Enhanced assimilation rate and water use efficiency with latitude through increased photosynthetic capacity and internal conductance in balsam poplar (Populus balsamifera L.)



    1. Department of Forest Sciences, University of British Columbia, Vancouver, British Columbia, Canada V6T 1Z4 and
    2. Agroforestry Division, Agriculture Agri-Food Canada, Indian Head, Saskatchewan, Canada S0G 2K0
    Search for more papers by this author

    Corresponding author
    1. Department of Forest Sciences, University of British Columbia, Vancouver, British Columbia, Canada V6T 1Z4 and
      R. Guy. Fax: +1 604 822 9102; e-mail: rob.guy@ubc.ca
    Search for more papers by this author

    1. Agroforestry Division, Agriculture Agri-Food Canada, Indian Head, Saskatchewan, Canada S0G 2K0
    Search for more papers by this author

    1. Department of Forest Sciences, University of British Columbia, Vancouver, British Columbia, Canada V6T 1Z4 and
    Search for more papers by this author

    1. Agroforestry Division, Agriculture Agri-Food Canada, Indian Head, Saskatchewan, Canada S0G 2K0
    Search for more papers by this author

R. Guy. Fax: +1 604 822 9102; e-mail: rob.guy@ubc.ca


In outdoor common gardens, high latitude populations of deciduous tree species often display higher assimilation rates (A) than low latitude populations, but they accomplish less height. To test whether trends in A reflect adaptation to growing season length or, alternatively, are garden growth artefacts, we examined variation in height increment and ecophysiological traits in a range-wide collection of Populus balsamifera L. populations from 21 provenances, during unconstrained growth in a greenhouse. Rooted cuttings, maintained without resource limitation under 21 h photoperiod for 90 d, displayed increasing height growth, A, leaf mass per area and leaf N per area with latitude whereas stomatal conductance (gs) showed no pattern. Water-use efficiency as indicated by both gas exchange and δ13C increased with latitude, whereas photosynthetic nitrogen-use efficiency decreased. Differences in δ13C were less than expected based on A/gs, suggesting coextensive variation in internal conductance (gm). Analysis of ACi curves on a subset of populations showed that high latitude genotypes had greater gm than low-latitude genotypes. We conclude that higher peak rates of height growth in high latitude genotypes of balsam poplar are supported by higher A, achieved partly through higher gm, to help compensate for a shorter growing season.


Variation in geographically-widespread species is produced either by phenotypic plasticity (acclimation to environmental conditions) or by adaptation of genotypes to specific environmental conditions (Conover & Schultz 1997). Understanding of these patterns has been advanced by numerous studies along resource gradients (Poorter 1999; Reich et al. 1999) and with latitude or altitude (see Farmer 1993; Howe et al. 1995; Jonas & Geber 1999). Most such studies are restricted to variation along a single transect or a limited number of populations. Species with extensive geographic ranges, however, have the potential to exhibit large intraspecific variation in physiology, morphology, phenology and growth rate, and thus constitute good models for the study of local and regional adaptation. In this regard there are numerous reports of differentiation in photosynthesis to contrasting environments in Populus species (e.g. Ceulemans et al. 1992). Such patterns have not, however, been explored in much detail in balsam poplar (Populus balsamifera L.), a transcontinental species with a wide range in the boreal zone across North America, from Colorado to Nunavut, and from Alaska to Newfoundland. Balsam poplar is very closely related to black cottonwood (P. trichocarpa Torr. & Gray); in fact, black cottonwood is considered to be a subspecies of P. balsamifera in many treatments. Black cottonwood is the first woody plant to have had its genome sequenced (Tuskan et al. 2006).

Latitude represents a complex environmental gradient, along which photoperiod, temperature, growing season length [frost-free days (FFDs)], moisture availability and soil nutrient status can all be expected to vary. Differences in growing season length have been reported to correlate with characters such as leaf nitrogen per area (Leaf N), specific leaf area and photosynthesis (see Reich, Walters & Ellsworth 1997; Diemer 1998). In a pot study using populations of Sitka alder [Alnus sinuata (Reg.) Rydb.] and paper birch (Betula papyrifera Marsh.) from British Columbia, Canada, Benowicz, Guy & El-Kassaby (2000b) found an intrinsic relationship between midsummer photosynthetic rate and subsequent levels of fall frost hardiness (as growing season length is reflected in the date of hardiness development). Similarly, Gornall & Guy (2007) reported that photosynthetic rates (A) increased with latitude of origin in five provenances of black cottonwood along with an increase in Leaf N, stomatal density (SD) and stomatal conductance (gs), with no trend in leaf mass per area (LMA). Intrinsic water use efficiency (WUEi) and carbon isotope discrimination did not vary with latitude, implying a common internal conductance (gm) to the diffusion of CO2 from the intercellular space to the site of carboxylation, but this was not assessed. Flexas et al. (2007) emphasized that comparisons of gm in different natural plant populations, and its possible influence over efficiencies of water-use and nitrogen-use, should be a research priority.

The trend towards increased photosynthesis with latitude may be a case of parallel evolution among deciduous trees. Genotypic differences found in common garden environments presumably reflect adaptive variation, but careful experimental work is necessary to distinguish traits under selection from other effects such as plasticity or artefacts of differential growth.

