UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Evaporation within and above a boreal aspen forest Blanken, Peter David 1997

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

Item Metadata

Download

Media
831-ubc_1997-195546.pdf [ 12.48MB ]
Metadata
JSON: 831-1.0088118.json
JSON-LD: 831-1.0088118-ld.json
RDF/XML (Pretty): 831-1.0088118-rdf.xml
RDF/JSON: 831-1.0088118-rdf.json
Turtle: 831-1.0088118-turtle.txt
N-Triples: 831-1.0088118-rdf-ntriples.txt
Original Record: 831-1.0088118-source.json
Full Text
831-1.0088118-fulltext.txt
Citation
831-1.0088118.ris

Full Text

EVAPORATION WITHIN AND ABOVE A BOREAL ASPEN FOREST by PETER DAVID BLANKEN B.Sc. (Honours) McMaster University, 1990 M.Sc. McMaster University, 1992 A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY in THE FACULTY OF GRADUATE STUDIES Department of Soil Science We accept this thesis as conforming to the required standard THE UNIVERSITY OF BRITISH COLUMBIA April 1997 © Peter David Blanken, 1997 In presenting this thesis in partial fulfilment of the requirements for an advanced degree at the University of British Columbia, I agree that the Library shall make it freely available for reference and study. I further agree that permission for extensive copying of this thesis for scholarly purposes may be granted by the head of my department or by his or her representatives. It is understood that copying or publication of this thesis for financial gain shall not be allowed without my written permission. Department of The University of British Columbia Vancouver, Canada Date for.'I DE-6 (2/88) A B S T R A C T As part of the Boreal Ecosystem-Atmosphere Study, water vapour, heat, C O 2 and momentum exchange between the atmosphere and a southern boreal aspen {Populus tremuloides Michx.) forest in central Saskatchewan, Canada (53.629 °N, 106.200 °W) were measured continuously throughout much of 1994 using the eddy-covariance method. Measurements were made both above the c. 21.5-m tall 70 year-old aspen stand and within the leafless trunk space above a lush c. 2-m tall hazelnut (Corylus cornuta Marsh.) understory. This research focused on the measurements of and processes controlling water vapour exchange within and above the aspen canopy. Above-canopy turbulent exchange was dominated by large, slowly rotating eddies whereas in-canopy exchange was dominated by the intermittent, downward penetration of gusts. A constant flux layer redeveloped beneath the aspen canopy making eddy-covariance measurements possible. Nocturnal eddy fluxes were often underestimated at both heights due to spatial heterogeneity in turbulence statistics caused by low wind speeds. These periods were identified from the height-independent similarity function normalized by that expected from Monin-Obukhov theory and were empirically corrected as a function of friction velocity. Erratic daytime flux behaviour was corrected on the basis of conservation of energy and partitioning of the missing energy using the original eddy fluxes of latent and sensible heat. Evapotranspiration from the forest accounted for 82-91% of the annual precipitation. Aspen, hazelnut transpiration and soil water evaporation were 68%, 27% and 5%, respectively, of the total annual evapotranspiration. Over the growing season, ii there was no net change in the soil water content and there was little drainage beyond the root zone. Understory radiation levels decreased exponentially with increasing aspen leaf area. Surface conductance to water vapour was a linear function of forest leaf area and was dominated by the aspen canopy. Aspen and hazelnut canopy conductances decreased non-linearly with increasing saturation deficit and were best parameterized by net assimilation divided by the product of the mole fractions of leaf-level saturation deficit and CO2 concentration. The accommodation of the transpiring vegetation by the atmosphere was quantified using the Priestley and Taylor a and the McNaughton and Jarvis Q parameters. iii TABLE OF CONTENTS Abstract ii Table of Contents iv List of Symbols & Acronyms viii List of Tables xvi List of Figures xviii Acknowledgments xxvii Dedication xxviii CHAPTER 1 - INTRODUCTION 1 1.1. References 4 CHAPTER 2 - CAN EDDY COVARIANCE MEASURE T U R B U L E N T FLUXES WITHIN A BOREAL A S P E N FOREST ? 5 2 . 1 . Introduction 5 2.2. Site Description 7 2.3. Material & Methods 8 2 .3 .1 . Measurement of Above- & Within-canopy Eddy Fluxes 8 2.3.2. Turbulence Statistics & Parameters 11 2.4. Results & Discussion 12 2 .4 .1 . Vertical Profiles of Turbulence Statistics 12 2.4.2. Spectral Analyses of Turbulence 17 2.4.3. Effect of Leafing on Within-canopy Horizontal Wind & Friction Velocity 21 2.4.4. Monin-Obukhov Similarity 25 2.4.5. Homogeneity of Turbulent Fluxes 28 iv 2.4.6. Closure of the Surface Energy Balance 36 2.4.7. Unexplained Short- term Variability in the Eddy Fluxes & Data Quality Checks 39 2.5. Summary & Conclusions 51 2.6. References 54 CHAPTER 3 - S E A S O N A L ENERGY & W A T E R E X C H A N G E A B O V E & WITHIN A B O R E A L A S P E N FOREST 59 3 . 1 . Introduction 59 3.2. Materials & Methods 61 3.2 .1 . Above & Within-canopy Eddy Flux Measurements 61 3.2.2. Above-canopy Radiat ion Measurements 61 3.2.3. Within-canopy Radiation Measurements 63 3.2.4. Energy Storage 64 3.2.5. Soil Heat Flux, Water Content & Evaporation 68 3.2.6. Precipitation 71 3.2.7. Leaf, Plant & Wood Area Indexes 72 3.2.8. Fetch or Flux Footprints 73 3.3. Results & Discussion 74 3.3 .1 . Site Condit ions During 1994 74 3.3.2. Footprint Predictions of the Scalar Flux Source Area 79 3.3.3. Within-canopy Radiat ion 84 3.3.4. Seasonal Progression of the Diurnal Energy Balance 89 3.3.4.1. Spring (April - May) 89 V 3.3.4.2. Summer (June - July - August) 95 3.3.4.3. Late Summer - Early Fall (September) 96 3.3.5. Partit ioning Overstory & Understory Turbulent Fluxes 97 3.3.6. The Water Balance 101 3.4. Summary & Conclusions 107 3.5. References 111 CHAPTER 4 - AN ANALYSIS OF BOREAL A S P E N O V E R S T O R Y & HAZELNUT U N D E R S T O R Y C A N O P Y C O N D U C T A N C E T O W A T E R V A P O U R 115 4 . 1 . Introduction 115 4.2. Materials & Methods 116 4 .2 .1 . Calculat ion of Canopy Conductance 116 4.2.2. Calculation of Regional Evapotranspirat ion Parameters 120 4.3. Results & Discussion 121 4 .3 .1 . The Seasonal Patterns of Surface & Canopy Conductances 121 4.3.2. Methods of Calculat ing Conductances 125 4.3.3. Seasonal Progression of the Diurnal Surface & Canopy Conductances 128 4.3.4. Controls on Aspen & Hazelnut Canopy Conductances: Calculating Responses f rom Observed Relationships 130 4 .3 .4 .1 . Relating Conductance to Humidity Stratified by Light 133 4.3.4.2. Calculating Aspen Conductance & Transpirat ion f rom Saturation Deficit & Light 137 4.3.4.3. Relating Conductance to Net Assimilat ion, Relative Humidity & C 0 2 Concentrat ion at the Leaf Surface: The Bal l -Woodrow-Berry Index 141 VI 4.3.4.4. Calculat ing Aspen Conductance & Transpirat ion from the Bal l -Woodrow-Berry Index 147 4.3.4.5. Relat ing Conductance to a Modif ied Form of the Bal l -Woodrow-Berry Index 149 4.3.4.6. Calculat ing Aspen Conductance & Transpirat ion f rom the Modif ied Bal l -Woodrow-Berry Index 151 4.3.5. Regional Evapotranspirat ion Parameters: The Priestley & Taylor a & the McNaughton & Jarvis Q Coefficients 153 4.4. Summary & Conclusions 160 4.5. References 162 CHAPTER 5 - S U M M A R Y & C O N C L U S I O N S 167 5 . 1 . References 173 APPENDIX A. C U R V E FIT P A R A M E T E R S & COEFFICIENTS 174 A . 1 . Introduction 174 A.2. References 178 APPENDIX B. M A P S & P H O T O G R A P H S 179 vii LIST O F S Y M B O L S & A C R O N Y M S In an attempt to adopt some standardization of symbols and notation, standards suggested by Reifsnyder et al. (1991) have been followed when possible. Sign convention for all fluxes was from an atmospheric point-of-view with positive denoting a flux away from the surface and negative denoting a flux towards the surface. The term flux implies a flux density. Symbol Definition Units An C a Cm Co C w C w x ( / ) D D0 E Fc net CO2 assimilation flux density (aspen An Fc(39 m) - Fc (4 m); hazelnut An = Fc(4 m) -•#soil) //mol or mmol m"2 s"' energy required to photosynthetically fix carbon J //mol" (0.469 J//mol" 1) air volumetric heat capacity mineral soil volumetric heat capacity organic soil volumetric heat capacity liquid water volumetric heat capacity cospectrum of variable x and w as a function of frequency (/) saturation deficit saturation deficit at the big-leaf surface saturation deficit in the air outside the leaf boundary layer J m"3 "C" 1 J m"3 ° C l J m"3 °C"' J m"3 ° C ' 1 dimensionless kPa or mol mol" kPa or mol mol" kPa or mol mol" evapotranspiration (transpiration & evaporation) mm d COaeddy flux density //mol m"2 s"1 or mg m"2 s"1 Ff tree form factor dimensionless viii G sensible heat flux density to the soil at depth Go sensible heat flux density to the soil at surface H sensible heat flux density to the atmosphere H' H corrected for energy balance closure (f5\E) Jx total canopy heat storage per unit ground area Jb rate of change of bole heat content per unit ground area Je rate of change of latent heat content per unit ground area yh rate of change of sensible heat content per unit ground area Ji rate of change of leaf heat content per unit ground area Jp rate of change of energy consumed for photosynthesis per unit ground area K turbulent or eddy diffusivity K3 dielectric constant for air Kw dielectric constant for water L scaling (Obukhov) length P precipitation P a ambient atmospheric pressure QI or t generic flux density (?plor T photosynthetic flux density (direction relative to the surface indicated by arrows) Q o l o r T above-canopy photosynthetic flux density (direction relative to the surface indicated by arrows) R autocorrelation function W m" W m" W m" W m - 2 Wm" - 2 W m Wm" Wm" W m" W m " m V dimensionless dimensionless m mm kPa W m"2 or /imol m"2 s"1 //mol m"2 s"1 //mol m"2 s"1 dimensionless ix Re R\ I or T K i or T •^ soil S St Sx(f) Skj T* Tc T\ Ts To U Wd available energy flux density (Rn - Go - Jt) available energy at the big-leaf surface Reynolds number longwave radiation flux density (direction relative to the surface indicated by arrows) net all-wave radiation flux density total solar radiation flux density (direct + diffuse) (direction relative to the surface indicated by arrows) respiratory CO2 flux from the soil liquid water contained in the soil stand live stem density power spectrum of variable* as a function of frequency (/) W m"2 W m"2 dimensionless Wm" 2 W m"2 -2 W m mg m"2 s"1 or //mol m"2 s"1 mm n trees per m 2 ground dimensionless skewness or third moment of velocity component dimensionless j air temperature canopy leaf temperature (a proper mean of the temperature of all the individual leaves) individual leaf temperature temperature of the soil at a depth of 2 cm temperature at the big-leaf surface drainage beyond the depth to which AS is calculated or the height-averaged constant wind speed (clear from context) water content on a dry-mass basis water content on a wet-mass basis °C °C °C °C °C mm or m s" % % x effective forest projected (one side) leaf area index (ae = a,(aspen)[gjgc{aspen)]) projected (one side) leaf area index ( f*c LAD(z)efe) Jo projected (one side) plant area index (aj + a w) projected (one side) wood area index specific heat of fresh tree boles specific heat of dry wood specific heat of glucose specific heat of fresh leaves specific heat of air at constant pressure specific heat of liquid water zero-plane displacement partial pressure of water vapour in the atmosphere frequency canopy conductance to water vapour excluding soil water evaporation surface conductance to water vapour including soil water evaporation stomatal conductance to water vapour mean canopy height turbulence intensity of velocity component j effective extinction coefficient von Karman constant (0.40) xi m 2 leaf per m 2 ground underneath m 2 leaf per m 2 ground underneath m 2 plant per m 2 ground underneath m 2 wood per m 2 ground underneath kJ k g 1 °C] kJ kg"1 °C"' kJ k g 1 ° C ' kJ kg"1 °C"' kJ kg"1 °C" 1 kJ k g 1 °C" 1 m kPa H z mmol m" 2 s"1 or mm s"1 mmol m" 2 s"1 or mm s"1 mmol m" 2 s"1 or mm s"1 m dimensionless dimensionless dimensionless n number of observations dimensionless n natural frequency, (fz /uhc) dimensionless p statistical probability level dimensionless q volume flow m3 s_1 r coefficient of determination dimensionless t time s, h, min or day u longitudinal (streamwise) velocity component m s'1 "* friction ve loc i ty / p A N m 2 v lateral (crosswind) velocity component m s'1 w vertical velocity component m s"1 w gravimetric soil water content kg water per kg soil solids x upwind distance m xmax peak footprint (upwind distance to which eddy m fluxes are most sensitive to) z height above ground m z\ height of the convective boundary layer m zr reference height above ground m z0 roughness length m AH' net canopy sensible heat flux density corrected W m" for energy conservation (aspen AH'= H'(39 m) -H'(4 m); hazelnut AH'= H'(4 m) - //(soil) AXE' net canopy latent heat flux density corrected for W m"2 energy conservation: aspen AXE'- XE'(39 m) -XE'(4 m); hazelnut AXE'= XE'{4 m)- ^(soil) AS change in soil water content mm d"1 x i i Q a P r 6 ox 6^ om x IE XE' XE, P A Pb Pi eq Eulerian integral length scale of velocity m component j Eulerian integral time scale of velocity s component j McNaughton & Jarvis Coefficient Priestley & Taylor Coefficient Bowen ratio (HI AE) psychrometric constant strength of the capping layer inversion soil volume fraction of water soil volume fraction of air soil volume fraction of minerals soil volume fraction of organics standard deviation of the mean (unless otherwise stated, ± one standard deviation reported) latent heat of vaporization J kg" evaporative heat flux density W m /^corrected for energy balance closure W m (Rja + P)) equilibrium evaporative heat flux density W m time lag s density kg m" density of dry air kg m density of dry soil kg m" specific leaf weight (g dry leaves m"2 ground) g m"2 dimensionless dimensionless dimensionless kPa °C °C m"1 m 3 water per m 3 soil m 3 air per m 3 soil m 3 minerals per m 3 soil m 3 organics per m 3 soil units of variable in question -2 -2 -2 - 3 X l l l Ps pv T V K xl Xo X a Xo ( ) ( )' A E S A W G B O R E A S B W B C 3 C 4 C B L C T C S T G C M density of soil solids (particles) density of water vapour shear stress or momentum flux density kinematic viscosity Monin-Obukhov similarity function for w C O 2 mole fraction in the air outside the leaf boundary layer C O 2 mole fraction at the big-leaf surface H 2 O mole fraction in the air outside the leaf boundary layer H 2 O mole fraction at the big-leaf surface time average deviation from time average Atmospheric Environment Service American Wire Gauge Boreal Ecosystem-Atmosphere Study Ball-Woodrow-Berry Index/l^o/xl species fixing C O 2 largely into 3-carbon phosphoglyeric acid species fixing C O 2 largely into 4-carbon malic and aspartic acids Convective Boundary Layer -3 kg m kg m" 3 N m - 2 m 2 s"1 dimensionless mol mol" 1 mol mol" 1 mol mol" 1 mol mol" 1 N / A N / A N / A N / A N / A mmol m"2 s"1 N / A N / A N / A 2 -1 Canopy Temperature equation for calculation gc mmol m" s (equation 3.5) Central Standard Time Global Climate Models s, min, h, day N / A xiv G M T Greenwich Mean Time I R G A Infrared Gas Analyzer IRT Infrared Thermometer L A D leaf area density L A I projected (one side) leaf area index ( f * ° L A D (z)dz) M O Monin-Obukhov surface-layer similarity hypothesis P A I projected (one side) plant area index (sum of W A I and L A I ) P M Penman-Monteith combination equation S iB2 Simple Biosphere Model , Version 2 T D R Time Domain Reflectrometry W A I projected (one side) wood area index (includes all stems and boles) s, min, h, day N / A °C m 2 leaf per m 3 volume of space 2 2 m leaf per m ground underneath N / A m 2 plant per m 2 ground underneath N / A N / A N / A m 2 wood per m 2 ground underneath Reference Reifsnyder, W . E . , McNaughton, K . G . & Milford, J.R. (1991) Symbols, units and notation. A statement of journal policy. Agricultural and Forest Meteorology, 54, 389-397. LIST O F T A B L E S Table 2.1. Minimum nocturnal friction velocities used by several investigators to reject CO2 fluxes due to heterogeneity resulting from insufficient turbulent mixing 30 Table 3.1 Approximations for total heat storage between 0-39 m or 0-4 m using a linear equation (W m"2) = a x + b where x is the 1/2-h change in air temperature (ATa °C) or net radiation (ARn W rrf2) measured above the aspen canopy and a and b are empirical coefficients (equations are independent of time of year or day) 69 Table 3.2. Comparison of leafless and leafed deciduous forest canopy's effective extinction coefficients (k) expressed on a leaf and plant area index basis for net radiation (Rn), photosynthetically active photon flux (Qpi) and solar radiation (Rsi) 88 Table 3.3a. Ensemble daytime (net radiation positive) means of monthly forest (z = 39 m) energy balance terms normalized by net radiation (R„). The last column is the sum of the latent heat (AE), sensible heat (//), soil heat (Go) and total heat storage (JJ terms... 92 Table 3.3b. Ensemble daytime (net radiation positive) means of monthly understory (z = 4 m) energy balance terms normalized by net radiation (Rn). The last column is the sum of the latent heat (AE), sensible heat (H), soil heat (G0) and total heat storage (7J terms. 93 Table 3.4. Evapotranspiration (E) totals for 1994 for the forest, the hazelnut understory and the soil, and the soil alone. See Figure 3.12 for measurement and calculation methods. Also shown are the ratios of annual E to annual P (ratios with no parenthesis are for P = 462.20 mm, not including January precipitation before gauge installation and ratios with parenthesis are for P = 488.35 mm, including missing January precipitation estimated from the 30-year mean) 104 Table 3.5. Estimation of the storage of water in the soil profile (S = Ox Az) from a depth of 0 to 123 cm over the period April 20 - September 20, 1994. OSP and MSP refer to organic and mineral 3-rod horizontal TDR probes, respectively, and Rod (segment number) refers to the vertical segmented TDR rod 106 Table 4.1a. Relationship between 1/2-hourly full-leaf (June 7 - September 10, 1994), dry canopy aspen canopy conductance and variables know to affect gc listed from least to highest scatter. Saturation deficits were expressed as mole fractions. Parameters were determined from means calculated over 20 equally spaced bins. Data were excluded when n I total n was less than 5% 131 Table 4.1b. Relationship between 1/2-hourly full-leaf (June 7 - September 10, 1994), dry canopy hazelnut canopy conductance and variables know to affect gc listed from least to highest scatter. Saturation deficits were expressed as mole fractions. Parameters were determined from means calculated over 20 equally spaced bins. Data were excluded when n I total n was less than 5% 132 xvi Table A . l . Parameters determined from a non-linear least squares fit to a 4-parameter logistic curve of the form f (x) =a /{\ + exp\b(x -c)ty + d used to fit the data presented in Figure 2.11 174 Table A.2. Parameters determined from a linear regression of the form f(x) = m x + k, where m is the slope and k is the y-intercept used to fit the energy balance closure data presented in Figure 2.14 175 Table A.3. Coefficients determined for the curve fit of the form Q4>(4 m)/Qvl(39 m) = a exp(-ba\) as shown in Figure 3.6 176 Table A.4. Parameters and defining the non-linear least squares determined equation gs or gc = .gcmax exp(-6Do) as shown in Figure 4.5. The high (H), medium (M) and low (L) photosynthetic photon flux density levels (Qpi) were Qpi > 1400 //mol m"2 s"\ 800 < Qpi < 1400 //mol m"2 s"1 and Qpi (200 < Qpi < 800 //mol m"2 s~\ respectively 177 xvii LIST OF FIGURES Figure 2.1. Mean 1/2 h profiles of leaf area density, skewness and turbulent intensity for a typical mid-summer day (August 4, 1994) during the daytime 6 am - 6 pm (CST) unstable period. Total (projected) leaf area indexes were 2.3 and 3.2 m 2 leaf per m2 ground for the aspen and hazelnut, respectively. Average standard errors of the means for the longitudinal (streamwise) (•), lateral (A) and vertical (•) velocity components were 0.08, 0.07 and 0.05, respectively, for the skewness and 0.12, 0.27 and 0.05, respectively, for the turbulent intensities. Profiles were normalized by the mean aspen canopy height (hc) of 21.5 m 13 Figure 2.2. Mean 1/2 h profiles of the Eulerian integral time and length scales for a typical mid-summer day (August 4, 1994) during the daytime 6 am - 6 pm (CST) unstable period. Average standard errors of the means for the longitudinal (streamwise) (•), lateral ( A ) and vertical (•) velocity components were 17, 3 and 6 s, respectively, for the time scale and 1.2, 0.9 and 0.2, respectively, for the length scales. Length scale profiles were normalized by the mean aspen canopy height (hc) of 21.5 m 15 Figure 2.3. Mean power spectra (Sx) for variablex wherex is the longitudinal (w), lateral (v) or vertical (w) wind velocity component or air temperature (7\) above (O, z = 39 m; z/hc = 1.81) and within (•, z = 4 m; zlhc = 0.19) the canopy for 8 1/2-h periods from 10 am - 2 pm, August 4, 1994. Power spectra were multiplied by frequency ( / ) and normalized by the variance of x. The solid lines are the -2/3 slope expected in the surface layer inertial subrange. Above and within canopy u were 2.04 and 0.31 m s"1, respectively 18 Figure 2.4. Mean cospectra (Cwx) between the vertical wind speed (w) and the variable x where JC is air temperature ( ^ a ) , H 2 0 mole fraction {%"), longitudinal wind velocity («) or C O 2 mole fraction (x\) above (O, z = 39 m; z/hc = 1.81) and within (•, z = 4 m; z/hc = 0.19) the canopy for 8 1/2-h periods from 10 am - 2 pm, August 4, 1994. Cospectra were multiplied by frequency if) and normalized by the covariance between w and x. Solid line is the -4/3 slope expected in the surface layer inertial subrange and the dashed line is the -1 slope. Above and within canopy u were 2.04 and 0.31 m s"1, z = 39 m and 4 m, respectively 20 Figure 2.5. Mean cospectra (Cwx) between the vertical wind speed (w) and variable x where* is air temperature (Ta), H 2 O mole fraction (xl )> longitudinal wind velocity (w) or C 0 2 mole fraction (x\) above (O; z = 39 m; z/hc = 1.81) and within (•; z = 4 m; zlhc = 0.19) the canopy for 8 1/2-h periods from 10 am - 2 pm, August 4, 1994. Cospectra were multiplied by frequency (/) and normalized by the covariance between w and*. The area under each curve equals one. Above and within canopy u were 2.04 and 0.31 m s'1, z = 39 m and 4 m, respectively 22 x v i n Figure 2.6. Relationship between friction velocity (u*) and horizontal wind speed (w), above (z = 39 m; z/hc = 1.81) and within (z = 4 m; z/hc = 0.19) the aspen canopy during pre-leaf (O) (before June 1, 1994) and leafed period (•) (June 2, 1994 to September 7, 1994). Mean values of u* calculated correspond to binned values of u with a bin width of 0.4 and 0.1 m s 1 (39 and 4 m heights, respectively). Mean standard errors of the mean were 0.07 and 0.04 m s"1 (39 m, pre-leaf and leafed, respectively), and 0.02 and 0.006 m s'1 (4 m, pre-leaf and leafed, respectively). Frequency distributions for the pre-leaf (clear) and leafed (shaded) canopies at both heights are shown by the vertical bars. Total number of 1/2-h observations were 5391 and 2691 (bare canopy 39 and 4 m, respectively) and 4533 and 4699 (leafed canopy 39 and 4 m, respectively) 24 Figure 2.7. Relationship between the standard deviation of vertical wind (cr ) normalized by the friction velocity («*) atz = 39 m; zlhc = 1.81 (•) and z = 4 m; zlhQ = 0.19 (•) and atmospheric stability, (z - d)IL. Triangles show crw(4 m)/«*(39 m). Mean values of oju* were calculated corresponding to binned values of (z - d)IL of 0.25 with a mean standard error of 0.11 (39 m) and 0.21 (4 m). Non-linear regression fitted equations (solid lines) are of the form expected by Monin-Obukhov similarity theory within the surface layer (a = 1.107, 1.2636, 0.3147 and b = 0.2120, 0.2750, 0.1460, • , • , • , respectively. Frequency distributions for (z - d)IL at the z = 39 m (Z//Jc = 1.81) (clear) and z = 4 m (z/Ac = 0.19) (shaded) levels are shown by the vertical bars (total n of 7283 and 5084 1/2 hours, 39 m and 4 m, respectively) 27 Figure 2.8. Dependence of the C O 2 flux (Fc, •) and turbulence homogeneity (O) on friction velocity («*) at night (available energy < 0 W rrf2). Means were calculated corresponding to binned values of u* at 0.05 (39 m) and 0.01 m s"1 (4 m) intervals. Mean standard errors were 0.26 (39 m) and 0.49 (4 m) //mol m"2 s"1 for Fc and 0.01 (39 m) and 0.02 (4 m) for turbulence homogeneity. Vertical bars show the u* frequency distributions (total n of 1804 and 1860 1/2 hours, 39 m and 4 m, respectively) 31 Figure 2.9. Dependence of the latent heat flux (AE,•) and turbulence homogeneity (O) on friction velocity (u*) at night (available energy < 0 W m"2). Means were calculated corresponding to binned values of u* at 0.05 (39 m) and 0.01 m s"1 (4 m) intervals. Mean standard errors were 1.32 (39 m) and 0.78 (4 m) W rrf2 for XE and 0.01 (39 m) and 0.02 (4 m) for turbulence homogeneity. Vertical bars show the u* frequency distributions (total n of 1804 and 1860 1/2 hours, 39 m and 4 m, respectively) 33 Figure 2.10. Dependence of the sensible heat flux (H, •) and turbulence homogeneity (O) on friction velocity (u*) at night (available energy < 0 W m"2). Means were calculated corresponding to binned values of u* at 0.05 (39 m) and 0.01 m s"1 (4 m) intervals. Mean standard errors were 2.07 (39 m) and 0.75 (4 m) W m'2 for Hand 0.01 (39 m) and 0.02 (4 m) for turbulence homogeneity. Vertical bars show the u* frequency distributions (total n of 1804 and 1860 1/2 hours, 39 m and 4 m, respectively) 34 x ix Figure 2.11. Relationship between the measured turbulent flux divided by the expected flux (• = C O 2 flux, (Fc); • = latent heat flux (AE); • = sensible heat flux (H) and the nighttime friction velocity («*) (binned at 0.015 (39 m) and 0.005 (4 m) m s"1 intervals). The expected flux was the mean 1/2-h flux that occurred when u* exceeded 0.30 (39 m) and 0.10 (4 m) m s"1 which were: 3.92 and 4.44 //mol m"2 s"1 (Fc); 15.8 and 4.9 W m"2 (AE) and -49.2 and -6.4 W m"2 (H) for the 39 and 4 m levels, respectively. The solid line is a non-linear least square sigmoidal 4-parameter logistic function fit (see Appendix A, Table A . l for parameters). Mean standard errors of the mean were 0.07 //mol m"2 s"1, 0.09 W m"2 and 0.05 W m"2 for the (39 m F C t AEand H, respectively) and 0.13 //mol m"2 s'\ 0.15 W m"2 and 0.07 W m"2 for the (4 m Fc, AEand H, respectively) 35 Figure 2.11. Nighttime mean measured (•) available energy (Ra), soil temperature at a depth of 2 cm (Ts), horizontal wind speed (u), friction velocity («*), C O 2 flux (Fc) and the sum of the latent (AE) and sensible (H) heat fluxes at the 39-m height. Turbulent fluxes corrected for heterogeneity due to low values of u* using the relationships shown in Figure 2.11 are shown by (•) 37 Figure 2.12. Nighttime mean measured (•) available energy (/?a), soil temperature at a depth of 2 cm (Ts), horizontal wind speed («), friction velocity («*), C O 2 flux (Fc) and the sum of the latent (AE) and sensible (H) heat fluxes at the 4-m height. Turbulent fluxes corrected for heterogeneity due to low values of u* using the relationships shown in Figure 2.11 are shown by (•) 38 Figure 2.13. Energy balance closure for above (A and B) and below (C and D) the aspen canopy separated into nighttime (A and C) and daytime (B and D) periods for the entire measurement period. Means of the measured sum of latent (AE) and sensible (H) 1/2 hourly fluxes (•) were based on binned measurements of available energy (Ra) with bin widths of 4, 15, 1.5 and 5 W m"2, for plots A, B, C and D, respectively. Estimates of the nighttime fluxes corrected for underestimation based on friction velocity (•) were closer to the 1:1 line (solid line). Linear regression (dashed lines) parameters are given in Appendix A, Table A.2. Frequency distributions of Ra are shown by vertical bars 40 Figure 2.15. Example of the recalculation of the overstory (39 m) latent (AE) and sensible (H) heat eddy fluxes (thick lines) on August 4, 1994 using the available energy and the ratio of the measured eddy fluxes. The original fluxes (dashed lines) show half-hourly variation not explained by changes in net radiation (upper thin solid line) even when some of the original points (•) were suspect (O) based on stationarity and homogeneity tests 41 Figure 2.16. Example of the recalculation of the understory (4 m) latent (AE) and sensible (H) heat eddy fluxes (thick lines) on August 4, 1994 using the available energy and the ratio of the measured eddy fluxes. The original fluxes (dashed lines) show half-hourly variation not explained by changes in net radiation (upper thin solid line) even when some of the original points (•) were suspect (O) based on stationarity and homogeneity tests 42 xx Figure 2.17. Effect of various integration periods on the 1/2-h latent heat flux (AE) at the 39-m height for a typical mid-summer day. Means for each integration period were determined either by a simple non-overlapping block average (•) or by non-overlapping linear detrending (•) 45 Figure 2.18. Effect of various integration periods on the 1/2-h latent heat flux (AE) at the 4-m height for a typical mid-summer day. Means for each integration period were determined either by a simple non-overlapping block average (•) or by non-overlapping linear detrending (•) 46 Figure 2.19. Mean daytime (6 am - 6 pm) latent heat flux density (AE) calculated as a function of various integration periods with either non-overlapping block averages (•) or non-overlapping linear detrends (•) normalized by flux calculated with a 30-min integration period (163 and 157 W m~2, block averaged and linear detrended, respectively, at the 39 m height; 59 and 54 W m' 2 , block averaged and linear detrended, respectively, at the 4 m height) 47 Figure 3.1. Comparison of the daily mean air temperature (Ta) measured above the aspen forest at z = 39 m (jagged line) and monthly total precipitation (P) measured in a clearing at the research site (solid vertical bars) to the 1951-1980 average Ta (smooth line) and monthly total precipitation (clear vertical bars) measured at Waskesiu Lake 75 Figure 3.2. Seasonal course of daily volumetric soil moisture content (6; lines) and precipitation (P; vertical bars). The organic layer (~ 0-10 cm) was represented by a 3-prong T D R probe inserted horizontally at a depth of 8 cm approximately in the middle of the organic horizon. The shallow mineral layer (= 10-30 cm) was represented by a 3-prong T D R probe inserted horizontally at a depth of 15 cm (thin line) and the 30-61 cm section of a segmented T D R vertical rod (thick line). The deep mineral layer (= 61-123 cm) was represented by the 61-92 cm (thin line) and 92-123 cm (thick line) sections of a segmented T D R vertical rod 77 Figure 3.3. Seasonal development of the forest (•), hazelnut understory (A) and aspen overstory (•) leaf area indices. Maximum leaf area indexes of 2.3, 3.3 and 5.6 m 2 leaf per rrf2 ground were obtained by the aspen, hazelnut and forest, respectively 80 Figure 3.4. The "flux footprint" or relative contribution to the flux measured at x = 0 m (l/Qo dQ(x)ldx) as a function of the upwind distance JC for z = 39 m and z = 4 m (area under each curve is unity). Neutral stability, typical daytime and nighttime footprints are shown. The latter were calculated by correcting for atmospheric stability using the median (z-d)/L during daytime (Rn > 0 W m"2) and nighttime (Rn < 0 W m' 2) periods.... 81 Figure 3.5. The cumulative flux footprint or cumulative relative contribution at an upwind distance* (Q(x)) to the total flux Q measured atx = 0 m forz = 39 m andz = 4 m. Neutral stability, typical daytime and nighttime footprints are shown. The latter were calculated by correcting for atmospheric stability using the median (z-d)IL during daytime (Rn > 0 W m~2) and nighttime (Rn < 0 W m"2) periods 83 xx i Figure 3.6. Ratio of net radiation (Rn), photosynthetically active photon flux (Qpi) and solar radiation (Rsi) measured above the hazelnut understory (z = 4 m) to that measured above the aspen (z = 39 m) plotted as a function of the seasonal development of aspen leaf area index (LAI). Points represent daytime means (Rn > 0 W m'2) and the solid line represents a curve fit of the form Qi(4 m)/()j.