigating tower-based VI (Jenkins et al., 2007; Richardson, et al., 2007) including this one, only use measurements made between 10:00 and 14:00 hrs (CST) during periods with no precipitation (P), to reduce the need for BRDF  27  corrections and concentrate analyses on seasonal changes largely controlled by the reflective properties of sunlit leaves. For RLVI analyses, the mean of both permanent cameras (C2 and C3) was used except where noted otherwise. Linear regression analysis was used to investigate the relationships between LAI and NDVI and between Amax and NDVI across the 2006 GS at HJP02. Furthermore, linear regression analysis was also used to investigate the relationships between LAI and RLVI and between NDVI and RLVI. All spectral VI were analysed using 5-day, 16-day and 30-day, average values centred on LAI sample dates calculated from daily-average values where possible.  28  3. RESULTS AND DISCUSSION 3.1. Site climate 3.1.1. Site climate in 2006 To coincide with the VI field campaign, the 2006 site climate will first be described in detail followed by a comparison of monthly and annual climate conditions for all years. Figure 4 shows the daily-daytime-mean climate values of Ta, Ts, D, daily meanθ and the cumulative daily P in 2006. Ts at the depth of 2 cm was closely coupled with Ta but was often higher during the GS. Ta rose from the annual minimum of –38 °C on DOY 53 to the annual maximum of 32 °C on DOY 187. As expected with the site’s sandy soil, θ was low and rapidly decreased after P events, the maximum value occurred in spring during snowmelt (0.22 m3 m-3, DOY 103). θ coverage was missing for DOY 172 to 182, a period of very little P (< 3.5 mm), suggesting the decrease in θ1 from 0.14 to 0.10 m3 m-3 during this time would not have been affected by this small P amount. On average during the GS, θ1 was about 0.03 m3 m-3 less than θ2. As a result of it being a highly permeable and rapidly draining soil both θ measurements responded roughly in synchrony with snowmelt and P inputs, with the lower-depth lagging somewhat in its response. The daily maximum P of 53 mm on DOY 249 caused θ1 to recover from a low of 0.10 m3 m-3 to a high of 0.20 m3 m-3. The largest daily-daytime-mean values of D over an extended period of time occurred between DOY 175 and 200 with early afternoon values exceeding 1.8 kPa. The maximum daytime-daily-mean D of 2.4 kPa occurred on DOY 203.  29  4 T a T  20  D  S1  S2  S3  S4  S5 3  o  0  2  -20  1  -40 0.3  0  θ1 θ2  50  P  40  θ (%)  0.2  30 0.1  P (mm)  Temp ( C)  s  D (kPa)  40  20 10  0  0  50  100  150  200  250  300  350  0  DOY  Figure 4. (a) Daily-daytime-mean soil (Ts, at 2 cm depth) and air (Ta) temperatures, and vapour pressure deficit (D). (b) Daily (24-hr) mean volumetric water content (θ1 and θ2 are the average of two probes (South and West of EC tower) at depths 0-15 cm and 15-30 cm, respectively) and daily cumulative P at HJP02 in 2006. The upper and lower limits of the lower grey band represent the wilting point for θ1and θ2, respectively. The upper and lower limits of the upper grey band represent the field capacity for θ2, and θ1, respectively. The vertical dotted lines indicate the date LAI samples were taken and are labelled accordingly (S1 to S5). Table 4 shows the pronounced differences in climate observed during the 5-day LAI sample periods. In 2006, the LAI sample period daily-daytime-mean Ta increased to a maximum (25 °C) in the July sample period (S3), along with high Qi (858 µmol m-2 s-1), D (1.84 kPa), albedo (0.15) and Bowen ratio (1.7). Over the same period, there was little P (1 mm). The dry sand being more reflective than when wet or a reduction in green vegetation cover likely caused the high albedo values while large Si input to areas with little water resulted in high Bowen ratios. In the May and September LAI sample periods (S1 and S5), temperatures were cool (daytime mean Ta < 15 °C). S1 had the highest Qi (877 µmol m-2 s-1), while for S5 it was low 30  (379 µmol m-2 s-1). P was high during the June LAI sample period (S2) (42 mm) and Qi was moderate (463 µmol m-2 s-1). Table 4. 5-day LAI sample period climate conditions in 2006 (S1, DOY 130 to 134; S2, DOY 163 to 167; S3, DOY 203 to 207; S4, DOY 227 to 231; S5, DOY 269 to 273). Albedo (So/Si) was calculated between 10:00 and 14:00 hrs (CST). Bowen ratio (H/LE) was estimated from daytime measurements. P is shown as the cumulative daily value measured using the Geonor precipitation gauge. Daily mean values θ are shown, while the remaining climate variables, Ta, Ts, Qi and D are daily-daytime-mean values. LAI Sample Period S1 S2 S3 S4 S5  Qi Bowen θ1 θ2 Ta (oC) Ts (oC) P (mm) (µmol D (kPa) 3 -3 3 -3 Albedo ratio (m m ) (m m ) m-2 s-1) 11 13 0 877 0.89 0.13 0.15 0.15 3.7 14 16 42 463 0.47 0.14 0.17 0.13 0.9 25 26 1 858 1.84 0.10 0.13 0.15 1.7 19 20 0 801 1.23 0.11 0.15 0.15 1.9 9 9 13 379 0.32 0.14 0.16 0.14 0.8  3.1.2. Site climate from 2004 to 2007 Figure 5 shows the monthly-daytime-mean values of Ta, H, LE, D, Bowen ratio and albedo for each year, monthly cumulative P and mean θ, while Table 5 shows the annual climate conditions including GS length (GSL) and the frost-free period (FFP) for each year. Generally, 2006 was a very warm and relatively dry year in comparison to the normals from 1961 to 1990 for nearby Waskesiu, SK, (http://berms.ccrp.ec.gc.ca/Data/Overview/graphs/OJP_2006.JPG, 2008-03-25). The mean (± 1 S.D.) GSL for all 4 years was 185 (± 9). The 2006 GSL was 179 days, 2005 and 2007 had the longest GSL, both being 192 days. 2004 and 2007 both had a late starts to the GS due to very cold temperatures in April (Figure 5). In 2007, May was generally warm however there was a late frost when nocturnal temperatures dropped below –5 °C on two days DOY 140 and 143. The mean (± 1 S.D.) FFP of the 4 years was 145 (± 12) days. The 2006 FFP (145 days) lasted from DOY 131 to 276. Unlike the GS, the FFP in 2005 was shorter than in 2007, which had the longest FFP (162 days). The large FFP and GSL in 2007 were due to a warm autumn, delaying 31  the onset of winter (Figure 5). HJP02 FFP were very similar to the general area averages from 1978-1995 (http://atlas.nrcan.gc.ca/site/english/maps, 22/07/08). 2006 had the highest daytime mean Qi (538 µmol m-2 s-1) and the highest daytime mean temperatures (6.8 and 8.8 °C for Ta and Ts, respectively) ( Table 5). GS mean θ in 2006 was the lowest (0.1 m3 m-3) of the 4 years, because of the warm weather and relatively dry summer, with 93 mm less P during the GS than in 2005 ( Table 5). 2007 was nearly a 1 °C cooler on average than 2006, starting the GS with a very cool spring and ending with a slightly cooler but extended autumn (Figure 5). The GS in 2007 was punctuated with a very warm July (Figure 5). The Bowen ratio during July, August, and September was higher in 2006 than 2007, as was the albedo. The Bowen ratio was also lowest in 2007 compared to other years. The steady reduction in the GS Bowen ratio and monthly-daytime average LE (Figure 5) indicates that LE as a proportion of the energy balance increased from 2004 to 2007.  32  160 20 T (oC) a  0 -10 -20  2004 2005 2006 2007  P (mm)  120  10  80 40 0  1 2 3 4 5 6 7 8 9 101112  1 2 3 4 5 6 7 8 9 101112  120 60 LE (W m-2)  H (W m-2)  80 40 0 -40  D (kPa)  -3  θ (m m )  3  0.05  1.0 0.5 0  1 2 3 4 5 6 7 8 9 101112  10  0.9  5  0.8 Albedo  0 -5 -10  1 2 3 4 5 6 7 8 9 101112  0.6 0.4 0.2  -15 -20  1 2 3 4 5 6 7 8 9 101112  1.5  0.10  Bowen ratio  20 0  1 2 3 4 5 6 7 8 9 101112  0.15  0  40  1 2 3 4 5 6 7 8 9 101112 Month  0  1 2 3 4 5 6 7 8 9 101112 Month  Figure 5. Monthly mean climate values for years 2004 to 2007. Ta, H, LE, and D were calculated from their daytime values. For 2004 and 2007 P was measured using a TBRG at HJP02 and OJP, respectively and is representative of GS P. For 2005 and 2006 P was measured using a Geonor model R100 precipitation gauge. Albedo (So/Si) was calculated between 10:00 and 14:00 hrs (CST). Bowen ratio (H/LE) was estimated from daytime measurements. Monthly mean θ was calculated from all available data.  33  Table 5. Annual mean climate values for years 2004 to 2007. Albedo (So/Si) was calculated between 10:00 and 14:00 hrs (CST). GS Bowen ratio (H/LE) was estimated from daytime measurements. For 2004 and 2007 P was measured using a TBRG at HJP02 and OJP, respectively and is representative of GS P. For 2005 and 2006, annual total P is shown in brackets, as measured by the Geonor model R100 precipitation gauge. θ is the annual mean of all GS values and all other climate variables (Ta, Ts and D) are the annual mean of daytime values. Year  Ta ( C)  Ts ( C)  GS (days)  FFP (days)  2004 2005 2006 2007  4.0 5.8 6.8 6.2  7.0 7.8 8.8 7.8  176 192 179 192  132 143 145 162  o  o  Qi GS GS D θ1 Bowen (µmol (kPa) (m3 m-3) -2 -1 Albedo Ratio m s ) 490 0.46 0.17 529 0.20 2.0 513(580) 0.49 0.14 526 0.20 2.1 420(548) 0.60 0.10 538 0.20 1.6 327 0.63 0.11 525 0.20 1.5 P (mm)  3.2. Carbon balance This section describes the HJP02 carbon balance from 2004 to 2007. It begins with a description of the impact of data screening. Then it reviews the annual model relationships for all years and Michaelis-Menten output parameters in more detail (5-day LAI sample period, 5-day movingwindow and monthly analyses) for 2006 and the r2 values at 5-day time scales across all years using a 5-day moving-window. The section is concluded with a discussion concerning the annual cumulative R, GPP and NEP calculated using daily values. 