By and large, height increment is greater in low latitude populations than in high latitude populations when both are growing in a common garden. Such was the case in the study reported by Gornall & Guy (2007) – although the northern black cottonwood populations had the highest photosynthetic rates, they were also smaller in size, and it is possible that one result may have spawned the other. Photosynthetic rates may have declined in the larger low latitude trees because of increased self-shading, or because of higher rates of soil nitrogen depletion. Furthermore, if adaptation to shorter growing seasons has indeed resulted in higher rates of carbon assimilation, then there should be a similar effect on height increment when plants are kept free of other limitations, most particularly the short days (or long nights) that trigger the cessation of active shoot elongation.

By making use of the extensive Agriculture Canada Balsam Poplar (AgCanBaP) collection we set out to test the hypothesis that during free growth (i.e. without photoperiod, nutrient and water limitations) genotypes native to regions with short growing seasons would grow more rapidly than genotypes adapted to longer growing seasons, commensurate with higher A. The study aimed to answer the following questions:

  • 1Do climate-related patterns in A and related ecophysiological variables occur under greenhouse conditions when all genotypes are of similar age and size, and are grown without resource limitation?
  • 2If population-level variation in A exists in balsam poplar, is it reflected in variation in height increment during free growth under extended days?
  • 3Are the proximal causes for population-level variation in A in balsam poplar the same as those previously reported for the closely related black cottonwood?


Plant material

This study makes use of the AgCanBaP collection of the Agriculture and Agri-Food Canada (AAFC), Agroforestry Division, Indian Head, Canada. Balsam poplar stem whips were collected from 46 provenances throughout the species range. The collection was done during November to March, 2005–2006 from the upper crowns of dormant trees aged 15 to 30 years. Fifteen trees were randomly sampled without phenotypic selection from each population. Care was taken to reduce the possibility of sampling clones more than once by collecting well separated individuals, at least 0.2 to 4 km apart. Branch cuttings were collected using a pole cutter, and stored at −4 °C in black plastic bags. For all the 690 sampled trees, Geographic Information System (GIS) coordinates and other site information were recorded (data provided upon request).

For the present study, we used a subset of 10 clones from each of 21 provenances that differ widely in respect to environmental conditions (Fig. 1 and Table 1), for a total of 210 clones. Stem cuttings of 6–9 cm length with a minimum of two buds were rooted in 965 mL plastic containers filled with a mixture of Sunshine-2 (Sun Gro Horticulture, Vancouver, Canada) growing mix (60%), peat (30%) and vermiculite (10%). The rooted cuttings (stecklings) were grown in a greenhouse beginning the second week of May under natural light supplemented by cool-white fluorescent lamps to provide a 21 h photoperiod and a minimum photosynthetic photon flux density (PPFD) of 400 µmol m−2 s−1 at plant level. Maximum day and night temperatures were maintained close to 25 and 18 °C, respectively, throughout the experimental period. The stecklings were kept well watered and fertilized using modified Hoagland's solutions (one-half strength micronutrients and one-fourth strength macronutrients) at a pH adjusted to 5.8–6.3 (Hoagland & Arnon 1950). The greenhouse was well ventilated and PPFD, humidity and air temperature were continuously monitored and recorded (Delta T Logger, Delta T Devices, Cambridge, UK). The potted stecklings were re-randomized weekly to minimize positional effects.

Figure 1.

Natural range of Populus balsamifera (shaded area) and provenances of 21 populations used in this study.

Table 1.  Geoclimatic data for 21 provenances of the AgCanBaP collection of Populus balsamifera
  1. LAT, latitude (°N); LON, longitude (°W); ELV, elevation (m); FFD, frost-free days (days); MAT, mean annual temperature (°C); AST, average summer temperature (°C); MTCM, mean temperature of the coldest month (°C); MTWM, mean temperature of the warmest month (°C); MAP, mean annual precipitation (mm); MSP, mean summer precipitation (mm); CONT, continentality; ADI, annual dryness index; SDI, summer dryness index.

Inuvik (INU)68.38°133.77°7107−8.88.1−27.614.224911541.81.281.42
Norman Wells (NWL)65.23°126.67°84133−5.511.8−26.517.029116143.51.411.22
Fairbanks (FBK)64.90°146.35°248125−2.912.4−23.216.926316740.11.891.16
Denali National Park (DEN)63.87°149.02°594122−3.211.8−22.316.123515938.42.081.16
Hay River (HAY)60.80°115.78°168144−1.111.5−23.115.932118339.01.781.00
Whitehorse (WHR)60.70°135.33°770138−0.211.2−18.414.828315733.22.141.08
Stony Rapids (STO)59.23°105.72°306153−0.713.0−20.416.945229337.31.290.66
Kuujjuaq (KUU)58.02°68.65°17115−5.77.0−24.311.552724335.80.770.56
Fort McMurray (FTM)56.92°111.50°3381570.713.3−18.816.845630835.61.430.63
Gillam (GIL)56.35°94.63°126129−4.210.4−25.815.349928941.10.910.61
Grande Prairie (GPR)54.75°118.63°7691671.913.1−15.015.944728330.91.580.64
Love (LOV)53.63°105.50°4191570.714.0−19.617.644030037.21.470.68
Stettler (STL)52.35°112.73°7951613.013.5−12.616.448134129.01.590.55
Mount Groulx (MGR)51.33°68.13°452152−4.08.8−23.313.279044336.50.580.35
Calgary (CGY)50.82°113.42°9151704.113.2−8.916.241330025.12.000.62
Sioux Lookout (SOU)50.08°91.90°3841681.614.6−18.618.671642437.20.970.51
Carnduff (CAR)49.18°101.83°5581733.815.9−14.819.552732134.31.530.71
Rouyn-Noranda (RNA)48.60°78.67°3101500.713.0−17.916.795051334.60.680.37
Matane (MTN)48.57°67.35°1191923.914.4−11.718.291541729.90.890.51
Roseville (ROS)47.33°64.37°251955.114.8−8.518.6110542227.10.810.51
Fredericton (FRE)46.40°67.25°1471955.615.8−9.519.3112446128.80.820.49