(39 m) = a exp(-baj) (see Appendix A, Table A.3. for curve-fit parameters) 85 Figure 3.7. Effective extinction coefficients (k) for net radiation (Rn), the photosynthetically active photon flux (Qpi) and solar radiation (Rsi) derived from the equation k = -ln(<2i(4 m)/(3i(39 m))/ap where ap is the aspen plant area index, the sum of aspen leaf and wood area indexes. Solid lines represent k calculated using the equation Qi(4 m)/(24,(39 m) = a exp(-ba\) obtained in Figure 3.6 87 Figure 3.8. Ensemble monthly averages of the forest (z = 39 m) diurnal net radiation (Rn, • ) , latent heat (AE, •) , sensible heat (H, • ) , soil heat (Go, • ) and total heat storage (Jt, •) flux densities. Each data point represents the mean of 2 1/2-h periods with ± one standard error of the mean shown by the vertical lines. The vertical bars represent the residual of energy balance closure, calculated as Rn - AE - H - Go - Jt 90 Figure 3.9. Ensemble monthly averages of the understory (z = 4 m) diurnal net radiation (Rn, • ) , latent heat (AE, •) , sensible heat (H, A) , soil heat (Go, • ) and total heat storage (Jt, •) flux densities. Each data point represents the mean of 2 1/2-h periods with ± one standard error of the mean shown by the vertical lines. The vertical bars represent the residual of energy balance closure, calculated as Rn - AE - H - Go-Jt 91 Figure 3.10. Seasonal progression of the daytime mean latent heat (thick line) and sensible heat (thin line) fluxes measured above the aspen canopy (39 m) and above the hazelnut understory (4 m). October and November measurements were made in 1993. Vertical dashed lines indicate the dates of changes in the surface conditions noted on the figure (April 15, May 21, September 21 and November 1) 98 Figure 3.11. The ratio of the daytime (net radiation positive) mean latent (thick line) and sensible (thin line) turbulent fluxes measured at the understory level (z = 4 m) relative to the overstory level (z = 39 m). Vertical dashed lines note the dates of changes in the surface conditions noted on the figure (April 15 and May 21) 100 Figure 3.12. Cumulative precipitation (P) and evapotranspiration (E) for 1994. The solid line for P does not include part of January preceding gauge installation whereas the dashed line is complete for 1994 with the missing January amount estimated from the 30-yr mean. Evapotranspiration was measured using the eddy-covariance method at the 39-m level (forest E) and 4-m level (hazelnut and soil) with cumulative 1/2-h totals calculated using three methods: 1) not corrected for nighttime underestimation (solid line); 2) corrected for nighttime underestimation (dashed line) and 3) corrected for energy balance closure (dash-dotted line). Soil water evaporation £(soil) (three overlapping lines) was estimated from mid-summer lysimeter measurements which indicated Zs(soil) = 0.05£(forest). Periods when measurements were not available were estimated using relationships between E and solar radiation 103 xx ii Figure 3.13. Soil water storage in the 0-123 cm layer relative to April 20 (initial measured soil water content of 292 mm) determined from volumetric soil moisture measurements (thick line) and cumulative precipitation-evapotranspiration (E) (thin lines) with the eddy flux E uncorrected for nocturnal heterogeneity (solid line), corrected for nocturnal heterogeneity (dashed line) or corrected for energy balance closure (dotted line). The differences between the thick and thin lines are estimates of drainage beyond 123cm (U = P-E- AS) 108 Figure 4.1. Seasonal course of the forest surface conductance,,^ (solid line) and aspen (dashed line) and hazelnut (dotted line) canopy conductance (gc). Panel A shows calculations based on the PM equation whereas panel B shows calculations based on canopy temperature (CT method). For comparison, the gs calculated from the PM equation is redrawn in Panel B. Daily values are plotted as a daytime (above-aspen Qpi > 200 //mol m"2 s"1), dry canopy running mean of 10 days 122 Figure 4.2. Relationship between forest surface conductance (gs) and forest LAI (Panel A) and aspen canopy conductance (gc) calculated using the PM equation and^s (Panel B). Mean values were calculated from the daytime mean gs corresponding to binned values of 1 m 2 m"2 (forest LAI) and the 1/2-h binned values of 32 mmol rrf2 s"1 (forest gs). Solid lines are linear regressions gs = 47.6ai + 56.2, rz - 0.98 (Panel A) and^c = 0.7Qgs + 4.9, r 2 = 0.99 (Panel B). Vertical lines represent ± one standard deviation and vertical bars represent the frequency distributions with 215 (days) and 1672 (1/2 hours) total number of samples, forest LAI and^ s. respectively 124 Figure 4.3. Comparison of 1/2-h aspen and hazelnut canopy conductances (gc) calculated either using the PM or CT equation. Mean gc = /(CT) was calculated corresponding to binned values of gc = /(PM) at 32 //mol m"2 s"1 intervals with ± one standard error of the mean represented by vertical bars. The linear regressions (solid lines) are^c/(CT) = 1.27 gcf(?M) - 12.3, r1 = 0.99 and gc /(CT) = 1.60 £ C/(PM) - 11.8, r = 0.93, aspen and hazelnut, respectively. The total number of observations were 1683 and 1171 1/2 hours, aspen and hazelnut, respectively 127 Figure 4.4. Ensemble monthly averages of the forest surface (gs; • ) , aspen canopy (•) and hazelnut (A) canopy conductances (gc) all calculated using the PM equation. Each data point represents the mean of 2 1/2-h periods with ± one standard error of the mean shown by vertical lines 129 Figure 4.5. Empirical relationships between the forest surface (gs) or canopy (gc) conductance and the 1/2-h saturation deficit at the big leaf surface (£>o) and for the full-leaf period when stratified by high (Qpi > 1400 //mol rrf2 s 1; • and solid line), medium (800 < Qpi < 1400 //mol m"2 s"1; • and dashed line) and low Qpi (200 < Qpi < 800 //mol rrf2 s"1; A and dotted line) photosynthetic photon flux density. Mean values of gs and gc were calculated at binned 0.25 kPa intervals of Do. Parameters defining the non-linear curve fits (lines) of the formes org c = g c m a x exp(-bD0) are given in Appendix A, Table A.4. Vertical lines are +, ± and - one standard deviation, (high, medium and low Qpi, respectively) 134 xxm Figure 4.6. Measured above aspen canopy (z = 39 m) 1/2-h photosynthetic photon flux density (Qpi) (thick line), air temperature (Ta) and saturation deficit (D3) (thin lines) representing a wide range in ambient conditions over which to test the empirical aspen gc calculations. Also shown are daytime, dry canopy air temperature (7b) and saturation deficit (Do) at the aspen canopy "big-leaf" surface (thick lines) 138 Figure 4.7. Measured (thin lines) and calculated (thick lines) daytime dry-canopy 1/2-h aspen canopy conductance (gc) and transpiration (AE) calculated using gc = a exp(-bDo) and AE using the PM equation. This is essentially a test of the Jarvis-Stewart approach using two variables, Do and light 140 Figure 4.8. Empirical relationship between the storage and low wind speed corrected C O 2 flux measured at the 4-m height (assumed to be a good estimate of soil respiration, ^ s o ii) and the mean 1/2-h nocturnal soil temperature at a depth of 2 cm (Ts). Mean soil respiration values were calculated at 1 °C Ts intervals with ± one standard error of the mean shown by the vertical lines. The solid line shows the best-fit exponential line given as Rsoii = 0.4349exp(0.20947's), r2 = 0.95. The 13.5 °C and 14.5 °C Ts bins were excluded from the curve fit. Vertical bars represent the T% frequency distribution with a total of 2611 1/2-h observations 143 Figure 4.9. Relationship between net C O 2 assimilation (An) by the aspen (•) and hazelnut ( O ) and incident photosynthetic photon flux density (Qpi) at the 39 and 4-m heights. Mean values of A n were calculated corresponding to a binned 1/2-h Qpi width of 100 (aspen) and 20 (hazelnut) //mol m~2 s*1. Solid lines represent a rectangular hyperbolic curve fits An = (aQpib)/(a Qpi + b) - R where a, b and R were 0.0259, 31.2123, -0.1646 (aspen) and 0.0786, 9,6638, -0.9381 (hazelnut), respectively. Aspen and hazelnut r2 were 0.99 and 0.88, respectively. Dashed line shows an extrapolation of the hazelnut Qpi to simulate overstory Qpi levels. The vertical lines are ± one standard error of the mean 145 Figure 4.10. The empirical relationship between the June 7 - September 10, 1994 aspen and hazelnut canopy conductance (gc) and AMX\ (BWB index) as suggested by Ball et al. (1987). A mean gc was calculated at binned BWB index values at 3 (aspen) and 1.5 (hazelnut) mmol m"2 s"1 intervals. The solid lines represents the linear regression £c(aspen) = 7.68 BWB index + 128.4 (r2 = 0.87) and ^ (hazelnut) = 2.84 BWB index + 86.1 (r2 = 0.36). Vertical lines represent ± one standard deviation (means of 144.2 and 104.8 mmol m"2 s"1, aspen and hazelnut, respectively). Vertical bars represent the frequency distributions with a total n of 1735 (aspen) and 1344 (hazelnut) 1/2 hours.. 146 Figure 4.11. Measured (thin lines) and calculated (thick lines) daytime dry-canopy 1/2-h aspen canopy conductance (gc) and transpiration (AE) with gc calculated using gc = a Anho/%1 + b and AE using the PM equation 148 xxiv Figure 4.12. The empirical relationship between the June 7 - September 10, 1994 aspen and hazelnut canopy conductance (gc) and a modified form of the BWB index where ho has been replaced by l/Do to improve the sensitivity of gc to humidity. A mean gc was calculated at binned modified BWB index values at 750 (aspen) and 300 (hazelnut) mmol 2 1 2 m" s~ intervals. The solid lines show linear regressions with slopes, y-intercepts and r of 0.035, 135.7 mmol rn"2 s\ 0.91 (aspen) and 0.031, 40.4 mmol m2 s"\ 0.95 (hazelnut), respectively. The vertical lines show ± one standard deviation (means of 152.1 and 76.1 mmol rrf2 s 1, aspen and hazelnut, respectively). Vertical bars represent the frequency distributions with a total n of 1717 (aspen) and 1278 (hazelnut) 1/2 hours 150 Figure 4.12. Measured (thin lines) and calculated (thick lines) daytime dry-canopy 1/2-h aspen canopy conductance (gc) and transpiration (AE) with calculated using gc = a AJiPo ZV>  + ° a r | d AE using the PM equation 152 Figure 4.13. The diurnal course of the dry-canopy 1/2-h Priestley & Taylor a during the daytime for August 13-18, 1994 (mean 1.11 ± a 0.31) 155 Figure 4.14. The seasonal pattern of the daytime mean, dry-canopy forest Priestley & Taylor a and the McNaughton & Jarvis decoupling coefficient 12. Thick lines show a 10-day running mean. Vertical dashed lines denote major changes in the surface conditions as noted on the figure 156 Figure 4.15. The relationship between the daytime mean forest dry canopy Priestley & Taylor a and the forest LAI (a\). The line represents the best-fit non-linear curve a= 1.19exp(l- (ai/1.70)), r 2 = 0.96 . The mean a values were calculated at a binned LAI (width 1 m 2 rrf2) with ± one standard deviation shown by vertical lines. Vertical bars show the forest LAI frequency distribution (total n = 223 days) 158 Figure 4.16. The relationship between the forest daytime mean, dry-canopy Priestley & Taylor a and the forest surface conductance (gs)- The line represents the best-fit non-linear curve a = 1.33exp(l- (gs/290.5)), r2 = 0.82, n = 142 days 159 Figure B . l . Plan view of the main tower study area showing major instrumentation locations 179 Figure B.2. Map of BOREAS southern study area showing OA (Old Aspen), YA (Young Aspen), OBS (Old Black Spruce), OJP (Old Jack Pine), YJP (Young Jack Pine) Fen sites and the BOREAS operations centre (Ops). Source: BOREAS Home Page on the World Wide Web 180 Figure B.3. View looking east standing on the main tower in the early-spring bare-canopy period. Note SRC tower near centre of photograph 181 Figure B.4. View looking east standing on main tower in the late-summer full-leaf period. Note SRC tower at bottom left of photograph 181 xxv Figure B.5. View of 37-m tall main tower from ground. Note sonic anemometer and infrared gas analyser (white box) at top of tower and temperature profile sensors (black cylinders) 182 Figure B.6. Understory canopy flux tower showing sonic anemometer 4-m above ground with heated sampling tube connecting to the infrared gas analyser (white box) 183 Figure B.7. Understory tram used to measure radiation levels along a 65-m transect. Visible are two net radiometers, two pyranometers and two quantum sensors 183 Figure B.8. Cross-sectional view of aspen 21.5-m tall, 70-yr old aspen stand and 2-m tall hazelnut understory. Such vegetation was typical around the study area 184 Figure B.9. Above aspen canopy eddy-covariance instrumentation 39-m above ground supported on top of the main tower. Visible are the Kaijo-Denki sonic anemometer (left), heated H 2 0-C0 2 (black), ozone (clear) and N 2 0-CH 4 (funnel) sampling tubes 185 xxvi A C K N O W L E D G M E N T S Any large project does not come to completion without the cooperation and support from many individuals. Their efforts are recognized here with apology for omitting anyone. I would like to thank my supervisor, Dr. T. Andy Black, for teaching me a great deal through his tireless energy and endless enthusiasm shown at all stages of this research. His momentum helped carry this research to its conclusion. This research was a collaborative project and could not be presented in its current form without the efforts of Drs. Harold H. Neumann and Gerry den Hartog of the Atmospheric Environment Service (AES), Downsview. They freely and openly gave access to their above-canopy data, expertise and equipment in the field and provided a valued friendship formed during the many hours in Saskatchewan. I hope I reciprocated in some way. My supervisor committee, Drs. Rob D. Guy, Mike D. Novak, Tim R. Oke and the external examiner, Dr. J. Harry McCaughey, sacrificed their time and gave their expertise to guide this research. Their efforts are sincerely appreciated. Financial support to myself was generously provided by the Natural Sciences and Engineering Research Council of Canada (NSERC) in the form a 2-year Postgraduate Scholarship and the University of British Columbia in the form of a 1-year Graduate Fellowship. Expert assistance with engineering and numerous technical issues was provided by Zoran Nesic. John Deary (AES) could always be counted on to fix anything and had a way of putting research into perspective. The friendship of Paul Yang in and out of the field was valued and appreciated. Additional field assistance was provided by Monica Eberle (UBC) and Tom Hertzog (AES). Logistics and site management was generously provide by Prince Albert National Park staff Mary Dahlman, Paula Pacholek and Murray Heap. Marian Breazu assisted with instrument construction and electronics, Mary Yang assisted with TDR data analysis and Ralph Adams assisted with the initial field setup. Jing Chen offered advice with the tram operation and LAI measurements. Craig Russell, Ralf Staebler, Isobel Simpson and Grant Edwards provided various forms of field assistance and companionship. Nigel Livingston and Bill Hook provided a segmented TDR rod, diode shorting units and advice on TDR measurement of soil water content. Friendship here at Annex 3 and a comfortable work environment was provided by Uwe Gramann, Siguo Chen, Xuhui Lee, Rick Ketler, Aisheng Wu, Alberto Orchansky, Wenjun Chen and Jon Warland. My family must also be thanked for their encouragement to pursue education throughout my academic career clear from kindergarten through Ph.D. It would have been easy to take many smoother roads in life. Finally and foremost I would like to thank my wife, Michele, for following me across the country in pursuit of my goals. Her encouragement and support made this work possible and enjoyable. The sacrifices she has made are truly appreciated. Our daughter, Claire, although presently too young to realize it, has also been an intangible motivation in both our lives. xxvn To M y Father xxvi i i CHAPTER 1 INTRODUCTION You see the gently rolling hills ahead of you, covered by the white-barked aspen with bright, trembling leaves growing on the glacial till covered by a hazelnut understory. Trying to reach the upland which would make travel much easier, you stumble over the hummocky, frost-heaved wet muskeg and step into the saturated black organic soil releasing a strong, pungent odour. Your goal is easily seen through the wide spaces between the drunken black spruce with their needles tightly clustered against their short, stunted branches. The cool breeze that easily passed through the sparse trees suddenly dies and you feel your constant companion, the mosquito, settle on your neck. P.D.B. This passage is easily related to by those who have walked through parts of the North American boreal forest which represents 25% of the world's forests (Salisbury & Ross 1978) and 29% of Canada's total land area (Rizzo & Wiken 1992). In the southern boreal forest, extensive stands of aspen are found (estimated to make up 10-20% of the southern boreal forest; Kabzems et al. 1986) often with a lush shrub understory permitted by the ample light penetration and relatively nutrient-rich soil. With the recent threat of global warming due to increasing atmospheric C O 2 levels, several studies have indicated that the mid- to high-latitude continental interiors {i.e. boreal forests) wi l l be most susceptible to warming (Davis & Botkin 1985, Tans et al. 1990, Denning et al. 1995). To improve our understanding of the boreal forest in global climate and improve process modelling of 1 Introduction 2 this ecosystem, the multidisiplinary, multiscale BOReal Ecosystem-Atmosphere Study (BOREAS) was undertaken (see Sellers et al. 1995 for an overview) of which this research was part. The objective of this research was to meet the prescribed BOREAS primary goal for tower-based measurements of quantifying the turbulent exchanges of energy and mass between the atmosphere and a specific surface type of the boreal forest (in this case, aspen stands) and to understand the processes controlling these exchanges. Two tower-flux research teams (TF1 and 2) enabled measurements to be made not only above but also within the aspen forest continuously throughout much of 1994, the main year for BOREAS field research. In addition, intensive measurements were made within the aspen canopy above a hazelnut understory in order to form a more complete understanding of the forest. This likely represents the most comprehensive study of boreal aspen stand physical climatology. The research presented here focused specifically on water vapour exchange, as C O 2 exchange is dealt with elsewhere (Black et al. 1996, Yang 1997). The eddy-covariance technique (also commonly referred to as eddy correlation; see Kanemasu et al. 1979 for a review) has become the method of choice of late to measure turbulent energy and mass exchange, especially over forests where high aerodynamic roughness weaken scalar gradients. A critical examination into the accuracy and validity of this technique to reliably measure turbulent fluxes not only within but also above the aspen canopy forms the basis of Chapter 2. While completing this examination, various aspects of turbulence above and within the forest are described, as turbulence forms the basis of eddy-covariance measurements. Corrections to eddy-Introduction 3 covariance measurements based on spatial heterogeneity and energy balance closure are also presented. Having dealt with the theory and technical issues of eddy-covariance, Chapter 3 presents measurements made with the eddy-covariance instruments, namely the seasonal energy and water exchange rates within and above the forest. After the climatic conditions experienced in 1994 are put into context, the source areas (fetch or footprints) at each measurement height are described. The effects of leafing both canopies on the surface energy balance then forms the majority of this chapter. The chapter concludes with an analysis of the seasonal forest water balance. The dominance of transpiration in the evapotranspiration stream is the rationale for the topic of Chapter 4, an analysis of the canopy conductance to water vapour for both the hazelnut and the aspen. Temporal trends are analyzed in terms of their relationship with leaf area and various combinations of environmental variables known to influence transpiration. After predicting canopy conductance from these relationships, this chapter ends by relating the forest's surface conductance to the regional scale meteorology. Chapter 5 concludes this thesis by providing a brief summary and general conclusions. Introduction 4 1.1. References Black, T. A., Den Hartog, G., Neumann, H. H., Blanken, P. D., Yang, P. C., Russell, C , Nesic, Z., Lee, X., Chen, S. C , Staebler, R. & Novak, M. D. (1996) Annual cycles of water vapor and carbon dioxide fluxes in and above a boreal aspen forest. Global Change Biology, 2, 219-229. Davis, M. B. & Botkin, D. B. (1985) Sensitivity of cool temperate forests and their fossil pollen to rapid climatic change. Quarterly Research, 23, 327-340. Denning, A. S., Fung, I. Y. & Randall, D. (1995) Latitudinal gradient of atmospheric C 0 2 due to seasonal exchange with land biota. Nature, 376, 240-243. Kabzems, A., Kosowan, A. L. & Harris, W. C. (1986) Mixedwood Section in an Ecological Perspective: Saskatchewan Technical Bulletin No. 8, Saskatchewan Parks and Renewable Resources, Saskatchewan. Kanemasu, E. T., Wesely, M . L., Hicks, B. B., & Heilman, J. L. (1979) Techniques for calculating energy and mass fluxes. In B. L. Barfield & J. F. Gerber (eds.) Modification of the Aerial Environment of Crops, American Society of Agricultural Engineers, St. Joseph, MI, pp. 156-182. Rizzo, B & Wiken, E. (1992) Assessing the sensitivity of Canada's ecosystems to climatic change, Climatic Change, 21, 37-55. Salisbury, F. B. & Ross, C. W. (1978) Plant Physiology, Second Edition, Wadsworth, Belmont. Sellers, P., Hall, F., Margolis, H., Baldocchi, D., den Hartog, G., Cihlar, J., Ryan, M . G., Goodison, B., Crill, P., Ranson, K. J., Lettenmaier, D. & Wickland, D. E. (1995) The Boreal Ecosystem-Atmosphere Study (BOREAS): An overview and early results from the 1994 field year. Bulletin of the American Meteorology Society, 76, 1549-1577. Tans, P. P., Fung, I. Y. & Takahashi, T. (1990) Observational constraints on the global atmospheric C O 2 budget. Science, 247, 1431-1438. Yang, P. C. (1997) Carbon Dioxide Exchange above a Hazelnut Understory in a Boreal Aspen Stand, Ph.D. thesis, University of British Columbia, Vancouver (in preparation). CHAPTER 2 CAN EDDY COVARIANCE MEASURE TURBULENT FLUXES WITHIN A BOREAL ASPEN FOREST ? 2.1. Introduction The economic importance and vast areal extent of the world's forests have made micrometeorological measurements of exchanges of heat, mass and momentum crucial for a complete understanding of global climatology. By virtue of the aerodynamic roughness of forests, above-canopy gradients of heat and mass are at best weak but can still be measured accurately enough to quantify the surface energy balance using gradient-based methods (e.g., Stewart & Thorn 1973, Denmead & Bradley 1985, Price & Black 1990) . Advances in instrumentation and the proliferation of computers has resulted in eddy-covariance (Kanemasu etal. 1979) as becoming a popular method for turbulent flux measurements. Comparisons between the conventional gradient and the eddy-covariance methods over forests typically reveal an underestimation of the surface energy balance by the latter (Spittlehouse & Black 1979, Shuttleworth et al. 1984, Barr et al. 1994). As our understanding of above-canopy processes has evolved, the need to better understand the role of the understory in forest micrometeorology has increased, prompted by several studies hinting at the possible importance of the understory and/or forest floor (e.g. Walker 1984, Kelliher & Black 1986, Black & Kelliher 1989, Baldocchi & Meyers 1991) . Coincidental with early within-canopy flux measurements, pioneering work by investigators such as Finnigan (1979a,b), Denmead & Bradley (1985), Raupach (1989) 5 Can Eddy Covariance Measure Turbulent Fluxes ? 6 and Thurtell (1989) and others all have shown flux measurements based on gradients often fail within canopies. Gradient-based flux measurements assume that steady, turbulent diffusion occurs along the mean concentration gradient (dc/dz) of the entity in question, i.e. F(z)= - K(z) dc/dz, where F is the flux density and K is the height and time dependent turbulent diffusivity for the entity in question. This linear relationship, also commonly referred to as /f-theory, fails within canopies because; /) length scales for vertical exchange exceed that for changes in the mean gradients and ii) mean concentrations are a combination not only of distant sources (far-field: independent of source) but also of concentration plumes arising from individual leaves (near-field: dependent on source) resulting in "bumps" on the concentration profiles and the resultant K more heavily weighted by near-field contributions and hence dependent on the source distribution (Wilson 1989). Since the eddy-covariance method does not rely on /C-theory, it is conceivably free to operate within plant canopies. A mean flux across a plane implies a correlation between the scalar in question and the wind component normal to that plane and the covariance between the two directly measures the flux with no assumptions about mixing properties of turbulence (Kaimal & Finnigan 1994). Despite this, there are several other assumptions related to the statistical properties of turbulence which may be questioned within the canopy space. The purpose of this chapter, therefore, is to thoroughly investigate these assumptions both above and especially within the boreal aspen canopy, to answer the question "can eddy-covariance measure turbulent fluxes within a boreal aspen forest ?". Can Eddy Covariance Measure Turbulent Fluxes ? 7 2.2. Site Description The study site (53.629 °N 106.200 °W) was located in Prince Albert National Park, approximately 50 km NNW of Prince Albert, Saskatchewan, Canada. An excellent and thorough description of the ecology of the region is presented by Peterson & Peterson (1992) but relevant site-specific details will be given here. The site lies near the southern limit of the boreal forest with the transition to the parkland region (rolling prairie with intermittent aspen groves) occurring approximately 15 km to the SW. Glacial erosion has left the region with a gently rolling topography (elevation of 400-700 m above sea level). Orthic gray luvisols with a loam to clay-loam texture have developed on Pleistocene deposits of glacial till, glaciolacustrine and glaciofluvial material. The site was moderately well-drained due to the elevated topography, coarse-grained glacial deposits and the frequent occurrence of intertill and surficial aquifers. A natural fire occurred approximately 70 years ago resulting in an even-aged stand of aspen (Populus tremuloides Michx.) with a mean canopy height (hc) of 21.5 m, a diameter at the 1.3 m height of 20 cm (standard deviation, cr± 4.5 cm) and a stem density of 830 stems ha"1. Canopy closure (bearing and transect length in parentheses) was 90% (270°, 500 m), 83% (120°, 300 m) and 94% (20°, 300 m) along three transects radiating away from the main tower (BOREAS Experimental Plan 1994). Crown space was limited to the upper 5-6 m beneath which was a branchless trunk space. The understory was dominated by a uniform cover of hazelnut (Corylus cornuta Marsh.) with a mean height of 2 m. Wild rose (Rosa woodsii) and alder (Alnus crispa) were also found intermittently. A variety of herbs and grasses were found along edges and clearings, but Can Eddy Covariance Measure Turbulent Fluxes ? 8 their occurrence was sparse beneath the dense hazelnut cover. This predominant aspen-hazelnut cover extended for at least 3 km in all directions. Infrastructure at the study site consisted of 37-m (main tower) and 4-m (understory tower) walk-up scaffold towers approximately 40 m apart, boardwalk access to the towers, 120-V AC power, and two heated huts in which computers, dataloggers and other sensitive instruments were housed. 2.3. Materials & Methods 2.3.1. Measurement of Above- & Within-canopy Eddy Fluxes The fluxes of latent heat (AE), sensible heat (H) and momentum (r) were measured above the aspen overstory (height above ground, z, = 39 m) and above the hazelnut understory (z = 4 m) using the eddy-covariance technique. Above-canopy sensors were supported by a vertical triangular mast directly on top of the main tower. For the understory flux measurements, sensors were supported by a 2.10 m horizontal boom fastened to the side of the understory tower at a bearing of 238°. The ratios of reference height-to-canopy height were 1.8 and 2.0 for the overstory and understory, respectively. Overstory instrumentation consisted of a 3-dimensional sonic anemometer (model DAT-310, Kaijo-Denki, Tokyo, Japan) with a 15-cm path length. Air was drawn at 6.5 L min"1 through a 6-m long 3.34-mm inner diameter sampling tube (model Bev-a-line, Thermoplastic Processes Inc., Sterling, NJ). Calculation of the Reynolds number (Re) at this flow rate (q) and sampling tube diameter (d) (Re = 4q/(iuiv), where v is the kinematic viscosity of dry air) gave 2800, above the critical Re of 2300 for turbulent flow in a pipe Can Eddy Covariance Measure Turbulent Fluxes ? 9 (Leuning & Judd 1996). Thus, air flow in the sampling tube was turbulent at all times. To prevent condensation in the sampling tube, heat (2-3 °C above ambient) was supplied by a voltage (= 18-V DC) passed through nichrome wire (20-AWG, « 15 ohms resistance) coiled around the exterior of the tube. Air was pushed through the sample cell of an infrared gas analyzer (IRGA) (model 6262, LI-COR Inc., Lincoln, NE) located in a temperature-controlled enclosure on the main tower by two diaphragm pumps (model TD-4X2N, Brailsford Co., Rye, NY) connected in parallel located upstream of the sample cell. This resulted in a sample cell pressure near atmospheric and a delay time of 1.2 s. The IRGA was operated in differential mode using the raw (non-linear) voltage output with 320 /umol mol"1 CO2 balanced in dry air flowing through the reference cell at 30 cm3 min"1. Understory instrumentation consisted of a 3-dimensional sonic anemometer (model 1012R2A Solent, Gill Instruments, Lymington, UK) with a 15 cm path length. Air passed through a heated (heat supplied in the same manner described for the 39-m flux measurement sampling tube) 3-m long 3.34-mm inside diameter sampling tube (model Bev-a-line, Thermoplastic Processes Inc.) before entering an IRGA (model 6262, LI-COR Inc.) located in a temperature-controlled enclosure on the understory tower. To raise the sampled air temperature to that of the IRGA sample cell, once inside the enclosure the air passed through 1.7 m of coiled copper tubing (3 mm inside diameter) sandwiched between heat-conducting aluminum plates before entering the IRGA's sample cell. Eliminating fluctuations in temperature before the air sample entered the IRGA reduced the undesirable effect of temperature fluctuations altering gas concentration calculations through changes in air density. A flow rate of 8.0 L min"1 Can Eddy Covariance Measure Turbulent Fluxes ? 10 produced by a diaphragm pump (model DOA-V191-AA, Gast Inc., Dayton, OH) located downstream of the sample cell resulted in a delay time of 0.8 s and a sample cell pressure 22 kPa below atmospheric. The Re with this flow rate and sampling tube diameter was 3400, well above the Re of 2300 at which turbulent flow is initiated. The IRGA operated in absolute mode with C02-free dry air flowing through the reference cell at 25 cm3 min"1. Voltage signals from the IRGA were obtained using the raw (non-linear) analyzer output. To determine delay times, evaluate possible attenuation of water vapour in the sampling tube (Leuning & King 1992) and provide an independent high frequency measurement of water vapour density (erv), a krypton open-path sensor (model KH20, Campbell Scientific Inc. (CSI), Logan, UT) was operated continuously with the understory IRGA. For both the overstory and understory systems, half-hourly fluxes were calculated as the covariance between the instantaneous vertical wind speed (w) and the instantaneous scalar quantity (r) (cov(wx) = w x =(w-w)(*-*)) where primes note deviations from the mean (overbars). Means were calculated as a block-average over a 1/2 h period with instantaneous measurements taken at 20 Hz with a analogue filter cut-off frequency of 10 Hz. Calculations were performed on-line by two 486 PCs with the raw data saved on tape drives. The above-canopy sensible heat flux was corrected for the effects of humidity on the sonic anemometer's calculation of air temperature (Schotanus et al. 1983). Since the IRGAs measure mole fractions (mol of H 2 0 or C 0 2 per total mol of all gases) and not mixing ratios (mol of H 2 0 or C 0 2 per mol of dry air), 1/2 h mean flux calculations included a correction for the effects of air density (Webb et al. 1980) Can Eddy Covariance Measure Turbulent Fluxes ? 1 1 and sample cell pressure. The Webb et al. (1980) correction was also applied to the open-path krypton hygrometer. At the overstory level, the mean vertical (iv) and lateral (v) velocity components and wV were rotated to zero following the procedure of Tanner and Thurtell (1969) {i.e. the anemometer signals were mathematically rotated so that there was no net upward or downward movement of air and the anemometer always faced the direction of the mean horizontal wind). At the understory level, however, only the mean lateral wind velocity component was rotated to zero under the suspicion that non-zero mean vertical velocities are possible within the trunk space (Baldocchi & Hutchison 1987). Measurement of the available energy (Ra) at both levels is given in Chapter 3, Section 3.2. 2.3.2. Turbulence Statistics & Parameters The third moment (skewness, Sk) for variable j was calculated as Sk j = j / ( c j where the overbar represents a 30-min temporal mean, prime denotes deviation from that mean and cr is the standard deviation of the mean. Turbulence intensity (/) for each velocity component was calculated as i. = G- / u . The Eulerian integral time and length scales (T and A , respectively) for each velocity component were calculated as where is the autocorrelation function between velocity component j and itself, t is time and £ is the time lag with respect to t. As integration to infinity was impossible, Ri} was (2.1) Can Eddy Covariance Measure Turbulent Fluxes ? 12 integrated until R„ reached zero as suggested by Kaimal and Finnigan (1994) with a § of 10 s. The power spectrum (S x(/)) and the cospectrum (C w x ( / ) ) of x and the vertical velocity, w, as a function of frequency (/) were calculated by MATLAB software (Math Works Inc., Natick, MA) using Welch's averages periodogram method (Krauss et al. 1993). Time series were divided into 12-min intervals and linearly detrended. Both the power and cospectra were multiplied by / to enhance the higher frequencies and Sx{ f) and C w x ( / ) were normalized by the variance of x and the covariance between w and x, respectively. Spectral analyses were performed 1/2-hourly from 10 am - 2 pm, August 4, 1994 (a typical clear full-leaf day) with the means from these 8 1/2 hours plotted. For clarity, means within 50 logarithmically equally spaced intervals were plotted. 2.4. Results & Discussion 2.4.1. Vertical Profiles of Turbulence Statistics Typical profiles of mid-summer leaf area density, turbulence skewness and turbulence intensity are shown in Figure 2.1. The small aspen LAI of 2.3 m 2 m"2 (see Chapter 3, Section 3.2.7. for determination of LAI) spread over a crown depth of 6 m resulted in a small leaf area density (LAD). In contrast, the slightly larger hazelnut LAI of 3.2 m2 m"2 contained over a shallow crown depth of 1.1 m resulted in a larger LAD. Above the canopy (z//?c = 1.81) skewness (Sk) for all three velocity components were close to zero (0.12, 0.09 and 0.003, Sku, Skv and Skw, respectively) indicative of a Gaussian distribution. Moving down past the sparse aspen canopy, Sku became positive by the Can Eddy Covariance Measure Turbulent Fluxes ? 13 2.0 1.5 -o 1.0 T N 0.5 -0.0 0 4-0.2 0.0 0.2 Leaf A r e a D e n s i t y S k e w n e s s ( m 2 leaf / m 2 g r o u n d per m height) 1.0 2.0 T u r b u l e n t Intensity Figure 2.1. Mean 1/2 h profiles of leaf area density, skewness and turbulent intensity for a typical mid-summer day (August 4, 1994) during the daytime 6 am - 6 pm (CST) unstable period. Total (projected) leaf area indexes were 2.3 and 3.2 m 2 leaf per m2 ground for the aspen and hazelnut, respectively. Average standard errors of the means for the longitudinal (streamwise) (•), lateral (A) and vertical (•) velocity components were 0.08, 0.07 and 0.05, respectively, for the skewness and 0.12, 0.27 and 0.05, respectively, for the turbulent intensities. Profiles were normalized by the mean aspen canopy height (Ac) of 21.5 m. Can Eddy Covariance Measure Turbulent Fluxes ? 14 roughly same amount that Skw became negative (0.20 and -0.19, Sku and Skw, respectively). A positive Sk indicates a frequency distribution with more occurrences above the mean rather than at the mean (Sk = 0) or below the mean (Sk < 0). Therefore, Figure 2.1 shows that in-canopy turbulence was dominated by intermittent downward (Skw < 0) penetrating gusts (Sku > 0). This observation of in-canopy turbulence dominated by fast, downward penetrating gusts agrees with the theoretical and measured findings of others in several canopy types (Baldocchi & Hutchison 1987, Shaw & Seginer 1987, Amiro 1990a, Lee & Black 1993, Kaimal & Finnigan 1994). The near-zero in-canopy Skv (-0.006) was expected (Seginer et al. 1976) and implies no occurrences of systematic wake circulation developed in the lee of the trunk spaces, as observed by Baldocchi & Hutchison (1987). Turbulent intensity (/') above the aspen canopy was 0.49 and 0.43 (u and v, respectively) and slightly less for w (0.26). Within the canopy, iu and iv increased almost identically, reaching values of 1.98 (/„) and 2.03 (z"v) atz/7zc = 0.19 while / w increased only to 0.69. Deflection of vertical motion into horizontal motion as the ground was approached was the probable cause of these increases in /. These values compare very well with the / profiles observed in a boreal aspen and pine forest (Amiro 1990a). Open-canopy boreal stands with a small LAD allow higher in-canopy wind speeds and hence a low i compared to denser stands with lower in-canopy wind speeds and hence high Ts. Despite the high / that occurs within the canopy, time and length scales are good indicators of the eddies that dominate turbulent exchange (Raupach et al. 1989) and are just as important as velocity moments in describing in-canopy turbulence (Kaimal & Finnigan 1994). Figure 2.2 shows typical mid-summer mean profiles of Eulerian integral Can Eddy Covariance Measure Turbulent Fluxes ? 15 0 0 ' 1 1 1 ' ' 1 i i I i i I i i . I . i i I i • • 0 30 60 90 0 4 8 12 Time Length Scale (s) Scale I hc Figure 2.2. Mean 1/2 h profiles of the Eulerian integral time and length scales for a typical mid-summer day (August 4, 1994) during the daytime 6 am - 6 pm (CST) unstable period. Average standard errors of the means for the longitudinal (streamwise) (•), lateral (A) and vertical (•) velocity components were 17, 3 and 6 s, respectively, for the time scale and 1.2, 0.9 and 0.2, respectively, for the length scales. Length scale profiles were normalized by the mean aspen canopy height (hc) of 21.5 m. Can Eddy Covariance Measure Turbulent Fluxes ? 