3.2.1. Data screening Table 6 shows the effects on data % coverage from the different data screening (DS) stages and techniques. Evidence from data point counts showed that the majority of missing raw NEE data were the result of problems with (1) the data acquisition (DAQ) uptime (i.e., the time that the system was actually on) and (2) incomplete high-frequency sampling (< 20,000 data points sampled per half-hour). Causes for this data loss were largely due to instrument maintenance, including daily calibrations (~ 2 %), and adverse weather conditions (~ 5 %), e.g., hoar frost in the middle of February 2006 lasting one week. Distortion of the CSAT signal due to frost and P on its transducers, causing errors to accumulate, was the most common cause of data loss. In 34  2006 much of this data loss was during winter months when fluxes were low in magnitude and diurnal variability (winter NEP mean (± 1 S.D.) = -0.05 (± 0.00) µmol m-2 s-1), indicating a net C loss but having little impact on annual totals. In 2005 two weeks of data were lost due to instrument failure in June, a time when fluxes are more variable (monthly NEP mean (± 1 S.D.) = - 0.60 (± 1.18) µmol m-2 s-1). On average, a week of data was lost each year and this combined with daily loss due to calibration amounted to about 5% data loss. 2006 had the highest fraction of high-quality flux data, but the highest fraction of data removed during data screening stage two (D2) (19, 25, 37 and 36% for 2004 to 2007, respectively). The application of the σwth filter removed more data than the outlier-removal technique. Both, however, removed more than a full week of data. Additionally there was an increasing trend from 2004 to 2007 in the amount of data removed by the σwth application (12, 19, 28 and 28% for 2004 to 2007, respectively). This was largely due to an increasing number of negative outliers in the nocturnal NEP data associated with inadequate turbulent mixing (see Section 2.4.2, Figure 1). Table 6. EC data coverage (%) during data screening (DS) stages one and two. For DS 2, (1) represents data left after outlier-removal, and (2) represents data left after σwth application for all four years. DS Component 2004 2005 2006 2007 1 DAQ uptime (>20,000 pts) 90 92 93 92 83 82 91 88 sc 88 91 93 92 w Fc 82 82 91 87 NEE 80 81 90 86 2 (1) NEP 73 75 81 78 (2) NEP 61 56 53 50 Histogram plots (not shown) for all years showed that nocturnal NEP (i.e., -R) data was negatively skewed regardless of the data screening technique used in DS 2. The 2006 nocturnal NEP measurements had a negatively skewed non-normal distribution with low median values of -0.0379 and -0.0657 µmol m-2 s-1 and low mean (± 1 S.D.) values of -0.39 (± 0.72) and -0.43 (± 35  0.93) µmol m2 s-1 using σwth and u*th, filters, respectively. The application of the σwth filter removed a greater proportion of these negative fluxes, which lowered both the median and mean values. All daytime data had normal distributions (not shown), making them statistically more suitable for use in comparative analysis. That nocturnal NEP outliers were negative suggests a venting of a CO2 previously built up beneath the sample-tube air sample inlet, which is not accounted for in the ∆Sc calculation (Section 2.3), is responsible. This may indicate a changing turbulence regime with stand age and other site characteristics (Baldocchi, 2003; Twine et al., 2000). The treatment of these points in this study adequately removed the majority of non-biologically realistic fluxes; however, this resulted in the loss of a large number of nocturnal measurements. For the purpose of carbon budgeting there is a real need for further research to determine how to best treat such points when comparing stands of different ages at different locations. This could be through detailed studies of CO2 advection beneath the sensor level or by defining a suitable filtering technique for all ecosystem types. 3.2.2. Carbon dynamics for 2004 to 2007 Table 7 shows the annual estimates of coefficients from the REL, RLT, GPP and NEP equations (Equation (Eq) 6, 7, 9, and 10, respectively) using un-binned (no bin), and 50-point binnedaveraging windows (bin). Figure 6 compares the 2006 raw-data line fits to the binned data points. Both the REL and RLT annual models based on half-hourly measurements explained a large amount of the variation in R during 2006 (r2 = 0.69, root mean squared error (RMSE) = 0.36). The GPP and NEP Michaelis-Menten annual light-response models based on half-hourly measurements poorly fit the data in 2006 (r2 = 0.22 and 0.21, and RMSE = 1.07 and 0.75 respectively). However, as part of an increasing trend with stand age, it fit better in 2007 (r2 = 0.27 and RMSE = 0.85), indicative of increasing photosynthesis resulting from increases in LAI. 36  Bin-averaging nocturnal measured NEP by Ts, and, daytime measured NEP and GPP by Qi using 100-point bins greatly improved all model fits (r2 > 0.77 and RMSE < 0.36), providing confidence in these relationships. Between the two R models, REL had the highest r2 of 0.95 and lowest RMSE of 0.13, the output of which is hereafter referred to as R.  37  Table 7. 2004 to 2007 annual estimates of coefficients from the REL, RLT, GPP and NEP Michaelis-Menten models using un-binned (no bin), and 50-point binned-averaging windows (bin). The model equations used are shown in brackets inside column one (Model). r2 RMSE r1 r2 r3 2.04 0.23 9.37 0.49 0.50 2.36 0.21 10.94 0.89 0.17 1.92 0.21 10.60 0.68 0.29 1.96 0.21 10.85 0.94 0.11 2.00 0.25 9.19 0.69 0.36 1.99 0.25 9.18 0.95 0.13 1.95 0.29 7.96 0.67 0.40 1.94 0.29 7.88 0.95 0.13 Rref* T0 (°C) r2 RMSE E0 2004 no 1.05 70.98 255.83 0.48 0.50 bin 1.04 96.88 252.18 0.90 0.17 2005 no 0.89 62.51 256.92 0.68 0.29 RLT bin 0.90 60.09 257.53 0.94 0.11 (Eq 7) 2006 no 1.08 41.44 261.28 0.69 0.36 bin 1.08 40.44 261.45 0.94 0.13 2007 no 1.15 34.37 263.00 0.67 0.41 bin 1.15 33.76 263.04 0.94 0.14 2 Amax* r RMSE α 2004 no 1.27 0.00 0.14 0.74 bin 2.09 0.00 0.64 0.29 2005 no 1.46 0.00 0.15 0.82 GS bin 1.66 0.00 0.61 0.30 GPP 2006 no 2.66 0.01 0.22 1.07 (Eq 9) bin 3.20 0.01 0.77 0.36 2007 no 3.09 0.01 0.27 1.15 bin 3.49 0.01 0.80 0.39 Amax* Rd* r2 RMSE α 2004 no 0.87 0.01 1.24 0.07 0.65 bin 0.87 0.01 1.24 0.39 0.22 2005 no 0.99 0.01 0.99 0.08 0.61 GS bin 0.99 0.01 0.99 0.36 0.24 NEP 2006 no 1.87 0.01 1.18 0.21 0.75 (Eq 10) bin 1.87 0.01 1.18 0.71 0.25 2007 no 2.44 0.01 1.31 0.27 0.85 bin 2.44 0.01 1.31 0.76 0.3 *units are in µmol CO2 m-2 s-1 a MAE is the mean absolute error of the model fit b n is the number of sampled points available for model line fitting. Model Year Binning 2004 no bin 2005 no REL bin (Eq 6) 2006 no bin 2007 no bin  MAE*a 0.25 0.03 0.09 0.01 0.13 0.02 0.16 0.02 MAE*a 0.25 0.03 0.08 0.01 0.13 0.02 0.17 0.02 MAE*a 0.55 0.09 0.68 0.09 1.15 0.13 1.32 0.15 MAE*a 0.42 0.05 0.38 0.06 0.56 0.06 0.73 0.09  nb 4181 3706 2966 2995 nb 4181 3706 2966 2995 nb 4842 4513 4748 4517 nb 4842 4513 4748 4517  38  -2  -1  Flux (µmol CO2 m s )  1.0  3  3 R EL R  0.5  LT  2  0.0 -0.5  2  1  1  -1.0 NEP -1.5  0 500 1000 1500 Q (µmol photons m-2 s-1) i  0  R 0  10 T (oC) s  20  0  GPP  0 500 1000 1500 Q (µmol photons m-2 s-1) i  Figure 6. Panel 1. 2006 GS daytime binned NEP vs. Qi with the Michaelis-Menten model line fit to half-hourly data. Panel 2. 2006 GS nocturnal binned NEP vs. Ts with REL and RLT model lines fit to the half-hourly data and panel 3. GS daytime binned GPP vs. Qi with Michaelis-Menten model line fit to half-hourly data. Each panel shows the measurements grouped into 50-point windows after sorting values in an ascending order of the independent variable. Figure 7 shows the mean daytime NEP, R and GPP; and nocturnal mean NEP and R for 2006. Maximum GS nocturnal R was ~ 0.6 g C m-2 d-1 and maximum non-GS nocturnal R was ~ 0.2 g C m-2 d-1. Maximum GS daytime R was ~ 1.8 g C m-2 d-1 occurring on DOY 181 at the same time as maximum GPP, which was ~ 2.5 g C m-2 d-1. Maximum daytime GS NEP was ~ 1.0 g C m-2 d-1. The maximum GS nocturnal R, non-GS nocturnal R and GS daytime NEP were approximately 100, 70 and 20%, respectively, of the corresponding values reported by Griffis et al. (2003) for OJP during a warm dry year (2000). Daytime NEP was intermittently a very small C sink (< 0.8 g C m-2 d-1) from DOY 120 onwards, with the sink strengthening around DOY 170, then decreasing through August and ending around DOY 280, shortly after the first frost on DOY 276. R increased more quickly after the GS onset than GPP and decreased more slowly after its GS maximum. Also R showed less daily variability than GPP, particularly in the middle of the GS (DOY 170 to 250). The GS distribution of daily (24 hr) R is controlled by the larger daytime R contribution. Lower LAI values early in the GS likely explain the  39  NEP  0 Nocturnal Daily Daytime  -2  2  -2  R  -1  -1  Flux (g C m d )  -2  -1  Flux (g C m d )  1  1  2  GPP  -2  -1  Flux (g C m d )  0  1  0  0  50  100  150  200 DOY  250  300  350  Figure 7. 2006 NEP and R nocturnal, daily(24-hr) and daytime values, and GPP values slower increase of GPP relative to R, Zha et al. (2008) found that GPP of boreal JP stands is more sensitive to a reduction in θ, which may, in combination with decreasing LAI, explain why GPP decreased more rapidly than R near the end of the GS. Variability in GPP is to be expected as its main forcing variable Qi can be highly variable within and between days, depending on cloudiness. Daily NEP was largely controlled by daily R resulting in HJP02 being a very small net sink (< 0.5 g C m-2 d-1) on only five days in 2006 (DOY 182 to 184, 240 and 260). 5-day moving-window Michaelis-Menten analyses (not shown but similar to the analyses shown in Fig. 6) indicated a maximum Rd of approximately 2.04 µmol m-2 s-1 (i.e., ~ 1.3 g C m-2 d-1) between DOY 165 and 180, generally in good agreement, for the same period, with the daytime R maximum of approximately 1.6 g C m-2 d-1 (Figure 7).  40  The biological and physical processes involved in controlling ecosystem C fluxes differ at a range of time scales (Hartley et al., 2006; Reichstein et al., 2007). The sensitivity of R and GPP to Ts and Qi, respectively, was largely determined by the annual relationship and slightly modified by their respective rw. The mean (± 1 S.D.) values of GS R and GPP rw were 0.96 (± 0.02) and 0.58 (± 0.02), respectively. This indicates that modelled R and GPP values were overestimated and other variables not accounted for in the annual models reduced the magnitude of the fluxes. A closer look at the monthly mean R rw during 2006 (not shown) showed that it was, generally, < 1 in April, May and August and > 1 during June, July and September. This indicates that at cooler temperatures (< 15 °C), when there was little P the model overestimated R, while during warmer low P conditions the annual model underestimated R. Hartley et al. (2006) found that over the course of weeks temperature is the biggest influence on R, while short-term changes in photosynthesis play a major role in affecting rates of R. Janssens et al. (2001) showed that productivity is the strongest determinant of soil and ecosystem respiration at European EC flux sites by constraining the supply of photosynthates to the roots. Table 8 shows the Amax, α, Rd, r2, RMSE and MAE values when using Equation 10 with daytime NEP data from the 5-day LAI sample periods. The cool wet S2 (June) period had the lowest Amax and Rd (0.92 and 1.03 µmol m-2 s-1, respectively), while S3 (July) with little P (1 mm), high daily mean Qi (858 µmol m2 s-1), and warm daily mean Ta (25°C) had the maximum Amax and Rd (4.25 and 3.78 µmol m2 s-1, respectively).  