Standard gas exchange measurements

A portable infrared gas analyser (LC Pro+, Analytical Development Co. Ltd., Hoddesdon, UK) was used for measuring gas exchange variables. Beginning 65 d after rooting, measurements were taken twice on all 210 clones: once during the third week of July; and again during the first week of August. Measurements were made on two fully expanded mature leaves per steckling between 0900 and 1130 h. This timing was based on a preliminary study of diurnal patterns in 45 clones (data not shown) that indicated maximum rates of A during the morning hours, declining uniformly across all populations in the afternoon (with some recovery in the evening). The CO2 concentration of the inlet air was set to 370–380 µL L−1 to achieve an ambient air CO2 concentration (Ca) of ∼345 µL L−1 inside the cuvette. Other conditions were set to: 25 °C chamber temperature, vapour pressure deficit (VPD) at 1.9 kPa, and PPFD of 1000 µmol m−2 s−1 supplied by a mixed red/blue light-emitting diode (LED) unit mounted on the top of the cuvette. Net CO2 assimilation rate (A), intercellular carbon dioxide concentration (Ci), and stomatal conductance (gs) were calculated according to von Caemmerer & Farquhar (1981). WUEi was calculated as A/gs (Farquhar, O'Leary & Berry 1982). Chlorophyll content index (CCI) was determined following the gas exchange measurements with an Opti-Sciences (Hudson, NH, USA) CCM-200 meter. The CCM-200 uses calibrated LEDs and receptors to calculate the CCI, which is defined as the ratio of percent transmission at 655–940 nm through a leaf sample.

Plant sampling and analysis

Because net carbon isotope discrimination during photosynthesis is directly related to the diffusion gradient for CO2 from the bulk atmosphere to the sites of carboxylation, the carbon isotope ratio of plant tissue provides a proxy measure of water-use efficiency over the time when the carbon was fixed (Farquhar et al. 1982). 13C/12C ratios are expressed as del (δ) values in parts per million (‰) with respect to the Vienna Peedee Belemnite (VPDB) carbonate standard:


Leaf punches were taken immediately after gas exchange measurements for δ13C and nitrogen analysis; wood for δ13C was harvested at the end of experiment from the basal 10 cm of the shoots. Tissue samples were oven dried for 48 h at 70 °C to constant mass. After grinding to fine powder in liquid N2 with a mortar and pestle, 2–2.5 mg homogenized sub-samples were packed in tin capsules and sent to the University of California, Davis, Stable Isotope Facility. The samples were combusted in an online continuous flow dual analyser coupled to an isotope ratio mass spectrometer (Europa Scientific Integra, Cheshire, England, UK). The overall sample preparation and analysis error between repeated analyses of the same ground tissue was less than ±0.11‰. As the δ13C of air (δ13Cair) can vary, particularly under greenhouse and growth chamber conditions (Guy, Reid & Krouse 1986), we confirmed that the δ13Cair in the greenhouse was near ambient by simultaneously growing maize. The δ13C of the maize was −11.6‰±0.01, which corresponds to an approximate δ13Cair of −8.3‰ (Marino & McElroy 1991). Net discrimination against 13CO2 (Δ) was calculated from tissue (both leaf and wood) δ13C values according to Farquhar, Ehleringer & Hubick (1989):


Expected discrimination based on gas exchange data (Δi) was calculated following Evans & von Caemmerer (1996):


where a is the fractionation occurring because of diffusion in air (4.4‰) and b is the net fractionation by ribulose 1·5-bisphosphate carboxylase/oxygenase (Rubisco) and phosphoenolpyruvate carboxylase, which is considered by various authors to be between 27 and 30‰ (Warren 2006), but is here taken to be 29‰ for purposes of calculation.

Leaf nitrogen was calculated per unit leaf area (Leaf N, µmol N cm−2). The photosynthetic nitrogen-use efficiency (PNUE; µmol CO2 mol−1 N s−1) was calculated as the ratio of assimilation rate to Leaf N. SD (per mm−2) was determined on fully expanded leaves on each clone by making impressions with a thin layer of clear nail polish applied to one side of the midrib near the middle of the abaxial and adaxial leaf surfaces (Gornall & Guy 2007). Dried nail-polish impressions were stripped from the leaves, mounted on a slide and viewed under a microscope at 150× magnification. Three randomly selected fields of view (300 × 300 µm each) were counted per leaf impression and averaged. Height was recorded on five dates, and is here reported as the increment incurred over peak growth between days 57 and 70. Total leaf mass per unit area (LMA, mg cm−2) was determined at the end of the experiment.