16 time and length scales for all three velocity components. The integral time scale ( T), representing the time scales over which turbulence remained correlated (a measure of the turbulence persistence or memory) were 91, 26 and 18 s for u, v and w, respectively, above the aspen. Within the canopy, Tu and Tw decreased to 50 and 12 s, respectively, while Tv slightly increased to 47 s. The autocorrelation function used to calculate time and length scales showed a clear periodicity in w at z/hc = 1.81 with a period of 90 s during 6-7 am CST (August 4, 1994). At z/hc = 0.19, a clear periodicity in w with § = 6 0 s was also observed during the evening (5-6 pm CST August 4, 1994). This consistent wave pattern in w likely indicates the passage of gravity waves in w along a thermal temperature stratification which existed higher above the ground in the early morning and formed from the ground up in the late afternoon. The Eulerian integral length scales (A) were of the order 9.0hc, 2.6hc and l.7hc (Au, Av and / l w , respectively) at zlhc = 1.81. This was consistent with observations of Raupach (1989) who reports A u of the order of several hc and A w of the order hc. Moving down through the canopy to zlhc = 0.19 shows a decrease in A for all three velocity components (to 0A6hc, 0A6hc and 0.11/ic for A u , Av and A w , respectively) due to the cascade of energy to smaller eddies and the production of smaller eddies due to turbulent wakes (Raupach 1989). Despite being relatively deep within the aspen trunk space, energy-containing eddies were still large and illustrate that no appreciable daytime vertical scalar gradients would be expected within 0.1-0.5/ic. Can Eddy Covariance Measure Turbulent Fluxes ? 17 2.4.2. Spec t ra l A n a l y s e s o f T u r b u l e n c e The important contribution of large-scale motions to energy and mass transfer within the canopy was confirmed by spectral analyses which examine the importance of various frequency contributions to the time series. The power spectrum (Sx), showing the frequency contributions for u, v, w and Ta is shown in Figure 2.3. The two salient features from this plot are the spectral peaks and the slopes. Although difficult to discern, the positions in the spectral peaks for u, v, and w are at approximately 0.01 Hz both above and within the canopy. The calculation u If gives 31 m (or 1.4hc) and 204 m (or 9.5hc) at z = 4 and 39 m, respectively, a rough approximation of the size of the peak energy-containing eddies (u = 0.31 and 2.04 m s_1 at the 4 and 39 m levels, respectively). This calculation should match A u given previously and does above the canopy (Au = 9.0hc) although it is higher for within the canopy (Au = 0.46hc). Scaling Su by the natural frequency, n - fzlu , gives longitudinal velocity component spectral peaks of 0.19 and 0.13 above and within the canopy, respectively, close to the 0.15 reported by Amiro (1990b) for three boreal forests independent of height. The independence of the spectral peak (when scaled by n) from height indicates that large eddies dominate energy and mass transport through the entire depth of the canopy (Kaimal & Finnigan 1994). In the surface layer where small-scale turbulence is isotropic (independent of rotation and reflection), theory supported by measurements shows an inertial subrange where energy is neither created or destroyed but rather passed down to progressively smaller scales following a power-law slope of -2/3 (Kaimal & Finnigan 1994). Above the aspen canopy, Sx slopes in the inertial subrange were very close to the expected -2/3 CM b X Can Eddy Covariance Measure Turbulent Fluxes ? 1 0.1 0.01 0.001 1 0.1 0.01 0.001 0.0001 18 - 1—IT 1 l l l i p v 1 1 M IIIIIJ 1 l l l l l l l | o o - u ° O O 1 1 i r i I I I | V I I I IIIIIJ 1 l l l l l l l | 1 l l l l l l l [ ; O o --V oo\ : o o - o ! w i i i m i l l i i i m i l l i i i m i l l 1 i i m i l l i i i m i l l i i i m i l l i i i m i l l i i i m i l l 0.01 0.1 1 10 0.01 0.1 1 10 f ( H z ) Figure 2.3. Mean power spectra (Sx) for variable x where x is the longitudinal (u), lateral (v) or vertical (w) wind velocity component or air temperature (T3) above (O, z = 39 m;zlh c = 1.81) and within (9,z = 4 m; zlhc = 0.19) the canopy for 8 1/2-h periods from 10 am - 2 pm, August 4, 1994. Power spectra were multiplied by frequency ( / ) and normalized by the variance of*. The solid lines are the -2/3 slope expected in the surface layer inertial subrange. Above and within canopy u were 2.04 and 0.31 m s"\ respectively. Can Eddy Covariance Measure Turbulent Fluxes ? 19 for all three velocity components and temperature for at least three decades (Figure 2.3). The sharp roll-off at/= 5 Hz for above-canopy velocity components may indicate a high-frequency sampling problem in the Kaijo-Denki sonic anemometer, although it is puzzling why the same roll-off was absent with Sya. Where similarity between above-canopy Sx slopes indicates an isotropic inertial subrange, within-canopy Sx slopes were not similar or constant within the inertial subrange, indicating anisotropic turbulence (Figure 2.3). Prominent dips in Su and 5W followed by a secondary peak at higher/ were indicative of a "spectral short cut" where form drag by the flexible forest elements transfers energy directly to higher frequencies and avoids the energy cascade. A t / > 1Hz, velocity Sx slopes became stable with a slope closer to -1 than -2/3, indicating a faster than expected cascade of energy to smaller eddies, a finding also reported by others working within forest canopies (e.g. Amiro 1990b, Baldocchi & Meyers 1988, Meyers & Baldocchi 1993). Cospectral analyses (Cw x), showing the joint frequency contributions between w and the scalar x to the time series of the covariance or flux of scalar x, should show a slope of -4/3 in the inertial subrange (Kaimal & Finnigan 1994). Although difficult to interpret, Figure 2.4 shows that both above and within the canopy, C w x slopes were less than -4/3 through much of the inertial subrange, similar to the findings of Amiro (1990b) and Baldocchi & Hutchison (1988). Within the canopy, the dip in C w x for both H 2 0 and C 0 2 at 0.01< / < 0.02 Hz separating two spectral peaks may be the result of turbulent wake generated by the forest elements (Seginer etal. 1976). Plotting C w x against/ on a semi-log plot has the advantage of maintaining an area proportional to the covariance between w and x, thus allowing a direct indication of Can Eddy Covariance Measure Turbulent Fluxes ? 20 1 0.1 0.01 " x & 0.001 X 1 0.1 0.01 0.001 0.0001 ; 1 1 1 n T T T f y 1 M l l l l ) | 1 I I IMIIJ 1 l l l l l l l | T T T I I I I I f i y l " I T I i n i | 1 TTMIIIT 1 T TTTTTTj—= W \ D = Y \ ° E AJ \ O " a \ o " I ^ E U \ °° i i i m i l l i i i 1 i 1 i i i m ™ X ^ { i 1 i i 1 i i i m i l l i i i ni™ 0.01 0.1 1 10 0.01 0.1 1 10 f(Hz) Figure 2.4. Mean cospectra (Cwx) between the vertical wind speed (w) and the variable x where x is air temperature (ra), H 2 O mole fraction (xl), longitudinal wind velocity (w) or C O 2 mole fraction above (O, z - 39 m; z/hc = 1.81) and within (•, z = 4 m; zlhc -0.19) the canopy for 8 1/2-h periods from 10 am - 2 pm, August 4, 1994. Cospectra were multiplied by frequency (/) and normalized by the covariance between w and x. Solid line is the -4/3 slope expected in the surface layer inertial subrange and the dashed line is the -1 slope. Above and within canopy u were 2.04 and 0.31 m s'\ z - 39 m and 4 m, respectively. Can Eddy Covariance Measure Turbulent Fluxes ? 21 frequency contributions to the flux (Figure 2.5). The most important feature of Figure 2.5 is the similarity between the traces both above and within canopy (CO2 at the 4-m level appears positive since the C O 2 flux at that level was upward). This similarity indicates that the same turbulent transfer mechanisms were responsible for vertical flux exchange and the peaks at low frequencies indicate that infrequent coherent structures were the dominant transfer mechanism. There was almost no contribution to the fluxes at frequencies greater than 1 Hz, indicating sampling frequencies were adequate with no loss of high-frequency fluxes at both levels. There was, however, a greater high-frequency contribution to the fluxes within the canopy than above. Close agreement between C w x for water vapour measured both with the open- and closed-path sensors within the canopy indicated little adsorption of water vapour on the walls of the sampling tube (see Figure 1 in Black etal. 1996). 2.4.3. Effect of Leafing on Within-canopy Horizontal Wind & Friction Velocity In the previous section, a typical day during the full-leaf period was chosen as this is the period when energy and mass exchange were most active (see Chapter 3, Section 3.3.4.). In this section, the effect of aspen leafing on u and the friction velocity (u*) is described. The pre-leaf period (pre June 1, 1994) aspen canopy was characterized by a wood area index (WAI) of 0.62 and, since there were no leaves, a plant area index (PAI) of 0.62, comparable to pre-leaf values measured in other deciduous forests (e.g. 0.50 mixed temperate forest (Neumann et al. 1989), 0.60 oak/hickory forest (Tennessee) (Hutchison etal. 1986), 0.48 oak forest (Russia) (Rauner 1976). The development of aspen leaves Figure 2.5. Mean cospectra (Cwx) between the vertical wind speed (w) and variable * where* is air temperature (Ta), H 2 0 mole fraction (%1), longitudinal wind velocity (w) or C0 2 mole fraction (x\) above ( o ; z = 39 m; z/hc = 1.81) and within (•; z = 4 m; zlhc = 0.19) the canopy for 8 1/2-h periods from 10 am - 2 pm, August 4, 1994. Cospectra were multiplied by frequency (/) and normalized by the covariance between w and*. The area under each curve equals one. Above and within canopy u were 2.04 and 0.31 m s"1, z = 39 m and 4 m, respectively. Can Eddy Covariance Measure Turbulent Fluxes ? 23 (LAI of 2.3) increased the aspen (PAI) to 2.92, a 371% increase over the pre-leaf value but still not large compared to more southerly forests. The mean u above the aspen canopy of 3.11 m s"1 (cr ± 1.49 m s"1) during the aspen pre-leaf period was similar to that (2.60 m s"1, cr± 1.14 m s"1) during the aspen leafed period (June 2 - September 7, 1994) with any differences probably not due to leaves, but simply due to stormier conditions during the springtime. During both periods, the skewness of u was slightly negative, indicating a higher occurrence of u below rather than above the mean (Figure 2.6). Within the canopy, however, pre-leaf mean u of 0.62 m s"1 (cr± 0.29 m s"1) fell to 0.39 m s"1 (cr± 0.19 m s"1) with aspen leaves above. Mzlhc = 0.19, the skewness of u was positive during both periods (even more positively skewed during the leafed period) indicating a dominance of u above rather than below the mean. This confirms the result given in Section 2.4.1. that gusts are prominent within the canopy and even more so during the leafed period. The pre-leaf and leafed period ratios of the mean u at z/hc = 0.19 to z/hc = 1.81 were 0.20 and 0.15, respectively, showing that even without leaves, the aspen canopy's branches and boles were sufficient to retard a substantial amount of u. Friction velocity above the canopy was not affected by the presence of leaves (mean u* = 0.50 m s"1 (cr± 0.30 m s"1) without leaves; mean u* = 0.41 m s"1 (cr± 0.26 m s"1) with leaves). A dominance of energy transfer by horizontal gusts was indicated by the positive skewness of u* during both periods. A substantial decrease in the mean «* at zlhc = 0.19 from 0.11 (cr ± 0.08 m s"1) to 0.06 (cr ± 0.05 m s"1), pre-and post-leafed, respectively, was observed as a function of aspen leafing, again with a positive skewness Can Eddy Covariance Measure Turbulent Fluxes ? 24 0.0 2.0 4.0 6.0 8.0 B o h -Figure 2.6. Relationship between friction velocity («*) and horizontal wind speed (w), above (z = 39 m; zlhc = 1.81) and within (z = 4 m; z / / i c = 0.19) the aspen canopy during pre-leaf (O) (before June 1, 1994) and leafed period (•) (June 2, 1994 to September 7, 1994). Mean values of u* calculated correspond to binned values of u with a bin width of 0.4 and 0.1 m s'1 (39 and 4 m heights, respectively). Mean standard errors of the mean were 0.07 and 0.04 m s"1 (39 m, pre-leaf and leafed, respectively), and 0.02 and 0.006 m s"1 (4 m, pre-leaf and leafed, respectively). Frequency distributions for the pre-leaf (clear) and leafed (shaded) canopies at both heights are shown by the vertical bars. Total number of 1/2-h observations were 5391 and 2691 (bare canopy 39 and 4 m, respectively) and 4533 and 4699 (leafed canopy 39 and 4 m, respectively). Can Eddy Covariance Measure Turbulent Fluxes ? 25 during both periods indicating the dominance of energy transport within the canopy by horizontal gusts. Ratios of the mean u* at z/hc = 0.19 to zlhc - 1.81 were nearly identical to that of the mean u, with values of 0.22 and 0.15 (pre-leaf and leafed periods, respectively). The relationship between u and u* and its dependence on overstory leaf development showed no difference in the magnitude of u* at a given u until a u of approximately 4.0 m s"1 was reached. Thereafter, aspen leaves were effective in generating more u* than a bare canopy. Within the aspen canopy, hazelnut leaves generated more u* when a u of approximately 0.8 m s"1 was reached. Between u of 0.4 and 0.8 m s"\ slightly more u* was generated without rather than with leaves. In general, regardless of the canopy's leafing condition, the slopes between u and u* above and within the canopy were similar. The ratio of the mean ulu* was 0.16 independent of zlhc or leaf development, and falls towards the lower end of the 0.16-0.25 range given as typical for forests (Douw Steyn, pers. comm. 1996). 2.4.4. Monin-Obukhov Similarity The atmospheric surface layer is defined as the layer immediately above the surface where vertical variations in fluxes of momentum, heat and mass are small. The height where the rate of decrease in the flux with height exceeds 10% is generally used to define the vertical extent of the surface layer (Panofsky & Dutton 1984). Within the surface layer, Monin and Obukhov (1954) hypothesized that several atmospheric parameters and statistics become universal functions of height (normalized by the scaling length, L) when normalized by the scaling velocity u* or temperature (Kaimal & Finnigan 1994). Can Eddy Covariance Measure Turbulent Fluxes ? 26 Moreover, eddy-covariance measurements of fluxes within the surface layer demand that turbulence statistical properties are independent of horizontal position (horizontal homogeneity) and do not change with time (stationarity: time averaging averages the process, not changes in the process over time). The former can be assessed by comparing the Monin-Obukhov (MO) surface layer similarity function for w (0W= crju*) as a function of atmospheric stability (z-d)/L where d is the zero-plane displacement, well approximated by 2/3/zc (Brutsaert 1984), to the measured fa, (Foken & Witchura 1996). Several empirical fits to this relationship have been found {e.g. Dyer 1974, Merry & Panofsky 1976, Panofsky et al. 1977, Hogstrom 1988) and the common and accepted form of this equation for unstable atmospheric stability is 0W = crw / = a(\ + ^(z-d)/ Z,|)* where a and b are empirical coefficients. Use of this equation both above and within the aspen canopy (Figure 2.7) described <p^ very well, not only for unstable conditions but also for stable conditions. This implies that measurements made at these heights were within the surface layer which was unique at z/hc = 0.19 since this indicates a constant-flux layer had indeed developed within the canopy making eddy-covariance measurement possible. A second constant-flux layer within a trunk space has also been reported by others (Denmead & Bradley 1985, Lee & Black 1993). Limits of -2 < (z - d)IL < +1 for 0W were imposed since the free convective asymptotic form is reached at (z - d)IL = -1 and 0w becomes poorly behaved and no longer responds to surface layer scaling at (z- d)IL > 1 (Kaimal & Finnigan 1994). The frequency distribution shows that the majority of (z - d)IL were near neutral at both levels and not many observations were outside the imposed range for <pw. Can Eddy Covariance Measure Turbulent Fluxes ? 27 (z-d)IL Figure 2.7. Relationship between the standard deviation of vertical wind (<7W) normalized by the friction velocity («*) at z = 39 m; zlhc = 1.81 (•) and z = 4 m; zlhc -0.19 (•) and atmospheric stability, (z - d)IL. Triangles show crw(4 m) /«*(39 m). Mean values of aju* were calculated corresponding to binned values of (z - d)IL of 0.25 with a mean standard error of 0.11 (39 m) and 0.21 (4 m). Non-linear regression fitted equations (solid lines) are of the form expected by Monin-Obukhov similarity theory within the surface layer {a = 1.107, 1.2636, 0.3147 and b = 0.2120, 0.2750, 0.1460, • . • . • , respectively. Frequency distributions for (z - d)IL at the z = 39 m {zlhc = 1.81) (clear) and z = 4 m (zlhz = 0.19) (shaded) levels are shown by the vertical bars (total n of 7283 and 5084 1/2 hours, 39 m and 4 m, respectively). Can Eddy Covariance Measure Turbulent Fluxes ? 28 The non-linear least squares determined values of the coefficient b for (b = 0.21 = 1/5 and 0.28 = 1/3, at 39 m and 4 m, respectively) were close to the most often reported value of 1/3 at z = 4 m, but below the expected value at z = 39 m. As reported by Roth & Oke (1996), this implies enhanced momentum transport above compared to within the aspen canopy due to increased surface roughness. The values of the coefficient a representing <^  in purely mechanical turbulence of 1.11 (39 m) and 1.26 (4 m) were close to the 1.25 value reported for flat terrain independent of the surface cover (Panofsky & Dutton 1984). Normalizing crw(4 m) by w*(39 m) still gave a well defined relationship described by the form given above with a and b coefficients of 0.32 and 0.15, respectively. This value of a agrees closely with other in-canopy results and model predictions for a wide variety of canopies (Kaimal & Finnigan 1994). 2.4.5. Homogeneity of Turbulent Fluxes As indicated earlier, the eddy-covariance technique has become a popular means of measuring surface fluxes. Over the last year or two, this technique has been widely adopted as a means of monitoring long-term CO2 exchange in order to better understand global climatic change. As nocturnal respiration can represent a substantial proportion of the net ecosystem exchange, the correct measure of nocturnal eddy fluxes is critical in annual carbon balance estimations. During the night, however, surface temperature variations are minimized and hence u and u* are suppressed, especially within the canopy. The assumption of homogeneity is then violated if there is insufficient turbulent mixing to create a horizontally homogenous flux. During periods when homogeneity did not Can Eddy Covariance Measure Turbulent Fluxes ? 29 occur, CO2 fluxes must often be recalculated as a biological function of soil surface or air temperature. Most often, heterogeneity has been assessed indirectly using the magnitude of u* as the criterion. For example, Table 2.1 shows the minimum value of w* various investigators have used to reject CO2 fluxes determined using eddy covariance (F c). It is important to note that all of these studies were above-canopy (z/hc > 1). Analysis of our data (Figure 2.8) shows that while «* > 0.30 m s"' was appropriate at zlhc = 1.81 for systematic underestimation of Fc, use of this w* rejection value would have resulted in the rejection of most of the within-canopy Fc measurements, when in fact systematic underestimation did not occur until u* < 0.10. When u* was low, Fc(4 m) did not fall to zero as it did at the 39-m level presumably due to erratic episodes of CO2 emission under nocturnal in-canopy temperature inversions. This analysis was only performed during the nighttime (available energy, Ra < 0 W rrf2) since during the daytime, turbulent fluxes may be systematically underestimated not only because of low u*, but also because of low R3. No dependence of any of the turbulent fluxes on Ra was found during the nighttime period. The discrepancy between the measured oju* and that predicted by M O theory fell to less than 20% at both heights when u* reached its threshold values of 0.30 m s"1 (39 m) and 0.10 m s"1 (4 m) (Figure 2.8). This shows the shortcoming of using a minimum u* to reject eddy fluxes, since the minimum u* depends on z. Use of the difference between aju* and that predicted by M O theory was independent of z and hence served as a more robust criterion. Moreover, if heterogeneity in Fc, leading to systematic underestimation Can Eddy Covariance Measure Turbulent Fluxes ? 30 Table 2.1. Minimum nocturnal friction velocities used by several investigators to reject CO2 fluxes due to heterogeneity resulting from insufficient turbulent mixing. Canopy Type Canopy Height (m) Measurement Minimum Height (m) u* (m s"1) Reference Boreal Black Spruce 10-11 27 0.40 Jarvis etal. (1997) Boreal Aspen 21.5 39.0 0.30 this study Boreal Black Spruce 10 29 0.24 Goulden etal. (1997) Temperate Deciduous 24 35 0.20 Greco & Baldocchi (1996) Temperate Deciduous 20-24 31.0 0.17 Goulden etal. (1996) Boreal Hazelnut 2.0 4.0 0.10 this study Can Eddy Covariance Measure Turbulent Fluxes ? 31 c/> CM - 0.4 0.2 0.0 o £ . o 4.0 -0.12 0.06 0.00 2.0 -0.0 0.24 0.12 -l 0.00 0.00 0.04 0.08 0.12 um (m s"1) c aj -t—> o Figure 2.8. Dependence of the C 0 2 flux (F c , • ) and turbulence homogeneity (O) on friction velocity («*) at night (available energy < 0 W m"2). Means were calculated corresponding to binned values of u* at 0.05 (39 m) and 0.01 m s"1 (4 m) intervals. Mean standard errors were 0.26 (39 m) and 0.49 (4 m) //mol m"2 s"1 for F c and 0.01 (39 m) and 0.02 (4 m) for turbulence homogeneity. The vertical bars show the u* frequency distributions (total n of 1804 and 1860 1/2 hours, 39 m and 4 m, respectively). Can Eddy Covariance Measure Turbulent Fluxes ? 32 occurs at low u*, then the other turbulent fluxes which depend on the same assumptions and theory must also be systematically underestimated. Figures 2.9 and 2.10 show that in agreement with Figure 2.8, systematic underestimation occurred when u* falls below 0.30 m s"' (39 m) and 0.10 m s"1 (4 m), corresponding to oju* within 20% of that predicted by M O theory for both latent (AE) and sensible (H) heat fluxes. Just as respiration from the soil can be recalculated with some confidence from soil surface or air temperature (see Chapter 4, Section 4.3.4.3.), AE and H can be recalculated from the available energy (/?J and the Bowen ratio /? = HlAE (see Chapter 2, Section 2.4.7.). Alternatively, all turbulent fluxes can be recalculated given some empirical quantification of the amount of underestimation due to lack of turbulent.mixing resulting in heterogeneity. Figure 2.11 shows the relationship between the ratio of the measured nocturnal turbulent fluxes to the expected flux (mean flux that occurs with u* above the minimum values of 0.30 (39 m) and 0.10 (4 m)) was best described by the empirical fits of 4-parameter sigmoidal logistic functions (see Appendix A , Table A . l for parameters). Above the aspen canopy, one curve could describe all three fluxes, while below the aspen canopy, a separate curve was needed for F c since some C 0 2 emissions occur due to wave turbulence along the in-canopy nocturnal temperature inversions. Given the relationships shown in Figure 2.11, the mean 1/2-h nocturnal flux underestimation during the full-leaf period due to heterogeneity (caused by insufficient turbulent mixing) was 28% (a± 33%) for F c , AE and H at the 39-m level, 32% (a± 27%) for Fc at the 4-m level and 55% (cr± 42%) for AE and H at the 4-m level. Can Eddy Covariance Measure Turbulent Fluxes ? 33 24.0 0.00 0.04 0.08 0.12 u. (m s" 1) Figure 2.9. Dependence of the latent heat flux {AE, • ) and turbulence homogeneity (O) on friction velocity (u*) at night (available energy < 0 W m"2). Means were calculated corresponding to binned values of u* at 0.05 (39 m) and 0.01 m s"1 (4 m) intervals. Mean standard errors were 1.32 (39 m) and 0.78 (4 m) W m" 2 for AE and 0.01 (39 m) and 0.02 (4 m) for turbulence homogeneity. Vertical bars show the u* frequency distributions (total n of 1804 and 1860 1/2 hours, 39 m and 4 m, respectively). Can Eddy Covariance Measure Turbulent Fluxes ? 34 -20.0 -CN -40.0 --60.0 - 0.4 - 0.2 0.0 O O 0.12 0.06 c 03 O !— 0.00 0.0 -4.0 -8.0 -0.00 0.0 -J 0.00 0.24 0.12 0.04 0.08 u* (m s" 1) 0.12 "TO -4—» o !— Figure 2.10. Dependence of the sensible heat flux (H, •) and turbulence homogeneity (O) on friction velocity («*) at night (available energy < 0 W m 2). Means were calculated corresponding to binned values of u* at 0.05 (39 m) and 0.01 m s"1 (4 m) intervals. Mean standard errors were 2.07 (39 m) and 0.75 (4 m) W rn 2 for H and 0.01 (39 m) and 0.02 (4 m) for turbulence homogeneity. Vertical bars show the u* frequency distributions (total n of 1804 and 1860 1/2 hours, 39 m and 4 m, respectively). Can Eddy Covariance Measure Turbulent Fluxes ? 35 0.00 0.03 0.06 -1-0.09 u, (m s" ) Figure 2.11. Relationship between the measured turbulent flux divided by the expected flux ( • = CO2 flux, (Fc); • = latent heat flux (AE); A = sensible heat flux (//)) and the nighttime friction velocity («*) (binned at 0.015 (39 m) and 0.005 (4 m) m s"1 intervals). The expected flux was the mean 1/2-h flux that occurred when u* exceeded 0.30 (39 m) and 0.10 (4 m) m s'1 which were: 3.92 and 4.44 //mol m"2 s"1 (Fc); 15.8 and 4.9 W m" 2 (AE) and -49.2 and -6.4 W m"2 (H) for the 39 and 4 m levels, respectively. The solid line is a non-linear least square sigmoidal 4-parameter logistic function fit (see Appendix A , Table A . l for parameters). Mean standard errors of the mean were 0.07 //mol m*2 s*1, 0.09 W m" 2 and 0.05 W m"2 for the (39 m Fc, AE and H, respectively) and 0.13 //mol m" 2 s~\ 0.15 W m"2 and 0.07 W m"2 for the (4 m Fc, AE and H, respectively). Can Eddy Covariance Measure Turbulent Fluxes ? 36 The effect of correcting underestimated nocturnal fluxes due to insufficient u* using the relationships shown in Figure 2.11 is shown in Figures 2.12 and 2.13. Over much of the consecutive 15-day period, the nocturnal average F c at both levels was slightly larger than the uncorrected Fc underestimated due to a low u*. Surprisingly, there appears to be no relationship between short-term soi! surface temperature (Ts) and Fc. Especially at the 4-m level, the low w* resulted in estimates of zero for Fc and the sum of AE and H. Correction for flux underestimation, however, adjusts the fluxes to make them more closely follow /? a (see following Section). 2.4.6. Closure of the Surface Energy Balance Perhaps one of the best assessments of the ability of eddy covariance to measure turbulent fluxes is to quantify the extent to which AE + H = Ra (energy balance closed when left-hand side equals right-hand side). Often, investigators present energy balance closure in the form of 24-h means which tend to give an unrealistically good assessment since energy storage terms and erratic fluctuations are cancelled. In fact, Baldocchi & Meyers (1991) state that "Hence we conclude that eddy flux measurements from a deciduous forest floor are valid and reliable if data from many runs (1/2-h) are lumped together and averaged." (p. 7276). Estimates of energy balance closure based on 24-h means (all in W m"2) over the entire measurement period were AE + H = 0.98 /? a + 3.4 (r2 = 0.93) and AE + H = 1.01 Ra + 1.3 (r2 = 0.88) at the 39- and 4-m levels, respectively. Plotting the half-hourly AE + H against Ra for both the 39-m (AE + H = 0 .87 Ra + 14.2; rz = 0.93) and 4-m (AE + H = 0.84 Can Eddy Covariance Measure Turbulent Fluxes ? 37 Nigh t t ime Ju ly 1994 Figure 2.12. Nighttime mean measured (•) available energy (/?a), soil temperature at a depth of 2 cm (Ts), horizontal wind speed (u), friction velocity («*), C 0 2 flux (Fc) and the sum of the latent (AE) and sensible (H) heat fluxes at the 39-m height. Turbulent fluxes corrected for heterogeneity due to low values of u* using the relationships shown in Figure 2.11 are shown by (•). Can Eddy Covariance Measure Turbulent Fluxes ? 38 CM 1 0 E -10 CO -20 T — 0.6 0.3 CT) CM O E . o 8 4 0 1 I 1 I . | 1 1 1 1 , , , , . , -- \ / • — • ^ » -1 . 1 1 1 . 1 7^ *—w-y f T T 1 10 12 14 16 18 20 22 24 N igh t t ime Ju ly 1994 14 12 0.6 0.3 O o to in E CM 1 20 10 0 3: + Figure 2.13. Nighttime mean measured (•) available energy (Ra), soil temperature at a depth of 2 cm (Ts), horizontal wind speed (w), friction velocity («*), CO2 flux (Fc) and the sum of the latent (AE) and sensible (H) heat fluxes at the 4-m height. Turbulent fluxes corrected for heterogeneity due to low values of u* using the relationships shown in Figure 2.11 are shown by (•). Can Eddy Covariance Measure Turbulent Fluxes ? 39 Ra + 5.2; r2 = 0.77) levels indicated an average underestimation of Ra by 13 and 16%, respectively. Plots of energy balance closure against wind direction showed no discernible pattern, indicating lack of adequate fetch was not an issue. Figure 2.14 shows that average daytime values of (AE + H)IRa (slope of the regression line) were 0.95 (39 m) and 1.09 (4 m) indicating energy balance closure within 10% at both levels. During the night, however, energy balance closure was poor at both levels with {AE + H)IRa values of 0.53 (39 m) and 0.19 (4 m). Application of the low u* corrections described above greatly improved the nocturnal energy balance with revised estimates of (AE + H)IRa being equal to 0.65 and 0.76 for the 39 and 4 m levels, respectively. The frequency distributions of Ra (Figure 2.14) show a bimodal distribution at night above the aspen canopy indicative of clear and cloudy skies, while the bimodal distribution was absent beneath the aspen canopy due to the constant presence of the canopy above. 2.4.7. Unexplained Short-term Variability in the Eddy Fluxes & Data Quality Checks Besides the lack of energy balance closure, inspection of the diurnal patterns of the eddy fluxes at both levels often revealed a pronounced "saw-blade" pattern in both AE and H, especially on clear days when Rn was steady during the day (Figure 2.15 and 2.16). The sudden decrease in both AE and H was physically unrealistic since there were no coincidental changes in any of the variables known to decrease stomatal conductance and if such a decrease in AE did occur, H should have increased due to an increase in leaf temperature. Poor energy balance closure was associated with the trough values of AE Can Eddy Covariance Measure Turbulent Fluxes ? 40 R a (W m"2) Figure 2.14. Energy balance closure for above (A and B) and below (C and D) the aspen canopy separated into nighttime (A and C) and daytime (B and D) periods for the entire measurement period. Means of the measured sum of latent (AE) and sensible iH) 1/2 hourly fluxes (•) were based on binned measurements of available energy (Ra) with bin widths of 4, 15, 1.5 and 5 W m"2, for plots A, B, C and D, respectively. Estimates of the nighttime fluxes corrected for underestimation based on friction velocity (•) were closer to the 1:1 line (solid line). Linear regression (dashed lines) parameters are given in Appendix A, Table A.2. Frequency distributions of Ra are shown by vertical bars. Can Eddy Covariance Measure Turbulent Fluxes ? 41 600 400 -200 -0 300 200 -100 -0 -100 0 12 18 24 Time August 4, 1994 CST (h) Figure 2.15. Example of the recalculation of the overstory (39 m) latent (AE) and sensible (H) heat eddy fluxes (thick lines) on August 4, 1994 using the available energy and the ratio of the measured eddy fluxes. The original fluxes (dashed lines) show half-hourly variation not explained by changes in net radiation (upper thin solid line) even when some of the original points (•) were suspect (O) based on stationarity and homogeneity tests. Can Eddy Covariance Measure Turbulent Fluxes ? 42 200 <r 1 0 0 E 0 = 40 -20 -0 -20 0 12 18 24 T ime A u g u s t 4, 1994 CST (h) Figure 2.16. Example of the recalculation of the understory (4 m) latent (AE) and sensible (H) heat eddy fluxes (thick lines) on August 4, 1994 using the available energy and the ratio of the measured eddy fluxes. The original fluxes (dashed lines) show half-hourly variation not explained by changes in net radiation (upper thin solid line) even when some of the original points (•) were suspect (O) based on stationarity and homogeneity tests. Can Eddy Covariance Measure Turbulent Fluxes ? 43 and H. During these events, however, the ratio H/AE remained relatively constant. This behaviour was also observed in 1996 with a Gill R3 sonic anemometer replacing the Kaijo-Denki unit used in 1994. To investigate this problem, a typical cloud-free day with good examples of the saw-blade pattern both above and within the canopy was chosen (August 4, 1994) for an in-depth examination of possible causes of these closure errors. This evaluation was performed by first checking for missing high-frequency data, incorrect latent and sensible storage measurements or abrupt changes in flux footprints due to changing wind direction or atmospheric stability. The effect of mathematical sonic anemometer axis coordinate rotation was also investigated. Compliance with the statistical assumptions for eddy-covariance measurements, namely stationarity and homogeneity, was then checked. Inspection of the high-frequency data showed continuous data without any gaps. Fluctuations in water vapour density measured above the understory by the open-krypton) and closed- (IRGA) path sensors were identical, indicating no problems with the IRGA's response. Half-hourly fluctuations in AE of 100 W m"2 at the 39-m height would have to be accompanied by a corresponding change in the average vapour pressure in the 39-m air column of 0.25 kPa per 1/2-h for storage changes to account for the AE fluctuations. Vapour pressure measurements at the 4- and 39-m heights changed in concert and did not show such a large change in vapour pressure, supporting the hypothesis that the variability in AE was at times unrealistic and not accounted for by a changes in storage. Variability in the eddy fluxes was not associated with changes in flux footprints (see Chapter 3, Section 3.3.2.). The Webb et al. (1980) density correction of Can Eddy Covariance Measure Turbulent Fluxes ? 44 AE was small, increasing the average understory closed-path 24-h flux by 2% compared to the uncorrected value. The effect of mathematical sonic anemometer coordinate rotation was also small (August 4 24-h understory AE means of 32 W rrf2 with and 30 W rrf2 without coordinate rotation) and confirms the indications from Skv that recirculating eddies developed within the lee of trunks were not experienced and the understory >v could have been rotated to zero. None of the half hourly variations in the fluxes could be explained by the checks described above. Stationarity of turbulence statistics was assessed by re-calculating AE and H from 6 am to 6 pm (CST) August 4, 1994 for integration periods of 1, 2, 5, 10, 15 and 30 minutes for both the 4- and 39-m levels with the half-hourly fluxes calculated as the average of the fluxes determined over the shorter intervals. In addition, the effect of linear detrending versus block averaging over each averaging period was investigated. An example using AE is given (Figures 2.17 and 2.18) with similar results found for//. At both levels, as the integration periods decreased both the 1/2-h to 1/2-h variability and magnitude of AE decreased, especially in the late afternoon. This suggests that the late afternoon decrease in AE and associated variability may have been caused by large coherent turbulent structures which were well-developed later in the day due to strong surface heating and a relatively deep convective boundary layer (CBL). As illustrated by Figure 2.19, integration periods longer than 15 min (regardless of block averaging or linear detrending) were necessary to capture 90% of the 6 am - 6 pm mean AE {i.e. to avoid significant loss of low-frequency contributions). By adopting the stationarity test that a flux calculation should be rejected when the difference between the flux calculated with an averaging period of 30 min and the flux Can Eddy Covariance Measure Turbulent Fluxes ? 45 i i I i i I i i I i i 6 9 12 15 18 T ime (h) CST A u g u s t 4, 1994 Figure 2.17. Effect of various integration periods on the 1/2-h latent heat flux (AE) at the 39-m height for a typical mid-summer day. Means for each integration period were determined either by a simple non-overlapping block average (•) or by non-overlapping linear detrending (•). Can Eddy Covariance Measure Turbulent Fluxes ? 