41  Table 8. Michaelis-Menten Amax, α, Rd, r2, RMSE and MAE (for Eq. 10) for 5-day LAI sample periods (SP) at HJP02 during 2006. Rd* SP Amax* α 1 1.25 0.04 1.21 2 0.92 0.01 1.03 3 4.25 0.11 3.78 4 2.74 0.01 1.89 5 1.20 0.01 0.66 Mean 2.18 0.02 1.52 ± 1 S.D. 1.46 0.02 1.16 -2 -1 *units are in µmol CO2 m s  r2 0.03 0.07 0.14 0.31 0.44  RMSE MAE* 0.37 0.14 0.76 0.58 0.78 0.60 0.66 0.44 0.28 0.08  Table 9 shows the Amax, α, Rd, r2, RMSE and MAE values when using Equation 10 during the GS months. Similar to the 5-day LAI sample period analysis, the maximum Amax, and Rd occurred in July (3.68 and 2.22 µmol m2 s-1, respectively). The difference between the two was likely due to the monthly estimate representing the cumulative climate effects on productivity. Table 9. Michaelis-Menten Amax, α, Rd, r2, RMSE and MAE (for Eq. 10) for GS months during 2006. Rd* Month Amax* α May 1.29 0.02 1.20 June 2.58 0.01 1.80 July 3.68 0.02 2.22 Aug 2.90 0.01 1.68 Sept 1.55 0.01 0.74 Mean 2.40 0.01 1.53 ± 1 S.D. 0.98 0.01 0.57 -2 -1 * units are in µmol CO2 m s  r2 0.23 0.22 0.35 0.41 0.31  RMSE MAE* 0.43 0.18 0.97 0.94 0.89 0.80 0.69 0.48 0.52 0.27  RMSE and MAE vs DOY (not shown, but deducible from Tables 8 and 9) had roughly positive bell-shaped distributions, reflecting the greater variability in NEP magnitude near the middle of the GS. Radiation interception, nutrient uptake and water relations are all processes that affect the NEP of forest stands and climate plays a significant part in controlling each of these processes (Landsberg and Gower, 1997). Five-day moving-window Michaelis-Menten (Eq. 10) analyses (not shown), analysed with concurrent climate data during 2006 DOY 100-300 (the approximate GS period) showed a weak positive correlation between Amax and Ta (r2 = 0.46, p < 0.05) (monthly across the 2006 GS r2 = 0.96, p < 0.05), as did Rd vs Ts (r2 = 0.66, p < 0.05) 42  (monthly across the 2006 GS r2 = 0.51, p < 0.05). Ta better explains the variation in Rd than Ts at monthly time scales (r2 = 0.55, p < 0.05), while variation in nocturnal NEP using REL was better explained by Ts. There was a very weak negative correlation between Amax with θ1 and θ2 (r2 = 0.21, p < 0.05; r2 = 0.25, p < 0.05, respectively) and very weak negative correlation for Rd with  θ2 (r2 = 0.13, p < 0.05). An observation that negative correlations between both Amax and Rd and θ can be explained by wet conditions being associated with cooler temperatures (Jassal et al., 2005) is supported by the weak negative correlation found between Ta and θ2 (r2 = 0.34, p < 0.05). There was a weak positive correlation between 5-day moving-window Amax and D (r2 = 0.31, p < 0.05) and a weak positive correlation between Rd and D (r2 = 0.46, p < 0.05), up to D of 1.5 kPa, indicating that as evaporative demand increases at the site, with increasing temperature, there would be an increase in photosynthetic activity resulting in an increase in both autotrophic and heterotrophic R. Comparison of the results for the S3 5-day LAI sample period climate in Table 4 with the Michaelis-Menten output parameters in Table 8 shows that the largest Amax and Rd were associated with the greatest D of 1.84 kPa, indicating that the aforementioned relationship is realistic up to this value. The weak to moderate positive correlations between the Michaelis-Menten parameters, Amax and Rd, and climate variables Ta, and D, and their weak negative correlation to θ, suggests that high temperatures and low water supply (down to 30 cm) were not strong limiting factors on site productivity during 2006. Rather cooler temperatures occurring at times of high θ, coinciding with low Qi was limiting to productivity. Results show September to have the largest r2 and the RMSE to be greatest for S3 and July (Tables 8 and 9) Michaelis-Menten analyses. From the 5-day LAI sample period analyses and monthly analyses, Amax was greater at the end of the GS compared to the beginning, excluding the month of September in Table 9. The increasing r2 values that can be observed for all Michaelis-Menten  43  analysis from the beginning to end of the 2006 GS could be explained by the increasing stand LAI and its photosynthetic interaction with Qi. The interannual variability of the annual values of Amax obtained using the GPP and NEP Michaelis-Menten equations were similar (Table 7) i.e., there was an increasing trend with stand age, and are consistent with results shown in Humphreys et al. (2005). Figure 8 shows that the value of r2, when fitting the NEP Michaelis-Menten model using 16-day moving-windows of daytime NEP data, increased annually. In 2006 and 2007 it was generally high throughout the GS, though slightly larger at the end, indicating an increase in the importance of Qi in controlling daytime and hence, annual NEP. Low r2 values are to be expected during earlier years when LAI would have been lower. The Amax values found here (mean value for 2004 and 2005, 0.93 µmol m-2 s-1), using Eq. 10 are nearly twice as large as the mean value reported for HJP02 in Zha et al. (2008), for 2004 and 2005 (0.48 µmol m-2 s-1). One possible cause for this could be the effect of the missing data in 2005 during late June when photosynthesis would potentially be at its greatest, combined with the moving-window technique used by Zha et al. (2008) when fitting the Amax parameter. Another possible cause could be the different data screening stage two techniques used, as these are known to have a significant impact on annual data sets (Moffat, et al., 2007; Reichstein et al., 2005). Differences between techniques include the outlier-removal technique used in this study but not by Zha et al. (2008); Also they used a 0.1 m s-1 u*th instead of the 0.12 m s-1 σwth used in this study.  44  2005  0.75  0.75 2  1.00  0.50  r  r  2  2004 1.00  0.25 0.00  0.50 0.25  0  0.00  50 100 150 200 250 300 350  0  2007 1.00  0.75  0.75 2  1.00  0.50  r  r  2  2006  0.25 0.00  50 100 150 200 250 300 350  0.50 0.25  0  50 100 150 200 250 300 350 DOY  0.00  0  50 100 150 200 250 300 350 DOY  Figure 8. r2 for Michaelis-Menten light response analysis for years 2004 to 2007 using 16-day moving-windows of daytime NEP measurements. The importance of these differences are strictly because Zha et al. (2008) applied stage two data screening techniques to only nighttime data, unlike this study, which applied stage two data screening to daytime and nighttime data. Also Zha et al. (2008) defined the GS differently, preferring to define it as the time period between the first and last occurrence of three consecutive days when daytime daily mean GPP exceeded 5% of average summertime (MayOctober) Amax, rather than the traditional FCRN technique used in this study (GS when Ta and Ts > 0 °C). The latter may be considered a less likely cause for the difference as it would increase Amax estimates by removing the influence of lower daytime NEP values in the GS shoulder months (May and October). When data screening was only performed during the night for this study, the mean Amax calculated for 2004 to 2005 became 0.76 µmol m2 s-2 (reduced by 18%). This is because data-screening NEP during the GS days tends to remove negative and low NEP 45  values. Furthermore this indicates that the choice of technique used for outlier removal within data screening stage two should be carefully considered for Amax analyses, but does not invalidate the removal of unrealistic NEP outliers at HJP02. The Amax and Rd values are < 24 % of all 22 values cited in Zang et al. (2006), for different and older forest stands, but are approximately the same size as those estimated by Humphreys et al. (2005) over the first year and ~ 50 % of those estimated for the last two of three years after harvesting a coastal Douglas-fir stand on Vancouver Island. 3.3.2. Annual ecosystem respiration R significantly increased in late March to early April (DOY 80 to 110), at the onset of the GS in all years and rapidly decreased near the end of September (~ DOY 275) (Figure 9). The lowest annual R total occurred in 2005 (221 g C m-2 yr-1); however, the value in 2004 was very similar (231 g C m-2 yr-1). There was a marked increase (47 g C m-2 yr-1) in R from 2005 to 2006, with 2006 being the highest of the 4 years though only ~ 5 g C m-2 yr-1 greater than in 2007 total (Figure 9). There was ~ 10 g C m-2 yr-1 difference between the annual totals of the two tested models (Eq. 6 and 7) with the RLT model giving consistently slightly higher values (~ 10 g C m-2 yr-1). The annual cumulative R traces show very similar seasonal dynamics, as all are largely constrained by Ts alone and the seasonal patterns in Ts were similar between years. R was significantly higher in 2006 and 2007 than in 2004 and 2005. During 2004 R increased slightly more rapidly than in 2005 around DOY 230. 2004 compared to 2005 had higher monthly θ and lower Ta at this time (late August). Zha et al. (2008) found that in 2004 and 2005 the soil water supply limited R late in the GS. During 2006, R increased more rapidly than in 2007 from DOY 170 onwards due to warmer temperatures in June; however, it caught up to 2006 values due to the warm July and long GS (Figure 5). The annual cumulative mean (± 1 S.D.) R for the four years was 246 (± 23) g C m-2 yr-1. Rd calculated using Eq.10 was largest in 2007 (Table 7) which 46  contradicts the fact that the largest cumulative R occurred in 2006, but which could be explained by the increased LAI having a greater impact on autotrophic R, rather than the moderate temperature difference favouring greater R in 2006. 3.2.4. Annual gross primary productivity Annual GPP estimates increased with stand age with 2004 having the smallest (77 g C m-2 yr-1) and 2007 the largest (200 g C m-2 yr-1) values (Figure 9). There was a ~10 g C m-2 yr-1 difference between the annual GPP derived from the REL, and RLT, with the RLT consistently giving higher values. The increases in annual GPP were consistent with increases in annual Amax. Seasonal patterns in the GPP accumulation indicate water supply limitation occurred in mid July and early September of 2004, and mid August 2005, and are shown as short plateaus in accumulation in Figure 9. After DOY 225, cumulative GPP in 2006 increased more slowly than in 2007. The Bowen ratio and D were both lower in 2007, suggesting water was less limited throughout the GS in 2007, in addition to a warmer October in 2007. The annual cumulative mean (± 1 S.D.) GPP for the four years was 144 (± 56) g C m-2 yr-1. 3.2.5. Annual net ecosystem productivity Annual NEP, like GPP, increased with stand age (Figure 9), with the net C loss in 2007 (-63 g C m-2 yr-1) being only 42% of that in 2004 (-153 g C m-2 yr-1). The ratio of annual R to annual GPP was 2.66, 1.85, 1.46 and 1.30 for 2004 to 2007, respectively; showing that as the stand aged GPP became a proportionally larger component of the stand C balance. Zha et al. (2008) calculated a R/GPP ratio of 2.51 for 2004 to 2005 at HJP02, and 0.95 for HJP94, a nearby JP stand 10-yearsold in 2004. The annual NEP at the site showed considerable interannual variability, with 20042007 mean (± 1 S.D.) of -101 (± 38) g C m-2 yr-1.  