ACi curve measurements

To gain better understanding of geographic trends in photosynthesis a subset of six populations was selected for ACi curve analysis. Two ramets for each of 10 clones (n = 120) from the populations Inuvik (INU), Denali National Park (DEN), Kuujjuaq (KUU) (northern), and Stettler (STL), Fredericton (FRE) and Matane (MTN) (southern) were raised in pots in a greenhouse under extended 21 h days as described earlier. ACi curves were constructed for all individuals with a LI-6400 gas exchange system (Li-Cor Instruments, Lincoln, NE, USA) equipped with a red LED unit. With the leaf chamber temperature set to 25 °C and the initial reference CO2 at 400 µL L−1, PPFD was ramped up to 1000 µmol m−2 s−1 over a period of 10 min, maintaining the VPD inside the leaf chamber close to 1.5 kPa. The reference CO2 was then changed in the following order: 400, 500, 630, 700, 800, 900, 1050, 1400, 1200, 970, 850, 750, 570, 450, 250, 100, 50, 200 and 370 µL L−1. These points were chosen to: (1) minimize the number of readings that would lead to a Ci between 200 and 300 µL L−1, which are avoided in the curve-fitting procedure, as there is a transition from the Rubisco-limited state (∼200 µL L−1) to the RuBP-regeneration-limited state (>300 µL L−1); and (2) ensure that there was no temporal drift in precision occurring in the LI-6400. A and gs were allowed to stabilize between readings. The A–Ci curve fitting model described by Sharkey et al. (2007) was used to estimate the maximum carboxylation rate allowed by Rubisco (Vcmax), the rate of photosynthetic electron transport (J), triose phosphate utilization (TPU), day respiration (Rd) and internal conductance (gm).

Climate data

Climate Normals (1971–2000) for nearby stations were obtained from Environment Canada (http://www.climate.weatheroffice.ec.gc.ca/climate_normals/index_e.html) and, for locations in Alaska, from the National Climatic Data Center (http://cdo.ncdc.noaa.gov/cgi-bin/climatenormals/climatenormals.pl), to provide FFDs (days), mean annual temperature (MAT; °C), average summer temperature (AST; °C), mean temperature of the coldest month (MTCM; °C), mean temperature of the warmest month (MTWM; °C), mean annual precipitation (MAP; mm) and mean summer precipitation (MSP; mm). The FFD is calculated based on number of days where minimum temperature was above zero, and is used here as a proxy for growing season length. Included in the analysis were indices of continentality (CONT), and annual (ADI) and summer (SDI) ‘dryness’ (Guy & Holowachuk 2001):


where precipitation is in millimetres, and es is the saturation vapour pressure in kilopascals at MAT and MTWM, respectively, calculated according to Buck (1981). Equations 5 and 6 are based on the knowledge that potential evapotranspiration is in large part determined by es, which is given by Yin (1998):


In Eqn 7, T is temperature and P is atmospheric pressure in kilopascals calculated from elevation (m) after Yin (1998):


Potential evapotranspiration over precipitation is the preferred measure of climate dryness, but the available data do not permit that calculation.

Statistical analysis

All statistical analyses used SAS version 9.1.3 (SAS Institute Inc. 2003). Pearson's correlations (r) were calculated to determine relationships between all variables on 210 genotypes. Canonical correlation analysis was done between physiological and environmental variables. For the A–Ci curve analysis, data were objectively screened by rejecting curves that did not yield a sum of squared deviations between observed and modelled points of <2. Screening caused the sample size to vary across populations; consequently, an unbalanced nested analysis of variance was performed within the general linear model with population (random) nested within geography (fixed).


Several growth and physiological traits in P. balsamifera are strongly related to geographic and climatic variables (Table 2), which by nature co-vary over North America for physiographic reasons. Treeline and the species range for P. balsamifera are at generally higher latitudes in the west (Fig. 1), resulting in similar relationships with longitude as with latitude. Geography is also confounded by elevation that tends to be greater for provenances from the west and southwestern part of the range. In this context, the preferred measure for growing season length is FFD. Measures of temperature (MAT, AST) and CONT are generally consistent with FFD. Indices of annual and summer drought tend to be highest in the northwest of the species range because precipitation decreases towards the north and west (Table 1).

Table 2.  Pearson correlations (r) between geographic, climatic and physiological variables for all 210 genotypes [those that are significant are set in bold (P < 0.05); bold * are significant after Bonferroni selection (P < 0.0003)]
  1. Height increment (cm); A, assimilation rate (µmol CO2 m−2 s−1); gs, stomatal conductance (mol H2O m−2 s−1); WUEi, intrinsic water use efficiency (µmol CO2 mmol−1 H2O); Leaf N, leaf nitrogen content (µmol N cm−2); PNUE, photosynthetic nitrogen-use efficiency (µmol CO2 mol−1 N s−1); δ13C, carbon isotope composition for leaf and wood (‰); SD, stomatal density (per mm−2); CCI, chlorophyll content index; LMA, leaf mass area (mg cm−2); LAT, latitude (°N); LON, longitude (°W); ELV, elevation (m); FFD, frost-free days (days); MAT, mean annual temperature (°C); AST, average summer temperature (°C); CONT, continentality; ADI, annual dryness index; SDI, summer dryness index.