46 160 80 0 80 0 80 CNJ i E 0 80 UJ 0 80 0 80 0 10 min 5 min - 2 min - 1 min 6 9 12 15 18 T ime (h) CST A u g u s t 4, 1994 Figure 2.18. Effect of various integration periods on the 1/2-h latent heat flux (AE) at the 4-m height for a typical mid-summer day. Means for each integration period were determined either by a simple non-overlapping block average (•) or by non-overlapping linear detrending (•). Can Eddy Covariance Measure Turbulent Fluxes ? 47 12 5 10 15 30 In tegrat ion Per iod (min) Figure 2.19. Mean daytime (6 am - 6 pm) latent heat flux density (AE) calculated as a function of various integration periods with either non-overlapping block averages (•) or non-overlapping linear detrends (•) normalized by flux calculated with a 30-min integration period (163 and 157 W m~2, block averaged and linear detrended, respectively, at the 39 m height; 59 and 54 W m"2, block averaged and linear detrended, respectively, at the 4 m height). Can Eddy Covariance Measure Turbulent Fluxes ? 48 calculated with an averaging period of 5 min exceeds 20% (Foken & Wichura 1996), some of the unrealistic flux measurements were objectively rejected (Figures 2.15 and 2.16). Several methods exist for determining time-averages of a scalar where the average deviations from this mean determine the flux {e.g. non-overlapping or overlapping (moving) block averages, non-overlapping or overlapping linear or non-linear detrending). It is, however, suggested that some form of detrending be used since any trends violate the stationarity assumption Kaimal & Finnigan (1994) and Panofsky & Dutton (1984) go as far as to recommend linear detrending over a 1-h period. In and above the aspen forest, Figures 2.17, 2.18 and 2.19 show that block averaging or linear detrending does not have a large effect on the calculation of the flux, especially when integration periods of 15 min or more were used. Linear detrending, however, greatly reduced unreasonable flux variations at the 4-m level and some at the 39-m level. Block averaging tended to give larger H fluxes than linear detrending (larger by 4 and 12% for the mean 6 am - 6 pm H on August 4 at 39 and 4 m, respectively) for example, by misinterpreting the morning steady trend of increasing Ta as a low frequency contribution to H. Homogeneity of turbulence statistics was assessed using the theory of flux-variance similarity. Spatial heterogeneity in turbulence caused either by an increase in mechanical turbulence due to obstacles (instrumentation, tower, undulating or uneven surface) and/or variability in surface temperature and water vapour sources will cause a departure in the ratio, oju*, from that predicted using similarity theory (Foken & Wichura 1996). Following Foken & Wichura (1996), a flux calculation was questioned and considered suspicious when the difference between aju* and that predicted by MO Can Eddy Covariance Measure Turbulent Fluxes ? 49 theory as illustrated in Figure 2.7 exceeded 20%. Use of the stationarity check above the canopy resulted in a lack of confidence in some flux measurements on August 4 (Figures 2.15 and 2.16). Over the entire 1994 measurement period, 13% (39 m) and 41% (4 m) of the AE and H eddy-flux 1/2-h measurements were deemed suspect as indicated by this stringent homogeneity test with the majority of the suspicious measurements occurring with low u*. Although the above data quality assessment procedure did eliminate many suspect flux measurements, many still remained and could not be objectively removed except for an adjustment based on energy balance closure. Since the worst erratic flux patterns appear in the afternoon on clear days, it was suspected that local, large turbulent structures recognizable by large temperature ramps (typical duration of 6 min) were not properly spatially averaged by eddy flux measurements taken at one fixed point. Similar structures in Ta with periods of 2-3 min with a clear saw-tooth appearance indicative of well-developed convective cells have been by observed by Roth & Oke (1996). Others have proposed objective methods to achieve energy balance closure when the energy balance is overestimated (e.g. Steyn 1985). This situation was one of turbulent flux underestimation and to correct for both energy balance closure and erratic eddy fluxes, it was assumed that the ratio of H to AE was correct (i.e. relatively constant Bowen ratio, /3) even if the individual magnitudes were not. This assumption is equivalent to the cospectral similarity approach (Hicks & McMillen 1988) which assumes the ratio of the two fluxes can be approximated by the ratio of their respective covariance over a restricted frequency band width. Hicks and McMillen (1988) used this approach to correct for the loss of high-frequency contributions undetected by slow response sensors. Can Eddy Covariance Measure Turbulent Fluxes ? 50 We, however, did not limit ourselves to restricted frequencies since we did not have slow response sensors or flux contribution from high frequencies. To recalculate the eddy fluxes, the ratio B = H/AE was calculated where H and AE are the half-hour fluxes as measured using eddy covariance. Using this ratio in the energy balance allows the fluxes to be recalculated as AE' = RJ{\ + 0) and H'= fiAE '. This is equivalent to partitioning the missing energy flux unaccounted for by the ratio of the original eddy-covariance measurements of H and AE. This approach assumes reliable estimates of Rn, Go and Jt and has been used for eddy flux measurements determined using aircraft (Banetal. 1997). This technique was applied to all of the daytime data and an example of the effect of the technique is shown in Figures 2.15 and 2.16. Clearly the correlated "saw-blade" behaviour in AE and H was reduced. During the daytime (Ra > 0) in this example, the mean forest AE and H were increased by 7 and 8%, respectively at z = 39 m, and decreased by 2 and 20%, respectively, at z = 4 m compared to the uncorrected block-averaged eddy fluxes. Over the entire measurement period in 1994, the daytime mean forest AE and H were increased by 9 and 12%, respectively, while the understory daytime mean AE and H were increased by 16 and 13%, respectively. In addition, the degree of scatter in physiological plots such as canopy conductance to water vapour against saturation deficit (see Chapter 4, Section 4.3.4.1.) was greatly reduced when this technique was used. Can Eddy Covariance Measure Turbulent Fluxes ? 51 2.5. S u m m a r y & C o n c l u s i o n s Comparison of turbulence statistics above and beneath the boreal aspen overstory has shown that despite the relatively sparse aspen canopy, the canopy and trunks were sufficient to produce typical beneath-canopy turbulence. In-canopy turbulence was characterized by large eddies generated from intermittent gusts of fast, downward penetrating air. Power spectral analyses confirmed the energy-containing structures are of the order 9-9.5/ic (above overstory) and 0.46-1.4hc (above understory). Above the aspen canopy, power spectra slopes followed the expected -2/3 slope through three decades in the inertial subrange, whereas above the hazelnut canopy, power spectra slopes were discontinuous and closer to -1 at high frequencies indicating a faster than expected cascade of energy. Cospectral analysis revealed that turbulent transfer mechanisms were similar above and below the aspen overstory, with virtually no contribution to scalar fluxes at frequencies greater than 1 Hz, and energy roll-offs were less than expected. It is these large, intermittent, energy and mass transferring eddy structures that both preclude the use of AT-theory gradient-diffusion methods within the canopy and at the same time give the potential for eddy-covariance flux measurements. Strong, long-term events are easier to detect than weak, short-term events making in-canopy eddy flux measurements easier. Despite the re-development of a constant-flux layer within the canopy trunk space, heterogeneity of turbulent statistics often develops at low wind speed where there is a lack of turbulent mixing within the canopy. Suspect data during such periods are best identified not by use of the friction velocity, but by the Monin-Obukhov stability function for the vertical wind velocity since the former is height-dependent while Can Eddy Covariance Measure Turbulent Fluxes ? 52 the latter is not. Suspect data may then be removed or re-calculated based on empirical functions. Alternatively, the surface energy balance can be re-calculated based on partitioning of available energy flux between sensible and latent heat fluxes. The answer to the question "can eddy-covariance measure turbulent fluxes within a boreal aspen canopy" is a conditional "yes". In addition to attempting to answer this question, issues concerning the measurement of eddy fluxes above the aspen canopy have also surfaced. It is often assumed that above-canopy eddy fluxes are accurate, but as shown here, issues such as energy balance closure and turbulence stationarity and homogeneity must be addressed. The conditions imposed on the response are that the measurements are made at a height where a constant-flux layer has re-developed (i.e. not too close to the aspen canopy above or hazelnut canopy below) and the heterogeneity caused by low wind speeds is at least recognized and corrected if possible. Other factors such as limited and variable flux source areas (flux footprints) must also be taken into account (see Chapter 3, Section 3.3.2.). It is recommended that understory flux measurements made at this particular boreal aspen site: /) be made at a height not less than 2 m above the hazelnut understory since below that height near-field effects from the hazelnut may be felt as well as the spatial heterogeneity of the roughness sublayer just above the hazelnut canopy. Turbulence intensities and length scales may also decrease under the influence of the forest floor. Using a z too large, say 0.5Ac(aspen) would take advantage of higher turbulent intensities but increase the chance of fetch advection (horizontal flux divergence); ii) be calculated using a non-overlapping linear detrending algorithm to determine means. Although the effect is minimal, this would still eliminate some overestimation of fluxes; iii) be averaged with an integration period of not less then Can Eddy Covariance Measure Turbulent Fluxes ? 53 15 min to avoid loss of low-frequency eddy contributions; iv) be corrected for nocturnal heterogeneity either by independent recalculation of the CO2 flux from soil or air temperatures or by recalculation of all turbulent fluxes based on the empirical estimate of flux underestimation based on the friction velocity or Monin-Obukhov similarity theory. At the very least, if not recalculated, fluxes during such periods should be taken as suspicious and v) be taken concurrently with measurements of available energy to allow the recalculation of the sensible and latent heat fluxes in a meaningful manner at a 1/2 h time resolution. Can Eddy Covariance Measure Turbulent Fluxes ? 54 2.6. Re fe rences Amiro, B . D. (1990a) Comparison of turbulence statistics within three boreal forest canopies. Boundary-Layer Meteorology, 51, 99-121. Amiro, B . D. (1990b) Drag coefficients and turbulence spectra within three boreal forest canopies. Boundary-Layer Meteorology, 52, 227-246. Baldocchi, D . D. & Hutchison, B . A . (1987) Turbulence in an almond orchard: vertical variations on turbulent statistics. Boundary-Layer Meteorology, 40, 127-146. Baldocchi, D . D . & Meyers, T. P. (1988) Turbulence structure in a deciduous forest. Boundary-Layer Meteorology, 43, 345-364. Baldocchi, D . D . & Meyers, T. P. (1991) Trace gas exchange above the floor of a deciduous forest 1. Evaporation and CO2 efflux. Journal of Geophysical Research, 96, 7,271-7,285. Barr, A . G. , Betts, A . K . & MacPherson, J. I. (1997) Boundary-layer budgets of sensible and latent heat above boreal forest. Submitted to Journal of Geophysical Research. Barr, A . G . , K i n g , K . M . , Gillespie, G . , den Hartog, G . & Neumann, H . H . (1994) A comparison of Bowen ratio and eddy correlation sensible and latent heat flux measurements above deciduous forest. Boundary-Layer Meteorology, 71, 21-41. Black, T. A . & Kelliher, F. M . (1989) Processes controlling understory evapotranspiration. Philosophical Transactions of the Royal Society of London, Series B, 324, 207-231. Black, T. A . , Yang, P. C , Blanken P. D. , Nesic, Z. , den Hartog, G . , Neumann, H . H . , Lee, X , Chen S. G . & Novak, M . D. (1996) Eddy-correlation meaurements of water vapor and CO2 fluxes above the understory of a boreal aspen forest. 2 2 n d Conference of Agricultural and Forest Meteorology with Symposium on Fire and Forest Meteorology, Jan. 28 - Feb. 2, 1996, Atlanta, 66-69. B O R E A S Experimental Plan, Chapters 1-3, Version 3.0 (1994) P. J . Sellers, F. G . Hal l , D . D . Baldocchi, J . Cihlar, P. C r i l l , G . den Hartog, B . Goodison, R. D. Kel ly, D . Lettenmeier, H . Margolis, J. Ranson, & M . Ryan, N A S A , Greenbelt. Brutsaert, W. (1984) Evaporation into the Atmosphere. Reidel, Dordrecht. Denmead, O. T. & Bradley, E . F. (1985) Flux-gradient relationships in a forest canopy. In B . A . Hutchison & B . B . Hicks (eds.) The Forest-Atmosphere Interaction, 421-442. Can Eddy Covariance Measure Turbulent Fluxes ? 55 Dyer, A . J. (1974) A review of flux-profile relationships. Boundary-Layer Meteorology, 7, 363-372. Finnigan, J. J. (1979a) Turbulence in waving wheat. I. Mean statistics and Honami. Boundary-Layer Meteorology, 16, 181-211. Finnigan, J . J . (1979b) Turbulence in waving wheat. II. Structure of momentum transfer. Boundary-Layer Meteorology, 16, 213-236. Foken, Th. & Wichura, B . (1996) Tools for quality assessment of surface-based flux measurements. Agricultural and Forest Meteorology, 78, 83-105. Greco, S. & Baldocchi, D . D . (1996) Seasonal variatios of CO2 and water vapour exchange rates over a temperate deciduous forest. Global Change Biology, 2, 183-197. Goulden, M . J. , Daube, B . C. , Fan, S - M . , Sutton, D. J., Bazzaz, A . , Munger, J . W . & Wofsy, S. C. (1997) Gross CO2 uptake by a black spruce forest Journal of Geophysical Research, (in press). Goulden, M . J. , Munger, J . W. , Fan, S - M . , Daube, B . C. & Wofsy, S. C . (1996) Measurement of carbon sequestration by long-term eddy covariance: methods and a critical evaluation of accuracy, Global Change Biology, 2, 169-182. Hogstrom, U . (1988) Nondimensional wind and temperature profiles. Boundary-Layer Meteorology, 42, 55-78. Hare, F. K . & Thomas, M . K . (1974) Climate Canada. Wiley, Toronto. Hicks, B . B . & M c M i l l e n , R. T. (1988) On the measurement of dry deposition using imperfect sensors and in non-ideal terrain. Boundary-Layer Meteorology, 42, 79-94. Hodges, G . B . & Smith, E . A . (1995) Optimal estimates of surface net radiation field over B O R E A S study-area from combination of net pyrradiometer point measurements and G E O S satellite retrievals. Poster Presentation, B O R E A S Workshop, Calverton, Maryland, October 17-20, 1995. Hutchison, B . A . , Matt, D . R., M c M i l l e n , R. T., Gross, L . J., Tajchman, S. J. & Norman, J . M . (1986) The architecture of a deciduous forest canopy in Eastern Tennessee, U . S . A . . Journal of Ecology, 74, 635-646. Jarvis, P .G. , Massheder, J. M . , Hale, S. E. , Moncrieff, J. B. , Rayment, M . & Scott, S. L . (1997) Seasonal variation of carbon dioxide, water vapour and energy exchanges of a boreal black spruce forest. Journal of Geophysical Research, (in press). Can Eddy Covariance Measure Turbulent Fluxes ? 56 Kaimal , J. C. & Finnigan, J . J. (1994) Atmospheric Boundary Layer Flows: Their Structure and Measurement. Oxford, New York. Kanemasu, E . T., Wesely, M . L . , Hicks, B . B. , & Heilman, J. L . (1979) Techniques for calculating energy and mass fluxes. In B . L . Barfield & J. F. Gerber (eds.) Modification of the Aerial Environment of Crops, American Society of Agricultural Engineers, St. Joseph, M I , pp. 156-182. Kelliher, F. M . & Black, T. A . (1986) Estimating the effects of understory removal from a Douglas fir forest using a two-layer canopy evapotranspiration model. Water Resources Research, 22, 1891-1899. Krauss, T. P., Shure, L . & Little, J . N . (1992) Signal Processing Toolbox for use with MATLAB. Math Works Inc., Natick. Lee, X & Black, T. A . (1993) Atmospheric turbulence within and above a Douglas-fir stand. Part I: Statistical properties of the velocity field. Boundary-Layer Meteorology, 64, 149-174. Lee, X & Black, T. A . (1993) Atmospheric turbulence within and above a Douglas-fir stand. Part II: Eddy fluxes of sensible and latent heat. Boundary-Layer Meteorology, 64, 369-389. Leuning, R. & Judd, M . J . (1996) The relative merits of open- and closed-path analyzers for measurement of eddy fluxes. Global Change Biology, 2,241-253. Leuning, R. & King , K . M . (1992) Comparison of eddy covariance measurements of C 0 2 fluxes by open- and closed-path CO2 analyzers. Boundary-Layer Meteorology, 59, 297-311. Merry, M . & Panofsky, H . A . (1976) Statistics of vertical motion over land and water. Quarterly Journal of the Royal Meteorological Society, 102, 255-260. Meyers, T. P., & Baldocchi, D . D . (1993) Trace gas exchange above the floor of a deciduous forest 2. SO2 and 0 3 deposition. Journal of Geophysical Research, 98,12,631-12, 638. Monin , A . S. & Obukhov, A . M . (1954) Basic laws of turbulent mixing in the ground layer of the atmosphere. Transactions of the Geophysical Institute Academy, Nauk USSR, 151, 163-187. Neumann, H . H . , den Hartog, G . & Shaw, R. H . (1989) Leaf area index measurements based on hemispheric photographs and leaf-litter collection in a deciduous forest during autumn leaf-fall. Agricultural and Forest Meteorology, 45, 325-345. Panofsky, H . A . & Dutton, J . A . (1984) Atmospheric Turbulence: Models and Methods for Engineering Applications. John Wiley and Sons, New York. Can Eddy Covariance Measure Turbulent Fluxes ? 57 Panofsky, H . A . , Tennekes, H . , Lenschow, D. H . & Wyngaard, J. C. (1977) The characteristics of turbulent velocity components in the surface layer under stable conditions. Boundary-Layer Meteorology, 11,355-361. Peterson, E . B . & Peterson, N . M . (1992). Ecology, management, and use of aspen and balsam poplar in the prairie provinces, Canada. Forestry Canada Northwest Region, Northern Forest Centre, Edmonton, Alberta, Special Report 1. Price, D . T. & Black, T. A . (1990) Effects of short-term variation in weather on diurnal canopy CO2 flux and evapotranspiration of a juvenile Douglas-fir stand. Agricultural and Forest Meteorology, 50, 139-158. Rauner, J . U . L . (1976) Deciduous Forests, In J . L . Monteith (ed.) Vegetation and the Atmosphere, Volume 2, Case Studies. Academic Press, London, pp. 241-264. Raupach, M . R. (1989) Turbulent transfer in plant canopies, In G . Russell, B . Marshall & P. G . Jarvis (eds.) Plant Canopies: Their Growth, Form and Function. Cambridge University Press, Cambridge, pp. 41-61. Raupach, M . R., Finnigan, J . J . & Brunet, Y . (1989) Coherent eddies in vegetation canopies. Fourth Australian Conference on Heat and Mass Transfer, 9-12 May 1989, Christchurch, New Zealand, pp. 75-90. Roth, M . & Oke, T. R. (1996) Relative efficiencies of turbulent transfer of heat, mass, and momentum over a patchy urban surface. Journal of the Atmospheric Sciences, 52, 1863-1874. Schotanus, P., Nieustadlt, F. T. M . & de Bruin, H . A . R. (1983) Temperature measurement with a sonic anemometer and its application to heat and moisture fluxes. Boundary-Layer Meteorology, 26, 81-93. Seginer, I, Mulhearn, P. J., Bradley, E . F. & Finnigan, J. J . (1976) Turbulent flow in a model plant canopy. Boundary-Layer Meteorology, 10, 423-453. Shaw, R. H . & Seginer, I. (1987) Calculation of velocity skewness in real and artificial plant canopies. Boundary-Layer Meteorology, 39, 315-332. Spittlehouse, D . L . & Black, T. A . (1979) Determination of forest evapotranspiration using Bowen ratio and eddy correlation measurements. Journal of Applied Meteorology, 18, 647-653. Stewart, J. B . & Thorn, A . S. (1973) Energy budgets in pine forest. Quarterly Journal of the Royal Meteorological Society, 99, 154-170. Steyn, D. G . (1985) A n objective method to achieve closure of overdetermined surface energy budgets. Boundary-Layer Meteorology, 33, 303-310. Can Eddy Covariance Measure Turbulent Fluxes ? 58 Shuttleworth, W . J. , Gash, J . H . C., Lloyd, C. R., Moore, C. J., Roberts, J., de O. Marques, A . , Fisch, G. , de P. Silva, V . , Ribeiro, M . N . G. , Mol ion , L . C. B . , de Abreu Sa, L . D. , Nobre, J. C , Cabral, O. M . R., Patel, S. R. & de Moraes, J . C . (1984) Eddy correlation measurements of energy partitioning for Amazonian forest. Quarterly Journal of the Royal Meteorological Society, 110, 1143-1162. Tanner, C. B . & Thurtell, G . W . (1969) Anemoclinometer Measurements of Reynolds Stress and Heat Transport in the Atmospheric Boundary Layer. Research and Development Technical Report ECOM-66-G22F , University of Wisconsin, Madison, Wisconsin. Thurtell, G . W . (1989) Comments on using K-theory within and above the plant canopy to model diffusion processes. In T. A . Black, D . L . Spittlehouse, M . D . Novak & D . T. Price (eds.) Estimation of Areal Evapotranspiration, I A H S Press, Wallingford, Publication No. 177, pp. 81-85. Walker, G . K . (1984) Evaporation from wet soil surfaces beneath plant canopies. Agricultural and Forest Meteorology, 33, 259-264. Webb, E . K . , Pearman, G . I. & Leuning, R. (1980) Correction of flux measurements for density effects due to heat and water vapor transfer. Quarterly Journal of the Royal Meteorological Society, 106, 85-100. Wilson, J. D . (1989) Turbulent transport within the plant canopy. In T. A . Black, D . L . Spittlehouse, M . D . Novak & D . T. Price (eds.) Estimation of Areal Evapotranspiration, I A H S Press, Wallingford, Publication No. 177, pp. 43-80. CHAPTER 3 S E A S O N A L ENERGY & W A T E R E X C H A N G E A B O V E & WITHIN A B O R E A L A S P E N FOREST 3 . 1 . I n t r o d u c t i o n Without water there is no forest. This statement summarizes the rationale for a study of the energy and water exchange of any forest. Forests are found only in regions with ample precipitation and the vigor of the forest is often reflected directly by the amount of precipitation. Forests affect the local water balance by transpiring water to the atmosphere, developing a high-porosity organic soil horizon and by intercepting precipitation with subsequent evaporation. The water exchange of a forest is a function of the available energy and the direct link between the energy and water exchange is made via the latent heat flux. A knowledge of forest energy and water exchange is fundamental for an understanding of the Earth's climate. Despite the areal expanse of the Canadian boreal forest, this landscape has not been extensively studied and warrants further research such as the B O R E A S project (Sellers et al. 1995). In addition, this region is sensitive to climate change (Rizzo & Wiken 1992) and may be the first ecosystem to respond to anthropogenic induced climatic changes. Deciduous aspen forests are found extensively throughout North America with a northern limit roughly corresponding to the 13 °C July isotherm (Peterson & Peterson 1994). Of the areas where Populus is the prominent genus, over 71% (ground area basis) occurs in the boreal forest region with 20-40% of Canada's aspen/poplar stands located in the prairie provinces (Peterson & Peterson 1994). Aspen has been 59 Seasonal Energy & Water Exchange 60 recognized as an economically important hardwood species with a variety of usages such as pulp, strandboard, lumber and plywood, fuel, shingle and shakes, etc. with harvesting recently increasing. The deciduous nature of aspen demands that the energy and water exchanges be studied, if possible, over an entire season. The sparse and trembling nature of aspen crowns which is exaggerated by northerly locations often allows a sufficient amount of light penetration and the proliferation of understory species such as the hazelnut found at this study site. Whether the lush shrub understory found in aspen stands, in contrast to the moss and lichen layer with few shrubs typical of boreal conifer stands, is a result of the ample light penetration or due to the relatively higher nutrient status of Populus ecosystems is not clear, however exceptionally well-developed shrub understories are typical under aspen stands (Peterson & Peterson 1992). The extensive areal cover and relatively large leaf area of this understory species raises the question of how important is the understory relative to the overstory in energy and water exchange. The deciduous nature of the understory hazelnut recommends that measurement be made here too throughout an entire growing season. The purpose, therefore, of this chapter is to describe the seasonal energy and water exchange both above and within a boreal aspen forest. The ambient conditions experienced in 1994 (air temperature, precipitation, soil water content, leaf growth) are first described since this study was specific to the conditions experienced in one year. This is followed by micrometeorological details concerning the areal extent of tower measurements (i.e. fetch or flux footprints). The effect of leafing on radiation penetration is shown followed by description of the energy balance above and within the canopy for a Seasonal Energy & Water Exchange 61 typical day in each month from Apr i l through September. The seasonal playoff between sensible and latent heat is illustrated and the chapter concludes with an analysis of the seasonal water balance. 3.2. Materials & Methods 3.2.1. Above- & Within-canopy Eddy Flux Measurements Measurements of the latent {AE) and sensible (H) heat fluxes above the aspen canopy (z = 39 m) and above the hazelnut understory canopy (z = 4 m) were made using the eddy-covariance technique which has been thoroughly described previously (see Chapter 2, Section 2.3.1. and overview by Baldocchi et al. 1988). To summarize, the three components of wind velocity (vertical, lateral and longitudinal) and air temperature were measured at each level at high frequency (20 Hz) with sonic anemometers-thermometers. The water vapour and CO2 mole fraction scalars were measured with infrared gas analyzers located at each level. Turbulent scalar eddy fluxes were calculated for each 1/2-h as the covariance between the vertical wind speed (w) and the scalar x (flux = cov(w,x) = w'x = (w - w ) ( * - x ) ) where the prime denotes deviations from the 1/2-h mean (overbars). Upward fluxes are considered positive and downward fluxes are considered negative. 3.2.2. Above-canopy Radiation Measurements Net radiation (Rn) (model S - l , Swissteco Instruments, Oberriet, Switzerland and model C N - 1 , Middleton Instruments, Melbourne, Australia), incident solar radiation (direct and Seasonal Energy & Water Exchange 62 diffuse, R%i) (model PSP, Eppley Inc., Newport, RI) and incident photosynthetically active radiation (Qpi) (model 190-SB, L I - C O R Inc.) were measured at 33-m above the ground from the main tower. The net radiometers were supported at the end of a 3-m long horizontal boom extending south of the tower. A l l instruments were recorded on a datalogger (model 21X, Campbell Scientific Inc. (CSI), Logan, UT) . Prior to installation in February 1994, both net radiometers were calibrated at the Atmospheric Environment Service (AES) radiation laboratory facility in Downsview, Ontario. A t the end of the 1994 field season, the Swissteco radiometer was sent back to the A E S laboratory where the sensitivity (W m" 2 m V ' 1 ) was found to be less than .2% above the pre-field season calibration. In-situ short-wave calibration was performed on Apri l 11 and September 19, 1994. During clear-sky conditions, the Eppley pyranometer and the net radiometers were simultaneously shaded with small shades 2 m from the sensors in order to minimize the effect of incident diffuse long- and short-wave radiation from the shade device itself. The ratio of the drop in Eppley short-wave radiation to the drop in the net radiometer's millivolt output was compared to the A E S calibrations. These experiments revealed that in contrast to the pre-field laboratory calibrations, the Swissteco was overestimating Rsi by a mean 3% while the Middleton was underestimating Rsi by a mean 9%. A n in-situ evaluation of the net radiometers was performed (Hodges & Smith 1995) by placing a net radiometer (model Q * l (Fritschen), Radiation Energy Balance Systems, Seattle, W A ) beside the Swissteco and Middleton net radiometers for a three-day period in July 1994. This evaluation showed that for R„ > 0, the Swissteco and Middleton Rn measurements were 3% and 8% lower than the R E B S , respectively, in good Seasonal Energy & Water Exchange 63 agreement with the in-situ Rsi calibration for the Middleton but slightly at odds for that of the Swissteco. Based on the results of the in-situ calibration, the consistency of the laboratory calibration as well as its short time constant, the overstory Rn reported here is from the Swissteco reduced by 3% from the original A E S laboratory calibration. 3.2.3. Within-canopy Radiation Measurements Measurement of the radiation regime beneath an overstory canopy is difficult due to spatial heterogeneity, yet its determination is crucial for a complete understanding of within-canopy energy balance processes. By the nature of a forest overstory canopy, radiation levels can change quickly from full above-canopy values to almost zero due to fluttering leaves, flexible trunks and canopy gaps, and vary diurnally due to changes in the radiation beam path length with solar azimuth. This often precludes the use of stationary radiation sensors beneath the canopy. For example, to detect to within 10% a forest floor solar irradiance of 200 W m" 2 (using statistical a and B parameters of 0.95 and 0.90, respectively) in a homogeneous overstory canopy (variance in forest floor Rsi of only 10 W m"2) would require 10 samples or 10 stationary solarimeters. The number of required samples would increase to 20 and 100 as the overstory canopy became more heterogeneous resulting in forest floor Rsi variances of 50 and 200 W m" , respectively. The variance in Rsi beneath the aspen canopy was typically 51 W irf2 (clear-sky conditions) so a minimum of 20 stationary solarimeters would be required under the aspen canopy to reasonably measure understory solar radiation. To overcome the inherent impractical problem of this large number of replicates, Rsl> Qpi and Rn were measured by individual sensors as the average of 900 samples taken Seasonal Energy & Water Exchange 64 on a tram platform moving along a horizontal distance of 60-m over a duration of 15 minutes (equivalent to 30 samples every metre at a velocity of 7 cm s"'). The tram was suspended on two parallel steel cables 4 m above the ground and carried two net radiometers (models S - l and S-14 miniature, Swissteco Instruments), up-facing and down-facing quantum sensors (model 190-SB, L I -COR, Lincoln, NB) and two up-facing pyranometers (one shaded) (model C M - 5 , Kipp and Zonen Laboratory, Delft, The Netherlands). Data were recorded by a datalogger (model CR10, CSI) carried onboard the tram. Magnetic switches located at both ends of the tramway were activated by magnets on the tram and sent a signal to reverse a 1/4 hp D C motor which pulled the tram with a fine-steel cable. Extinction coefficients derived from tram measurements were used to extrapolate to periods when the tram was not operating (see Chapter 3, Section 3.3.3. ). 