47  240 180 120 60  R g C m-2 yr-1 2004 231 2005 221 2006 268 2007 263  -2  Flux (g C m )  0 240 180 120 60  -2  -1  -2  -1  GPP g C m yr 2004 78 2005 121 2006 179 2007 200  0 -40 NEP g C m yr -80 -120 -160  2004 2005 2006 2007 50  -153 -100 -89 -63 100  150  200 DOY  250  300  350  Figure 9. Cumulative R (upper panel) and GPP (middle panel) and NEP (lower panel) for 2004 to 2007. Annual total values quoted beside the years were calculated using REL. The total RLT values were < 10 g C m-2 yr-1 different from the REL. The NEP totals reported here compare well with estimates by Zha et al. (2008) using a similar technique for HJP02 during 2004 (-152 (± 11) g C m-2 yr-1), but are less negative by 23 g C m-2 yr-1 for 2005. Compared to other stand types of similar age (Humphreys et al., 2005; Humphreys et al., 2006; Kowalski et al., 2003) NEP, R and GPP found here are small, further supporting the Humphreys et al. (2005) conclusion that variability in NEP between different recently disturbed stands is large and depends upon site specific characteristics like microclimate, nutrient supply, management practices, differences between partitioning aboveand-below-ground respiration, and vegetation succession rates and composition.  48  3.3. Vegetation indices The results presented in this section describe LAI measured at HJP02 over the 2006 GS and how they compare with other relevant studies in Canada. Following this is a discussion of the results of the RNIR and RVIS estimates and their impact on NDVI daily and seasonal estimation for years 2004 to 2007. Finally the relationships of RNIR and RVIS with RLVI, and the seasonal pattern of RLVI are discussed. 3.3.1. Leaf area index Figure 10 shows the progression of the LAI and its components due to different vegetation types during the GS in 2006. LAI of the deciduous components (LAIdc), i.e., shrub, sub-shrub (predominantly bearberry), grasses and herbs, varied markedly during the GS. As expected mean (± 1 S.D.) JP values were relatively constant (0.29 (± 0.02) m2 m-2) (Chen, 2003). LAI shortly after snowmelt was mainly due to the JP contribution (0.27 m-2 m2). LAI increased most during spring and early summer (DOY 132 to 164) from 0.40 to 0.72 m2 m-2 reaching a maximum of 0.91 m2 m-2 in mid-summer (DOY 204), then slowly decreasing to 0.85 m2 m-2 in August (DOY 229) and 0.43 m2 m-2 in September (DOY 271). The spring to mid-summer LAI increase (0.51 m2 m-2) was largely due to the development of the sub-shrub (bearberry) (0.1 to 0.39 m2 m-2). LAI remained relatively constant over the summer period (DOY 164 to 229), with the main change being the increase in the herb component (0.07 m2 m-2). The sub-shrub contributed the most to mid-summer maximum LAI (43%), followed by JP (34%), with grasses, shrubs and herb components making up the remainder (23%). The reduction in LAI between the mid-summer peak and September (- 0.48 m2 m-2) was mainly due to the reduction in cover by the sub-shrub component (- 0.29 m2 m-2). The standard errors of the measurements indicate that there was a significant difference between the spring and autumn minimums and the summer maximum total LAI estimates. The mean (± 1 S.D.) GS LAI for 2006 was 0.67 (± 0.24) m2 m-2, significantly 49  lower than the LAI range (1-4 m2 m-2) for older JP stands described by Chen et al. (1997). In their study the youngest stand investigated was SYJP (now called HJP75 as part of the BERMS JP chronosequence, a middle aged, 34-year-old jack pine stand), located within 5 km of HJP02. The LAI values also compare well with Amiro et al. (2005) optically-based estimates for a young JP stand (HJP94), also within 5 km of HJP02, which was 0.6-1 m2 m-2 during 2001 and 2002 (7-years-old). These results suggest that LAI development associated with stand regeneration after harvesting boreal JP stands in this area is relatively similar for a particular age.  1.2  1.0  Jack pine Grass Shrub Sub-shrub Herb Total  LAI (m 2 m-2)  0.8  0.6  0.4  0.2  0.0  50  100  150  200 DOY  250  300  350  Figure 10. LAI components during the 2006 GS. Values of the JP component were derived from allometric relationships as described in Section 2.7.2. Also shown are the standard errors of the measurement of the deciduous component. Where LAIdc components are very similar in magnitude, values on the x-axis have been adjusted by ± 3 days to separate symbols for clarity. 3.3.2. Annual pattern of NDVI at HJP02 Figure 11 shows the daily mean HJP02 reflectivities in the near-infrared and visible wavebands in solar radiation that were used to calculate daily NDVI values during the GS of years 2004 to 50  2007. Significant downward spikes in both traces are associated with the dark soil, lichen and CWD background after rainfall. RNIR was more variable between days as it is more affected by changing soil moisture content than RVIS (Campbell and Norman, 1998) in addition to changing plant pigmentation properties. During all years RVIS decreased in the early to middle of the GS as the perennial and annual deciduous vegetation (e.g., dwarf-shrub, shrubs, herbs, grasses) leafed out at the site. It then increased slowly towards the end of the GS to a value similar to that at the beginning of the GS (0.07-0.09) as the deciduous vegetation senesced in response to dropping temperatures and the associated reflective properties changed, i.e., an increase in carotenoid and anthocyanin pigmentation. During the years 2004 to 2006 there was a reduction in RNIR at the beginning and end of the GS (~DOY 150-170 and 250-270, respectively). The end of the 2007 GS was different from other years, i.e., RNIR remained low up to DOY 300, possibly because the frequency of the occurrence of P (though not magnitude) was greater than previous years thus holding down RNIR. Furthermore, a decrease in RNIR was observed during 2006 from DOY 210 to 235 when there was a high frequency of P events (See Section 3.1.1 Figure 4) and a measured decrease in the bearberry contribution to total LAI (See Section 3.3.1 Figure 10). Linear regression analyses of daily mean values concluded that NDVI was negatively correlated to RVIS which explained more of the variance in NDVI (Figure 12) than RNIR. NDVI had an insignificantly positive correlation with RNIR for 2004 to 2005 and a significant but weak correlation (r2 = 0.25, 0.16, P < 0.05) for 2006 to 2007, respectively. Clearly much of the variation in NDVI was not accounted for, and could potentially be the result of BRDF effects in addition to effects of a darkened soil, lichen and CWD background not being filtered out when removing just the half hour when P was occurring. After removing all days when there was P, all years showed increases in r2 for the relationships between NDVI and RVIS (r2 = 0.51, 0.65, 0.81 and 0.65, P < 0.05 for 2004 to 2007, respectively). In addition, this increased the positive correlation between NDVI and RNIR particularly in the last two years (r2 = 0.18, 0.07, 0.41 and 51  0.41, and P < 0.05 for 2004 to 2007, respectively). 2006 was consistently the year with the strongest linear relationship between NDVI and RVIS, indicating that during that GS vegetation cover was particularly abundant and green. 2004  2005  0.3  0.3  Reflectivity  R  R  NIR  0.2  0.2  0.1  0.1  0 100  150  200  250  300  0 100  150  2006 0.3  0.2  0.2  0.1  0.1  150  200 DOY  250  300  250  300  2007  0.3  0 100  200  VIS  250  300  0 100  150  200 DOY  Figure 11. Daily mean RNIR and RVIS between 10:00 and 14:00 hrs (CST) during the GS. Using the available RVIS and RNIR half-hour values on days with no P to estimate daily NDVI, showed a similar seasonal pattern of NDVI across the GS at HJP02 for years 2004 to 2006; unlike other years, 2007 was different at the end of the GS as the low RNIR ultimately caused a decrease in the estimated NDVI (Figure 13). DOY 100 to 150 is referred to as spring here for simplicity and represents a time when JP was usually the largest vegetation component contributing to total LAI. As such, it serves as a useful time period for investigating the combined NDVI of the largely JP, soil, lichen and CWD background components. The mean NDVI (± 1 S.D.) value for this time period increased each year from 0.407 (± 0.020) in spring 2004 to 0.436 (± 0.016) in spring 2007 (Figure 13). 52  GS 2005  0.6  0.6  0.4  0.4  NDVI  NDVI  GS 2004  0.2  0.2 2  2  r = 0.40; y = -2.32x +0.61 0  0  0.05  0.10 R  r = 0.55; y = -3.12x +0.70 0  0.15  0  0.05  vis  0.15  vis  GS 2007  0.6  0.6  0.4  0.4  NDVI  NDVI  GS 2006  0.2  0.2 2  2  r = 0.62; y = -4.00x +0.77 0  0.10 R  0  0.05  0.10 R  vis  r = 0.57; y = -3.66x +0.72 0.15  0  0  0.05  0.10 R  0.15  vis  Figure 12. NDVI vs. RVIS from daytime estimates calculated as described in Section 2.8.1 for 2004 to 2007. All r2 are significant at the 5% level and the estimated linear models are shown. The dashed lines represent the mean GS values of NDVI and RVIS for the respective year. After spring, there was a rapid summer greening (~DOY 150 to 200) followed by gradual senescence (~DOY 200 to 300). The summer (DOY 160 to 230) mean NDVI (± 1 S.D.) value had an increasing trend from 2004 to 2006, 0.451 (± 0.050), 0.506 (± 0.045) and 0.540 (± 0.037), respectively. 2007 had a summer mean NDVI (± 1 S.D.) of 0.483 (± 0.032), which was lower than the values for 2005 and 2006. The summer mean NDVI was largely representative of the deciduous component LAI contribution in combination with the JP and soil, lichen and CWD background components and its magnitude was likely controlled annually by different early to mid-GS climate conditions. The cold start to the GS in 2007 and late frost in May, as well as the very high July daytime temperatures could have had adverse effects on the deciduous component GS development.  53  GS 2005  0.6  0.6  0.4  0.4  0.2  NDVI  NDVI  GS 2004  DOY 100 to 150 DOY 160 to 230 0.406 0.460  0 100  150  200  250  0.2  GS 0.436  DOY 100 to 150 DOY 160 to 230 0.415 0.507  0 100  300  150  0.6  0.6  0.4  0.4  0.2  DOY 100 to 150 DOY 160 to 230 0.428 0.540  0 100  150  200 DOY  250  300  GS 2007  NDVI  NDVI  GS 2006  200  GS 0.462  250  GS 0.483  300  0.2  DOY 100 to 150 DOY 160 to 230 0.437 0.486  0 100  150  200 DOY  250  GS 0.443  300  Figure 13. GS daytime mean NDVI values calculated as specified in Section 2.8.1, with all days with P removed for 2004 to 2007. Values quoted inside the panels are the average of the spring (DOY 100 to 150), summer (DOY 160 to 230) and the GS mean (DOY 100 to 300) NDVI values calculated from the daytime mean values. Figure 14 shows the aforementioned spring, summer and GS mean NDVI values and clearly illustrates the lack of any linear annual increase in NDVI for all except the spring mean NDVI values (r2 = 0.97, P < 0.05). The GS distributions of daily mean NDVI values at the site were unlike those reported for a 17-m-tall ponderosa pine forest (BH) by Wilson and Meyers (2007); HJP02, however, shares similarities in shape but not the timing of GS greening up and senescence with the deciduous forest (MO) and grassland (SF) sites.  