Height increment0.622*0.557*−0.083−0.501*−0.486*−0.254*0.547*0.309*0.560*
Leaf N0.465*0.203−0.210−0.572*−0.524*−0.558*0.308*0.0490.355*

During rapid growth in the greenhouse, plant height increment was positively correlated with both latitude and longitude of origin, along with CONT, ADI and SDI (Table 2). Similarly, height increment was negatively correlated with FFD and measures of temperature. The population KUU deviated from this pattern, for reasons unknown (Fig. 2a). Final height (not shown) had a similar pattern. Irrespective of their height increment, INU, KUU and DEN had the highest rates of photosynthesis (Fig. 2b). Variation in A was highly consistent with variation in LMA (Fig. 2c) and strongly correlated with Leaf N (r = 0.744, P = 0.0001). Indeed, there was no relationship between A and FFD if photosynthesis was expressed relative to leaf mass (not shown). Across all 210 genotypes, light-saturated photosynthetic assimilation (A) was significantly correlated with FFD and several other co-varying climate parameters (Table 2). PNUE showed opposite correlations with geographic and climatic variables relative to A, LMA, Leaf N, WUEi and δ13C. Among the physiological variables considered (Table 3), A was positively correlated with gs, Leaf N and CCI, but was inversely correlated with SD.

Figure 2.

Mean values (±standard deviation) for (a) height increment over 18 d during peak growth (b) assimilation rate (A) and (c) leaf mass per unit area (LMA) during free growth across 21 populations of Populus balsamifera. Provenances are arranged from left to right according to increasing frost-free days. INU, Inuvik; KUU, Kuujjuaq; DEN, Denali National Park; FBK, Fairbanks; GIL, Gillam; NWL, Norman Wells; WHR, Whitehorse; HAY, Hay River; RNA, Rouyn-Noranda; MGR, Mount Groulx; STO, Stony Rapids; FTM, Fort McMurray; LOV, Love; STL, Stettler; GPR, Grande Prairie; SOU, Sioux Lookout; CGY, Calgary; CAR, Carnduff; MTN, Matane; ROS, Roseville; FRE, Fredericton.

Table 3.  Pearson correlations (r) among physiological variables for all 210 genotypes. [those that are significant are set in bold (P < 0.05); bold * are significant after Bonferroni selection (P < 0.001)]
 AgsWUEiLeaf NPNUEδ13Cleafδ13CwoodSDCCILMA
  1. Height increment (cm); A, assimilation rate (µmol CO2 m−2 s−1); gs, stomatal conductance (mol H2O m−2 s−1); WUEi, intrinsic water use efficiency (µmol CO2 mmol−1 H2O); Leaf N, leaf nitrogen content (µmol N cm−2); PNUE, photosynthetic nitrogen-use efficiency (µmol CO2 mol−1 N s−1); δ13C, carbon isotope composition for leaf and wood (‰); SD, stomatal density (per mm−2); CCI, chlorophyll content index; LMA, leaf mass area (mg cm−2).

Height increment0.019−0.1060.1050.107−0.1990.439*0.452*−0.0570.1690.427*
gs 1−0.638*0.0550.059−0.277*−0.401*0.1380.023−0.233*
WUEi  10.358*−0.1250.226*0.300*−0.2080.267*0.292*
Leaf N   1−0.831*−0.087−0.050−0.0960.475*0.488*
PNUE    10.040−0.063−0.008−0.354*−0.464*
δ13Cleaf     10.656*−0.1650.1530.204
δ13Cwood      1−0.0890.0850.242*
SD       1−0.063−0.134
CCI        10.268*

Even though SD was higher at lower latitudes, gs was not clearly related to any geographic or climatic variables (Table 2). In contrast, correlations between intrinsic water-use efficiency (WUEi) and the geographic and climatic variables closely paralleled A alone (Table 2), even though variation in both A and gs contributed strongly to WUEi (Table 3). Across all 210 genotypes, A and gs were also positively correlated with each other (r = 0.236, P < 0.001). Trends were different at the population level (not shown) in that gs was in this case not significantly correlated with WUEi, whereas A was (r = 0.831, P < 0.0007). Consequently, population differences in WUEi were largely determined by A. Population means for WUEi ranged from 28.6 to 39.0 µmol CO2 mmol−1 H2O (Fig. 3). Populations from the extreme northwest had higher WUEi than those from the southeast. For example, populations DEN and INU had greater WUEi (39.0 and 36.2 µmol CO2 mmol−1 H2O, respectively) than FRE (29.1 µmol CO2 mmol−1 H2O).

Figure 3.

Mean values (±standard deviation) for intrinsic water use efficiency (WUEi) during free growth across 21 populations of Populus balsamifera. Provenances are arranged from left to right according to increasing frost-free days. INU, Inuvik; KUU, Kuujjuaq; DEN, Denali National Park; FBK, Fairbanks; GIL, Gillam; NWL, Norman Wells; WHR, Whitehorse; HAY, Hay River; RNA, Rouyn-Noranda; MGR, Mount Groulx; STO, Stony Rapids; FTM, Fort McMurray; LOV, Love; STL, Stettler; GPR, Grande Prairie; SOU, Sioux Lookout; CGY, Calgary; CAR, Carnduff; MTN, Matane; ROS, Roseville; FRE, Fredericton.

There was variation in δ13C among population means (∼0.98‰ foliar, and ∼1.1‰ wood). Wood and foliar δ13C values were correlated with FFD, LAT and LON but not with summer temperature (Table 2). Wood and foliar δ13C were, as expected, strongly inter-correlated across all 210 genotypes (Table 3) and across populations (r = 0.799, P < 0.0007, not shown). In contrast, δ13C values were correlated with WUEi across all genotypes (Table 3), but not across populations (r = 0.185, not significant).