3.2.4. Energy Storage The total rate of energy storage in a column extending from the ground surface to z r was calculated as Jt = Jb + J\ + Jh + Je + Jp where subscripts b, 1, h, e and p are the rates of change of heat content in the boles (stems), leaves, sensible heat content in the air column, latent heat content in the air column and energy consumed for photosynthesis, respectively. The rate of bole heat storage was calculated as n Jb= SdFfhc7[^{r? ~r^)pbicbi(ATbi I At) where SA is the stand live stem density (830 1=1 trees per m 2 ground) determined from a 50 m by 50 m stand survey near the main tower, Seasonal Energy & Water Exchange 65 F{ is the tree form factor (= 0.45), the ratio of the actual bole volume determined from measurements of a 16-cm diameter felled aspen to that assuming the bole is a cylinder with a diameter equal to that at the 1.3 m height, n is the number of bole annuli (3 for aspen, 1 for hazelnut) and r„ c b , and ATbi/At are the respective values of the radius, density, specific heat and temperature change per half-hour for a particular bole annulus /'. For Jb up to z r = 39 m, terms were included for both aspen and hazelnut, while Jb up to z r = 4 m included hazelnut boles as well as aspen boles up to the 4-m height. Aspen 7b was measured with 0.254 mm diameter fine-wire chromel-constantan thermocouples installed in February 1994 in a north-to-south transect through a 16-cm diameter aspen tree (north side bark-2 mm, 1/2 way between north side and tree center-4 cm, center-8 cm, 1/2 way between south side and tree center-4 cm, south side bark-2mm) at a height of 3 m above the ground recorded on a datalogger (model 2 I X , CSI). Hazelnut Tb was measured with a thermocouple of the same design positioned 73 cm above the ground (75% of hazelnut height from ground to the base of the crown) inserted at the center of the bole over a period from September 1 to 20, 1994. The linear relationship between the change in air temperature at the 4-m height (7^(4 m)) and the change in TbOiazelnut) (ATb = 1.0247; + 0.0015; rz = 0.93, n = 915) was used to extrapolate to periods before the thermocouple was installed. Fresh aspen bole cross-sections were collected during the winter, spring and fall at various positions along the height of a felled aspen to determine bole densities and water contents. Hazelnut boles were harvested from a 3 m by 3 m plot during early August. The bole specific heat was calculated as c b = (c d + Wdcv) I ( l + Wa) (Marshall 1958) where Cd is the specific heat of dry wood (given as 1.37 kJ kg"1 ° C for all dry wood Seasonal Energy & Water Exchange 66 independent of species; Dunlap 1912), Wa is the water content on a dry mass basis and c w is the specific heat of liquid water (4.19 kJ kg"' ° C ' ) . The rate of leaf heat storage was calculated as / , = a^crf^AT^ I At) I ( l - Wv) where a t is the L A I , o\ is the specific leaf weight (mass of dry leaves per m 2 of leaf), c\ is the leaf specific heat, Ww is the water content on a wet mass basis (= 80%) and AT\IAt is the leaf temperature change per half-hour. The aspen a\(7\ was calculated from the 900 kg C ha"1 estimated from destructive sampling of the aspen (Tom Gower, pers. comm. 1995). Converting this value gives an aspen a\<j\ of 225 g dry aspen leaves per m 2 ground (900 kg C haVlO4 m 2 ha"1 x 30g CH20 /12g C). Destructive sampling of hazelnut leaves in the 3m by 3m plot gave a a\G\ of 117 g dry hazelnut leaves per m 2 of ground. Dividing a\0\ by the maximum L A I of 2.3 (aspen) and 3.3 (hazelnut) gave Oj of 111 and 36 g dry leaf per m 2 leaf for the aspen and hazelnut, respectively, comparable with the oj determined from independent leaf measurements (Betsy Middleton, pers. comm. 1996). Both the aspen and hazelnut leaf specific heat were calculated as c\ - 0.8 c w + 0.2 c g where c g is the specific heat of glucose (1.26 kJ kg"1 ' 'C" 1 ). Aspen and hazelnut T\ were measured using infrared thermometers (IRT) (model 4000, Everest Interscience Inc., Fullerton, C A ) each enclosed in a constant temperature container mounted at the 30-m (aspen) and 4-m heights (hazelnut) oriented at 45° to the normal. The IRT at the 4-m height was calibrated with a laboratory standard hand-held IRT (model 112, Everest Interscience Inc., Fullerton, C A ) at the end of the field season (hand held IRT = 0.955 IRT(4 m) + 0.94, r 2 = 0.99) and adjusted for spatial heterogeneity Seasonal Energy & Water Exchange 67 by in-situ scans of the hazelnut canopy made throughout the daytime on several days during the full-leaf period (hand held IRT = 0.733 IRT(4 m) + 5.16, r2 = 0.87). The rate of sensible (Jn) and latent heat (Je) storage in the air within the column was calculated as Jh + Je = pacpzrATa /At + (pacp I y)zIAea I At where pa is the air density, c p is the specific heat of air at constant pressure, y is the psychrometric constant and ATa I At and Aea I At are the average air temperature and vapour pressure changes per half-hour, respectively. For the forest (zr = 39 m), the latter were calculated by weighting the measurements at 39 m and 4 m by 0.8 and 0.2, respectively. For the hazelnut (zr = 4 m), only measurements at 4 m were used. A i r temperature and vapour pressure at 39 m were measured with a ventilated platinum resistance thermometer and a dewpoint hygrometer (models M l and D2, respectively, General Instruments, Waterdown, M A ) . A t 4 m, Ta and e a were measured with a shielded thermistor and humidity (model H M P - 3 5 C , Vaisala Inc., Woburn, M A ) sensor, respectively. The rate of energy consumption during photosynthesis in the chemical formation 2 I of carbon bonds was calculated as Jp = - FCC where Fc is the C 0 2 flux in /umo\ m" s" and C is the photosynthetic energy conversion factor (0.469 J ^mol" 1) with a sign convention of negative Fc representing C 0 2 uptake (i.e. downward C 0 2 flux). For the aspen vegetation, Jp(aspen) was calculated with Fc equal to the difference between F c (39 m) and F c (4 m). For the hazelnut vegetation, / p(hazelnut) was calculated with Fc equal to the difference between F c (4 m) and the soil respiration (see Chapter 4, Section 4.3.4.3.). Energy consumption during photosynthesis over the 0-39 m height was taken as the sum of yp(aspen) and J p(hazelnut), whereas Jp over the 0-4 m height was set equal to Seasonal Energy & Water Exchange 68 y p(hazelnut). Calculation of any Jp term was only performed when Fc was negative (net carbon uptake) and with positive Rn (39 m). Any release of sensible heat during respiration was already accounted for with calculation of heat storage in the leaves and boles. Due to the complexity involved in fully instrumenting all components of the total storage terms, the linear equations derived from linear regressions are given as useful approximations using readily available measurements of Ta or even Rn (Table 3.1) 3.2.5. Soil Heat Flux, Water Content & Evaporation The soil heat flux (G) at a depth of 3 cm was measured with a 20-m transect of nine heat flux plates (two model F, Middleton Instruments; seven home-made following Fuchs and Tanner 1968) recorded on a datalogger (model C R 7 , CSI). The surface soil heat flux (Go) was calculated by adding G at the 3 cm depth to the 0-3 cm rate of heat storage (i.e. Go = G + CSATS/At where C s is the volumetric heat capacity in the top 3 cm of organic soil and ATJAt is the 1/2-hourly temperature change in the top 3 cm). The volumetric heat capacity was calculated as Cs = 9mCm + 60C0 + 6^C W + (9aCa = 0OCO + 6^C W where 0m, 6>0, 6k and 6a are volumetric fractions of minerals, organic matter, water and air. A typical C 0 and C w of 2.50 and 4.18 M J m"3 ° C ] (respectively) were used (Hillel 1982) and 6>0 was calculated as <90 = Px,lp0 where Po (measured as 160 kg m 3) and p0 (typically 1300 kg m" 3; Hil le l 1982) are the organic soil bulk and particle densities, respectively. The volumetric water content in the 0-3 cm layer (6or #w = w / V A - ) was determined from the gravimetric soil water content w (see next paragraph). The average temperature change per 1/2 hour Seasonal Energy & Water Exchange 69 Table 3.1. Approximations for total heat storage between 0-39 m or 0-4 m using a linear equation J t (W m' 2) = a x + b where x is the 1/2-h change in air temperature (ATa °C) or net radiation (ARn W m"2) measured above the aspen canopy and a and b are empirical coefficients (equations are independent of time of year or day). Height a (W m ' 2 /°C or x (°C or W m' 2) b (W m"2) r2 n Interval (m) W m ' 2 / W m"2) 0 - 3 9 44.5 ATa 1.66 0.85 7962 0 - 3 9 0.10 ARn -7.4 0.71 7962 0 - 4 7.0 ATa 0.70 0.83 8053 0 - 4 0.02 ARn -1.59 0.78 8053 Seasonal Energy & Water Exchange 70 in the 0-3 cm layer was determined by two home-made integrating thermometers (nickel wire wound around a glass cylinder and coated with epoxy) inserted diagonally across the 0-3 cm depth. The soil water content was determined at several depths through the soil profile by two methods. To ensure correct measurement of C s, w was measured gravimetrically every 2-3 days with soil samples obtained over the 0-3 cm, 3-6 cm and start of mineral soil to 10 cm layers. Each sample was oven-dried at the site at 105 °C for at least 24-h with w calculated as the difference between the fresh and oven-dried weights relative to the dry weight. This method was complimented with the time-domain reflectrometry (TDR) technique (Topp et al. 1980) which permits measurements of 6 with absolute errors of 0.023-0.034 m 3 m" 3 (Hook & Livingston 1996). T D R relies on the different dielectric constants for water (ATW = 80 at 20 °C) and air (Ka ~ 1) to detect a reflection of a high-frequency electromagnetic pulse sent down along a transmission line or wave guide, with 0 related to the apparent dielectric constant by a model described by Hook & Livingston (1996). In the fall of 1993, two probes consisting of three stainless steel rods (3 mm diameter, 30 cm long, 2 cm apart) were positioned horizontally in the middle of the local organic (8 cm depth) and mineral (15 cm depth) horizons. A 123-cm rod (two stainless steel strips 2 cm apart separated by epoxy) was also installed to measure 6 at depths integrating over 30.5 cm segments. Both the probes and the rod used the shorting-diode technique (Hook et al. 1992) and 8 was calculated from the delay times using the algorithm of Hook and Livingston (1996). The delay times were measured using cable-length testers (models 1502B and 1502C, Tektronix, Beaverton, OR) with a resolution of Seasonal Energy & Water Exchange 71 0.004 m 3 rrf3 (Hook & Livingston 1996) located inside one of the huts approximately 10 m from the T D R probes and rod. The gravimetrically determined 9 for the 3-6 cm layer was used to calibrate the 8 cm T D R probe (Orm = 0.48 0 g r a v . + 0.049; r2 = 0.69; n = 52). Soil water potential was measured continuously throughout the growing season using a gypsum block at each of the 6-, 16- and 46-cm depths. While installing the T D R probes and rod and while making bulk density measurements and gravimetric measurements, the average rooting depth was found to be approximately 60 cm. To determine soil evaporation, two small thin-walled plastic lysimeters (15 cm diameter and 15 cm deep) were installed in late July. The lysimeters were weighed manually every 2 hours to the nearest 0.1 g on a balance with a range of 3000 g (model P3, Mettler Instrument Corp., Princeton, NJ) during the day for a 15 day rain-free period. To prevent drying out of the soil in the lysimeters and hence decoupling from the surrounding soil, the soil in the lysimeters was replaced after 5-7 days following the suggestion of Boast & Robertson (1982). Measurements during this 15-day period were used to estimate soil evaporation at other times using the ratio of soil evaporation to forest evapotranspiration. 3.2.6. Precipitation Precipitation was measured 1/2-hourly in the centre of a clearing (approximate diameter 75 m) located roughly 200 m from the main tower. Precipitation in the form of snow was measured with a Nipher-shielded weighing precipitation gauge (Belfort Instruments, Baltimore, M D ) . Precipitation in the form of rain was also measured with a tipping-Seasonal Energy & Water Exchange 72 bucket rain gauge (model TE525, CSI). These instruments were maintained by the Saskatchewan Research Council under contract to B O R E A S (Shewchuck 1996). Periods of missing data were replaced by measurements obtained by an A E S tipping-bucket or manual rain gauge located in an adjacent clearing. 3.2.7. Leaf, P lant & W o o d A r e a Indexes One-side (projected) leaf area index (LAI) was measured optically throughout the year using a L I - C O R Plant Canopy Analyzer (model LAI-2000, L I - C O R Inc., Lincoln, N B ) . Since this instrument requires diffuse solar radiation conditions, readings were taken either on overcast days or just before or after sunset. To isolate the aspen canopy, readings were made at the 25-m height from the walk-up tower and at the 3-m height (i.e. above the hazelnut) at two pre-determined locations. Hazelnut canopy measurements consisted of pairs of below and above hazelnut readings taken at six predetermined locations along a 100-m east-west transect within the flux footprints of the towers. The same locations were used throughout the year. A wood (boles and branches) area index (WAI) of 0.62 (aspen) and 0.34 (hazelnut) was determined from measurements taken before leafing and the L A I was calculated by subtracting W A I from the plant area index (PAI). L A I values were also corrected for clumping (seasonally adjusted for the aspen canopy; see Chen et al. 1997 for details) and were adjusted for the pre-leaf wood area index. Forest L A I was calculated as the sum of aspen and hazelnut L A I . Leaf wetness sensors (model 237, CSI) were positioned in the aspen and hazelnut canopies. Seasonal Energy & Water Exchange 73 3.2.8. Fetch or Flux Footprints A n analysis of the upwind land surface area that contributes to a scalar flux measurement, often referred to as "fetch" or "footprint", is crucial in understanding the origins of the flux and any possible influences of spatial heterogeneity. The aspen stand extended for at least 3 km in all directions, but it was still important to determine the extent and behaviour of the flux footprints, especially in the case of measurements made within the canopy. Stemming from the original works of Pasquill (1972), Gash (1986) and Schmid & Oke (1990), the footprint algorithms adopted here are those presented by Schuepp et al. (1990). More complex footprint models are available {e.g. Schmid 1994, Amiro 1996), but the relatively straightforward, robust model adopted here has been successfully tested over boreal forests (J. M . Chen, pers. comm 1996). The one-dimensional footprint from a measurement made at a height z at an upwind distance* normalized by the flux Q is given as (Schuepp et al. 1990 equation 9) where d is the zero-plane displacement (well approximated by 2/3A c; Brutsaert 1984), k is the von Karman constant, u* is the friction velocity and U is height-averaged (39 or 4 m) wind speed given as (3.1) U = u* [ln((z-d)zQ-l + z0/(z-d)] k(\-zj(z-d)) (3.2) Seasonal Energy & Water Exchange 74 (Schuepp et al. 1990 equation 6). The first derivative of equation (3.1) equated to zero yields the upwind horizontal distance to which the flux measurement is most sensitive (peak footprint) and is given as (Schuepp et al. 1990 equations 10 and 13, respectively). Since flux footprints are affected by atmospheric stability (Leclerc & Thurtell 1990), equation (3.1) was modified as suggested by Schuepp et al. (1990) to include these effects (U/u* ratios multiplied by the stability correction for momentum of the forms suggested by Dyer 1972). The stability corrections were calculated using the median (z-d)/L (where L is the Obukhov stability length) during the daytime (-0.11 atz = 39 m and -0.28 atz = 4 m) and nighttime (0.26 at z = 39 m and 0.29 at z = 4 m) periods. 3.3. Results & Discussion 3.3.1. Site Conditions During 1994 The study site falls on the northern edge of the prairie climate region with mild, wet summers and cold, dry winters (Hare & Thomas 1974). Figure 3.1 shows this extreme seasonality in both air temperature (7"a) and precipitation (P). Compared with the "normal" or 30-year mean T3 and P recorded at the nearest centre to the site with long-term (1951-1980) climate records available (Waskesiu Lake, 53.917° N , 106.083 °W; Canadian Climate Normals 1982), January and February were overall colder than the (3.3) and the cumulative normalized contribution is the integration of (3.1), (3.4) Seasonal Energy & Water Exchange 75 J F M A M J J A S O N D Month 1994 Figure 3.1. Comparison of the daily mean air temperature (r a ) measured above the aspen forest at z - 39 m (jagged line) and monthly total precipitation (P) measured in a clearing at the research site (solid vertical bars) to the 1951-1980 average Ta (smooth line) and monthly total precipitation (clear vertical bars) measured at Waskesiu Lake. Seasonal Energy & Water Exchange 76 normal, with several contiguous days well below the normal, indicative of well-developed stationary high-pressure systems. The spring (March-April-May) was, however, slightly warmer than normal with the trend of above-average temperatures continuing right though the summer and into early fall (June-July-August-September). The remainder of the fall and winter experienced temperatures close to the normal except for December which was much warmer (not an uncommon occurrence). On an annual basis, the mean T a of 1.2 °C was significantly different from the normal -0.2 °C at the p = 0.11 level (Student's t-test). The annual precipitation cycle (Figure 3.1) shows that during a normal year, 45% of the annual P is delivered during the summer months (June, July and August) with 69% in the form of rain. During 1994, the normal annual P pattern was followed with the exception of a very wet May and June and a drier than normal August and September. March and October also received more P than expected. The high May and June P was partially attributable to intense, local convective storms. On an annual basis, the February-December total P of 462.2 mm was above the normal 433.8 mm during the same period (unfortunately, P instrumentation was not installed until late January). If it is assumed that the January P in 1994 was equal to the normal amount for that month (28.8 mm - 2.65 mm measured in late January), then the total P in 1994 (488.35 mm) was 25.8 mm above the normal 462.6 mm. The seasonal P patterns are somewhat reflected by the changes in the volumetric soil moisture content (9) observed from T D R measurements made at several depths in the soil profile (Figure 3.2) except during the spring thaw period. During the frozen winter period, measurements of 6 remained stable at 0.06-0.13 in the organic, shallow mineral Seasonal Energy & Water Exchange 11 0.6 0.4 0.2 o 0.0 <J) CO 1 0.6 E 0.4 o X 0.2 ^ J E 0.0 CD 0.6 0.4 0.2 0.0 60 E 40 20 a. 0 i r i r i i Organic Shallow Mineral <4 Deep Mineral Precipitation J F M A M J J A S O N D Month 1994 Figure 3.2. Seasonal course of daily volumetric soil moisture content (6; lines) and precipitation {P; vertical bars). The organic layer (= 0-10 cm) was represented by a 3-prong T D R probe inserted horizontally at a depth of 8 cm approximately in the middle of the organic horizon. The shallow mineral layer (= 10-30 cm) was represented by a 3-prong T D R probe inserted horizontally at a depth of 15 cm (thin line) and the 30-61 cm section of a segmented T D R vertical rod (thick line). The deep mineral layer (~ 61-123 cm) was represented by the 61-92 cm (thin line) and 92-123 cm (thick line) sections of a segmented T D R vertical rod. Seasonal Energy & Water Exchange 78 and 61-92 cm deep mineral layers, magnitudes similar to those reported by Hayhoe et al. (1983) in frozen soils. A substantially higher winter dot 0.25 occurred deep in the profile where temperatures remained positive. Since T D R measurements made in frozen soils indicate the unfrozen water content (Hayhoe et al. 1983; Spaans & Baker, 1996), this suggests that there was some liquid water present, even during the coldest time of the year. In Apr i l , # increased markedly in the Organic and shallow mineral horizons despite relatively little precipitation. The relative contribution to this increase from the infiltration of snowmelt water (all snow melted by Apr i l 15) and the melting of water in the soil (soil temperature at the 2, 5 10, 20 and 50-cm depths reached 0 °C by Apr i l 17, 18, 21, 23 and 29, respectively) is not known. The fact that there was no spring-time increase in 6 in the deep mineral soil and soil temperature at 100-cm depth was always positive (annual minimum of 0.05 °C reached between Apr i l 1-21) confirms that there was little frozen water at this depth over the winter and suggests that infiltration of snowmelt water did not reach deep into the soil profile. Following the spring increase, 6 gradually decreased through the season in the organic layer with intermittent responses to P events. In the shallow mineral layer, 0 maintained its spring melt value of approximately 0.35, showed little response to P events and did not decrease substantially until late July. The deep mineral layer showed a large increase following the moderately heavy July rainfall and slowly decreased thereafter. Both measurements of 8 in the deep mineral layer showed a marked response to the substantial May rainfall after the relatively dry Apr i l . Seasonal Energy & Water Exchange 79 Following snow melt, bud burst began and lasted until leaf emergence in the third week of May (Figure 3.3). Aspen leaf emergence lasted until early June while leaf growth continued until mid-July with a maximum leaf area index (LAI) of 2.3 m2 m"2. Branch loss and possible leaf curling slightly decreased the aspen LAI until senescence began the second week of September. Similar to the aspen, hazelnut bud-burst began right after snowmelt with leaf emergence beginning the third week of May. In contrast to the aspen, hazelnut leaf growth continued until late June with a maximum LAI of 3.3 m2 m"2 exceeding that of the aspen attained by July. Hazelnut senescence began approximately ten days before the aspen. Overall, the forest reached a maximum LAI of 5.6 m2 m"2 by mid July. 3.3.2. Footprint Predictions of the Scalar Flux Source Area The relative scalar flux contribution as a function of upwind distance (flux footprint or fetch) at both measurement levels above and within the aspen canopy is shown in Figure 3.4. The important feature of Figure 3.4 is the distance at which the peak footprint occurs (*max.' upwind distance to which the flux measurements are most sensitive) and how this distance changes with typical diurnal variations in atmospheric stability. Above the aspen canopy at z = 39 m, xmax under neutral conditions was 129 m. This contracted to 99 m and extended to 296 m under typical daytime and nighttime atmospheric stability conditions, respectively. Within the aspen canopy above the hazelnut understory (z = 4 m), thexmax was 15 m in neutral conditions and ranged from 10 to 37 m under typical daytime and nighttime atmospheric stabilities, respectively. Although the upwind distances to the peak footprint appear rather close given the Seasonal Energy & Water Exchange 80 6.0 Month 1994 Figure 3.3. Seasonal development of the forest (•), hazelnut understory (A) and aspen overstory (•) leaf area indices. Maximum leaf area indexes of 2.3, 3.3 and 5.6 m leaf per m"2 ground were obtained by the aspen, hazelnut and forest, respectively. Seasonal Energy & Water Exchange 81 3 0.005 0.004 0.003 0.002 0.000 0.000 D a y t i m e Z = 39 m .Neutral % 0.05 0 200 400 600 800 1000 ° 0.04 -- D a y t i m e Z = 4 m 100 Upwind Horizontal Distance (m) Figure 3.4. The "flux footprint" or relative contribution to the flux measured at x = 0 m (l/Qo dQix)ldx) as a function of the upwind distance x for z = 39 m and z = 4 m (area under each curve is unity). Neutral stability, typical daytime and nighttime footprints are shown. The latter were calculated by correcting for atmospheric stability using the median (z-d)IL during daytime (R„ > 0 W r r f 2 ) and nighttime (i?„<0W m"2) periods. Seasonal Energy & Water Exchange 82 height and cost of the main tower, it is the integration of the flux footprint that is most important in determining the importance of upwind source areas (Figure 3.5). Despite the relatively small values of * m a x , the cumulative flux reached 80% of the total flux at an upwind distance of 1160 m at z - 39 m under neutral conditions, 900 m under typical daytime stability conditions and 2660 m under typical nighttime stability conditions. Thus the fetch requirements (fetch : (z - d)) were relaxed (47:1 and 36:1, during neutral and daytime conditions, respectively) and slightly extended at night (106:1) compared to the micrometeorological rule-of-thumb of 100:1, but the horizontal distances were quite large because of the high measurement height. Within the canopy atz = 4 m, the upwind distance required to reach 80% of the total flux was also well beyond the peak footprints. Distances of 135 m (neutral), 90 m (daytime) and 333 m (nighttime) or fetch:height ratios of 51:1, 34:1 and 125:1 (neutral, daytime and nighttime, respectively) were comparable to those above the forest. The micrometeorological rule-of-thumb of a minimum 100 m of fetch for every 1 m of height {i.e. 2500 and 270 m for the 39 and 4 m measurement heights, respectively) is almost sufficient at both heights to capture 80% of the upwind source area regardless of atmospheric stability. The aerodynamically rough forest promotes mixing and acts to contract the footprints compared to smoother surfaces which require more fetch. Since the aspen and hazelnut stands were the dominant cover in all compass directions throughout the footprint areas, misrepresentative flux measurements from unintended surfaces was not a problem. When differencing, for example, the latent heat flux to isolate the transpiration from each level, problems arising from taking the difference Seasonal Energy & Water Exchange 83 1.0 , , r D a y t i m e E 1.0 O 0.8 0.6 0.4 0.2 0.0 z = 39 m . 4000 1 I 1 I 1 I i i • - / / / Neu t ra l -.// / N i g h t t i m e — / / . i . i . i z = 4 m . i 0 200 400 600 800 1000 Upwind Hor izonta l Dis tance (m) Figure 3.5. The cumulative flux footprint or cumulative relative contribution at an upwind distancex (Q(x)) to the total flux Q measured atx = 0 m forz = 39 m andz = 4 m. Neutral stability, typical daytime and nighttime footprints are shown. The latter were calculated by correcting for atmospheric stability using the median (z-d)/L during daytime (Rn > 0 W itf2) and nighttime {Rn < 0 W m-2) periods. Seasonal Energy & Water Exchange 84 between fluxes with different footprints were not anticipated to be significant because of the extensive uniform vegetation coverage. 3.3.3. Within-canopy Radiation To quantify the within-canopy radiation regime and to develop a simple empirical model of radiation extinction, the daytime means of the ratios of understory to overstory Rn, Qpi and Rsi were plotted against aspen L A I (Figure 3.6). Extinction of radiation through the canopy was largely a function of aspen L A I , and an exponential equation of the form Qi (4 m)/<2i(39 m) = a exp(-ba{) best described this relationship (see Appendix A , Table A.3.) . Without leaves, the ratios of below to above fluxes were greatest for/? sj. (0.58) and Qpi (0.57) followed by Rn (0.47). With leaves, ratios of Qi(4 m)/Qi(39 m) decreased significantly falling to 0.27 (Rsi), 0.23 (Rn) and 0.20 (Qpi). The exponential decrease in all three radiation components with increasing L A I implies that attenuation could be expressed using a form of the Beer-Bouger Law, Qi(z) = Qlo exp(-kap), where Qio is the top of the canopy incident radiation, k is the effective extinction coefficient and ap is the plant area index (ap = a\ + a w where a\ is the leaf area index and a w is the wood area index equal to 0.62 for the aspen). In strict terms, the Beer-Bouger Law describes the exponential attenuation of single-wavelength radiation with depth in a homogenous medium with complete absorption and no scattering. Its application with depth replaced by accumulated leaf or plant area (Monsi & Saeki 1953) has surprisingly resulted in good descriptions of attenuation of broad radiation band widths in several plant canopies. Seasonal Energy & Water Exchange 85 0.6 0.4 0.2 'E 0.0 £ 0 . 6 O 0.4 E 0.2 7 J 0 . 0 0.6 0.4 0.2 0.0 • • • *\ 0.0 0.5 1.0 1.5 2.0 2.5 Aspen LAI (m 2 rrf 2) Figure 3.6. Ratio of net radiation (Rn), photosynthetically active photon flux (Qpi) and solar radiation (Rsi) measured above the hazelnut understory (z = 4 m) to that measured above the aspen (z = 39 m) plotted as a function of the seasonal development of aspen leaf area index (LAI). Points represent daytime means (Rn > 0 W rrf2) and the solid line represents a curve fit of the form Qi(4 m)/Qi(39 m) = a exp(-&<2i) (see Appendix A, Table A.3. for curve-fit parameters). Seasonal Energy & Water Exchange 86 Calculating k and plotting it as a function PAI is shown in Figure 3.7. P A I was chosen over L A I since k cannot be calculated when a\ = 0. In general, k decreases suddenly following leaf out and then drops only slightly during the full-leaf period. The fact that the best fit between Qi(4 m)/Q/i(39 m) and L A I was exponential implies that a constant b {i.e. effective extinction coefficient) worked well as shown in Figure 3.7. The decrease in k for all three radiation streams as P A I increases implies that although there was less radiation reaching the understory when the aspen canopy was mature, the radiation that did reach the understory did so more efficiently. This counter-intuitive result, also found in other deciduous forests (Rauner 1976, Baldocchi et al. 1984), was likely due to the shorter radiation path lengths through the canopy in summertime when solar zenith angles are smaller and also due to beam enrichment resulting from forward scattering of light in the leafed canopy. Comparing the effective extinction coefficients from the boreal aspen forest to other deciduous forests (Table 3.2), some general conclusions may be drawn. During the winter period with leafless canopies, k(Rn) > k(Qpi) ~ k(Rsi). Large solar zenith angles and long path lengths severely attenuated the short-wave component of Rn in addition to long-wave attenuation by the opaque branches and trunks while the absence of leaves precluded any possible trapping and re-radiating of long-wave radiation. During the summer, k(Rn) was less since leaves are excellent absorbers and emitters of long-wave radiation (Idso et al. 1969) so only the short-wave component of Rn was significantly attenuated (Baldocchi et al. 1984). For the bare canopy, k(Qpi) ~ k(Rsi) since there were no leaves present to absorb Qpi so both radiation streams were attenuated equally. Seasonal Energy & Water Exchange 87 0.0 1 • 1 • 1 • '— 0.0 1.0 2.0 3.0 Aspen PAI (m 2 r r f 2 ) Figure 3.7. Effective extinction coefficients (k) for net radiation (Rn), the photosynthetically active photon flux (Qpi) and solar radiation {Rsi) derived from the equation k = -\n(Qi(4 m)/£>i(39 m))/a p where ap is the aspen plant area index, the sum of aspen leaf and wood area indexes. Solid lines represent k calculated using the equation Ql(4 m)/0\t(39 m) - a exp(-ba\) obtained in Figure 3.6. Seasonal Energy & Water Exchange 88 Table 3.2. Comparison of leafless and leafed deciduous forest canopy's effective extinction coefficients (k) expressed on a leaf and plant area index basis for net radiation (Rn), photosynthetically active photon flux (Qpi) and solar radiation (Rsi). Radiation Leafless Leafed L A I PAI Reference Stream Canopy k Canopy k 1.75 0.54 4.9 5.5 Baldocchi et al. (1984) Rn 1.62 0.47 3.5 4.0 Rauner (1972) 1.22 0.49 2.3 2.9 This Study 1.12 0.51 4.9 5.5 Baldocchi etal. (1984) Rsl 1.00 0.42 3.5 4.0 Rauner (1972) 0.88 0.44 2.3 2.9 This Study 1.06 0.66 4.9 5.5 Baldocchi etal. (1984) Qpi N / A , 0.6 - 1.0 3.5 4.0 Rauner (1972) 0.90 0.54 2.3 2.9 This Study Seasonal Energy & Water Exchange 89 As the leaves develop and reach maturity, k(Qpi) > k(Rn) ~ k(Rsi). Since leaves have a high absorptivity in the waveband 400-700 nm, Qpi was attenuated more than Rsi. As leaves are almost blackbodies, the long-wave portion of Rn suffered little attenuation but the leaves absorb and emit long-wave radiation decreasing k(Rn) from the winter values. The influence of this trapping, however, was probably not large since k(Rn) values were similar to k(Rsi) values. 3.3.4. Seasonal Progression of the Diurnal Energy Balance To summarize large amounts of data while still retaining a fine time resolution of one day, ensemble averages of the energy balance components for each month for the forest (Figure 3.8) and for the hazelnut understory (Figure 3.9) were calculated at hourly intervals. Note that AE and H were not corrected for energy balance closure (see Chapter 2, Section 2.4.7.) in order to reveal the residual energy balance term. Correcting for energy balance closure before ensemble averaging had virtually no effect (AE' or H' less than 5% greater than AE or H for all months) on the diurnal plots. To focus on the daytime period, energy balance components for the forest and understory normalized by Rn are given in Tables 3.3a and 3.3b. 3.3.4.1. Spring (April - May) During the April-May leafless period, the striking feature of Figure 3.8 is the large consumption of Rn by H above the aspen canopy with typical daytime HIRn ratios of 0.63-0.64. At the understory level, H was also the dominant consumer of R„, but not as large as above the canopy and did not peak until May. The lack of leaves meant minimal AE Seasonal Energy & Water Exchange 90 ' • I ' I I I I I I I I I 1 I 1 I 1 1 1 1 1 I 0 6 12 18 0 6 12 18 24 T i m e C S T (h) Figure 3.8. Ensemble monthly averages of the forest (z = 39 m) diurnal net radiation (Rn, • ) , latent heat (AE, • ) , sensible heat (H, A) , soil heat (Go, • ) and total heat storage (Ji, • ) flux densities. Each data point represents the mean of 2 1/2-h periods with ± one standard error of the mean shown by the vertical lines. The vertical bars represent the residual of energy balance closure, calculated as Rn - AE - H - Go - Jt-Seasonal Energy & Water Exchange 91 eg 2 0 0 1 0 0 0 ^ - 1 0 0 5 2 0 0 1 1 0 0 CD s 0 E - 1 0 0 P 2 0 0 (D iS 1 0 0 0 - 1 0 0 JUULIULT - June -o—nnnnn- u u u_ nnn_ u u u t -August nnnnn,,1,_r1nn. u cr • i i i i ,_,u_ nnnn u l J ^ L | t J nnn u u ^ v Ju ly " - n n n - u u [ J u u n n " L J U u u -September n[][]nn,nnnn Tporj-_l I I I I 1 I I I I L 0 6 12 18 0 6 12 18 24 T ime CST (h) Figure 3.9. Ensemble monthly averages of the understory (z = 4 m) diurnal net radiation (/?„, •) , latent heat (AE, •), sensible heat (H, A) , soil heat (Go, • ) and total heat storage (JT, •) flux densities. Each data point represents the mean of 2 1/2-h periods with ± one standard error of the mean shown by the vertical lines. The vertical bars represent the residual of energy balance closure, calculated as ^ n - AE - H - GQ - J { . Seasonal Energy & Water Exchange 92 Table 3.3a. Ensemble daytime (net radiation positive) means of monthly forest (z = 39 m) energy balance terms normalized by net radiation (Rn). The last column is the sum of the latent heat (AE), sensible heat (H), soil heat (Go) and total heat storage (/,) ratios. Month Rn(Wm~2) AE/Rn H/Rn G0/Rn JJRn E Apri l 257.0 0.09 0.63 0.04 0.06 0.82 May 247.6 0.20 0.64 0.08 0.05 0.97 June 247.3 0.54 0.31 0.05 0.07 0.97 July 268.7 0.65 0.20 0.05 0.08 0.98 August 243.9 0.59 0.24 0.03 0.09 0.95 September 220.6 0.47 0.32 0.03 0.10 0.92 Seasonal Energy & Water Exchange 93 Table 3.3b. Ensemble daytime (net radiation positive) means of monthly understory (z = 4 m) energy balance terms normalized by net radiation (Rn). The last column is the sum of the latent heat (AE), sensible heat (H), soil heat (Go) and total heat storage (Jt) ratios. Month / ? n ( W m 2 ) AEIRn H/Rn G0/Rn Jt/Rn E Apri l 118.4 0.12 0.36 0.09 0.02 0.59 May 95.4 0.23 0.51 0.20 0.03 0.97 June 64.8 0.53 0.12 0.20 0.06 0.91 July 61.6 0.62 0.09 0.20 0.07 0.98 August 58.8 0.57 0.08 0.14 0.08 0.87 September 63.5 0.31 0.10 0.11 0.06 0.58 Seasonal Energy & Water Exchange 94 contributions at both levels, with some soil water evaporation. With frozen surface soil, Go during Apr i l was limited since energy was used for the phase-change from solid to liquid water before soil warming could occur (soil temperature at the 5-cm depth hovered around 0 °C between Apr i l 1-18 but increased dramatically afterwards, thus indicating the period of melting frozen soil). The soil heat flux reached its peak in May when the bare canopies allowed maximum Rsi penetration resulting in steep soil temperature gradients (as much as 10 °C over a depth of 2 cm) occurring in the leaf litter layer acting as a surface mulch. At the understory level, Go was comparable to AE during the month of May. The total storage of heat over the 0-39 m height during the spring months was comparable to Go. The daytime mean 0-39 m / , was partitioned equally between heat storage in the air column and boles. A t the understory level, the short height (0-4 m) precluded any great amount of heat storage, but JJRn ratios still managed to reach 0.02-0.03. Energy balance closure was worst in Apr i l for both the forest and understory compared to any other month. The pattern of nighttime overestimation and daytime underestimation was pronounced in Apr i l , while in May this pattern became less clear. For the understory, the difficulties in estimating Rn with changing solar zenith angles and path lengths were apparent with the sunrise/sunset changes in the energy balance residual. It is not clear why closure was poor in Apr i l . It was not strictly due to the presence or absence of leaves, since May had much smaller residuals than Apr i l yet the canopies were bare for both months. Although heat storage in the snow pack was not calculated, this should not account for much heat storage given the relatively thin cover and disappearance by Apri l 15. Seasonal Energy & Water Exchange 95 3.3.4.2. S u m m e r (June - J u l y - A u g u s t ) During the June-July-August full-leaf period, there was no appreciable change in the daytime mean Rn compared to the leafless period indicating the increase in upward solar (Rst) (leafless with snow, leafless without snow and leafed RsxlRsi ratios were 0.20, 0.12 and 0.16, respectively) and long-wave (/?it) radiation due to leaves offset the increase in downwards solar radiation (Rsi) towards the summer solstice. Beneath the aspen canopy and above the hazelnut, Rn decreased substantially with the development of aspen leaves. The seasonal increase in Rsi towards the summer solstice was negated by the attenuation of Rsi through the aspen canopy resulting in an overall decrease in R„ at the hazelnut level. As expected, the development of transpiring leaf surfaces resulted in an increase in AE largely at the expense of H. Leafed canopy AE/Rn ratios climbed from their bare canopy counterparts to 0.54 to 0.65 for the forest while similarly climbing to 0.53 to 0.62 for the understory. Despite near similar AEIH ratio at both levels, HIRn ratios were small beneath the canopy (0.08-0.12) approximately half that measured above the aspen canopy (0.20-0.31). Gradual warming of the soil and shading from the two canopies resulted in a decrease of Go with typical daytime overstory G0/Rn ratios of 0.03-0.05. Relative to the diminished R„ reaching the understory, G0//?n appears more important, with daytime Go/Rn ratios of 0.14-0.20. The total daytime heat storage in the air and biomass relative to R„ was a consistent 0.06-0.09 for both the forest and hazelnut. These values rivaled the overstory GQIR„ ratio and the understory HIRN ratio and illustrate the importance of the Seasonal Energy & Water Exchange 96 storage calculation for complete energy balance analysis and closure. These values are consistent with the daytime JJRn ratio of 0.06 reported by McCaughey & Saxton (1988) in a mixed temperate forest. Greatest heat storage occurred shortly after sunrise and was comprised largely of heat storage in the boles followed by Jn, Jp, / e and finally J\ (e.g. Jt = Jb +Jn+Jp+Je+ / i = 8 + 6 + 5 + 2 + 1 W m"2). Heat storage in the woody biomass was important especially in the open-canopy structure typical of boreal forest and although a tedious measurement to make, it must be considered as noted by Saxton & McCaughey (1988). Energy balance closure was good during the summer periods, keeping in mind ensemble averaging removes a lot of unacceptable variation in the diurnal patterns of the energy balance components. Summation of the daytime energy balance components relative to Rn ranged from 0.95-0.98 for the forest to 0.87-0.98 for the understory. 