54  0.55  Mean spring NDVI Mean summer NDVI Mean GS NDVI 2 Spring 2004 to 2007 r = 0.99; P = 0.00; y = 0.01x 0.39 Spring linear regression 95% CI  NDVI  0.50  0.45  0.40  0.35  2004  2005  2006  2007  Year  Figure 14. NDVI mean values for the spring, summer and GS of 2004 to 2007 with the standard errors shown as black horizontal lines within the respective markers. The 95% confidence intervals (95% CI) are also shown with the linear regression line fit for spring NDVI values. 3.3.3. Growing season RLVI at HJP02 Figure 15 shows the daily seasonal development of the stationary RLVI (C2 and C3), 5-day LAI sample period RLVI (calculated from stationary cameras (mean of C2 and C3) interval images over the five days centered on LAI sampling) and the RLVI quadrat (the average RLVI from images taken on each LAI sampling date from >7 quadrats) values for comparison. A clear increase was apparent in all RLVI estimates from an early GS low of ~ 0.1 (DOY 132) to the maximum of ~ 0.37 that was reached around DOY 175 to 204, followed by a decrease to 0.15 by DOY 277. Differences between C2 and C3 were associated with differences in fractional cover of the various vegetation and soil, lichen and CWD background components in their respective fov. Linear regression analysis showed that, like NDVI, RLVI was negatively correlated to RVIS (r2 = 76 and 71, P < 0.00, for C2 and C3, respectively), which explained more of the variance in 55  both NDVI (Figure 12) and RLVI than RNIR (r2 = 0.49 and 0.43, P < 0.00 for C2 and C3, respectively). The larger variability during the mid-summer period for both cameras, C2 in particular, was attributed to differences in light conditions striking the vegetation and soil surface and differences in structure of the vegetation due to changes in wind speed and direction. When shade created by a green surface, i.e., foliage, onto the soil surface or lichen increased, there were more pixels classified as green due to color casting as a result of inaccurate auto-white balancing by the digital cameras. This condition was more apparent for C2 as there was taller vegetation, i.e., more shading, within the fov compared to that of C3 (See A4). Increasing the GIth reduced this effect, but at the cost of not classifying some green pixels as green, e.g., highly exposed sunlit leaves. The soil, lichen and CWD when in direct sunlight (wet or dry) showed higher DN intensities of R. The RLVI quadrat mean (± 1 S.D.) values showed that there was significant spatial variability in RLVI at HJP02 (Figure 15), which is indicative of a patchy vegetation cover, typical of a naturally regenerating forest stand after clearcut harvesting. Some of the differences between the 5-day LAI sample period RLVI and the RLVI quadrat mean values were due to the RLVI quadrat values being biased towards the sky conditions and soil, lichen and CWD background moisture conditions at the time of sampling. Regardless of these problems, 5-day LAI sample period RLVI and the RLVI quadrat mean values showed that there was a significant difference between GS maximum and minimum values.  56  0.5 5-day RLVI quadrat C2 C3 0.4  RLVI  0.3  0.2  0.1  0  50  100  150  200 DOY  250  300  350  Figure 15. Seasonal pattern of daily mean (all images taken between 10:00 and 14:00 hrs (CST)) values of RLVI at HJP02 in 2006. All days when P occurred were removed, excluding the RLVI quadrat samples taken in September (DOY 271). Daily mean and 5-day LAI sample period RLVI values for the permanent cameras (C2 and C3), and the mean RLVI quadrat values are shown. All five LAI sample periods were centred on the date of LAI sampling When making NDVI measurements over areas where bare soil is exposed, the effect of absorption by the soil of Si in the visible and near infrared wavebands can be taken into account, resulting in a marginal but consistent difference to NDVI estimates (Campbell and Norman, 1998). Although not investigated further in this study it is worth mentioning, that the RLVI approach has potential use as an indirect measurement of the sunlit leaf area, a necessary measurement for correcting NDVI over areas where bare soil is exposed (Campbell and Norman, 1998). Improvements would, however, have to be made with regards to camera calibration, image stability, BDRF correction and the spatial sampling technique used for both NDVI and RLVI estimates at this particular site.  57  3.4. Relationships between Amax and vegetation indices In this section, the results from previous sections are used and the results of linear regression analyses between Amax and LAI, LAI and NDVI, LAI and RLVI, NDVI and RLVI, and finally Amax and NDVI are discussed in view of climate effects. 3.4.1. Relationship between Amax and LAI Figure 16 shows that there was a strong positive correlation between LAI and Amax (r2 = 0.91) for the 30-day LAI sample periods, a moderate correlation (r2 = 0.74) for the16-day LAI sample periods, and an insignificant correlation (r2 = 0.58) for the 5-day LAI sample periods. S2 (June) had low 5-day and 16-day LAI sample period Amax values (0.92 and 1.94 µmol m-2 s-1, respectively) corresponding to the third largest LAI (0.73 m2 m-2). S2 was a particularly wet period with low Qi and cool temperatures. In contrast S3 and S4 were warm periods and showed similarly high 5-day LAI sampling period Amax values (2.83 and 2.74 µmol m-2 s-1, respectively). Linear regression analysis indicated that Ta explained a moderate amount of the variability in Amax at both the 5-day and 30-day time scales (r2 = 0.76 and 0.78, respectively, with P < 0.05). These results and the weak positive correlation (0.46) between 5-day running-window calculated Amax and Ta described in Section 3.2.2 suggest that at time scales less than one month the relationship between LAI and climate, particularly Ta, is most important in controlling Amax at HJP02. The smaller r2 for the relationship between 5-day LAI sample period Amax and LAI suggests that LAI integrates weather conditions over periods longer than 5-days. These results are consistent with those for more mature temperate and boreal forest ecosystems (Falge et al., 2002; Griffis et al., 2003). Humphreys et al. (2005) determined the relationship between Amax derived using Equation 9 and available monthly measurements of LAI over three years after harvesting a Douglas-fir stand to be Amax = 3.8LAI + 0.6  (r2 = 0.75)  (25) 58  In this study, a similar relationship was derived from 30-day LAI sample period Amax and LAI during the 2006 GS to be: Amax = 3.47LAI + 0.24  (r2 = 0.91, P < 0.05)  (26)  The slope in Equation 25 is greater than that in Equation 26, indicating that photosynthetic activity recovers more rapidly in a coastal Douglas-fir stand than it does in a harvested boreal jack pine stand, or, that the measures of LAI differed. The larger r2 value suggests that more of the variance in Amax is explained by LAI at HJP02 than for the regenerating Douglas-fir stand, of which some of the samples were collected during a fairly severe water stressed period in 2003. The exclusion of the June values of both the 5-day and 16-day LAI sample period values from the 95% CI estimated for the monthly values in Figure 16 suggests that weather, in particular Ta, is likely to be more important during the early GS than other factors in controlling Amax at HJP02. 3.4.2. Relationships of LAI to NDVI and RLVI In this study, r2 values for the relationships between LAI and NDVI for the 5-, 16- and 30-day sample periods decreased with increasing window width (r2 = 0.86, 0.85 and 0.81, respectively, and P < 0.05) (Figure 17). Observing that the August LAI value falls outside the 95% CI suggests that vegetations pigment properties may have changed (greater carotenoid to chlorophyll ratio), resulting in greater RVIS and less RNIR  59  4 2  5-day r = 0.58; P = 0.13; y = 2.96x - 0.18 2 16-day r = 0.74; P = 0.06; y = 3.14x + 0.25 2 30-day r = 0.91; P = 0.01; y = 3.47x + 0.24 30-day linear regression 95% CI  -2  -1  (µmol m s )  3  16-day  2  max  June  A  30-day  1 5-day  0  0  0.2  0.4  0.6 2  0.8  1  -2  LAI (m m )  Figure 16. Relationship between Amax and LAI at HJP02 for 2006. The r2 and lines of best fit for linear regressions between 5-day (dotted line), 16-day (dashed line), and 30-day (solid) LAI sample period Amax values vs. LAI are presented in the figure panel as well as the 95% CI (dot dashed line) of the 30-day Amax vs. LAI linear regression. as was described for Figure 11 (Section 3.3.2.). The large r2 values do, however, strongly suggest that NDVI well represents LAI at HJP02 up to an LAI of 0.91 m2 m-2. Extrapolating the linear fits to an LAI of zero gives a mean (± 1 S.D.) NDVI of 0.349 (± 0.015). This is a reasonable estimate of NDVI in 2002 immediately after scarification, when the necessary instruments for measurements were not installed. Pontailler et al. (2003) use a negative exponential function to describe the relationship between LAI and NDVI at seven contrasting plots up to an LAI of 3.5 m2 m-2, finding that NDVI remained constant at 0.8 when LAI exceeded ~ 3 m2 m-2. From those results it can also be seen that the relationship was linear from 0 to 1 m2 m-2, which is consistent with the results shown here (Figure 17).  