A canonical structure for the provenances under study was obtained using nine geoclimatic and seven ecophysiological traits, including height increment (Table 4). Two significant canonical variables were extracted (CLIM1 and CLIM2) that explained 17% and 26% of the variance in plant traits, respectively, during redundancy analysis. All climate predictor variables loaded highly on CLIM1 along with latitude and longitude, whereas measures of summer temperature (AST and CONT) were most strongly related to CLIM2. The strongest canonical loadings were seen for height increment and leaf N on CLIM1 and for δ13Cwood on CLIM2. PNUE had negative loading on CLIM1 whereas height increment, δ13Cleaf and δ13Cwood, were positively related to both CLIM1 and CLIM2. In contrast, although Leaf N, A, and WUEi were also positively related to CLIM1, they were negatively related to CLIM2. These tendencies indicate that trees from areas with short growing seasons and/or low summer temperatures had higher photosynthetic rates driven in part by elevated Leaf N. Figure 4 presents A as a function of FFD and Leaf N for all 210 genotypes. As shown by this figure, variation in Leaf N can account for considerable variation in A, both within and between populations, but FFD remains a significant predictor even when Leaf N is accounted for.

Table 4.  Canonical structure between geoclimatic parameters and plant traits with their first two canonical variables, CLIM1 and CLIM2
Geoclimatic variablesCLIM 1CLIM 2Response variablesCLIM 1CLIM 2
  1. LAT, latitude (°N); LON, longitude (°W); ELV, elevation (m); FFD , frost free days (days); MAT, mean annual temperature (°C); AST, average summer temperature; CONT, continentality; ADI, annual dryness index; SDI, summer dryness index; A, assimilation rate (µmol CO2 m−2 s−1); WUEi, intrinsic water use efficiency (µmol CO2 mmol−1 H2O); δ13C, carbon isotope composition for leaf and wood (‰); Leaf N, leaf nitrogen content (µmol N cm−2); PNUE, photosynthetic nitrogen-use efficiency (µmol CO2 mol−1 N s−1); height increment (cm).

MAT−0.860.11Leaf N0.55−0.32
CONT0.720.35Height increment0.590.31
Figure 4.

Assimilation rate (A) as a function of Leaf N and frost-free days (FFDs) across all 210 genotypes of Populus balsamifera. The dependent variable can be predicted from a linear combination of the independent variables using the equation A = 15.865 + (0.019 × Leaf N) − (0.019 × FFD).

The relationship between WUE as determined by gas exchange and WUE as indicated by δ13C is related to the conductance of CO2 diffusion within the leaf (Evans & von Caemmerer 1996). Carbon isotope discrimination during photosynthesis calculated from both gas exchange data (Δi) and from δ13C values of leaf (Δleaf) and wood (Δwood) tissue are plotted together in Fig. 5. Wood is typically more enriched in 13C than leaf tissue, hence Δleaf and Δwood are not the same but parallel each other closely. They differ in elevation by roughly 0.5‰. In contrast, Δi is ∼2‰ greater than Δwood, but this difference depends largely on the value of b used in Eqn 3. Some offset is expected because of the diffusion gradient for CO2 from the intercellular space to the sites of carboxylation in the chloroplast, inversely proportional to the internal transfer conductance (gm). Regardless of the value chosen for b, there is a significant increase in Δi as a function of FFD, but not in Δwood and Δleaf, the latter determined on the exact same leaves as used in the gas exchange analysis. Put another way, the discrepancy between predicted and observed isotope discrimination decreases with latitude. The implication is that gm is increased in populations adapted to shorter growing seasons.

Figure 5.

Carbon isotope discrimination calculated from δ13Cwood (□), δ13Cleaf (Δ) and intrinsic water use efficiency (○) across 21 populations of Populus balsamifera plotted against frost-free days (FFD).

ACi curves were constructed for six of the populations shown in Fig. 5; three approaching the northern edge of the species range (≤122 FFD) and three from the south (≥161 FFD). Consistent with the previous gas exchange measurements, photosynthesis was higher in genotypes of northern provenance than in genotypes of southern provenance, a difference that was reflected in somewhat higher J and TPU, and possibly Vcmax and gs (Table 5). Day respiration did not differ. The most obvious contrast was a greater than twofold difference in gm, whereby increased gm was again associated with fewer FFD. Population differences in gm (nested within geography) were also nearly significant (P = 0.0521, not shown).

Table 5.  Fitted ACi curve parameters (±standard error) estimated at 27°C on populations representative of North (INU, DEN, KUU) and South (STL, FRE, MTN) geography. Values reported for A and gs are at ambient CO2 (380 µL L−1). P is the probability of a difference between North and South
  1. A, assimilation rate (µmol CO2 m−2 s−1); gs, stomatal conductance (mol H2O m−2 s−1); Vcmax, maximum carboxylation rate allowed by Rubisco; J, rate of photosynthetic electron transport (based on NADPH requirement); TPU, triose phosphate use; Rd, day respiration (µmol m−2 s−1); gm, internal conductance (mol CO2 m−2 s−1); INU, Inuvik; DEN, Denali National Park; KUU, Kuujjuaq; STL, Stettler; FRE, Fredericton; MTN, Matane.