3.3.4.3. Late Summer - Early Fall (September) During the late summer, early fall (September), the decrease in Rsi and leaf loss resulted in a decrease in Rn at the overstory level. At the understory level, Rn slightly increased due to the removal of leaves from the aspen overstory. As leaf senescence was not as abrupt as leaf emergence, AE decreased somewhat from the summer time values (both levels) while H increased slightly (both levels) lessening the disparity between the two fluxes and returning to energy partitioning patterns similar to June. Continued weakening of soil temperature gradients and resulting near isothermal temperature profiles kept G0 small and J, remained an important component of the energy balance at both levels with typical daytime JJRn ratios of 0.10 (39 m) and 0.06 (4 m). As during April, energy Seasonal Energy & Water Exchange 97 balance closure worsened in September especially at the 4-m level, and it is unclear why this should be. 3.3.5. Partitioning Overstory & Understory Turbulent Fluxes The energy balance signature of this deciduous boreal forest was the complete reversal in energy partitioning between H and AE as the leaves developed. This is shown dramatically by the seasonal courses of the daytime mean H and AE not only for the forest as a whole, but also for the understory and forest floor (Figure 3.10). With bare canopies and a snow covered forest floor, sensible heat (daytime means of 111 and 20 W m"2, forest and understory, respectively) exceeded latent heat (daytime means of 11 and 9 Wm"2, forest and understory, respectively) and the magnitude of H steadily increased as the winter season progressed. Surprisingly, even with a snow-covered forest floor, a substantial amount of sensible heat was generated from the forest. The steady increase of the sensible heat flux from the forest level (daytime mean 172 W m"2) continued into the snow-free, leaf-free period, during which it was six times the latent heat flux (daytime mean 28 W m"2). This same pattern in the sensible heat flux at the understory level (daytime mean 61 W m"2) shows that much of the sensible heat was originating from the forest floor. Following snow melt, some of the radiation reaching the forest floor was used to evaporate water from the surface organic layer, moist from melted snow and ice (daytime mean soil water evaporation 19 W m"2). This was evident from the step-increase in AE from the forest floor. The development of leaves saw a concomitant decrease in forest H and an increase in forest AE (daytime means 70 and 139 W m"2, respectively). The wet Seasonal Energy & Water Exchange 98 400 no snow, no leaves no snow, no leaves J F M A M J J A S O N D Month 1994 Figure 3.10. Seasonal progression of the daytime mean latent heat (thick line) and sensible heat (thin line) fluxes measured above the aspen canopy (39 m) and above the hazelnut understory (4 m). October and November measurements were made in 1993. Vertical dashed lines indicate the dates of changes in the surface conditions noted on the figure (April 15, May 21, September 21 and November 1). Seasonal Energy & Water Exchange 99 conditions of 1994 which meant virtually no soil water supply limitation on forest evapotranspiration resulted in a large AE at the expense of H during the leafed period. Maximum 24-h rates of evapotranspiration equivalent to 5.0-6.0 mm d"1 were far above those of nearby boreal mature black spruce (3.0-3.5 mm d"'; Jarvis et al. 1997) and jack pine forests (2.0-2.5 mm d"1; Baldocchi et al. 1997) indicating the large water use by the aspen forest. During the leafed period, the mean daytime AE originating from the hazelnut understory was 33 Wm" 2 and the lysimeter measurements indicated that approximately 7 W m" 2 (5% of the daytime mean forest AE; 21% of the daytime mean understory AE) was soil water evaporation. The sensible heat flux from the understory during the leafed period was minimal (daytime mean of 8 W m"2). A positive understory H often occurred when mean air temperatures increased with height (inversions) in the trunk space, indicating counter-gradient flow, not an uncommon occurrence in forests (Denmead & Bradley 1985). Measurements during the fall snow-free, leaf-free period were unfortunately sparse, but it was clear that a repeat of the large springtime H fluxes would not occur given the decreases in solar radiation further away from the summer solstice. The steady decline in AE and H as solar radiation decreased was not much affected by a blanketing of snow. The magnitude of the sensible and latent heat fluxes originating from the hazelnut understory and forest floor relative to the forest can be inferred from Figure 3.10, but the role of the understory is more clearly shown by the ratio of the daytime understory-to-overstory turbulent fluxes (Figure 3.11). During the leafless period, most of the sensible Seasonal Energy & Water Exchange 100 co O 1.4 1.2 1.0 0.8 0.6 snow, no leaves E O 0.4 0.2 0.0 i no snow, no leaves T no snow, leaves A M A Month 1994 Figure 3.11. The ratio of the daytime (net radiation positive) mean latent (thick line) and sensible (thin line) turbulent fluxes measured at the understory level (z = 4 m) relative to the overstory level (z = 39 m). Vertical dashed lines note the dates of changes in the surface conditions noted on the figure (April 15 and May 21). Seasonal Energy & Water Exchange 101 and latent heat production beneath the overstory originated at the forest floor, whereas after leaf development, most was from the overstory canopy. After snow melt, virtually all of the forest AE was from the forest floor, as AE(4 m)/AE(39 m) approached unity peaking during the snow-free, leaf-free period. The H contribution from the forest floor also peaked during this period (H{4 m)///(39 m) = 0.40) as forest floor surface temperatures rose in response to the high solar radiation transmitted through the overstory and drying of the litter layer. The large transpiration rates from the leafed aspen canopy far exceeded that from the understory, decreasing AE{4 ra)IAE{39 m) to a steady 0.24 throughout the summer months. The dampening of gust penetration (see Chapter 2, Section 2.4.3.) together with an increase in Ta at the aspen canopy level produced weak and sometimes inverted air temperature profiles (mentioned earlier) and hence small sensible heat fluxes from the understory (mean daytime H(4 m)///(39 m) = 0.10). 3.3.6. The Water Balance The water balance of the forest is given as P = E + U + AS (3.5) (all in unit of mm d 1 ) where U is the drainage beyond the depth to which the rate of soil water storage, AS, was measured (123 cm). The analysis was performed on a daily basis for all of 1994, with extrapolation to periods when measurements were not made accomplished using the relationships between measured AE and Rsi during a period in close seasonal proximity to the required extrapolation period. As the extrapolation periods were short and occurred during the winter and late fall when latent heat fluxes were low, errors using these approximations were estimated to be small. Seasonal Energy & Water Exchange 102 Cumulative totals of the 1/2-hourly measured P and E from the forest (aspen + hazelnut + soil), understory (hazelnut + soil) and soil for 1994 are shown in Figure 3.12 and summarized in Table 3.4. As mentioned previously (see Chapter 3, Section, 3.3.1) missing measurements of P for much of January resulted in two estimates of the annual water balance. For each of the P scenarios, E from each level was expressed with and without nighttime low wind speed turbulent heterogeneity and energy balance closure corrections (see Chapter 2, Sections 2.4.5. and 2.4.7). Soil water evaporation was assumed to be 5% of the forest E based on summer lysimeter measurements for all three methods of calculating E. Several important points concerning the annual water balance can be made. From a methodological standpoint, it really does not matter on an annual basis which if any corrections were used to calculate E given the range in the possible total P. Correcting for low nocturnal wind speeds only added 11 mm to the annual forest total and 13 mm at the understory level because E was small at night anyway. The correction for energy balance closure added 21 mm to the annual total forest E but was still within the 20-30 mm uncertainty range in the annual P. In terms of partitioning the annual E from the forest between the three components, aspen, hazelnut and soil accounted for 67-68%, 26-28% and 4-7%, respectively. The fraction of P accounted for by E for the forest as a whole was large (82-91%), however, E represented a combination of transpired water, soil water evaporation and the evaporation of intercepted water. The evaporation of intercepted water was unknown but can be large in forests, commonly 10-25% and 15-40% of the annual P in deciduous and coniferous forests, respectively (Rutter 1975, Shuttleworth 1993). The Seasonal Energy & Water Exchange 103 J F M A M J J A S O N D Month 1994 Figure 3.12. Cumulative precipitation (P) and evapotranspiration CE) for 1994. The solid line for P does not include part of January preceding gauge installation whereas the dashed line is complete for 1994 with the missing January amount estimated from the 30-yr mean. Evapotranspiration was measured using the eddy-covariance method at the 39-m level (forest E) and 4-m level (hazelnut and soil) with cumulative 1/2-h totals calculated using three methods: 1) not corrected for nighttime underestimation (solid line); 2) corrected for nighttime underestimation (dashed line) and 3) corrected for energy balance closure (dash-dotted line). Soil water evaporation £(soil) (three overlapping lines) was estimated from mid-summer lysimeter measurements which indicated jE'(soil) = 0.05£"(forest). Periods when measurements were not available were estimated using relationships between E and solar radiation. Seasonal Energy & Water Exchange 104 Table 3.4. Evapotranspiration (E) totals for 1994 for the forest, the hazelnut understory and soil , and the soil alone. See Figure 3.12 for measurement and calculation methods. Also shown are the ratios of annual E to annual P (ratios with no parenthesis are for P = 462.2 mm, not including January precipitation before gauge installation and ratios with parenthesis are for P = 488.4 mm, including missing January precipitation estimated from the 30-year mean). Parameter Source Calculation Method Forest 401.1 412.1 422.8 E (mm yr' 1) Hazelnut 124.2 132.0 137.6 Soil 20.1 20.6 21.1 Forest 0.87 (0.82) 0.89 (0.84) 0.91 (0.87) EIP Hazelnut 0.27 (0.25) 0.29 (0.27) 0.30 (0.28) Soil 0.04 (0.04) 0.04 (0.04) 0.05 (0.04) Seasonal Energy & Water Exchange 105 relative transpiration rate (the ratio of the dry-canopy transpiration relative to the wet-canopy evaporation of intercepted water) originally defined by Monteith (1965) is today commonly referred to as the atmosphere-surface decoupling coefficient Q (McNaughton & Jarvis 1983) (see Chapter 4, Section 4.3.5.). The full-leaf aspen forest Q of 0.36 (± o 0.18) implied that for the same net radiation and saturation deficits, the evapotranspiration rate of the aspen forest with wet leaves could by be almost 3 times higher than that with dry leaves. Over the growing season, starting when most of the frozen water was melted from the soil (April 20) until senescence (September 20), the change in soil water content was calculated from the volumetric water content (9) measurements with assignments of 0 measurement locations to depths given in Table 3.5. The surface layers (0-30 cm) showed a net decrease in water content while the deeper layers showed a net increase resulting in an overall net loss of 34.7 mm over the 0-123 cm layer (i.e. AS = -34.7 mm). Calculating \P-E for the forest over the same period gave values of -7.0, -18.1 and -25.1 mm (uncorrected for nocturnal heterogeneity, corrected for nocturnal heterogeneity and corrected for energy balance closure, respectively; see Chapter 2, Sections 2.4.5. and 2.4.7.). The difference between the above estimates of \P-E and the AS of -34.7 mm suggests drainage beyond 123 cm over the growing season was small (27.7, 16.6 or 9.6 mm, uncorrected for nocturnal heterogeneity, corrected for nocturnal heterogeneity and corrected for energy balance closure, respectively) or 3-8% of the total growing season P of 347 mm. Considering the uncertainty in the ]P-E estimate and the measurement of AS Seasonal Energy & Water Exchange 106 Table 3.5. Estimation of the storage of water in the soil profile (S = 6 x Az) from a depth of 0 to 123 cm over the period Apr i l 20 - September 20, 1994. OSP and M S P refer to organic and mineral 3-rod horizontal T D R probes, respectively, and R O D (segment number) refers to the vertical segmented T D R rod. T D R Depth Az Apr. 20 G Initial S Sept. 20 0 Final S AS Probe Assignment (cm) (m 3 m"3) (mm) ( m 3 m"3) (mm) (mm) (cm) OSP1 0 - 10 10 0.50 50.0 0.09 9.0 -41.0 M S P 1 10- 30 20 0.28 56.0 0.11 22.0 -34.0 R O D 1 (3) 30 - 61 31 0.19 58.9 0.1.9 58.9 0.0 R O D 1 (4) 61 - 92 31 0.15 46.5 0.20 62.0 15.5 R O D 1 (5) 92 - 123 31 0.26 80.6 0.34 105.4 24.8 Z 0 - 123 123 - 292.0 - 257.3 -34.7 Seasonal Energy & Water Exchange 107 using one location, it can be concluded that there was little water movement at the 123-cm depth. The comparison and courses of measured daily soil water storage and JP-E over the April 20-September 20 period is a useful way to examine changes in the soil water balance (Figure 3.13). There was remarkably good agreement between the time course of the measured JP-E and S confirming little water flow at the 123-cm depth. Both JP-E and S increased up to mid-June with the persistent May-June rainfall, thereafter decreasing though the remainder of the season. The heavy 13-June rainfall of 61 mm resulted in the measured storage change exceeding the JP-E, presumably due to horizontal movement of water since capillary rise was probably unlikely at this time. During the drying phase, the greater decrease in S than JP-E indicated that there was some water drainage at the 123 cm depth. 3.4. S u m m a r y & C o n c l u s i o n s Anyone asked to picture the Canadian boreal forest likely thinks of a sparse black spruce forest growing on wet, hummocky terrain covered by lichens and moss with the occasional shrub such as Labrador tea, willow or birch. This image is for the most part correct, however it should be remembered that some areas of the southern boreal forest are covered by extensive aspen stands such as the one described in this study. It is the deciduous nature of these aspen forests that make them unique from their coniferous counterparts and the seasonal dynamics of leaf growth and loss have a marked affect on the energy and water exchange. Seasonal Energy & Water Exchange 108 A M J J A S Month 1994 Figure 3.13. Soil water storage in the 0-123 cm layer relative to April 20 (initial measured soil water content of 292 mm) determined from volumetric soil moisture measurements (thick line) and cumulative precipitation-evapotranspiration (E) (thin lines) with the eddy flux E uncorrected for nocturnal heterogeneity (solid line), corrected for nocturnal heterogeneity (dashed line) or corrected for energy balance closure (dotted line). The differences between the thick and thin lines are estimates of drainage beyond 123cm (U = P-E- AS). Seasonal Energy & Water Exchange 109 Although the results presented here are for a specific year which experienced a wet spring followed by a relatively dry and warm summer, the 371% (aspen) and 971% (hazelnut) increase in the plant area index had large effects on the energy exchange above and within the forest. The amount of net, photosynthetically active, and solar radiation received at the understory level decreased exponentially as the aspen leaf area increased. Simple effective extinction coefficients calculated from this relationship implied that radiation passed more easily through the leafed aspen canopy due to smaller solar zenith angles, shorter solar-beam path lengths and beam enrichment from forward scattering. The development of leaves was concomitant with an increase in the latent heat flux at the expense of the sensible heat flux. The latent and sensible heat fluxes measured above the hazelnut understory before leafing averaged 67% and 30% (respectively) relative to the forest as a whole falling to 24% and 10% (respectively) with the presence of leaves. On an annual basis, aspen and hazelnut transpiration accounted for 68% and 27%, respectively, of the total evapotranspiration. The seasonal water balance was also dependent on leaf development since the relatively high transpiration rates provided a means of soil water depletion. The spring thaw saw a sharp increase in the soil water content in the organic and near-surface mineral horizons. Persistent and often heavy May and June rainfalls were realized as an increase in soil water content as deep as 123 cm. On an annual basis, 82-91% of the precipitation was lost as evapotranspiration from the forest as a whole, while 25-30% and approximately 4% of the annual precipitation was evaporated from the understory and soil, respectively. The agreement between the measured soil water storage and the cumulative difference between precipitation and eddy-flux measured evapotranspiration Seasonal Energy & Water Exchange 110 indicated little drainage beyond the rooting zone. Over 200 mm of water was stored in the 0-123 cm soil layer at mid-June, after which soil water was depleted until again reaching the springtime value of 292 mm. Seasonal Energy & Water Exchange 111 3.5. Re fe rences Amiro, B. D. (1996) A practical use of simple footprints when measuring evapotranspiration fluxes. 22 n d Conference on Agricultural and Forest Meteorology with Symposium on Fire and Forest Meteorology, Jan. 28 - Feb. 2, 1996, Atlanta, 200-203. Baldocchi, D. D., Hicks, B. B. & Myers, T. P. (1988) Measuring biosphere-atmosphere exchanges of biometeorological related gases with micrometeorological methods. Ecology, 69, 1331-1340. Baldocchi, D. D., Matt, D. R., Hutchison, B. A . & McMillen, R. T . (1984) Solar radiation within an oak-hickory forest: A n evaluation of the extinction coefficients for several radiation components during fully-leafed and leafless period. Agricultural and Forest Meteorology, 32, 307-322. Baldocchi, D. D., Vogel, C . A . & Hall, B. (1997) Seasonal variation of energy and water vapor exchange rates above and below a boreal jack pine forest canopy. Journal of Geophysical Research, (in press). Boast, C . W . & Robertson, T . M . (1982) A "micro-lysimeter" method for determining evaporation from bare soil: Description and laboratory evaluation. Soil Science Society of America Journal, 46, 689-696. Brutsaert, W. (1984) Evaporation into the Atmosphere. Reidel, Dordrecht. Canadian Climate Normals (1982) Canadian Climate Normals 1951-1980: Temperature and Precipitation, Prairie Provinces. Environment Canada, Ottawa. Chen, J. M . , Blanken, P. D., Black, T. A. , Guilbeault, M . & Chen, S. (1997) Radiation regime and canopy architecture in a boreal aspen forest. Agricultural and Forest Meteorology, (in press). Denmead, O. T . & Bradley, E . F. (1985) Flux-gradient relationships in a forest canopy. In B. A . Hutchison & B. B. Hicks (eds.) The Forest-Atmosphere Interaction, 421-442. Dunlap, F. (1912) The specific heat of wood. U S D A Forest Service Bulletin, 110, 28 p. Dyer, A . J. (1972) A review of flux-profile relationships. Boundary-Layer Meteorology, 7, 363-372. Fuchs, M & Tanner, C . B. (1968) Calibration and field tests of soil heat flux plates. Soil Science Society of America Proceedings, 32, 326-328. Seasonal Energy & Water Exchange 112 Gash, J . H . C. (1986) A note on estimating the effect of limited fetch on micrometeorological evaporation measurements. Boundary-Layer Meteorology, 35, 409-413. Hare, F. K . & Thomas, M . K . (1974) Climate Canada. Wiley, Toronto. Hayhoe H . N . , Topp, G . C . & Bailey, W . G . (1983) Measurement of soil water contents and frozen soil depth during a thaw using time-domain reflectometry. Atmosphere-Ocean, 21, 299-311. Hi l le l , D . (1982) Introduction to Soil Physics. Academic Press Inc., San Diego. Hodges, G . B . & Smith, E . A . (1995) Optimal estimates of surface net radiation field over B O R E A S study-area from combination of net pyrradiometer point measurements and G E O S satellite retrievals. Poster Presentation, B O R E A S Workshop, Calverton, Maryland, October 17-20, 1995. Hook, W . R. & Livingston, N . J. (1996) Errors in converting time domain reflectrometery measurements of propagation velocity to estimates of soil water content. Soil Science Society of America Journal, 60, 35-41. Hook, W . R., Livingston, N . J., Sun, Z . J. & Hook, P. B . (1992) Remote diode shorting improves measurement of soil water by time domain reflectrometry. Soil Science Society of America Journal, 56, 1384-1391. Idso, S. B . , Jackson, R. D. , Ehrler, W . L . & Mitchell , S. T. (1969) A method for determination of infrared emmitances of leaves. Ecology, 50, 899-902. Jarvis, P. G. , Massheder, J . M . Hale, S. E . , Moncrieff, J. B. , Rayment, M . & Scott, S. L . (1997) Seasonal variation of carbon dioxide, water vapour and energy exchanges of a boreal black spruce forest. Journal of Geophysical Research, In press. Leclerc, M . Y . & Thurtell, G . W . (1990) Footprint prediction of scalar fluxes using a Markovian analysis. Boundary-Layer Meteorology, 52, 247-258. Marshall, D . C. (1958) Measurement of sap flow in conifers by heat transport. Plant Physiology, 33, 385-396. McCaughey, J. H . & Saxton, W . L . (1988) Energy balance storage terms in a mixed forest. Agricultural and Forest Meteorology, 44, 1-18. McNaughton, K . G . & Jarvis, P. G . (1983) Predicting the effects of vegetation changes on transpiration and evaporation. In T. T. Kozlowski (ed.) Water Deficits and Plant Growth, Volume VII . Academic Press, New York, pp. 1-47. Monsi , M . & Saeki, T. (1953) Uber der lichtfaktor in den Pfllanzengesell-schaften und seine Bedeutung fur die Stoffprodktion. Japan Journal of Botany, 14, 22-52. Seasonal Energy & Water Exchange 113 Monteith, J L . (1965) The state and movement of water in living organisms. In G . E . Fogg (ed.) Society of Experimental Biology Symposium Volume 19, Cambridge University Press, pp. 205-234. Pasquill, F. (1972) Some aspects of boundary layer description. Quarterly Journal of the Royal Meteorological Society, 98, 469-494. Peterson, E . B . & Peterson, N . M . (1992). Ecology, management, and use of aspen and balsam poplar in the prairie provinces, Canada. Forestry Canada Northwest Region, Northern Forest Centre, Edmonton, Alberta, Special Report 1. Rauner, J . L . (1975) Deciduous forests. In J . L . Monteith (ed.) Vegetation and the Atmosphere, Volume II, Case Studies. Academic Press, London, pp. 241-264. Rizzo, B . & Wiken, E . (1992) Assessing the sensitivity of Canada's ecosystems to climate change. Climatic Change, 21: 37-55. Rutter, A . J . (1975) The hydrologic cycle in vegetation. In J . L . Monteith (ed.) Vegetation and the Atmosphere, Volume II, Case Studies. Academic Press, London, pp. 111-154. Saxton, W . L . & McCaughey, J . H . (1988) Measurement considerations and trends in biomass heat storage of a mixed forest. Canadian Journal of Forest Research, 18, 143-149. Schmid, H . P. & Oke, T. R. (1990) A model to estimate the source area contributing to turbulent exchange in the surface layer over patchy terrain. Quarterly Journal of the Royal Meteorological Society, 116, 965-988. Schmid, H . P. (1994) Source areas for scalars and scalar fluxes. Boundary-Layer Meteorology, 67, 293-318. Schuepp, P. H . , Leclerc, M . Y . , MacPerson, J . I. & Desjardins, R. L . (1990) Footprint prediction of scalar fluxes from analytical solutions of the diffusion equation. Boundary-Layer Meteorology, 50, 355-373. Sellers, P., Hal l , F., Margolis, H . , Baldocchi, D. , den Hartog, G. , Cihlar, J., Ryan, M . G. , Goodison, B . , C r i l l , P., Ranson, K . J. , Lettenmaier, D. & Wickland, D. E . (1995) The Boreal Ecosystem-Atmosphere Study ( B O R E A S ) : A n overview and early results from the 1994 field year. Bulletin of the American Meteorology Society, 76, 1549-1577. Shewchuck, S. R. (1996) Surface mesoscale meteorological system for B O R E A S . 2 2 n d Conference on Agricultural and Forest Meteorology with Symposium on Fire and Forest Meteorology, Jan. 28 - Feb. 2, 1996, Atlanta, 42-44. Seasonal Energy & Water Exchange 114 Shuttleworth, W. J. (1993) Evaporation. In D . R. Maidment, (ed.) Handbook of Hydrology, McGraw-Hil l , Inc., New York Spaans, E . A . & Baker, J. M . (1995) The soil freezing characteristic: Its measurement and similarity to the soil moisture characteristic. Soil Science Society of America Journal, 60, 13-19. Topp, G . C . , Davis, J . L . & Annan, A . P. (1980) Electromagnetic determination of soil water content: Measurements in coaxial transmission lines. Water Resources Research, 16, 574-582. CHAPTER 4 AN ANALYSIS OF BOREAL ASPEN OVERSTORY & HAZELNUT UNDERSTORY CANOPY CONDUCTANCE TO WATER VAPOUR 4.1. Introduction Whereas the link between the energy and water balance is the latent heat flux, the link between water contained in vascular plants and the atmosphere is the microscopic pores on the leaf surfaces called stomata. The stomata therefore also represent the link between the disciplines of plant physiology/ecology and meteorology/climatology. These openings (typically 18 / /m in length with 10-170 stomata mm"2 in hypostomatous aspen; Peterson & Peterson 1992) provide a pathway for the absorption of carbon in the form of CO2 at the wet mesophyll cell walls, but as a consequence of carbon uptake, water is evaporated (transpiration). This water loss is regulated by two guard cells flanking the stoma which adjust their turgor osmotically by the uptake or release of potassium. Guard cell turgor is regulated by intercellular CO2 concentration (controlling photosynthesis) and abscisic acid (controlling water stress) (Taiz & Zeiger 1991). With proper techniques, theory and concepts, one can scale the behaviour of individual stoma to the leaf (Jarvis & McNaughton 1986), canopy (Baldocchi et al. 1991) and regional levels (McNaughton & Jarvis 1991). The correct scaling of multi-scale measurements is a primary B O R E A S scientific issue (Sellers etal. 1995). The previous chapter has shown that the transpiration component of the evapotranspiration stream was large indicating that the majority of the water vapour 115 An Analysis of Canopy Conductance 116 reaching the atmosphere passed through the stomata. Therefore, the purpose of this chapter is to analyze the aspen and hazelnut canopy conductance to water vapour in terms of its diurnal and seasonal temporal patterns and relationships to various environmental variables at both the stand and regional scales. 4.2. Materials & Methods 4.2.1. Calculation of Canopy Conductance The concept of canopy conductance is based on the "big-leaf" concept (i.e. that the canopy can be treated as a single, extensive isothermal leaf; Monteith 1981, see Lhomme 1991 or Wallace 1994 for reviews). The canopy conductance (gc) is generally defined as the simple, unweighted total of the stomatal conductances of all the leaves above a representative unit area of ground (Jarvis & McNaughton 1986). The canopy conductance was calculated using two methods. The first involved measurements of canopy temperature (Tc) using infrared thermometry while the second used the Penman-Monteith (PM) equation. Both methods have been used over forests (e.g. Jensen & Hummelsh0J (1995) for Tc method over a spruce forest and Dolman et al. (1991) for P M method in tropical rainforest). The first method, referred to here as the canopy temperature (CT) method, recognizes that the total water vapour conductance from the canopy to the reference height above it (gt) can be expressed as ygt = Vge+ygb+ygt (3.D where gb is the average boundary layer conductance of the canopy leaves and ge is the eddy diffusive conductance of the turbulent air beneath the reference height. The former, An Analysis of Canopy Conductance 117 usually referred to as the excess resistance because the momentum flux is virtually unaffected by this resistance owing to the dominant effect of form drag, is given by where u* is the friction velocity measured by eddy covariance and B" is the dimensionless sublayer Stanton number (Owen & Thompson 1963; Verma 1989). The eddy diffusive conductance is given by where u is the mean wind speed at the reference height. The total conductance can be obtained using where e*(Tc) is the saturation vapour pressure at canopy temperature (measured by an infrared thermometer, IRT), <?a is the vapour pressure at the reference height, AE' is the latent heat flux measured by eddy covariance and corrected for energy balance closure (see Chapter 2, Section 2.4.7.), /?a and cp are density and specific heat of air, respectively, and y is the psychrometric constant. Using the uncorrected AE resulted in an unreasonable amount of scatter in physiological plots such as gc as a function of Qpi and D, for example, decreasing the r2 from 0.33 to 0.21 and from 0.39 to 0.25 with Qpi > 1400 //mol m"2 s"\ aspen and hazelnut, respectively, when AE instead of AE' was used (note /3 is independent of the corrected or uncorrected fluxes since this ratio was maintained when correcting the fluxes). Combining equation (3.2) through (3.4) with (3.1) gives the C T equation for determininggc, (3.2) (3.3) (3.4) An Analysis of Canopy Conductance 118 pacp jy\e* {Tc )-ea]/AE'-u/u*2-B-'/i u* (3.5) During the leaf-less and full-leaf period, B" was set equal to 2.75 and 2.50, respectively, indicative of bare soil (Stathers et al. 1988) and aspen forest leafed conditions (Brutsaert 1984). The above values of B" 1 were used for both calculations of aspen and hazelnutg c. In the second method, gc was calculated using a rearranged form of the P M equation where ft is the Bowen ratio (H/AE), g3 is the aerodynamic conductance (l/g^ = l/gb +l/ge), s is the rate of increase of saturation vapour pressure with air temperature and Da is the saturation deficit at reference height. Use of the P M equation requires the measurement of (3 instead of Tc, which is required in the first method. Calculation of gc for the surface as a whole (referred to as the forest) including aspen transpiration, hazelnut transpiration and soil water evaporation is designated and was determined with equation (3.6) with all variables measured at z = 39 m. The C T method was considered inappropriate since Tc was determined with the upper IRT aimed at the aspen canopy and therefore presumably representing Tc of the aspen and not the forest (see Results & Discussion, Section 4.3.2.). To determine gc from the aspen canopy alone, H' and AE' measured below the aspen or hazlenut canopy (z = 4 m) were subtracted from those measured above, thus V*.= - £ - i ( V * . ) + y AE' (3.6) /UE'(aspen) = AE'(39m) - AE'(4m) zUE'(hazelnut) = AE'(4m) - AE(soil) (3.7) An Analysis of Canopy Conductance 119 and AH'(aspen) = / / ' (39m) - / / ' (4m) ^// '(hazelnut) = / / ' (4m) - / / (soi l) ( 3 " 8 ) where /? in equation (3.6) is AH'IAAE' when (3.6) is used for each individual canopy. When H'(4 m) and AE'(4 m) measurements were unavailable in the late winter, the ratios of the 4-m to 39-m fluxes were used to estimate the 4-m fluxes. The soil water latent heat flux was measured using lysimeters (see Chapter 3, Section 3.2.5.) and was set equal to 23% of AE'(4 m) during the leafed period. During the bare canopy period, AE(soil) was assumed to equal 90% of AE'(4 m). The sensible heat flux at the soil surface was assumed to equal 0 W rn"2, which was probably reasonable given that IRT readings of the soil surfaces at and near the lysimeters were always within 1 °C of the hazelnut Tc. Values of gc for both the aspen and the hazelnut canopies both calculated using the P M method were compared with values calculated using the C T method with the Tc in equation (3.5) determined with IRT thermometers aimed at the respective canopies. The daytime mean.gc was calculated using the daytime means of all the variables in equations (3.5) or (3.6) when Qpi(39 m) exceeded 200 //mol m" s" and only when the canopies were dry as indicated by leaf wetness sensors (model 237, CSI). Variables showing a relationship with gc often need to be defined at the leaf surface (subscript 0) rather than in the ambient atmosphere (subscript a) (Bunce 1985). The temperature (7o), saturation deficit (Do), relative humidity (ho) and CO2 mole fraction (XQ) all at the big-leaf surface were calculated as: (3.9) An Analysis of Canopy Conductance 120 D0={AAE'/gc)[y/(pacp)] (3.10) h0 = \-D0/e.{T0) (3.11) 0 =Zl -KIS: (3.12) where xl l s t n e COz mole fraction in the ambient air outside the leaf boundary layer. The canopy net CO2 assimilation flux density, An, was calculated as An (hazelnut) = Fc (4m) - Rsoil where Rsoi\ is the soil respiration CO2 flux (see Chapter 4, Section 4.3.4.3.). 4.2.2. Calculation of Regional Evapotranspiration Parameters Priestley & Taylor (1972) noted that the regional energy-limited evapotranspiration rate for a vegetated surface was proportional to the equilibrium evaporation rate (AEeq = s/(s + y) i? a) where R3 is the available energy (see Chapter 3, Section 3.2.) with the constant of proportionality, a, equal to 1.26. This 26% increase in AE above AE^ has been shown to be due to entrainment of warm air from above the convective boundary layer (CBL) capping inversion (de Bruin 1983). Stated explicitly, Also at the regional level, McNaughton & Jarvis (1983) quantitatively defined the degree of coupling between transpiration and the saturation deficit of air in the C B L by the decoupling coefficient, Q, as An (aspen) = Fc (39m) - Fc (4m) (3.13) (3.14) n= 1 + 7 8, as (3.15) s + y g An Analysis of Canopy Conductance • 121 where gas is the aerodynamic conductance of the C B L , approximated by g a . Rough surfaces such as forests with a large g a tend to be well-coupled to the air within the C B L and have a low Q (< 0.5) whereas smooth surfaces with a small g a tend to poorly coupled resulting in a large Q (> 0.5). The daytime mean values of a and Q were calculated using the arithmetic means of the variables in equations (3.14) and (3.15). 4.3. Results & Discussion 4.3.1. The Seasonal Patterns of Surface & Canopy Conductances The seasonal patterns of the daytime, dry-canopy gs and gc show marked seasonal variation ranging from near zero to 600 mmol m ' 2 s"1 (Figure 4.1). During the snow-covered, bare-canopy period, the relatively large aspen gc during February may be indicative of some water loss during this period when the aspens were translocating water and nutrients from the roots to the branches in anticipation of leaf development. Although a portion of this gc may be due to the sublimation of intercepted snow, this would likely be small given the small branch area of the aspen. Both gs and aspen gc steadily decrease through March and Apr i l following the February maxima. The trends in gs and aspen g c calculated using the P M equation were matched by those calculated using the C T method. The snow-free, bare-canopy spring period continued to experience the downward trend in gs and g c with the seasonal minimums of both being reached during this time. The P M and C T estimates of hazelnut g c exceeded that of aspen g c and it is difficult to An Analysis of Canopy Conductance 1 2 2 no snow, no leaves 600 *T 400 if) CM O E o LA snow 200 0 600 % 400 200 0 T no leaves no snow, leaves 16 12 8 F M A M J J A S Month 1994 Figure 4.1. Seasonal course of the forest surface conductance, gs (solid line) and aspen (dashed line) and hazelnut (dotted line) canopy conductance (gc). Panel A shows calculations based on the P M equation whereas panel B shows calculations based on canopy temperature (CT method). For comparison, the gs calculated from the P M equation is redrawn in Panel B . Daily values are plotted as a daytime (above-aspen Qpi > 200 //mol m"2 s"1), dry-canopy running mean of 10 days. An Analysis of Canopy Conductance 123 ascribe this difference as correct or due to an overestimate of hazelnut g c because of an underestimate of soil water evaporation. The C T estimate of hazelnut gc exceeded the P M estimate of gs suggesting that the IRT temperature recorded at the 4-m level was that of the soil surface and not the sparse hazelnut branches and stems. Assuming that the hazelnut gc calculated with the C T method is in fact indicative of the soil conductance during this bare-canopy period, soil conductance was highest following snow melt and decreased thereafter reaching a minimum in mid-May. Leaf development was concomitant with an increase in gs and g c (Figure 4.2). The increase ings with leaf development was linear and showed no hysteresis during the decrease in gs with senescence. Such a linear dependence of g c on L A I has also been observed in a deciduous oak forest (Granier & Breda 1996). Once at full-leaf, however, gs andgc decreased steadily from the maximum in early July throughout the remainder of the measurement period. This pattern is similar to the seasonal pattern in soil water content (see Chapter 3, Figure 3.13) but it is unclear whether the post July 1 decline ings and aspen g c was a result or the cause of the decrease in soil water content. If the decline ings and aspeng c was not due to low soil water content (likely, given the high drought-tolerance capability of aspen (Peterson & Peterson 1992) and the wet spring-time conditions), then this pattern must be due to the symmetrical decrease in Qpi about its peak at the summer solstice coupled with the phenological changes in the aspen canopy. The hazelnut g c did not show as strong a decline in g c as did the aspen given that Qpi at the understory level peaked approximately on May 1 when the aspen canopy was still bare. The forest gs was largely dominated by the behaviour of the aspen. The full-leaf daytime mean (± 1 d) gs, aspen g c and hazelnut g c calculated with the P M equation were An Analysis of Canopy Conductance 124 01 0 c/3 eg o D 5 C Q . W < 600 - 400 -o E 200 0 - 0.4 • 0.2 0.0 0 6 Forest LAI ( m 2 m"2) .