60  1.2  August  1.0  2  -2  LAI (m m )  0.8  0.6  30-day  5-day 16-day  0.4  0.2  0.0  2  5-day r = 0.86; P = 0.02; y = 5.43x -1.96 5-day linear regression 95% CI 2 16-day r = 0.85; P = 0.02; y = 5.32x -1.88 2 30-day r = 0.81; P = 0.04; y = 4.46x -1.48 0.45  0.50  0.55  NDVI  Figure 17. Relationship of LAI to NDVI for 2006. The r2 and line of best fit for linear regressions between LAI and 5-day (dotted line), 16-day (dashed line), and 30-day (solid line) LAI sample periods NDVI mean values are shown. Also shown are the 95% CI for the LAI vs. 5-day NDVI linear regression. There was a strong positive correlation between LAI and 5-day LAI sample period mean RLVI (r2 = 0.92, P < 0.05), and an insignificant positive correlation between LAI with 16day LAI sample period mean RLVI and between LAI and destructive quadrat RLVI (r2 = 0.93 and 0.74, P = 0.14 and 0.14, respectively) (30-day analyses was not possible due to a lack of available data in September) (Figure 18). These results suggest that RLVI well represented LAI at the site when a short averaging period was performed (using daily mean RLVI estimates to reduce problems due to BRDF effects) and when P days were removed. In the absence of direct measurements of LAI at site inception (i.e. in 2002) one would reasonably force this relationship through zero because when the fractional cover of vegetation is zero, there is no vegetation  61  greenness. In combination with the NDVI results, these results suggest that changes in deciduous vegetation LAI during the GS had a significant influence on vegetation indices at HJP02. Examining the relationship between daily mean NDVI and daily mean RLVI values for each camera showed that there was a significant positive correlation between the two indices (r2 = 0.78 and 0.71, for C2 and C3, respectively with P < 0.00). Figure 19 shows that when examining the relationship between NDVI and RLVI calculated as the average of the values from the two cameras, there was a marginally insignificant positive correlation between 5-day (r2 = 0.73, P = 0.06) and 16-day (r2 = 0.72, P = 0.07) LAI sample period NDVI values and corresponding RLVI, and an insignificant positive correlation between 5-day LAI sample period NDVI values and the mean RLVI quadrat values (r2 = 83, P = 0.09). The large r2 value for daily mean NDVI vs. RLVI suggests that RLVI well represented NDVI and greenness at the site, at the daily time scale, up to an LAI of 0.91 m2 m-2, and could be used as a surrogate for NDVI below this LAI.  62  1.2  1.0  2  -2  LAI (m m )  0.8  0.6 5-day RLVI quadrat  0.4  0.2  0.0  2  5-day r = 0.92; P = 0.01; y = 1.99x + 0.14 5-day linear regression 95% CI 2 16-day r = 0.93; P = 0.14; y = 2.14x + 0.14 2 RLVI quadrat r = 0.74; P = 0.14; y = 1.92x + 0.20  16-day  0  0.2  0.4  0.6  RLVI  Figure 18. Relationship of LAI to RLVI for 2006. The r2 and line of best fit drawn for linear regressions between 5-day (dotted line), 16-day (dashed line) LAI sample period RLVI and RLVI quadrat (solid line) mean values are shown. Also shown are the 95% CI for the LAI vs. 5day LAI sample period RLVI linear regression. The mean (± 1 S.D.) value of the y-intercept from the NDVI vs. RLVI linear regressions was 0.403 (± 0.006) (Figure 19), which is significantly greater than the bare soil, lichen and CWD NDVI calculated from the LAI vs. NDVI relationships (Figure 17) and is almost identical to the calculated mean NDVI for the spring of 2004 (0.407) (Section 3.3.2).  63  0.6  NDVI  0.5  5-day RLVI quadrat  0.4  16-day  2  5-day r = 0.73; P = 0.06; y = 0.30x + 0.41 5-day linear regression 95% CI 2 16-day r = 0.72; P = 0.07; y = 0.33x + 0.40 2 RLVI quadrat r = 0.83; P = 0.09; y = 0.33x + 0.40 0.3  0  0.1  0.2  0.3  0.4  0.5  RLVI  Figure 19. r2 values and lines of best fit for linear regression analysis between NDVI and RLVI 5-day and 16-day LAI sample period values (dotted line and dashed line, respectively), and, 5day LAI sample period NDVI vs. RLVI quadrat mean values (solid line). Also shown are the 95% CI for the NDVI 5-day LAI sample period vs. 5-day LAI sample period RLVI linear regression. Using the relationship (LAI = 5.43NDVI – 1.96), derived from 2006 GS 5-day LAI sample period NDVI values (Figure 17), to estimate the largely JP LAI in the spring of each GS gave 0.25, 0.29, 0.38 (compared to 0.40 m2 m-2 from measurements) and 0.41 m2 m-2 for 2004 to 2007, respectively, indicating a detectable increase in LAI over the four years. Using the relationship to estimate LAI values for previous years from available summer mean NDVI, gave 0.49, 0.79, 0.98 (compared to 0.89 m2 m-2 from measurements), and 0.66 m2 m-2 for 2004 to 2007, respectively. Finally, when using the relationship to estimate LAI values for previous years from available GS mean NDVI, gave 0.36, 0.51, 0.62 (compared to 0.67 m2 m-2 from measurements), and 0.39 m2 m-2 for 2004 to 2007, respectively. It is possible that the LAI values are overestimated as the relationship is biased towards values in the one GS (i.e. 2006) when the 64  deciduous component became a very large and a major component of the total LAI from early to middle of the GS, then decreasingly so from the middle to the end of the GS. Considering the lack of a significant linear relationship for annual summer and GS mean NDVI values (See Section 3.3.2 Figure 14), it is suggested that the technique only be used to estimate mean spring LAI, while the deciduous component remains a large part of the GS LAI. 3.5.1. Relationship between Amax and NDVI There was a very weak positive correlation between monthly averages of Amax and NDVI for all years (r2 = 0.26, P = 0.08), and strong positive correlation between monthly average Amax and NDVI for the three years 2004 to 2006 (r2 = 0.87, P < 0.05) (Figure 20). The 2007 monthly Amax values for July, August and September were unexpectedly high, and deviated from the strong linear relationship from previous years (Figure 20). Previous climate analysis (Section 3.2.2) and Michaelis-Menten analysis (see Section 3.2.2, Figure 8) suggests that this was likely caused by warm weather coupled with an adequate water supply and the increased JP LAI at the end of the 2007 GS. Additionally, the NDVI estimates were more sensitive to changes in the deciduous component, which appeared to be lower in 2007. Previous years LAI throughout the GS would, however, have been largely deciduous as the JP slowly established at the site after scarification.  65  2  6  Including 2007 r = 0.26; P = 0.06; y = 15.65x - 5.08 2 Excluding 2007 r = 0.87; P = 0.03; y = 20.50x - 7.89 Excluding 2007 linear regression 95% CI  2007  5  -2  -1  (µmol m s )  2006 4 2005  A  max  3  2 2004 1  0 0.30  0.35  0.40  0.45 NDVI  0.50  0.55  0.60  Figure 20. r2 values and line of best fit shown for linear regressions between monthly Amax and NDVI, including and excluding data from 2007. The July Amax vs. NDVI values have been labelled by year. Also shown are the 95% CI for the Amax vs. NDVI excluding 2007 in the linear regression. Extrapolating the two linear fits through zero on the y-axis in Figure 20 (i.e. the xintercept) shows that at zero Amax, NDVI would be 0.391 for the 2004 to 2006, and 0.321 for the 2004 to 2007. When using the 9 NDVI estimates of bare soil, lichen and CWD background, from intercepts derived from linear regression analyses (Section 3.4.2 and this Section 3.5.1), the results suggest that bare soil, lichen and CWD background mean NDVI (± S.D.) would have been approximately 0.371(± 0.033) immediately after scarification.  66  4. CONCLUSIONS 1. During the first four years following clearcut harvesting, NEP increased from -153 to -63 g m-2 yr-1, mainly as a result of GPP increasing significantly from 78 to 200 g m-2 yr-1, while R increased only slightly from 231 to 263 g m-2 yr-1. With this increase in GPP, there was also an increasingly strong annual relationship (increasing r2) of both GPP and daytime NEP to Qi. 2. Direct leaf area measurement (i.e., destructive sampling with scanning or volumetric displacement for deciduous and coniferous vegetation, respectively) indicated that the GS mean LAI at HJP02 in 2006 was 0.67 ± 0.24 with the jack pine component accounting for 34% ± 7%. The normal distribution of LAI during the GS was largely controlled by leaf emergence and senescence of the deciduous component (i.e., dwarf-shrubs, shrubs, herbs, grasses). 3. LAI was linearly related to NDVI and RLVI, although the relationship with RLVI was only significant at the 5 % level for 5-day sample period averaging from the mean value of the permanent cameras. The significant r2 between NDVI and RLVI (from daily values) indicated that RLVI well represents site greenness and can be used as a surrogate for NDVI up to this stage in stand development. Furthermore, the strong relationship found between LAI and NDVI permitted the estimation of mean spring LAI for years without direct measurement. Mean spring NDVI values were 0.41, 0.42, 0.43, and 0.44 for 2004 to 2007, respectively and the corresponding estimates of mean spring LAI were 0.25, 0.29, 0.38 (compared to measured 0.40 m2 m-2) and 0.41 m2 m-2. 4. Estimated mean spring NDVI and LAI values for all years suggested that the increasing trend found in GPP and Amax was the result of the developing JP and to a lesser degree the deciduous component.  67  5. At the monthly time scale during the 2006 GS, LAI was the main factor affecting standscale Amax; however, it was also strongly controlled by climate and, in particular, Ta at shorter time scales. 6. The strong relationships between Amax and NDVI for 2004 to 2006 and between Amax and LAI in 2006 suggest the increasing importance of LAI in controlling annual NEP with increasing stand age up to six years. The reduced r2 when data from 2007 was included was caused by the low mean summer and late GS NDVI values in 2007. This conclusion in conjunction with increasing r2 during the GS and with stand age when fitting the Michaelis-Menten function to daytime NEP measurements, suggests that there was less deciduous growth during 2007 than in previous years, while the increased JP LAI combined with warm temperatures and an adequate water supply resulted in large Amax values later in the 2007 GS.  68  REFERENCES Amiro, B.D., Orchansky, A.L., Barr, A.G., Black, T.A., Chambers, S.D., Chapin III, F.S., Goulden, M.L., Litvak, M., Liu, H.P., McCaughey, J.H., McMillan, A., and Randerson, J.T. 2006. The effect of post-fire stand age on the boreal forest energy balance, Agricultural and Forest Meteorology, 140: 41-50. Baldocchi, D.D. 2003. Assessing the eddy covariance technique for evaluating carbon dioxide exchange rates of ecosystems: past, present and future, Global Change Biology, 9: 479-492. Baldocchi, D.D., Vogel, C.A., and Hall, B. 1997. Seasonal variation of carbon dioxide exchange rates above and below a boreal jack pine forest, Agricultural and Forest Meteorology, 83: 147170. Barr, A.G., Morgenstern, K., Black, T.A., McCaughey, J.H., and Nesic, Z. 2006. Surface energy balance closure by the eddy-covariance method above three boreal forest stands and implications for the measurement of the CO2 flux, Agricultural and Forest Meteorology, 140: 322-337. Barr, A.G., Black, T.A., Hogg, E.H., Kljun, N., Morgenstern, K., and Nesic, Z. 2004. Inter-annual variability in the leaf area index of a boreal aspen-hazelnut forest in relation to net ecosystem production, Agricultural and Forest Meteorology, 126: 237-255. Brand, D.G. 1987. Estimating the surface area of spruce and pine foliage from displaced volume and length. Canadian Journal of Forest Research 17: 1305-1308. Boreal Ecosystems Research and Monitoring Sites: Data, Climate Maps, Old Jack Pine, 2006. Retrieved on March 25, 2008, from http://berms.ccrp.ec.gc.ca/Data/Overview/graphs/OJP_2006.JPG. Burdette, A. N. 1979. A nondestructictive method for measuring the volume of intact plant parts. Canadian Journal of Forest Research 9: 120-122. Campbell, G.S. and Norman, J.M. 1998. An Introduction to Environmental Biophysics. Springer, New York. Chen, J.M., and Black, T.A.1992. Defining leaf area index for non-flat leaves, Plant and cell environment, 15: 421-429. Chen, J.M. 1999. Spatial scaling of a remotely sensed surface parameter by contexture, Remote Sensing of Environment, 69: 30-42. Chen, J.M. 1996. Optically-based methods for measuring seasonal variation of leaf area index in boreal conifer stands, Agricultural and Forest Meteorology, 80: 135-163. Chen, J.M., Govind, A., Sonnentag, O., Zhang, Y., Barr, A., and Amiro, B. 2006. Leaf area index measurements at fluxnet-canada forest sites, Agricultural and Forest Meteorology, 140: 257-268.  