North14.89 ± 0.820.266 ± 0.0187.95 ± 6.26115.44 ± 6.978.12 ± 0.532.50 ± 0.190.447 ± 0.124
South10.84 ± 0.320.218 ± 0.0176.89 ± 4.3093.20 ± 1.886.56 ± 0.112.83 ± 0.190.173 ± 0.030


Height growth and photosynthesis

As commonly observed in many tree species when planted into outdoor common gardens [e.g. Acer saccharum Marsh. (Kriebel 1957), Salix pentandra L. (Junttila & Kaurin 1985) and Picea abies (L.) Karst. (Junttila & Skaret 1990)], balsam poplar genotypes originating from high latitude achieve much less height growth than those from low latitude (R. Y. Soolanayakanahally unpublished results). Similar trends in plant size with latitude are also commonly observed in herbaceous plants when grown outdoors [e.g. Carex aquatilis Wanlenb. ssp. aquatilis (Chapin & Chapin 1981), Verbascum thapsus L. (Reinartz 1984)] and in Arabidopsis in a greenhouse under short (12 h) days (Li, Suzuki & Hara 1998). In balsam poplar during free growth under long days, however, plant height increment was positively correlated with latitude (r = 0.622, P < 0.0003, Table 2). Similarly, when growing 24 populations of Scots pine (Pinus sylvestris L.) under conditions mimicking changes in photoperiod at 50 and 60°N, Oleksyn, Tjoelker & Reich (1992) found that secondary needle length, plant height and dry mass at harvest were greater under the longer photoperiod. Height growth rate was lowest among southern provenances regardless of photoperiod.

Although trees representative of northern populations generally do not grow as much as those from the south over any given summer, they can in fact possess higher photosynthetic rates, even in outdoor common gardens. Despite a decline in plant height, Norway spruce from colder habitats had higher photosynthetic rates associated with higher leaf nitrogen concentrations (Oleksyn et al. 1998). Likewise, latitudinal patterns in growth rate and photosynthesis are inversely related in paper birch (Benowicz et al. 2000b, 2001), Sitka alder (Benowicz et al. 2001, 2000a), black cottonwood (Gornall & Guy 2007) and Eurasian aspen (Populus tremula L.; R. Y. Soolanayakanahally unpublished results). In these studies, photosynthetic rate was always measured at the height of the growing season, well before buds had set. In the present study, bud set was avoided altogether by maintaining all clones under a uniform extended photoperiod. Higher rates of photosynthesis were then directly reflected by enhanced height increment. Growth, therefore, does not appear to be limited by photosynthesis in high latitude genotypes, but rather by the length of the realized growing season – that portion of the available growing season when extension growth is active. Gornall & Guy (2007) proposed that northern provenances may have inherently higher photosynthesis to compensate for shorter realized growing seasons.

Intrinsic growth capacity varies inversely with realized growth across environmental gradients in many animal species, a phenomenon commonly called ‘counter-gradient variation’ (Levins 1969). For example, in the fish species Menidia menidia L., high latitude genotypes grow more rapidly than low latitude genotypes within the brief period of the year when temperatures are favourable (Conover & Present 1990). Growth occurs within similar thermal boundaries independent of latitude, but it is the length of growing season that changes greatly with latitude. We find the same phenomenon in balsam poplar.

Canonical correlation analysis, an efficient approach for dealing with large inter-correlated datasets, revealed how much of the variation in key traits may be attributed to geography and climate (Table 4). The two significant canonical variables we extracted describe somewhat different geoclimatic terms; CLIM1 may be seen as an index of overall ‘borealness’ whereas CLIM2 may account for further differences in ‘growing season heat’. Latitude and frost free days loaded very highly on CLIM1, explaining up to 94% of its variance. Redundancy analysis indicated that CLIM1 and CLIM2 together accounted for 43% of the variance in plant traits. Wright et al. (2004) examined leaf economics across 2548 species worldwide and concluded that climatic variables explain only a small portion of the overall variation, but this of course included very different co-occurring species. Dang et al. (1994) studied geographic variation in ecophysiological traits among 40 populations of red alder (Alnus rubra Bong.) in British Columbia. They found a weakly positive correlation between A and latitude, but the latter co-varied with longitude and was confounded by elevation. In that study, however, canonical correlation analysis also established a clear relationship between photosynthetic capacity and geography. The consistency of this pattern, whereby A increases with latitude of origin in at least seven tree species from three families, is strong evidence for its adaptive importance.

Differences in Ci that result from variation in gs give rise to an inverse relationship between WUEi and PNUE (Field, Merino & Mooney 1983). In contrast, variation in gm should cause WUEi and PNUE to respond in parallel, in the same direction as gm. Consistent with the possibility of a trade-off, the efficiency of water-use decreased with FFD whereas PNUE increased (Table 2), and there was a similar reverse trend on CLIM1 (Table 4). There was, however, no direct relationship between PNUE and any of our measures of WUE (Table 3), as might be expected if gm, gs and other factors influencing A, vary independently.

We found strong positive correlations between LMA, Leaf N and A (Table 3). Positive correlations between foliar N and rates of photosynthesis have been observed across many species. Reich & Oleksyn (2004) reported that, globally, Leaf N tends to vary inversely with MAT. A large proportion of foliar N is found in proteins involved in photosynthesis, not least of which is Rubisco. Higher A in balsam poplar populations originating from high latitudes or low FFD (Fig. 4) can therefore be ascribed in part to enhanced photosynthetic capacity per unit area. Consistent with this interpretation, the ACi curve analysis revealed that representative northern populations had significantly higher rates of photosynthetic electron transport (J) and TPU than southern populations (Table 5). On the other hand, Vcmax did not differ significantly, though it may have been somewhat higher in the northern genotypes. Increased photosynthetic demand was supported by an increased supply of CO2 through both an increase in gs and, most particularly, an increase in gm. Although there was no significant relationship between gs and FFD, or indeed any of the geographic or climatic parameters listed in Table 2, we argue in the following discussion that enhanced gm contributes to higher A in northern genotypes of balsam poplar.