- 0.08 - 0.04 0.00 0 200 400 600 800 c o Forest g_ (mmol rrf 2 s"1) Figure 4.2. Relationship between forest surface conductance igs) and forest L A I (Panel A) and aspen canopy conductance (gc) calculated using the P M equation and gs (Panel B) . Mean values were calculated from the daytime mean corresponding to binned values of 1 m 2 m"2 (forest LAI) and the 1/2-h binned values of 32 mmol m ' 2 s"1 (forest gs)- Solid lines are linear regressions gs = 47.6a\ + 56.2, r2 = 0.98 (Panel A) and gc = 0.70^ s + 4.9, r2 = 0.99 (Panel B). Vertical lines represent ± one standard deviation and vertical bars represent the frequency distributions with 215 (days) and 1672 (1/2 hours) total number of samples, forest L A I and^s. respectively. An Analysis of Canopy Conductance 125 360 (104), 256 (80) and 122 (43) mmol m" 2 s"1, respectively, yielding g c/gs ratios of 71 ± 6% and 34 ± 8%, for aspen and hazelnut, respectively. The role of the aspen gc in the overall forest gs was confirmed by the linear relationship between 1/2-h aspen gc and forest gs (Figure 4.2). The linear regression slope of 0.70 shows that as mentioned above, the 1/2-h aspen gc was 70% of the forest gc throughout the near Gaussian frequency distribution of the forest gs-Given the dominance of the aspen g c over the hazelnut g c and the similarity in their physiological stomatal responses to the environment (see Results & Discussion, Section 4.3.4.), the forest could be treated as a single-layered aspen canopy with an effective L A I (ae) 43% larger than the aspen L A I (e.g. at full-leaf, ae = ai(aspen) gs/gc(aspen) = 2.3(1/0.70) = 3.29 m 2 m"2). This option is attractive for modelling the boreal aspen forest as a single-layered aspen canopy since current state-of-the-art models such as the Simple Biosphere model, Version 2 (SiB2, Sellers et al. 1996) have abandoned a two-layered approach in favour of a single-layer approach in order to incorporate realistic photosynthesis-conductance modelling and data obtained from satellite-based sensors. 4.3.2. Methods of Calculating Conductances While the general seasonal patterns in gs and g c whether calculated via the P M or C T equations were similar for both species, the magnitudes, however, were larger when the C T equation was used (Figure 4.1). For example, full-leaf means g c for the aspen and hazelnut (± 1 a) calculated with the C T method were 341 ± 136 and 168 ± 52 mmol m" s'\ respectively. Moreover, the aspen g c calculated with the C T method more closely An Analysis of Canopy Conductance 126 resembled the forest gs calculated using the P M equation. This indicates that the apparent aspen Tc determined with the IRT was more indicative of a composite forest Tc. With a mean aspen canopy closure ranging from 83-94% at full-leaf ( B O R E A S Experimental Plan 1994), the IRT mounted above the aspen canopy could have easily sensed some of the hazelnut (and possible soil) vegetation beneath the aspen. The overestimation of the hazelnut gc by the C T equation compared to the P M equation was even larger than that of the aspen overestimation, suggesting that the hazelnut IRT was underestimating the hazelnut big-leaf Tc by seeing a disproportionally large area of shaded understory. The effect of calculating the 1/2-h gc either from the P M or C T equation for both canopies is explicitly shown by Figure 4.3. Linear regression through the points show t h a t £ c calculated from the C T equation typically exceeds that using the P M equation by 27 and 60% for the aspen and hazelnut, respectively. As suggested above, this discrepancy may be due to the IRTs not viewing only the canopy in question, but were "seeing" through the canopy to the substrates below. Alternatively, the assumptions concerning the P M equation may not be valid. There may not be an effective single big-leaf source for water vapour responding to ambient conditions but a rather poorly-defined source distribution through the canopies. The big-leaf assumption, however, should be reasonable in the well-ventilated, small leafed aspen canopy. Regardless, within the range of the majority of the gQ values for both canopies, the differences between gc calculated with either method were not large and the discrepancy between the two methods was systematic and should be interpreted as a possible range in errors in determining gc. An Analysis of Canopy Conductance 127 CO CM 400 800 1:1-Hazelnut HI . . . 0.12 - 0.08 0.04 0.00 0.12 0.08 0.04 0.00 "co •4—» o I— 0 100 200 300 400 g = f(PM) (mmol m"2 s"1) Figure 4.3. Comparison of 1/2-h aspen and hazelnut canopy conductances (gc) calculated either using the P M or C T equation. Mean gc = / ( C T ) was calculated corresponding to binned values of gc = / ( P M ) at 32 //mol m"2 s"1 intervals with ± one standard error of the mean represented by vertical bars. The linear regressions (solid lines) a r e £ c / ( C T ) = 1 .27^ C / (PM) - 12.3, r2= 0.99 a n d ^ c / ( C T ) = 1 .60^ C / (PM) - 11.8, r 2 = 0.93, aspen and hazelnut, respectively. The total number of observations were 1683 and 1171 1/2 hours, aspen and hazelnut, respectively. An Analysis of Canopy Conductance 128 4.3.3. Seasonal Progression of the Diurnal Surface & Canopy Conductances The pronounced seasonal pattern in gs and gc were exemplified by the monthly plots of the ensemble mean diurnal gs and gc (Figure 4.4). For consistency and comparison purposes all conductances shown were calculated with the P M equation. In addition to reiterating the seasonal progressions ings andg c and the relative importance of aspen and hazelnutg c relative togs. Figure 4.4 illustrates two important conductance features. First, both the aspen and hazelnut tended to have their highest g c shortly after sunrise when water use efficiency can be high given the ample Qpi and low saturation deficits. Thereafter, g c decreased steadily though the remainder of the day as D increased and finally Qpi levels fell at sunset. There was no decrease in the aspen g c at the time of maximum D (mid afternoon) followed by a subsequent recovery. Such a diurnal pattern is evidence of a mid-day stomatal closure water conservation strategy (Blanken & Rouse 1996) and the absence of this is consistent with the wet 1994 conditions and high drought tolerance capacity of the aspen. There was a hint at a mid-day decrease in the hazelnut g c (July and August), suggesting that the hazelnut did experience partial mid-day stomatal closure to conserve water and was more sensitive to saturation deficits. Second, although subject to large errors due to small fluxes, there was the suggestion that there was some water loss at night from the leafed aspen canopy (also found using sap-flow methods by Hogg & Hurdle 1997). Since the stomates should be closed in the absence of light, these nocturnal g c estimates may be an indication of the aspen cuticular conductance. These cuticular conductances in July, August and An Analysis of Canopy Conductance 129 CM O E E^  o 600 400 200 0 600 400 200 0 600 400 200 0 . 1 1 I I , . . , . , - Apr i l - 1 1 1 1 1 1 1 1 1 1 1 May " - J u n e September -. 1 1 . . 1 1 1 1 1 1 • i i i i i i i i i i 0 6 12 18 0 6 12 18 24 T ime (h) CST Figure 4.4. Ensemble monthly averages of the forest surface (gs; • ) , aspen canopy (•) and hazelnut (•) canopy conductances (gc) all calculated using the P M equation. Each data point represents the mean of 2 1/2-h periods with ± one standard error of the mean shown by vertical lines. An Analysis of Canopy Conductance 130 September were larger that the typical 4-20 mmol m" 2 s"1 range (Jones 1992) implying a relatively thin wax layer on the aspen leaves. In contrast, the nocturnal hazelnut gc was often negative indicating condensation on the hazelnut leaves. 4.3.4. Controls on Aspen & Hazelnut Canopy Conductances: Calculating Responses from Observed Relationships While the general seasonal trends in gc are largely a function of L A I , the diurnal patterns are largely a function of the ambient environmental conditions. It is well known that the stomatal conductance of almost all terrestrial plants responds to Qpi, CO2 concentration, water status, humidity, temperature, pollutants, leaf phenology and nutrition (Jones 1992). Forest species are generally especially sensitive at the diurnal scale to Qpi and humidity with the latter possibly due to the high aerodynamic roughness of forests. Tables 4.1a and 4.1b shows several popular forms of the relationships between gc and several variables. Simple relationships between gc and Z) a or Do were chosen on the basis of the likelihood of a strong response in forests well coupled to the surrounding atmosphere and follows the non-linear least-squares approach of Stewart (1988). These empirical relationships between gc and D are often criticized for their empiricism and lack of physical interpretation (i.e. physiological mechanisms) (Collatz et al. 1991). In response to these criticisms, advances have been made in a more process-oriented relationship, relating gc to AN, humidity and^Jj. Arranging these variables as AJIQ/ZO I S referred to as the Ball-Woodrow-Berry index ( B W B ; Bal l , Woodrow & Berry 1987). The B W B index is claimed to be universal amongst C3, C4 or conifer species (parameterized by controlled leaf chamber measurements of gc against AJIQ/%1 with linear regressions An Analysis of Canopy Conductance 131 Table 4.1a. Relationship between 1/2-hourly full-leaf (June 7 - September 10, 1994) dry canopy aspen canopy conductance and variables known to affect gc, listed from least to highest scatter. Saturation deficits were expressed as mole fractions. Parameters were determined from means calculated over 20 equally spaced bins. Data were excluded when the n I total n was less than 5%. Functional Relationship igc in mmol m" 2 s"1) a b r 2 Mean Standard Deviation (mmol m" 2 s"1) Total n (1/2 h) gc = aAJ (DozD +b 0.038 142.1 0.93 118.6 1717 gc = aAn/D0+b 99.1 150.8 0.93 121.1 1729 gc = aAnI Da + b 89.8 157.3 0.94 127.7 1730 gc=aAnh0lxl +b 9.20 105.0 0.97 127.9 1735 gc=aAnh0 + b 26556 101.5 0.98 128.1 1734 gc = a A J (JD.xl) +b 0.032 155.7 0.94 128.9 1713 gc = aAn + b 11372 133.0 0.92 139.0 1738 gc = aAnhJxl + b 2.57 129.4 0.96 144.8 1738 gc = aD0 + b -11146 413.5 0.98 145.1 1758 gc = a exp {-b D0) 438.2 38.8 0.98 145.1 1758 gc=aAnha+ b 6964 136.4 0.98 150.0 1734 gc = aDa + b -10243 394.6 0.99 151.8 1774 gc = a exp (-b Da) 414.6 36.4 0.98 151.8 1774 An Analysis of Canopy Conductance 132 Table 4.1b. Relationship between 1/2-hourly full-leaf (June 7 - September 10, 1994) dry canopy hazelnut canopy conductance and variables known to affect gc listed from least to highest scatter. Saturation deficits were expressed as mole fractions. Parameters were determined from means calculated over 20 equally spaced bins. Data were excluded when the n I total n was less than 5%. Functional Relationship (g c in mmol m"2 s*1) a b 2 r Mean Standard Deviation (mmol m" 2 s"1) Total n (1/2 h) gc = a An / (DoZc0) +b 0.031 38.8 0.94 58.2 1427 gc = aAn/D0 + b 78.2 39.5 0.97 62.2 1474 gc = aD0 + b -9695.1 204.6 0.84 65.7 1510 gc = a exp {-b D0) 247.7 92.3 0.93 65.7 1510 gc = aDa + b -9063.7 170.6 0.85 109.8 1617 gc = a exp (-b D a ) 193.4 59.8 0.89 109.8 1617 gc = aAnh0/zl +b 1.30 105.4 0.11 112.8 1495 gc = a Anh0 + b 4944.0 102.0 0.29 115.0 1539 gc = aAnha/xl +b -0.54 133.2 0.94 115.1 1597 gc = aAnl (A>Xa) + b 0.021 63.0 0.95 116.9 1561 gc = aAn + b -5414.9 152.0 0.89 118.5 1598 gc = aAnl Da + b 54.0 65.4 0.99 119.0 1563 g c - a A n h a + b -1699.3 136.7 0.76 121.9 1597 An Analysis of Canopy Conductance 133 yielding slopes of 9, 4 and 6, for C3, C4 and conifers, respectively, and y-intercepts of 10 and 40 mmol m" 2 s"1, C3 and C4, respectively). Use of this index to model gc has been widely accepted and is currently incorporated into Global Climate Models (GCMs) through SiB2 (Sellers etal. 1996). What follows is a discussion of the various relationships given in Table 4.1a and 4.1b and how these relationships may or may not be employed as useful predictors of gc. Of note, the least scatter (lowest mean standard deviation) was observed when the aspen or hazelnut gc was plotted against AJ(DoXo>- In general, An served as a good predictor for the aspen gc and the use of Do instead of Z) a decreased the scatter for the hazelnut gc. 4.3.4.1. Relating Conductance to Humidity Stratified by Light The relationship between the forest gs and aspen and hazelnut gc to Qpi and humidity, with humidity expressed as a vapour pressure deficit at the leaf surface (Do), is shown in Figure 4.5. Relating conductance to Do rather than D& decreased the amount of scatter in gs or gc-D plots, especially in the case of the hazelnut which was less coupled to the air outside the leaf boundary layer than the aspen (Table 4.1). Because of the compounding influence of light on the relationship between gc and D0, the data were grouped according to high Qpi (Qpi > 1400 //mol m" 2 s"1), medium Qpi (800 < Qpi < 1400 //mol m" 2 s'1) and low Qpi (200 < Qpi < 800 //mol m"2 s"1). For both species and for the forest as a whole, there was a non-linear decrease in gc with Do (described best by the non-linear curve fit gs or gc = gmax exp(-bD0) ) with the relationships maintained as Qpi decreased. Similar non-linear responses have been observed for temperate (McCaughey & Iacobelli 1993) and subalpine aspen (Massman & An Analysis of Canopy Conductance 1200 134 (J) C M 800 400 0 - 600 o | 400 d? 200 ° 0 CO 300 200 100 0 Forest g S -*• 4-Aspen g, Hazelnut g 30 20 10 20 10 10 0 0.0 0.5 1.0 1.5 2.0 2.5 3.0 D 0 (kPa) Figure 4.5. Empirical relationships between the forest surface (gs) or canopy (gc) conductance and the 1/2-h saturation deficit at the big leaf surface (Do) and for the full-leaf period when stratified by high (Qpi > 1400 /tfnol m"2 s"1; • and solid line), medium (800 < Qpi < 1400 jimol m"2 s'1; • and dashed line) and low Qpi (200 < Qpi < 800 //mol m"2 s"1; A and dotted line) photosynthetic photon flux density. Mean values of gs a n d g c were calculated at binned 0.25 kPa intervals of DQ. Parameters defining the non-linear curve fits (lines) of the form gs o r g c = g c max exp(-bD0) are given in Appendix A , Table A.4 . Vertical lines are +, + and - one standard deviation, (high, medium and low Qpi, respectively). An Analysis of Canopy Conductance 135 Kaufmann 1991). In contrast, linear responses of gc to Do in aspen have been reported using leaf chambers where a non-linear function could justifiably be used (see Fredeen et al. 1997 Figure 6b). A linear response is unusual and most studies regardless of species report strongly non-linear responses such as those found here (Bunce 1985, Monteith 1995a). In addition, non-linear relationships expressed as gc = /(1/Z)) support the conclusion that stomata respond to the transpiration rate itself, not humidity, as shown by Mott & Parkhurst (1991) and reiterated by Monteith (1995a) since this implies that E is constant and independent of D (however expressed) with all other variables held constant. Moreover, a non-linear response of gc to D suggests a constant dAJ6El and therefore a high water-use efficiency (Lloyd 1991). In contrast, g c decreasing linearly with D or decreasing as exp(-bD) implies a maximum E occurring at an optimum D, with all other variables held constant. In the case of gc = a exp(-bDo), the maximum E occurred when Do = lib or 1.89, 1.29 and 1.08 kPa at high, medium and low Qpi, respectively. Independent measurement of aspen E at this site using sap-flow methods (Hogg et al. 1997) and leaf chamber (Fredeen etal. 1997) both show a maximum in E. Whereas those believing g c = -a D + b or gc = a exp{-bD) agree that a maximum E occurs at a certain D, their disagreement with the linear or non-linear response has serious implications for predicting , g c as wi l l be shown latter. This discrepancy between non-linear and linear D responses may stem from leaf chambers failing to take into account the wide range in leaf temperatures experienced in the real canopy or maybe due to a single leaf not responding to D as the canopy does as a whole. While a mechanistic causal interpretation of this relationship may indeed be weak (a mechanism(s) for the sensing of D or relative humidity at the leaf surface (ho) has not An Analysis of Canopy Conductance 136 yet been identified; see Sheriff 1984) the strong empirical relationship is still instructive at hinting where causal relationships may lie. As suggested previously and indicated by the b parameters, the sensitivity of gc to Do (i.e. dgc/dEt) increased as Qpi decreased and the sensitivity of the hazelnut gc to D0 was greater than that of the aspen for all Qpi levels. Estimating the maximum gc (gcmax) from the non-linear curve fits gave values of roughly 825 mmol m"2 s"1 (21 mm s'1) and 350 mmol m" 2 s"1 (9 mm s"1) for the aspen and hazelnut, respectively. These estimates are lower than the 1200 mmol m"2 s"1 (30 mm s"1) and 910 mmol m" 2 s"1 (23 mm s"1) originally given by Blanken et al. (1997) wheng c was calculated using the C T equation. Which g c m a x estimates are correct depends on whether the P M or the C T equation was more appropriate and for the reasons given previously, it appears that the P M equation was more suitable. Dividing thegCmax by the full-leaf LAIs gives an estimated maximum stomatal conductance of 359 mmol m"2 s'1 (9.1 mm s"1) and 106 mmol m"2 s"1 (2.7 mm s'1) for the aspen and hazelnut, respectively. These values are comparable to the 10-15 mm s"1 (aspen) and 4-8 mm s"1 (hazelnut) maximum stomatal conductances estimated from porometer measurements made in 1996. The aspen maximum canopy and stomatal conductances compare well with the 19.7 mm s"1 (± 1.5 standard error) and 5.5 mm s"1 (7% uncertainty), canopy and stomatal conductances, respectively, given in a review of 16 woody species by Kelliher et al. (1995). It is, however, difficult to compare the maximum hazelnut canopy and stomatal conductances with values given by Kelliher et al. (1995) because forest understory vegetation was not given. An Analysis of Canopy Conductance 137 4.3.4.2. Calculating Aspen Conductance & Transpiration from Saturation Deficit & Light It is instructive to calculate diurnal patterns of gc using empirical gc relationships and parameters. Although the procedure does not use independent data, it does indicate how reliably variations in day-to-day canopy conductance can be calculated. Given the dominance of the aspen over the hazelnut gc in terms of the total forest gs and the desirability of single-layered modelling approaches, only the aspen gc was calculated from the empirical relationships. The parameters and relationships determined over the entire full-leaf period (June 7 - September 10, 1994) from 1/2-hourly dry canopy, daytime measurements were tested against the 1/2-hourly measured gc calculated using the P M equation and measured aspen AE from August 13-18, 1994. These days represent a range in ambient conditions from clear to cloudy to overcast and a range in saturation deficits and air temperatures peaking at the maximum observed D3 over the entire year (August 15) (Figure 4.6). The close correspondence between ambient and leaf-surface air temperature and saturation deficits (Figure 4.6) illustrates the close coupling between the air at the leaf surface and in the forest surface layer. The aspen gc was first calculated as gc - a exp(-£D a ) and then as gc = a exp(-bD0) with a and b as a function of Qpi in both cases. To move from the air outside the leaf boundary layer to inside the boundary layer, an iterative procedure was used. First, the aspen AE was calculated using the Penman-Monteith equation in its original form with the initial gc = a exp(-kDa): An Analysis of Canopy Conductance 138 A u g u s t 1994 (CST) Figure 4.6. Measured above aspen canopy (z = 39 m) 1/2-h photosynthetic photon flux density (Qpi) (thick line), air temperature (Ta) and saturation deficit (Da) (thin lines) representing a wide range in ambient conditions over which to test the empirical aspen gc calculations. Also shown are daytime, dry canopy air temperature (7b) and saturation deficit (Do) at the aspen canopy "big-leaf" surface (thick lines). An Analysis of Canopy Conductance 139 AE = (3.16) and Ho was estimated as H0=RM-AE (3.17) where /?ao is the available energy at the aspen "big leaf" surface. To estimate Rao, the 1/2-h measured Ra for the forest as a whole was plotted against the measured aspen Rao where Rao = AAE'(aspen) + AH'(aspen). This yielded the linear relationship (not shown) Ra0 = 0.78/?a+ 1.6 W m ' 2 (r2 = 0.99, n - 10,259) which was used to estimate Ra0 for the aspen canopy. If Ra measurements were not available, Ra could be estimated from above-aspen R„ and the soil heat flux density (see Chapter 3, Section 3.2.4.). Equations (3.9) through (3.11) were then used to calculate temperature and humidity at the leaf surface. The "new" g c recalculated as gc = a exp(-kDo) was compared to the initial gc and if the absolute value of the difference was less than a specified value (e.g. 1 mmol m" 2 s"1), equations (in order) (3.16), (3.17), (3.9), (3.10) and (3.11) were repeated until convergence was obtained. The results of this calculation w i t h g c = ftDo, Qpi) are shown in Figure 4.7. This is essentially the Jarvis-Stewart approach to estimating gc using independent multiplicative functions (Stewart 1988). The diurnal course of the measured gc was well represented by the calculated gc, even responding to changes in D 0 induced by cloudy periods. In general, the calculated gc tended to overestimate the measured gc, especially Z 1 when Do was low, with a mean and calculated gc of 268 and 324 mmol m" s" , respectively, over the six-day period. Using D a instead of Do resulted in only a slight difference (mean gc of 316 mmol m"2 s"1). A n average of three iterations was required to An Analysis of Canopy Conductance 140 i 1 1 1 r A u g u s t 1994 (CST) Figure 4.7. Measured (thin lines) and calculated (thick lines) daytime dry-canopy 1/2-h aspen canopy conductance (gc) and transpiration (AE) calculated using gc = a exp(-bD0) and AE using the PM equation. This is essentially a test of the Jarvis-Stewart approach using two variables, Do and light. An Analysis of Canopy Conductance 141 move from D a to D 0 . The close agreement in the diurnal patterns and the close agreement i n g c calculated either with Do o r D a implies that the aspen gc was largely a function of D and was closely coupled to the air in the forest surface layer. Calculating the aspen AE using the P M equation and gc = / ( D 0 , Qpi) produced a similar diurnal pattern to the measured AE, again tending to overestimate when D was low, stemming from the overestimation of gc. Over the six-day period the mean calculated aspen AE of 137.5 W m" 2 calculated w i t h g c = /(Do, QPi) was almost identical to the 138.2 W m" 2 calculated from gc = f(Da, Qpi) and both agreed well with the measured mean AE of 134.1 W m" 2. 4.3.4.3. Relating Conductance to Net Assimilation, Relative Humidity & CO2 Concentration at the Leaf Surface: The Ball-Woodrow-Berry Index Currently, canopy conductance is being parameterized in SiB2, a G C M land surface sub-model (Sellers et al. 1996) using gc = m An ho Ix\ + ba\ where m and b are the slopes and intercepts of this linear relationship, An h0 I%\ is the Ball-Woodrow-Berry ( B W B ; Ba l l , Woodrow & Berry 1987) index. Determining the relationship between gc and the B W B index from canopy-scale measurements presumably does not require Fc measurements to be separated between aspen overstory and hazelnut understory since both are C3 species so the relationships should be the same (m = 9, b = 1 0 mmol m" 2 s"1; Sellers et al. 1996). Only the soil respiration needs be removed from Fc to obtain the forest An. Overstory and understory An Analysis of Canopy Conductance 142 eddy-covariance measurements, however, facilitated the isolation of the aspen canopy An (i.e. Fc at the 4-m level subtracted from that at the 39-m level), but to determine the An of the hazelnut (CO2 respiration from the soil subtracted from Fc at the 4-m level), the soil respiration had to first be determined. The daytime soil respiration was estimated from the biological relationship between soil temperature at the 2-cm depth and the nighttime CO2 storage-corrected 4-m Fc (Figure 4.8). The relationship was best described by an exponential function similar to that reported by Black et al. (1996). It was unclear why the high temperature bins did not fit this relationship. In addition to the storage (z Ax\IAt), a correction was also made for the lack of spatially homogenous turbulence at low wind speeds (see Chapter 2, Section 2.4.5.). Failure to make this correction decreased the soil respiration flux shown in Figure 4.8 by about 50% and resulted in the hazelnut^,, being less than the aspen at low light levels (see Figure 4.9), a result seriously inconsistent with the.4 n- Qpi relationship observed for shade-intolerant (aspen) and shade-tolerant (hazelnut) species (Taiz & Zeiger 1991). The ratio of the soil respiration at a given soil temperature to that 10 °C lower (the Q\0 ratio) was 8.1, higher than the mean Q\0 of 3.08 reported for several hardwood forest soils (Kicklighter et al. 1994). The Q10 value reported here, however, may not be unreasonable given that root respiration comprises a large component of soil respiration (Paul & Clark 1989) coupled with the high root growth of aspen (i.e. aspen are notorious for root vegetative clonal reproduction; Peterson & Peterson 1992) and the fact that the aspen are nearing the end of their life cycle and are entering a degenerative stage. An Analysis of Canopy Conductance 143 Nocturnal 2-cm Soil Temperature (°C) Figure 4.8. Empirical relationship between the storage and low wind speed corrected C 0 2 flux measured at the 4-m height (assumed to be a good estimate of soil respiration) and the mean 1/2-h nocturnal soil temperature at a depth of 2 cm (7/s). Mean soil respiration values were calculated at 1 °C 7^  intervals with ± one standard error of the mean shown by the vertical lines. The solid line shows the best-fit exponential line given as soil respiration = 0.4349exp(0.2094rs). r2 = 0.95. The 13.5 °C and 14.5 °C Ts bins were excluded from the curve fit. Vertical bars represent the 7^ frequency distribution with a total of 2611 1/2-h observations. An Analysis of Canopy Conductance 144 The net assimilation must also be calculated when using the B W B index. The photosynthesis model used in S iB2 is that of Collatz et al. (1991) in which An is the minimum of three limiting rates: /') Rubisco efficiency; ii) light or iii) capacity to export C (C3 species) or utilize PEP-Carboxylase (C4 species). Since the focus here is not to model An, it was estimated simply using a function of Qpi (Figure 4.9). Besides serving as a crude but effective empirical model, Figure 4.9 also reveals a well-defined relationship from which several salient physiological parameters may be determined. The.4n- Qpi relationship for both species was non-linear and best described by a rectangular hyperbolic function. The hazelnut had a higher quantum yield (slope at the origin) than the aspen (0.0786 and 0.0259 mol mol" 1, hazelnut and aspen, respectively), and a lower photosynthetic capacity (An at Qpi = 1800 //mol m"2 s"1) of 10.0 //mol m" 2 s'1 2 1 compared to the aspen's 18.8 //mol m" s" , the latter characteristic of shade-tolerant species (Salisbury & Ross 1978). Dark respiration rates (= -An when Qpi = 0) which were very close to zero, were higher for the aspen (0.16 //mol m"2 s"1) than the hazelnut (0.094 //mol m" 2 s"1). Comparing these results to an exhaustive review of An- Qpi relationships by Ruimy et al. (1995) shows that the non-linear relationships reported here were typical for forests. The boreal aspen quantum yield, photosynthetic capacity and dark respiration rate were all lower than the 0.037 mol mol" 1, 21.39 //mol m"2 s"1, and 3.34 //mol m"2 s"\ respectively, reported as means for several temperature deciduous forests. There were no understory species given for comparison. Figure 4.10 shows the relationship between gc and the B W B index, in which An was calculated from the difference between eddy-covariance CO2 fluxes at the 39- and 4-m heights. The relationship between the aspen gL and the B W B index was linear, with a An Analysis of Canopy Conductance 145 Figure 4.9. Relationship between net CO2 assimilation (An) by the aspen (•) and hazelnut ( O ) and incident photosynthetic photon flux density (Qpi) at the 39 and 4-m heights. Mean values of An were calculated corresponding to a binned 1/2-h Qpi width 2 1 of 100 (aspen) and 20 (hazelnut) pmo\ m" s" . Solid lines represent rectangular hyperbolic curve fits An = (aQpib)/(a Qpi + b) - R where a, b and R were 0.0259, 31.2123, -0.1646 (aspen) and 0.0786, 9,6638, -0.9381 (hazelnut), respectively. Aspen and hazelnut r 2 were 0.99 and 0.88, respectively. Dashed line shows an extrapolation of the hazelnut Qpi to simulate overstory Qpi levels. The vertical lines are ± one standard error of the mean. An Analysis of Canopy Conductance 146 10 20 30 40 50 Anh0li£(mmo\ m" 2 s' 1) c - 4 — ' o Figure 4.10. The empirical relationship between the June 7 - September 10, 1994 aspen and hazelnut canopy conductance (gc) and AM x\ ( B W B index) as suggested by Ball et al. (1987). A mean gc was calculated at binned B W B index values at 3 (aspen) and 1.5 (hazelnut) mmol m"2 s"' intervals. The solid lines represents the linear regression gc(aspen) = 7.68 B W B index + 128.4 (r 2 = 0.87) and gc(hazelnut) = 2.84 B W B index + 86.1 (r 2 = 0.36). Vertical lines represent ± one standard deviation (means of 144.2 and 104.8 mmol m"2 s"1, aspen and hazelnut, respectively). Vertical bars represent the frequency distributions with a total n of 1735 (aspen) and 1344 (hazelnut) 1/2 hours. An Analysis of Canopy Conductance 147 2 1 2 slope of 7.68, a y-intercept of 128.4 mmol rrf s" and an r of 0.87. This slope was less than the slope of 9 given for all C3 species and the 128 mmol rrf s" intercept was higher 2 \ than the expected 23 mmol m" s" . Restricting the data to include only those means from bins containing 95% of the data did, however, give a slope of 9.20 (Table 4.1a) close to the suggested slope of 9. In the case of the hazelnut gc, the slope and y-intercept and r2 of the linear regression were 2.84, 86.1 and 0.36, respectively, inconsistent both with the aspen and expected values. Restricting data to include only those means from bins containing 95% of the data resulted in a slope of 1.30, further from the expected value of 9 (Table 4.1b). 4.3.4.4. Calculating Aspen Conductance & Transpiration from the Ball-Woodrow-Berry Index The calculation of g c using the B W B index was performed identically to the procedure described above for g c = /(Do,Qpi) using parameters given in Table 4.1a. The diurnal plots of the measured and calculated gc with gc = / ( B W B index) (Figure 4.11) show an underestimation of the early morning high and an overestimation of the late afternoon low gc (i.e. a lack of sensitivity to Do). This pattern was not unique to these six test days, but was characteristic of the entire full-leaf period. The meangc, however, was not seriously 2 -1 affected with the calculated mean of 284 mmol m" s" slightly higher than the measured mean of 268 mmol m"2 s"1. Without iterating (mean number of iterations = 2) and just using ambient relative humidity and C 0 2 mol fraction gave a mean gc of 278 mmol m" s"1. The resultant 6-day daytime mean aspen AE from the B W B calculated gc of 139 and An Analysis of Canopy Conductance 148 Figure 4.11. Measured (thin lines) and calculated (thick lines) daytime dry-canopy 1/2-h aspen canopy conductance (gc) and transpiration (AE) with gc calculated using gc = a AMX\ + b and AE using the P M equation. An Analysis of Canopy Conductance 149 134 W m ' 2 (with h a n d ^ at and above the leaf surface, respectively) still compared favorably with the measured AE of 133 W m' 2 . 