69  Chen, J.M., Rich, P.M., Gower, S.T., Norman, J.M. and Plummer, S.P. 1997. Leaf area index measurements of boreal forests: Theory, techniques, and measurements, Journal of Geophysical Research, 102: 29,429-29,443. Cheng, Y., Gamon, J.A., Fuentes, D.A., Mao, Z., Sims, D.A., Qiu, H., Claudio, H., Huete, A., and Rahman, A.F. 2006. A multi-scale analysis of dynamic optical signals in a southern california chaparral ecosystem: A comparison of field, AVIRIS and MODIS data, Remote Sensing of Environment, 103: 369-378. Driese, K.L. and Reiners, W.A. 1997. Aerodynamic roughness parameters for semi-arid natural shrub communities of wyoming, USA, Agricultural and Forest Meteorology, 88: 1-14. Epson: 3100z PhotoPC. Support (2008). Retrieved on 2 September, 2008, from http://www.epson.com/cgi-bin/Store/support/supDetail.jsp?oid=14094 Falge, E., Baldocchi, D., Olson, R., Anthoni, P., Aubinet, M., Bernhofer, C., Burba, G., Ceulemans, R., Clement, R., Dolman, H., Granier, A., Gross, P., Grünwald, T., Hollinger, D., Jensen, N., Katul, G., Keronen, P., Kowalski, A., Lai, C.T., Law, B.E., Meyers, T., Moncrieff, J., Moors, E., Munger, J.W., Pilegaard, K., Rannik, Ü, Rebmann, C., Suyker, A., Tenhunen, J., Tu, K., Verma, S., Vesala, T., Wilson, K., and Wofsy, S. 2001. Gap filling strategies for defensible annual sums of net ecosystem exchange, Agricultural and Forest Meteorology, 107: 43-69. FastStone: Image Viewer 3.5 Freeware (15/01/2008). Retrieved on July 28, 2008, from http://www.faststone.org/ Gamon, J.A., Huemmrich, K.F., Peddle, D.R., Chen, J., Fuentes, D., Hall, F.G., Kimball, J.S., Goetz, S., Gu, J., McDonald, K.C., Miller, J.R., Moghaddam, M., Rahman, A.F., Roujean, J.-., Smith, E.A., Walthall, C.L., Zarco-Tejada, P., Hu, B., Fernandes, R., and Cihlar, J. 2004. Remote sensing in BOREAS: Lessons learned, Remote Sensing of Environment, 89: 139-162. Gamon, J.A., Rahman, A.F., Dungan, J.L., Schildhauer, M., and Huemmrich, K.F. 2006. Spectral network (SpecNet)—What is it and why do we need it? Remote Sensing of Environment, 103: 227-235. Gamon, J.A., Cheng, Y., Claudio, H., MacKinney, L., and Sims, D.A. 2006. A mobile tram system for systematic sampling of ecosystem optical properties, Remote Sensing of Environment, 103: 246-254. Griffis, T.J., Black, T.A., Morgenstern, K., Barr, A.G., Nesic, Z., Drewitt, G.B., Gaumont-Guay, D., and McCaughey, J.H. 2003. Ecophysiological controls on the carbon balances of three southern boreal forests, Agricultural and Forest Meteorology, 117: 53-71. Hartley, I.P., Armstrong, A.F., Murthy, R., Barron-Gafford, G., Ineson, P., and Atkin, O.K. 2006. The dependence of respiration on photosynthetic substrate supply and temperature: Integrating leaf, soil and ecosystem measurements, Global Change Biology, 12: 1954-1968. Humphreys, E.R., Black, T.A., Morgenstern, K., Li, Z. and Nesic, Z. 2005. Net ecosystem production of a Douglas-fir stand for 3 years following clearcut harvesting, Global Change Biology, 11: 450-464.  70  Humphreys, E.R., Black, T.A., Morgenstern, K., Cai, T., Drewitt, G.B., Nesic, Z. and Trofymow, J.A. 2006. Carbon dioxide fluxes in coastal douglas-fir stands at different stages of development after clearcut harvesting, Agricultural and Forest Meteorology, 140: 6-22. Janssens, I.A., Lankreijer, H., Matteucci, G., Kowalski, A.S., Buchmann, N., Epron, D., Pilegaard, K., Kutsch, W., Longdoz, B., Grunwald, T., Montagnani, L., Dore, S., Rebmann, C., Moors, E.J., Grelle, A., Rannik, U., Morgenstern, K., Oltchev, S., Clement, R., Gudmundsson, J., Minerbi, S., Berbigier, P., Ibrom, A., Moncrieff, J., Aubinet, M., Bernhofer, C., Jensen, N.O., Vesala, T., Granier, A., Schulze, E.-., Lindroth, A., Dolman, A.J., Jarvis, P.G., Ceulemans, R., and Valentini, R. 2001. Productivity overshadows temperature in determining soil and ecosystem respiration across european forests, Global Change Biology, 7: 269-278. Jenkins, J.P., Richardson, A.D., Braswell, B.H., Ollinger, S.V., Hollinger, D.Y., and Smith, M. 2007. Refining light-use efficiency calculations for a deciduous forest canopy using simultaneous tower-based carbon flux and radiometric measurements, Agricultural and Forest Meteorology, 143: 64-79. Jonckheere, I., Fleck, S., Nackaerts, K., Muys, B., Coppin, P., Weiss, M., and Baret, F. 2004. Review of methods for in situ leaf area index determination: Part I. theories, sensors and hemispherical photography, Agricultural and Forest Meteorology, 121: 19-35. Kidston, J. 2006. Energy balance closure in a boreal old-growth jack pine stand and clearcut, and implications for eddy-covariance CO2 flux measurements. MSc Thesis. University of BC (unpublished). Kowalski, S., Sartore, M., Burlett, R., Berbigier, P., and Loustau, D. 2003. The annual carbon budget of a french pine forest (pinus pinaster) following harvest, Global Change Biology, 9: 1051-1065. Lee, X., Neumann, H.H., Hartog, G., Mickle, R.E., Fuentes, J.D., Black, T.A., Yang, P.C., and Blanken, P.D. 1997. Observation of gravity waves in a boreal forest, Boundary-Layer Meteorology, 84: 383-398. Maitani, T., Hiramatsu, Y., and Seo, T. 1984. Wave-like wind fluctuations observed in the stable surface layer over a plant canopy, Boundary-Layer Meteorology, 29: 273-283. Maitani, T. 1989. Wave-like wind fluctuations observed in the stable surface layer over a plant canopy(2), Boundary-Layer Meteorology, 48: 19-31. Massman, W.J. and Lee, X. 2002. Eddy covariance flux corrections and uncertainties in longterm studies of carbon and energy exchanges, Agricultural and Forest Meteorology, 113: 121144. Moffat, A.M., Papale, D., Reichstein, M., Hollinger, D.Y., Richardson, A.D., Barr, A.G., Beckstein, C., Braswell, B.H., Churkina, G., Desai, A.R., Falge, E., Gove, J.H., Heimann, M., Hui, D., Jarvis, A.J., Kattge, J., Noormets, A., and Stauch, V.J. 2007. Comprehensive comparison of gap-filling techniques for eddy covariance net carbon fluxes, Agricultural and Forest Meteorology, 147: 209-232. Moncrieff, J.B., Massheder, J.M., de Bruin, H., Elbers, J., Friborg, T., Heusinkveld, B., Kabat, P., Scott, S., Soegaard, H., and Verhoef, A. 1997. A system to measure surface fluxes of 71  momentum, sensible heat, water vapour and carbon dioxide, Journal of Hydrology, 189: 589611. National Aeronautics and Space Administration (2008): About MODIS. Retrieved on June 4, 2008, from http://modis.gsfc.nasa.gov/about/ Natural Resources Canada (2004): The Atlas of Canada, Frost-Free Period. Retrieved on July 22, 2008, from http://atlas.nrcan.gc.ca/site/english/maps/archives/5thedition/environment/climate/mcr4037 Papale, D., Reichstein, M., Canfora, E., Aubinet, M., Bernhofer, C., Longdoz, B., Kutsch, W., Rambal, S., Valentini, R., Vesala, T., Yakir, D. 2006. Towards a more harmonized processing of eddy covariance CO2 fluxes: algorithms and uncertainty estimation, Biogeosciences, 3: 571-583. Reichstein, M., Falge, E., Baldocchi, D.D., Papale, D., Aubinet, M., Berbigier, P., Bernhofer, C., Buchmann, N., Gilmanov, T., Granier, A., Grünwald, T., Havránkov, K., Ilvesniemi, H., Janous, D., Knohl, A., Laurila, T., Lohila, A., Loustau, D., Matteucci, G., Meyers, T. 2005. On the separation of net ecosystem exchange into assimilation and ecosystem respiration: review and improved algorithm, Global Change Biology, 11: 1424-1439. Richardson, A.D., Braswell, B.H., Hollinger, D.Y., Burman, P., Davidson, E.A., Evans, R.S., Flanagan, L.B., Munger, J.W., Savage, K., Urbanski, S.P., and Wofsy, S.C. 2006. Comparing simple respiration models for eddy flux and dynamic chamber data, Agricultural and Forest Meteorology, 141: 219-234. Richardson, A.D. and Hollinger, D.Y. 2005. Statistical modeling of ecosystem respiration using eddy covariance data: Maximum likelihood parameter estimation, and monte carlo simulation of model and parameter uncertainty, applied to three simple models, Agricultural and Forest Meteorology, 131: 191-208. Richardson, A.D., Jenkins, J.P., Braswell, B.H., Hollinger, D.Y., Ollinger, S.V. and Smith, M-L. 2007. Use of digital webcam images to track spring green-up in a deciduous broadleaf forest, 152: 323-334. Stull, R.B. 1988. An Introduction to Boundary Layer Meteorology. Kluwer Academic Publishers, Dordrecht ; Boston. Suyker, A.E. and Verma, S.B. 2001. Year-round observations of the net ecosystem exchange of carbon dioxide in a native tallgrass prairie, Global Change Biology, 7: 279-289. Tanner, C.B. and Thurtell, G.W. 1969. Anemoclinometer measurements of Reynolds Stress and heat transport in the atmospheric surface layer. ECOM 66-G22_F, University of Wisconsin, Madison, Wisconsin. Turner, D.P., Ritts, ,W.D., Cohen, W.B., Gower, S.T., Zhao, M.S., Running, S.W., Wofsy, S.C., Urbanski, S., Dunn, A.L., Munger, J.W. 2003. Scaling Gross Primary Production (GPP) over boreal and deciduous forest landscapes in support of MODIS GPP product validation, Remote Sensing of Environment, 88: 256-270.  72  Twine, T.E., Kustas, W.P., Norman, J.M., Cook, D.R., Houser, P.R., Meyers, T.P., Prueger, J.H., Starks, P.J., and Wesely, M.L. 2000. Correcting eddy-covariance flux underestimates over a grassland, Agricultural and Forest Meteorology, 103: 279-300. Wang, S., Grant, R.F., Verseghy, D.L., and Black, T.A. 2001. Modelling plant carbon and nitrogen dynamics of a boreal aspen forest in CLASS — the canadian land surface scheme, Ecological Modelling, 142: 135-154. Weiss, A. and Norman, J.M. 1985. Partitioning solar radiation into direct and diffuse, visible and near-infrared components, Agricultural and Forest Meteorology, 34: 205-213. Weiss, M., Baret, F., Smith, G.J., Jonckheere, I., and Coppin, P. 2004. Review of methods for in situ leaf area index (LAI) determination: Part II. estimation of LAI, errors and sampling, Agricultural and Forest Meteorology, 121: 37-53. Wilson, K., Goldstein, A., Falge, E., Aubinet, M., Baldocchi, D., Berbigier, P., Bernhofer, C., Ceulemans, R., Dolman, H., Field, C., Grelle, A., Ibrom, A., Law, B.E., Kowalski, A., Meyers, T., Moncrieff, J., Monson, R., Oechel, W., Tenhunen, J., Valentini, R., and Verma, S. 2002. Energy balance closure at FLUXNET sites, Agricultural and Forest Meteorology, 113: 223-243. Wilson, T.B. and Meyers, T.P. 2007. Determining vegetation indices from solar and photosynthetically active radiation fluxes, Agricultural and Forest Meteorology, 144: 160-179. Wofsy, S.C., Goulden, M.L. and Munger, J.W. 1993. Net exchange of CO2 in a mid-latitude forest, Science. 