δ13C and internal conductance

Because of a mutual dependence on the diffusion gradient for CO2 into leaves, plant δ13C is widely used as a relative index of WUEi. Foliar δ13C integrates WUE during leaf formation and recent photosynthetic activity, but can be biased towards lighter values because of the foliar lipid content, which is depleted in 13C relative to carbohydrate exported in the phloem (Hobbie & Werner 2004). Wood δ13C should provide a better indicator of WUE than δ13Cleaf not only because it is closer in isotopic composition to the whole plant but also because it integrates over the crown as well as the full interval of growth. Discrimination (Δ) can be calculated from either δ13Cleaf or δ13Cwood, but the true value for Δ during CO2 assimilation probably resides between the two. The δ13C of both leaf and wood were well correlated with height increment and WUEi (Table 3). There was a tendency towards higher δ13C values (reflecting higher WUE) towards the north, but not as strong as might be expected based on WUEi (Fig. 5).

On-line isotope discrimination (Δobs), in comparison with gas exchange data, is often used to estimate gm. Although such estimates are very sensitive to the value chosen for b in Eqn 3, differences in Δi versus Δobs between different genotypes or treatment may be ascribed to differences in gm. Likewise, because trends in Δleaf and Δwood should reflect trends in Δobs, it is possible to compare these patterns with Δi calculated from the gas exchange data to explore putative patterns in gm. Because δ13Cleaf and δ13Cwood were highly correlated and trends in Δleaf and Δwood were exactly parallel, we can ignore the possibility of variation in post-photosynthetic isotope fractionation. Figure 5 shows that Δi increases with FFD but that Δleaf and Δwood do not. This disparity indicates a tendency towards higher gm at low FFD, resulting in increased A and an enhancement of WUE not captured by Δ. Using three tree species, Warren & Adams (2006) demonstrated that Δ may vary by up to 3‰ at a common WUE. The corollary must also be true, as seen here in balsam poplar.

We do not at this point know what accounts for the observed differences in gm. One possibility is that low latitude populations may invest in tougher leaves with thicker cell walls or other structural components that might increase path lengths or otherwise limit diffusion of CO2 into chloroplasts (Warren 2008). This is in keeping with the observation that species with longer leaf longevity allocate more resources to structural components, as reflected by LMA, to provide physical resistance to abiotic and biotic stress factors (Wright, Westoby & Reich 2002; Wright et al. 2004). In contrast, however, within balsam poplar we find a strong tendency towards higher LMA in populations from high latitudes, where leaf longevity is comparatively brief. Other possibilities may be aquaporins involved in CO2 transfer across cellular membranes (Terashima & Ono 2002; Flexas et al. 2006) or differences in cytoplasmic streaming and the positioning of chloroplasts within protoplasts (Tholen et al. 2008).

Trait convergence in Populus

Balsam poplar and black cottonwood are very similar in many respects and hybridize extensively where their ranges overlap (Farrar 1995). It was surprising to us, therefore, that although they both show increased A and Leaf N with latitude of origin, they achieve an increase in leaf conductance in different ways. In black cottonwood, gs and SD co-varied with A and Leaf N, resulting in no apparent population differences in A/gs or δ13C (Gornall & Guy 2007). Higher SD in high latitude genotypes was fully accounted for by the appearance of stomata on the upper leaf surface. In contrast, our balsam poplar accessions are almost entirely hypostomatous regardless of origin, and their gs was not related to any of the geoclimatic variables (Table 2). Consequently, in trees from regions with short growing seasons, water-use efficiency was enhanced in one species (balsam poplar) but not in the other (black cottonwood). Gornall & Guy (2007) did not assess gm in black cottonwood, but the absence of any change in gs in balsam poplar appears to be partially offset by enhanced gm, further contributing to a higher WUE.


We conclude that climate-related patterns in photosynthesis are not artefacts of photoperiodic response and differential growth in common garden environments. Rather, these patterns represent true adaptive variation in response to growing season length. Commensurate with higher A, and contrary to the typical view that trees adapted to short growing seasons will exhibit lower growth rates, high latitude genotypes of balsam poplar are in fact capable of equal or greater growth than their low latitude counterparts when photoperiodic restrictions are removed. The proximal causes for population-level variation in A in balsam poplar appear to be similar to those previously reported for black cottonwood, at least in terms of investments in Leaf N and resulting photosynthetic capacity. In balsam poplar much of the increase in Leaf N can be attributed to higher LMA, whereas in black cottonwood LMA did not vary. Another difference is the means by which they supply their increased photosynthetic capacity with an enhanced CO2 flux, in one case through gs (black cottonwood) and in the other through gm (balsam poplar); two convergent but not equivalent solutions to the same problem.


The authors would like to thank Agroforestry Division, Agriculture Agri-Food Canada (AAFC), Indian Head, Saskatchewan, Canada for providing access to the AgCanBaP collection and financial support to R. Y. Soolanayakanahally under the Student Research Affiliate Program. This work was supported by grants from AAFC to S. Silim and W. B. Schroeder, and by a Natural Sciences and Engineering Research Council (Canada) Discovery Grant to R. D. Guy. Special thanks to Don Reynard, Senior Technician, AAFC for help with instrumentation.