4.3.4.5. Relating Conductance to a Modified Form of the Ball-Woodrow-Berry Index The poor agreement between the measured and calculated aspen gc using the B W B index was puzzling. A thorough investigation into this problem showed that as indicated by the large amount of scatter shown in Figure 4.10, the slope of t h e g c - B W B index relationship which controls the sensitivity varied widely from as high as 18 (August 13) to as low as 7 (August 17) with underestimation of the sensitivity of gc to ho when ho was high (low saturation deficits). The previous discussion (see Chapter 4, Section 4.3.4.1.) has shown that gc was largely a function of Do in this aerodynamically rough forest, so underestimating the sensitivity of gc to humidity becomes apparent. In an attempt to impart greater sensitivity of gc to humidity while retaining the general form of the B W B index, ho was moved from the numerator to the denominator where it was expressed as Do (i.e. ho replaced by 1/Do). This gave a modified form of the B W B index as An/(DoZl) with the saturation deficit expressed as a mole fraction in order to maintain the same units of the index and gc. This modification substantially decreased the scatter in plots of gc against the B W B index (now modified) for both the aspen and hazelnut (least scatter of all theg c relationships tested) and gave more consistent slopes for both species (0.035 and 0.031, aspen and hazelnut, respectively; Figure 4.12). Lloyd (1991) also obtained better results in predicting theg c of Macadamia integrifolia when ho was replaced by 1/Do- We were unaware of Lloyd's identical modification until after theg c analyses were complete. An Analysis of Canopy Conductance 150 C/) 800 - n „ CM £ 6 0 0 - n E400 E, 0 200 0 1 0 ^400 - o § 3 0 0 O £ 2 0 0 c5100 O 0 0.00 0 3000 6000 9000 12000 15000 AJ(DX) (mmol rrf2 s"1) Figure 4.12. The empirical relationship between the June 7 - September 10, 1994 aspen and hazelnut canopy conductance (gc) and a modified form of the B W B index where ho has been replaced by \/DQ to improve the sensitivity of gc to humidity. A mean gc was calculated at binned modified B W B index values at 750 (aspen) and 300 (hazelnut) mmol rrf2 s"1 intervals. The solid lines show linear regressions with slopes, y-intercepts and r 2 of 0.035, 135.7 mmol rrf2 s'1, 0.91 (aspen) and 0.031, 40.4 mmol m"2 s \ 0.95 (hazelnut), respectively. The vertical lines show ± one standard deviation (means of 152.1 and 76.1 mmol rrf2 s"\ aspen and hazelnut, respectively). Vertical bars represent the frequency distributions with a total n of 1717 (aspen) and 1278 (hazelnut) 1/2 hours. An Analysis of Canopy Conductance 151 4.3.4.6. Calculating Aspen Conductance & Transpiration from the Modified Ball-Woodrow-Berry Index The seemingly simple modification of the B W B index altered the calculation of the aspen gc and AE (Figure 4.13) (again using the procedure described previously with parameters from Table 4.1a). The incorporation of the non-linear Do response was apparent by a l lowingg c to obtain its early morning high and decrease more during the afternoon. The early morning gc was often overestimated, resulting in a daytime mean calculated gc of 298 mmol m" 2 s"\ above the measured mean of 268 mmol m"2 s"1. Not using D and x c at 2 1 the leaf surface resulted in a slightly higher meang c of 300 mmol rrf s" , further from the measured value (a mean of three iterations were required to obtain gc from Do and To). The tendency to underestimate the mid-day gc resulted in an underestimation of the aspen AE at these times when AE was large giving slightly worse performance on the diurnal AE compared to that determined with the unmodified B W B . Compared with the other simulations, the 6-day daytime mean AE of 132 W m"2 (131 W m" 2 without iteration), was, however, closest to the measured AE of 133 W r r f 2 . The improvement made in the 1/2-hourly gc calculation worsened the 1/2-hourly AE predictions (although hidden by daily means) since the improvements in gc were made at times when AE was small and the underestimation of gc near mid-day was made at times when AE was large. The rational for making this modification to the B W B index lies in the relationship between gc and humidity. The B W B index in its original form assumes thatg c is a linear function of ho- Al lowing T\ to vary through a wide range may demonstrate the expected non-linear relationship, as leaf chamber measurements are often operated within a narrow An Analysis of Canopy Conductance 152 Figure 4.13. Measured (thin lines) and calculated (thick lines) daytime dry-canopy 1/2-h aspen canopy conductance (gc) and transpiration (AE) with calculated using gc = a AJ(D0 zl) + b and AE using the P M equation. An Analysis of Canopy Conductance 153 T\ range where a non-linear curve may be approximated by a straight line. Especially in forests where Z)0 plays a large role in g c , the non-linear response is demanded in gc calculations. There is considerable debate as to whether plants actually respond to ho at all, but are actually responding to Z>0, as shown by Aphalo 8c Jarvis (1991). Further, ho is a composite of both a humidity and temperature response (although in the aspen, To drives DQ) and T\ may be required as a variable regulating gc since gc sometimes responds to T\ at constant D0 (Aphalo & Jarvis 1993). Mott & Parkhurst (1991) further show that the apparent response of stomata to Do is actually a response to the transpiration rate itself. This conclusion was supported by Monteith (1995a) who shows that in 52 sets of measurements on 16 species, stomata responded to the transpiration rate, not humidity. As stated by Monteith (1995a) "If the true response is to transpiration rate, however, predictions from these models must be flawed." (p. 357) where "models" refers to predictinggc as an empirical function of D or h. 4.3.5. Regional Evapotranspiration Parameters: The Priestley & Taylor a & the McNaughton & Jarvis Q Coefficients The importance of the accommodation between transpiring vegetation and the CBL has been recognized by several authors (e.g. McNaughton & Spriggs 1989, McNaughton & Jarvis 1991, Monteith 1995b). This accommodation is maintained by a positive feedback between g c and Da. Increasing g c decreases H and decreases the height of the CBL (the height of the CBL, z-„ can be approximated by the rate of change of z, with time from an initial early morning zx as dzjdt = (H+0.07AE)/(pacpziyv) where yv is the strength of the An Analysis of Canopy Conductance 154 inversion at the top of the C B L (McNaughton & Spriggs 1989) and vigor of entrainment of warm, dry air from above the capping inversion. This in turn decreases the C B L saturation deficits (less volume of air to be humidified) and, as shown in Figure 4.5, increases ,g c (hencegs when soil water evaporation is small) thus completing the positive feedback cycle. If the initial response was a decrease in g c , H would increase promoting growth of a drier, deeper C B L with the increasing saturation deficits and a decreasing g c ; again positive feedback since the direction of the initial g c response was maintained. This must not be confused with the larger-scale regional negative feedback between evapotranspiration and saturation deficit in the C B L , where evapotranspiration becomes insensitive to the C B L saturation deficit when evapotranspiration rates are high (as discussed in McNaughton & Jarvis 1991). The Priestley & Taylor a coefficient has been used to predict AE from many types of surfaces (e.g. Spittlehouse 1989). Considering the same six days used to test the estimation of g c from the various relationships describe above, a shows a strong diurnal pattern (Figure 4.14) similar to that reported by de Bruin (1983) in part due to the diurnal variation in g c . The daytime mean a over these six days was 1.11 (± cr0.31). A well-mixed C B L , however, is insensitive to this diurnal variation hence allowing a estimates of AE to be good rough approximations (de Bruin 1989). On a seasonal basis (Figure 4.15) the mean daytime a of 0.27 ( ± c 0 . 2 4 ) and 0.30 (±.<J0.29) with a bare canopy with and without snow, respectively, indicates that AE was below the equilibrium rate due to restrictions on the water supply. In contrast, the presence of a transpiring canopy removed most of the restrictions on the water supply An Analysis of Canopy Conductance 155 3.0 | 1 1 1 ' 1 1 T 0 0 1 1 ' 1 1 1 1 — 1 — 1 — 1 — L " 13 14 15 16 17 18 August 1994 (CST) Figure 4.14. The diurnal course of the dry-canopy 1/2-h Priestley & Taylor a during the daytime for August 13-18, 1994 (mean 1.11 ± a 0.31). An Analysis of Canopy Conductance 156 no snow, Month 1994 Figure 4.15. The seasonal pattern of the daytime mean, dry-canopy forest Priestley & Taylor or and the McNaughton & Jarvis decoupling coefficient Q. Thick lines show a 10-day running mean. Vertical dashed lines denote major changes in the surface conditions as noted on the figure. An Analysis of Canopy Conductance 157 increasing the daytime mean a during the snow-free, leafed period to 0.99 (± o 0.28). During this period, AE was limited largely by Ra, with only slight stomatal limitation on water supply. The relationship between a and L A I was well defined by a non-linear relationship (Figure 4.16) and shows the importance of g c via the linear relationship with L A I in regulating a. The similarity between the air inside the aspen leaf boundary layer and the air in the surface layer and the C B L has been shown through the g c analysis described previously. Figure 4.15 shows the seasonal progression of this similarity expressed quantitatively with the McNaughton & Jarvis Q coefficient. The bare canopy Q of 0.09 (± cr 0.15) and 0.08 (± cr 0.08) with and without snow, respectively, increased to 0.36 (± cr 0.18) with a leafed canopy indicating the forest AE was less coupled to the C B L when it had leaves. As Q generally increases as surface roughness decreases, the implication was that the forest was aerodynamically smoother with leaves than without. This was confirmed by the daytime mean aerodynamic conductances of 84.8 and 71.0 mm s"\ without and with leaves, respectively, both slightly less than the 100-200 mm s"1 range typical for forests (Oke 1987). Hence, the increase in Q was the result of both an increase in g c and a decrease in g a . The decrease in coupling as L A I increases also concurs with the gc-D positive feedback effect describe above. Increasing gc (by adding more leaves) lessens the influence of D in the C B L on gc since the C B L is shallow with poor entrainment of dry, warm air above the inversion layer. The relationship between gs- gc, L A I , and D is illustrated by Figure 4.17. Using a mixed-layer model, McNaughton and Spriggs (1989) showed that a is relatively insensitive to changes i n g c for conductances larger than approximately 800 mmol m" 2 s"1 An Analysis of Canopy Conductance 158 Figure 4.16. The relationship between the daytime mean forest dry canopy Priestley & Taylor or and the forest LAI (a\). The line represents the best-fit non-linear curve a -1.19exp(l- (a|/1.70)), r2 = 0.96 . Mean or values were calculated at a binned LAI (width 1 m 2 m"2) with ± one standard deviation shown by vertical lines. Vertical bars show the forest LAI frequency distribution (total n - 223 days). An Analysis of Canopy Conductance 159 Figure 4.17. The relationship between the forest daytime mean, dry-canopy Priestley & Taylor a and the forest surface conductance (gs)- The line represents the best-fit non-linear curve a = 1.33exp(l- (g s/290.5)), r2 = 0.82, « = 142 days. An Analysis of Canopy Conductance 160 (20 mm s"1). This is due to the negative feedback at the regional scale between AE and the saturation deficit in the C B L (McNaughton & Jarvis 1991). In agreement with this, Figure 4.17 shows that when thegs reached 870 mmol rrf2 s"1 (22 mm s"1) a had adjusted to within 95% of the asymptotic a of 1.33, within the 1.1 to 1.4 range given regardless of surface wetness (Monteith 1995b). Hence because of the high gs at full leaf, the C B L was insensitive to the diurnal changes in gs- The deciduous nature of the canopy in contrast to boreal coniferous forests, however, resulted in a seasonal variation ings which corresponded to a wide range in a below the asymptotic value. During the period of leaf development (or during the full-leaf period if water stress imposes a decrease in gc), the strength of the regional negative feedback would increase and the C B L would become sensitive togs- Thus, the timing, duration and maximum forest L A I (through its directly proportional relationship to gs) limits a below the expected value for much of the year and influences the strength of the regional surface-CBL negative feedback. A strongly seasonally variable g s , mediated through the timing and maximum L A I also has implications for C B L growth via the positive feedback between conductance and saturation deficits described previously. 4.4. Summary & Conclusions This chapter has attempted to quantify some of the processes involved in regulating canopy conductance to water vapour and hence transpirational water loss from a boreal aspen forest. The currently popular Ball-Woodrow-Berry (BWB) parameterization of canopy conductance as a linear function of relative humidity at the leaf surface did not work well. This was in agreement with others and the B W B parameterization was An Analysis of Canopy Conductance 161 improved when conductance was expressed as a non-linear function of saturation deficit. Although there are still unknown complexities involved in parameterizing canopy conductance, the modified form of the B W B index is attractive given the consistent slopes and y-intercepts of the shade-intolerant aspen and shade-tolerant hazelnut species. Whether the slopes and y-intercepts remain constant across vegetation types with the modified B W B approach remains to be seen. The Jarvis-Stewart approach at calculating canopy conductance as a function of saturation deficit and light also works reasonably well (better than the B W B parameterization for the hazelnut). This approach, however, does suffer when applied to other vegetated sites where species-specific parameters have to be determined, so the modified B W B approach still remains attractive (assuming net assimilation can be calculated accurately). The wet conditions in 1994 resulted in a high evapotranspiration rate during the full-leaf period with rates approaching the equilibrium rate indicating that transpiration was largely an energy rather than water supply limited (hence stomatal controlled) process. This result hides the tremendous complexities in predicting canopy conductance since transpiration estimates often match the measured transpiration when calculated in any reasonable fashion using models such as the simple Priestley-Taylor a or a more complex, physiological model. This, however, does not excuse the researcher from understanding the stomatal regulation of atmospheric water vapour at this site since in drier years, stomatal control may be more important. An Analysis of Canopy Conductance 162 4.5. References Aphalo, P. J. & Jarvis, P. G. (1991) Do stomata respond to relative humidity ? Plant, Cell and Environment, 14, 127-132. Aphalo, P. J. 8t Jarvis, P. G. (1993) An analysis of Ball's empirical model of stomatal conductance. Annals of Botany, 72, 321-327. Baldocchi, D. D., Luxmoore, R. J. & Hatfield, J. L. (1991) Discerning the forest from the trees: an essay on scaling canopy stomatal conductance. Agricultural and Forest Meteorology, 54, 197-226. Ball, J. T., Woodrow, I. E. & Berry, J. A. (1987) A model predicting stomatal conductance and its contribution to the control of photosynthesis under different environmental conditions, In J. Biggins (ed.) Progress in Photosynthesis Research, Martinus Nijhoff Publishers, Dordrecht, pp. 221-224. Black, T. A., Den Hartog, G., Neumann, H. H., Blanken, P. D., Yang, P. C., Russell, C., Nesic, Z., Lee, X., Chen, S. C , Staebler, R. & Novak, M. D. (1996) Annual cycles of water vapor and carbon dioxide fluxes in and above a boreal aspen forest. Global Change Biology, 2, 219-229. Blanken, P. D. & Rouse, W. R. (1996) Evidence of water conservation mechanisms in several subarctic wetland species, Journal of Applied Ecology, 33, 842-850. Blanken, P. D., Black, T. A., Yang, P., Neumann, H. H., Staebler, R., Nesic, Z., Den Hartog, G., Novak, M. D., & Lee, X. (1997). The energy balance and canopy conductance of a boreal aspen forest: Partioning overstory and understory components, Journal of Geophysical Research, (in press). BOREAS Experimental Plan, Chapters 1-3, Version 3.0 (1994) Editors: Sellers, P.J., Hall, F.G., Baldocchi, D., Cihlar, J., Crill, P., Den Hartog, J., Goodison, B., Kelly, R.D., Lettenmeier, D., Margolis, H., Ranson, J. & Ryan, M., NASA, Greenbelt. Brutsaert, W. (1984) Evaporation into the Atmosphere, Reidel, Dordrecht. Bunce J. A. (1985) Effect of boundary layer conductance on the response of stomata to humidity. Plant, Cell and Environment, 8, 55-57. Collatz, G. J., Ball, J. T., Grivet, C. & Berry, J. A. (1991) Physiological and environmental regulation of stomatal conductance, photosynthesis and transpiration: a model that includes a laminar boundary layer. Agricultural and Forest Meteorology, 54, 107-136. de Bruin, H. A. R. (1983) A model for the Priestley-Taylor parameter a. Journal of Applied Meteorology, 22, 572-578. An Analysis of Canopy Conductance 163 de Bruin, H. A. R. (1989) Physical aspects of the planetary boundary layer with special reference to regional evapotranspiration. In T. A. Black, D. L. Spittlehouse, M. D. Novak & D. T. Price (eds.) Estimation of Areal Evapotranspiration, IAHS Press, Wallingford, Publication No. 177, pp. 117-132. Dolman, A. J., Gash, J. H. C , Roberts, J. & Shuttleworth, W. J. (1991) Stomatal and surface conductance of tropical rainforest, Agricultural and Forest Meteorology, 54, 303-318. Fredeen, A. L., Fu, W., Collatz, G. J. & Berry, J. A. (1997) An analysis of diurnal patterns in stomatal conductance and photosynthesis in tree species at the southern edge of the boreal forest zone in central Saskatchewan. Tree Physiology (in review). Granier, A. & Breda, N. (1996) Modelling canopy conductance and stand transpiration of an oak forest from sap flow measurements. Annates des Science Forestieres (in press). Hogg, E. H., Black, T. A., den Hartog, G., Neumann, H. H., Zimmerman, R., Hurdle, P. A., Blanken, P. D., Nesic, Z., Yang, P. C., Staebler, R., McDonald, K. C. & Oren, R. (1997) A comparison of sap flow and eddy flux of water vapor from a boreal deciduous forest, Journal of Geophysical Research (in press). Hogg, E. H. & Hurdle, P. A. (1997) Diurnal and nighttime sap flow in aspen in relation to vapour pressure deficit, Tree Physiology (in review). Jarvis, P. G. & McNaughton, K. G. (1986) Stomatal control of transpiration: Scaling up from keaf to region. Advances in Ecological Research, 15, 1-49. Jensen, N. O. & Hummelshoj, P. (1995) Derivation of canopy resistance for water vapour fluxes over a spruce forest, using a new technique for the viscous sublayer resistance, Agricultural and Forest Meteorology, 73, 339-352. Jones, H. G. (1992) Plants and Microclimate. Cambridge University Press, Cambridge. Massman, W. J. & Kaufmann, M. R. (1991) Stomatal response to certain environmental factors: a comparison of models for subalpine trees in the Rocky Mountains. Agricultural and Forest Meteorology, 54, 155-167. Kelliher, F. M., Leuning, R., Raupach, M. R. & Schulze, E.-D. (1995) Maximum conductances for evaporation from global vegetation types. Agricultural and Forest Meteorology, 73, 1-16. Kicklighter D. W., Melillo, J. M. Peterjohn, W. T., Rastetter, E. B. McGuire, A. D. 8c Steudler, P. A. (1994) Aspects of spatial and temporal aggregation in estimating regional carbon dioxide fluxes from temperate forest soils, Journal of Geophysical Research, 99, 1303-1315. An Analysis of Canopy Conductance 164 Lhomme, J-P. (1991) The concept of canopy resistance: historical survey and comparison of different approaches, Agricultural and Forest Meteorology, 54, 227-240. Lloyd, J. (1991) Modell ing stomatal response to environment in Macadamia integrifolia. Australian Journal of Plant Physiology, 18, 649-660. McCaughey, J . H . & Iacobelli, A . (1994) Modell ing stomatal conductance in a northern deciduous forest, Chalk River, Ontario. Canadian Journal of Forest Research, 24, 904-910. McNaughton, K . G . & Jarvis, P. G . (1983) Predicting the effects of vegetation changes on transpiration and evaporation. In T. T. Kozlowski (ed.) Water Deficits and Plant Growth, Volume VII . Academic Press, New York, pp. 1-47. McNaughton, K . G . & Jarvis, P. G . (1991) Effects of spatial scale on stomatal control of transpiration, Agricultural and Forest Meteorology, 54, 279-301. McNaughton, K . G . & Spriggs, T. W . (1989) A n evaluation of the Priestley and Taylor equation and the complementary relationship using results from a mixed-layer model of the convective boundary layer. In T. A . Black, D . L . Spittlehouse, M . D . Novak & D . T. Price (eds.) Estimation of Areal Evapotranspiration, I A H S Press, Wallingford, Publication No.177, pp. 89-104. Monteith, J . L . (1981) Evaporation and surface temperature. Quarterly Journal of the Royal Meteorological Society, 107, 7-27. Monteith, J. L . (1995a) A reinterpretation of stomatal responses to humidity: theoretical paper. Plant, Cell & Environment, 18, 357-364. Monteith, J . L . (1995b) Accommodation between transpiring vegetation and the convective boundary layer. Journal of Hydrology, 116, 251-263. Mott, K . A . & Parkhurst, D . F. (1991) Stomatal responses to humidity in air and helox. Plant, Cell & Environment, 14, 509-515. Oke, T. R. (1987) Boundary Layer Climates, Methuen, London. Owen, P. R. & Thompson, W . R. (1963) Heat transfer across rough surfaces. Journal of Fluid Mechanics, 15, 321-334. Paul, E . A . & Clark, F. E . (1989) Soil Microbiology and Biochemistry, Academic Press Inc., San Diego. Penman, H . L . (1948) Natural evaporation from open water, bare soil and grass. Proceedings of the Royal Society of London, Series A , 193, 120-146. An Analysis of Canopy Conductance 165 Peterson, E . B . & Peterson, N . M . (1992) Ecology, Management, and Use of Aspen and Balsam Poplar in the Prairie Provinces, Canada. Special Report 1, Forestry Canada, Northwest Region, Northern Forestry Centre, Edmonton. Priestley, C. H . B . & Taylor, R. J . (1972) On the assessment of surface heat flux and evaporation using large-scale parameters. Monthly Weather Review, 100, 81-92. Ruimy, A . , Jarvis, P. G . , Baldocchi, D . D. & Saugier, B . (1995) CO2 fluxes over plant canopies and solar radiation: a review. Advances in Ecological Research, 26, 1-68. Salisbury, F. B . & Ross, C. W . (1978) Plant Physiology, Second Edition, Wadsworth, Belmont. Sellers, P., Ha l l , F., Margolis, H . , Baldocchi, D. , den Hartog, G. , Cihlar, J. , Ryan, M . G. , Goodison, B . , C r i l l , P., Ranson, K . J., Lettenmaier, D . & Wickland, D. E . (1995) The Boreal Ecosystem-Atmosphere Study ( B O R E A S ) : A n overview and early results from the 1994 field year. Bulletin of the American Meteorology Society, 76, 1549-1577. Sellers, P. J., Randall, D . A . , Collatz, G . J., Berry, J. A . , Field, C. B . , Dazlich, D . A . , Zhang, C , Collelo, G . D. , & Bounoua, L . (1996) A revised land surface parameterization (SiB2) for atmospheric G C M s . Part I: Model Formulation. Journal of Climate, 9, 676-705. Sheriff, D . W . (1984) Epidermal transpiration and stomatal responses to humidity: Some hypothesis explored. Plant, Cell and Environment, 7, 669-677. Spittlehouse, D . L . (1989) Estimating evapotranspiration from land surfaces in British Columbia. In T. A . Black, D . L . Spittlehouse, M . D. Novak & D . T. Price (eds.) Estimation of Areal Evapotranspiration, I A H S Press, Wallingford, Publication No. 177, pp. 245-256. Stathers, R. J., Black, T. A . , Novak, M . D. & Bailey, W. G . (1988) Modell ing surface energy fluxes and temperatures in dry and wet bare soils. Atmosphere-Ocean 26, 59-73. Stewart, J . B . (1988) Modell ing surface conductance of a pine forest. Agricultural and Forest Meteorology, 43, 19-35. Taiz, L . & Zeiger, E . (1991) Plant Physiology, Benjamin/Cumings, Redwood City. Verma, S. B . (1989) Aerodynamic resistances to transfers of heat, mass and momentum. In T. A . Black, D . L . Spittlehouse, M . D . Novak & D. T. Price (eds.) Estimation of Areal Evapotranspiration, I A H S Press, Wallingford, Publication No.177, pp. 13-20. An Analysis of Canopy Conductance 166 Wallace, J. S. (1995) Calculating evaporation: resistance to factors. Agricultural and Forest Meteorology, 73, 353-366. CHAPTER 5 SUMMARY & CONCLUSIONS This study has investigated aspects of the evaporation within and above a southern boreal aspen forest. The primary goal of this research was to quantify and understand the processes involved in evaporation within and above the forest as part of B O R E A S . A n overview of the processes controlling southern boreal forest evaporation wil l be presented here in a "top-down" approach beginning with the forest as a whole, followed by aspen overstory, hazelnut understory and finally the soil. The forest experiences a dramatic change in surface conditions with the growth and shedding of leaves, a characteristic which sets this forest apart from its boreal coniferous counterparts. The aerodynamic characteristics above the aspen canopy were not much affected by the presence or absence of leaves. Slightly more friction velocity was generated with leaves present at wind speeds greater than 4.0 m s"1 and the mean daytime aerodynamic conductance decreased slightly from a bare-canopy 84.4 mm s' x to a full-leaf value of 71.0 mm s"1. Leafing was concomitant with a decrease in the pre-leaf forest sensible heat flux from a pre-leaf daytime mean of 172 W m~2 to a post-leaf 70 W m"2 and an increase in forest latent heat flux from 28 W m" 2 (pre-leaf) to 139 W m" 2 (post-leaf). During the relatively wet year of 1994, 90% of the annual precipitation was lost as evapotranspiration. Evapotranspiration rates approached the equilibrium rate with daytime mean Priestley and Taylor a of near unity at full leaf. Forest evapotranspiration (aspen and hazelnut evapotranspiration was 95% of the total forest evapotranspiration) and surface conductance to water vapour (daytime mean 360 mmol m"2 s"1) were large, 167 Summary & Conclusions 168 relative to coniferous boreal forests. The high surface conductance at full-leaf diminished the sensitivity of the convective boundary layer to the canopy conductance. The rough aerodynamic nature of the forest meant close coupling between the air inside the leaf boundary layers and the surface layer with a full-leaf daytime mean McNaughton and Jarvis decoupling coefficient of 0.36. Surface conductance and the Priestley and Taylor a were well-defined functions of forest leaf area. The aspen canopy played a dominant role in determining the forest evapotranspiration, despite having a smaller maximum leaf area index (2.3 m 2 m"2) than the hazelnut understory (3.3 m 2 m"2). The ratio of the horizontal wind speed above the aspen canopy to that below decreased from 0.20 before leafing to 0.15 after leafing. With leaves, 71% of the total forest conductance to water vapour originated from the aspen (daytime full-leaf mean of 256 mmol rrf2 s"1). There was a well-defined, non-linear relationship between aspen canopy conductance and the saturation deficit when stratified by light. This non-linear relationship, coupled with other physiological studies showing a stomatal response to saturation deficit, not relative humidity, meant that the aspen canopy conductance was best parameterized by net assimilation divided by the product of the mole fractions of saturation deficit and CO2 at the leaf surface. Because of the high transpiration rate, the calculation of aspen transpiration using the conductance-saturation deficit relationship or the more physiologically based relationship using net assimilation were both acceptable. The latter, however, better predicted aspen canopy conductance. Within the aspen canopy above the hazelnut understory, turbulent statistics and spectral analyses showed that large, intermittent, horizontal gusts similar to those above the aspen canopy were responsible for turbulent mass and energy exchange. Winds above Summary & Conclusions 169 the hazelnut understory were otherwise light, with means of 0.62 and 0.39 m s'1 without and with aspen leaves above. The amount of radiation reaching the understory was a function of aspen leaf area, with full-leaf simple extinction coefficients of 0.49, 0.44 and 0.54 for net, photosynthetically active and solar radiation, respectively. With leaves, the daytime mean understory latent and sensible heat fluxes were 33 and 8 W m" 2, respectively, 24% and 10% of the corresponding fluxes measured above the forest. Roughly 25% of the annual precipitation was evaporated from the hazelnut understory. Hazelnut canopy conductance to water vapour (daytime mean of 122 mmol m~2 s"1) represented 34% of the total forest surface conductance. As with the aspen, this conductance was also well defined by a non-linear relationship with saturation deficit stratified by light. Less coupling to the atmosphere above meant that using the saturation deficit at the leaf surface rather than in the air above improved this relationship. The hazelnut canopy conductance was best parameterized by net assimilation divided by the product of the mole fractions of saturation deficit and CO2 at the leaf surface. Soil water evaporation was low, totaling only 4% of the total annual precipitation. The daytime mean latent heat flux from the soil decreased from 19 W m"2 with no leaves above to 7 W m"2 with leaves above. Soil water content in the 8-10 cm deep surface organic horizon increased markedly in response to the spring snow melt and thawing of frozen water and responded to precipitation events throughout the year while the deeper mineral horizons responded only to large rainfall events. Over the growing season, drainage past a depth of 123 cm (deeper than the 60 cm rooting depth) was estimated as 3-8% of the total 347 mm of precipitation. Respiration from soil biota was large indicative of high aspen root growth and the degeneration of the aspen nearing the end of Summary & Conclusions 170 its life cycle. With the absence of white spruce emerging through the dense hazelnut understory, it appears that the death of the current aspen will be succeeded by another aspen stand from clonal growth from root stocks. This succession is classified as "stable" and is characterized by high levels of aspen stocking, low mortality and few conifers (Peterson & Peterson 1992). The majority of the measurements mentioned above were obtained using the eddy-covariance method. Although it has advantages for use over rough forests and especially within canopies, a critical analysis of this method revealed that the technique did suffer from failure to close the surface energy balance at low wind speeds and occasional erratic behaviour. A t friction velocities below 0.30 and 0.10 m s"1 (above and below aspen canopy, respectively), violation of the assumption of spatially homogenous turbulent statistics resulted in turbulent flux underestimation. Suspect eddy-flux measurements during these periods were best identified when the difference between the measured similarity function for vertical wind speed and that expected by Monin-Obukhov theory exceeded 20% of the latter. Fluxes measured during these periods were corrected using a function based on friction velocity. Use of eddy-covariance flux measurements to satisfy the conservation of energy and infer physiological relationships required a correction to the flux measurements based on the partitioning of the missing energy flux using the ratio of the measured sensible to latent heat flux. Scientific investigations usually generate more questions than answers. This study was no exception and a few pertinent suggestions for future work are as given here. The issue of energy balance closure and erratic flux behaviour remains largely unresolved and may be in part due to the natural spatial heterogeneity of the forest. This may be Summary & Conclusions 171 quantified, for example, placing another eddy-covariance system on one (or both) of the existing tall towers adjacent to the main tower. The eddy-covariance instrumentation need not be complex {e.g. a one-dimensional sonic anemometer) yet could still determine if measurements taken at one location above/within the forest are sufficient. A better assessment of energy balance closure and improvement to the correction of the turbulent fluxes using available energy and the ratio of sensible to latent heat could be made by again by better quantifying the spatial heterogeneity of above-canopy net radiation and forest biomass heat storage. The former can be assessed by mounting the understory tram system on the existing support cables located above the aspen canopy. The latter can be assessed by placing thermocouples in aspen boles in several trees and heights. Photosynthesis was an important variable in the prediction of aspen conductance to water vapour. The isolation of the aspen canopy by CO2 flux measurements both above and below the aspen canopy effectively measured the aspen photosynthesis, but such extensive instrumentation is rare. Therefore, if photosynthesis is in the future to be used to predict conductance, a practical yet effective model of photosynthesis needs to be developed. As a final conclusion, a great deal was learned not only about this boreal forest ecosystem, but also the benefits, difficulties and complexities of being involved in a large project such as B O R E A S . The cooperation of the many people from different backgrounds and disciplines working together and sharing data enabled the success of this highly scrutinized publicly funded project. As resources become increasingly limited, perhaps projects such as this will become more popular as expertise can be Summary & Conclusions 172 appropriately placed allowing much more knowledge to be obtained and exchanged than if projects are undertaken individually. Summary & Conclusions 173 5.1. References Peterson, E. B . & Peterson, N . M . (1992) Ecology, Management, and Use of Aspen and Balsam Poplar in the Prairie Provinces, Canada. Special Report 1, Forestry Canada, Northwest Region, Northern Forestry Centre, Edmonton. APPENDIX A. CURVE FIT PARAMETERS & COEFFICIENTS A.1. Introduction This appendix presents in tabular format various parameters and coefficients determined from linear and non-linear regression analyses which were performed using SigmaPlot software (SigmaPlot 1994). Table A . l . Parameters determined from a non-linear least squares fit to a 4-parameter logistic curve of the form f(x)=a /(l + exp[b(x-c)]) + d used to fit the data presented in Figure 2.11. Parameter 39-m Fc, AE and H 4-mFc 4-m AE and H a 0.9884 0.8226 42.4317 b -17.1900 -40.8245 -99.5168 c 0.1575 0.0657 0.1418 d -0.0370 0.2755 0.1203 2 r 0.9783 0.8731 0.8676 174 Curve Fit Parameters & Coefficients 175 Table A . 2 . Parameters determined from a linear regression of the form fix) -m x + k, where m is the slope and k is the y-intercept used to fit the energy balance closure data presented in Figure 2.14. Instrument Height Period Corrected for low u* m k r2 n (1/2-h) (m) underestimation ? 39.0 Day N / A 0.95 -7.77 0.99 5500 39.0 Night No 0.53 4.03 0.97 4216 39.0 Night Yes 0.65 2.21 0.94 4216 4.0 Day N / A 1.09 -14.93 0.98 4002 4.0 Night No 0.19 1.71 0.85 3012 4.0 Night Yes 0.76 5.56 0.84 3012 Curve Fit Parameters & Coefficients 176 Table A . 3 . Coefficients determined for the curve fit of the form Qi(4 m)/G\|.(39 m) a exp(-ba\) as shown in Figure 3.6. Radiation a b r2 n (days) Stream Qi Rn 0.4682 0.3038 0.74 44 Qpi 0.5719 0.4450 0.76 44 Rsi 0.2910 0.3283 0.72 44 Curve Fit Parameters & Coefficients 177 Table A.4 . Parameters and defining the non-linear least squares determined equation gs or gc - gcmax exp(-Z?£>o) as shown in Figure 4.5. The high (H), medium (M) and low (L) photosynthetic photon flux density levels (Qpi) were Q„i > 1400 jumol rrf2 s'1, 800 < Qpi < 1400 //mol rrf2 s"1 and Qpi (200 < Qpi < 800 jumol m s'\ respectively. Qpi gmax (mmol rrf2 s"1) b (kPa) r2 Mean n (1/2 Standard hours) Deviation (mmol rrf2 s"') H 1107 0.5781 0.99 70.0 279 Fores tg s M 1177 0.8310 0.95 161.2 769 L 633 0.9252 0.99 258.2 766 H 760 0.5302 0.98 60.9 277 Aspen g c M 826 0.7730 0.90 203.5 761 L 476 0.9221 0.94 195.5 758 H 271 0.6836 0.98 107.9 270 Hazelnutg c M 353 1.1071 0.97 83.8 659 L 209 1.8219 0.97 131.5 595 Curve Fit Parameters & Coefficients 178 A.2. References SigmaPlot Transforms & Curve Fitting (1994) Jandel Scientific Software 2.0, San Rafael. APPENDIX B. MAPS & PHOTOGRAPHS 3 0 - m A tall Rohn tower 65-m long Tram cable 20-m long Soil Heat Flux Transect Aspen with Gravimetric 0 Thermocouples Sampling area TDR Probes ^ Understory O Infrared O Thermometer^ 37-m tall Main tower Hut A N (approx.) Soil Temp. Profile + ^ 0 6 0 1 0 2 0 3 Hut B TDR Rods Forest Opening with Precipitation Gauge # x X Neutron^ Probes 1-m wide Boardwalk Soil Lysimeters Trail (1.5 km to Fish Lake Road) 4-m tall Understory tower A 30-m tall SRC tower Not to Scale Figure B . l . locations. Plan view of the main tower study area showing major instrumentation 179 Maps & Photographs 180 Figure B.2. Map of B O R E A S southern study area showing O A (Old Aspen), Y A (Young Aspen), O B S (Old Black Spruce), OJP (Old Jack Pine), Y J P (Young Jack Pine) Fen sites and the B O R E A S operations centre (Ops). Source: B O R E A S Home Page on the World Wide Web. Maps & Photographs 181 Figure B.3. View looking east standing on the main tower in the early-spring bare-canopy period. Note SRC tower near centre of photograph. Figure B . 4 . View looking east standing on main tower in the late-summer full-leaf period. Note SRC tower at bottom left of photograph. Maps & Photographs 1 Figure B.5. View of 37-m tall main tower from ground. Note sonic anemometer and infrared gas analyzer (white box) at top of tower and temperature profile sensors (black cylinders). Maps & Photographs 183 Figure B.6. Understory canopy flux tower showing sonic anemometer 4-m above ground with heated sampling tube connecting to the infrared gas analyser (white box). Figure B.7. Understory tram used to measure radiation levels along a 65-m transect. Visible are two net radiometers, two pyranometers and two quantum sensors. Figure B.8. Cross-sectional view of aspen 21 5-m tall, 70-yr old aspen stand and 2-m tall hazelnut understory. Such vegetation was typical around the study area. Maps & Photographs 185 Figure B.9. Above aspen canopy eddy-covariance instrumentation 39-m above ground supported on top of the main tower. Visible are the Kaijo-Denki sonic anemometer (left), heated H2O-CO2 (black), ozone (clear) and N2O-CH4 (funnel) sampling tubes. 

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.831.1-0088118/manifest

Comment

Related Items