260: 1314-1318. Yuan, W., Liu, S., Zhou, G., Zhou, G., Tieszen, L.L., Baldocchi, D., Bernhofer, C., Gholz, H., Goldstein, A.H., Goulden, M.L., Hollinger, D.Y., Hu, Y., Law, B.E., Stoy, P.C., Vesala, T., and Wofsy, S.C. 2007. Deriving a light use efficiency model from eddy covariance flux data for predicting daily gross primary production across biomes, Agricultural and Forest Meteorology, 143: 189-207. World Wildlife Foundation Canada (2005): Education Boreal Forest 101. Retrieved on January 23, 2008, from http://www.wwf.ca/satellite/wwfkids/Boreal/Default.asp Zha, T., Barr, A.G., Black, T.A., McCaughey, J.H., Bhatti, J., Hawthorne, I., Krishnan, P., Kidston, J., Saigusa, N., Shashkov, A., Nesic, Z. 2008. Carbon Sequestration in boreal jack pine stands following harvesting. Global Change Biology (submitted).  73  APPENDICES Appendix 1: Site photographs from the air and ground  Figure 21. Overhead digital photograph taken before 2004 looking northeast over HJP02, showing stand boundaries, access routes and the windrows of CWD created by post-harvest debris management.  74  Figure 22. Overhead digital photograph taken in 2005looking from the east over the site, showing stand boundaries, access routes and windrows of CWD made by post harvest debris management. The site hut is located at the black square, EC tower at the red square and C2 and C3 at the green squares.  75  Figure 23. Site overview May 2006.  Figure 24. Site overview July 2006.  Figure 25. Site overview October 2006.  76  Figure 26. Greyscale digital elevation map, showing small changes in elevation at the site (< 7 m) with the EC tower location displayed (red square).  77  Appendix 2: Cumulative footprint contributions  1000  Distance from tower S−N (m)  800 600 99  400 99 95  200 0  99 95  −200  70 80 90  −400  99  −600 −800 −1000 −1000 −800 −600 −400 −200  0  200  400  600  800 1000  Distance from tower W−E (m)  Figure 27. Annual cumulative daytime footprint. Values beside the lines inside the figure represent the % of the total flux that is measured within the bounds of that area. 1000  Distance from tower S−N (m)  800 600  99  99  400 200  90  0  99  70 80  −200 −400  95  95 99  −600 −800 −1000 −1000 −800 −600 −400 −200  0  200  400  600  800 1000  Distance from tower W−E (m)  Figure 28. Annual cumulative night footprint. Values beside the lines inside the figure represent the % of the total flux that is measured within the bounds of that area.  78  Appendix 3: Volumetric displacement technique The total surface area of needles is inaccurately estimated by the projected surface area method known to work well for broad leaves (Burdett, 1977; Johnson, 1984). Burdett 1977 developed a fast, accurate method for measuring total needle area using the volume displacement of water. The technique applies the Archimedes Principle:  “a body immersed in fluid is buoyed up by a force equal to the weight of the displaced fluid, and the relationship that the weight of the displaced water in grams is equal to the volume of that body in cubic centimetres”.  Mathematically this is described as:  Fb = gρ H 2OV  (27)  where Fb, is the buoyant force; g, is gravity; ρ H 2O , is the density of the water being used, which is often assumed to be one with units of g m-3; V, is the volume of displaced water in cm3, therefore  V =  m ρH 2 O  (28)  where m, is the mass of displaced water in g. For pine species measurements of the needle volume, length and number can be converted to needle surface area using simple geometry. For JP a hemi-cylindrical geometry is used to describe the shape of the needles (Brand, 1987, Chen  et al 1997).  79  A3.1 Equipment •  1 kg top loading balance sensitive to 0.01 g  •  A clean measuring cylinder, 30 cm high and 10 cm in diameter  •  Thin malleable wire  •  4 L deionized/distilled water of a known temperature  •  Washing up liquid  A3.2 Technique 1. Place two litres of the distilled water into the measuring cylinder. 2. Add a small amount of washing up liquid to avoid minute bubbles adhering to shoot surfaces. 3. Place the full cylinder on the balance and tare. 4. Select a single shoot. 5. Rinse the shoot thoroughly with water to remove any dirt and debris. 6. Dry the shoot with a tissue making sure that no water is trapped. 7. Using the wire to hold the shoot, fully submerge it into the cylinder to a depth marked on the wire. Take great care not to touch the sides or bottom of the measuring cylinder with the shoot. 8. Record the weight indicated by the balance 9. Record the number of needles on the shoot. 10. Record the length of each needle. 11. Remove the fascicles from the shoot and re-submerge the shoot branch as above 12. Record the weight indicated by the balance  80  A3.3 Total needle area of shoot The total needle area of the shoot (Ln) can be calculated using the equation, describing hemicylindrical shape, shown below (Chen, 1997). Ln = 4.1 VnL  (29)  where 4.10 represents the conversion factor necessary to account for the hemi-cylinder shape of JP needles, V is the displacement volume of needles (whole shoot volume – shoot branch volume – wire volume), n is the number of needles, and L is the average length of the needles.  81  Appendix 4: Median-absolute-deviation outlier removal technique This appendix describes an outlier detection and removal technique adapted from Papale et al., (2006). A double difference time series is created then the median of absolute deviations about the median (MAD) is used as a robust outlier estimator (Papale et al., 2003):  (  ) (  di = xi − x i −1 − x i +1 − xi  )  (30)  where di is the double difference time series between xi and xi-1, and xi+1 and xi and: Md j = median∑ d i  (31)  where Mdj is the median sum of differences within each data window and:  (  MAD j = median d i − Md j  )  (32)  where MAD j is the median of the absolute deviation about the median difference (Mdj) between sample half-hours and the median sum of differences for the data window they are within. Single points are considered as an outlier when: ⎛ Md j − threshold × MAD j d i < ⎜⎜ 0.6745 ⎝  ⎞ ⎟⎟ ⎠  (33)  ⎞ ⎟⎟ ⎠  (34)  or ⎛ Md j + threshold × MAD j d i > ⎜⎜ 0.6745 ⎝  Papale et al (2006) show that as the chosen ‘threshold’ increases the number of points considered erroneous is decreased as their deviation from the median average difference (MAD j) is allowed to be greater. Computation was carried out using MATLAB 6.5.1. (Code available, ihawth81@interchange.ubc.ca). For the purpose of this study the window length used to define the MAD was 13 full days using the complete diurnal cycle of values, creating 28 MAD values for one year.  82  Appendix 5: Epson 3100z PhotoPC image capture details The EPSON PhotoPC 3100z is a 3.34 -mega pixel camera with a 1/1.8 inch 2048 x 1536 pixel CCD sensor. A continuous power supply was provided to the permanent cameras using an AC\DC converter (Nextexx), with battery backup power (4 x AAA). The roaming cameras were battery powered (4 x AAA). Table 10 shows the camera settings used for all of the images captured. The user manual is available online at http://www.epson.com, 2008-09-02. Table 10. EPSON 3100z PhotoPC settings Setting Name Resolution Compression  Setting Super Fine JPEG 2048 x 1536 pixels Low  White Balance (W/B) ISO  Auto 100  Flash Program Subject Settings  No • Normal • Matrix metering  Justification Finest resolution setting possible when using interval shooting. Saves all of the C-Data for an image. Varies with light conditions This Number indicating the sensitivity of cameras CCD to light. Values can range from 50400. Lower settings require more light to attain the correct exposure. The setting of 100 was used to allow for extra sensitivity without greatly reducing image quality. Images at night are discarded. • 8:1 pixel dimension • Allows the camera to sample light from entire area to determine the correct exposure.  Figure 29 is an example of one of the images taken from C2 during the month of July; clearly visible are five of the JP, one green alder shrub, grass, exposed soil and remaining CWD. It is apparent that at this time, under the specific light (i.e., Si) and moisture conditions, there is not a clear visual definition between the vegetation and the soil, lichen and CWD background.  83  y Pixel Number  0  500  1000  1500 200  600  1000 1400 x Pixel Number  1800  Pixel Number  Figure 29. Example image from C2 image bundle (08/07/2006). Red markers indicate areas where 1 x 50 sample pixels have been taken for further analyses (See Figures 30 and 31).  1200  450  200  475  500  800  400  825 Pixel Number  850  1300  1325  1350  Figure 30. 1 x 50 sample pixels taken from the locations of the red markers in Figure 29. The right-hand panel is the exposed soil, the centre panel is the green leaf and the right-hand panel is the shaded area.  84  DN Intensity/100 & GI  2.0  2.0  2.0  1.0  1.0  1.0  0.0 450  475  500  0.0 800  825 x-Pixel Number  850  0.0 1300  Red Green Blue GI GIth  1325  1350  Figure 31. DN and RLVI values from Figure 30 sample pixels, showing GIth values (0.58) above, which pixels would be classified G. Panel one shows that R has the strongest intensity reading from the soil and senesced vegetation surfaces. Panel three shows that some pixels are classified G as a result of colour casting green onto a light coloured shaded area, where there was no vegetation.  85  Appendix 6:Growing season nocturnal NEP vs. σw 1 2004 2006  NEP (µmol m−2 s−1)  0  −1  −2  0  0.1  0.2  0.3 0.4 σw (m s−1)  0.5  0.6  0.7  Figure 32. GS nighttime NEP vs. σw for 2004 and 2006. The points and associated vertical lines represent 100-point bin-averages with their standard errors. The dot-dashed and dashed lines indicate the conservative estimates of σwth, 0.12 and 0.15 m s-1 for 2004 and 2006, respectively, below which data were removed. An increasing number of large –ve values with large standard errors can be seen during 2006 at low σw.  86  

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}]}"
                            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:
http://iiif.library.ubc.ca/presentation/dsp.24.1-0066759/manifest

Comment

Related Items