Conservative leaf economic traits correlate with fast growth of genotypes of a foundation riparian species near the thermal maximum extent of its geographic range


Correspondence author. E-mail:


  1. Plant functional traits involved in carbon and water acquisition are likely to be adaptive across the range of a species if the availability of these resources varies across this range and are limiting to growth or fitness. At the interspecific level, leaf economic traits associated with rapid resource capture are correlated with fast growth rates. However, relationships between leaf traits and growth are poorly understood at the intraspecific level.
  2. We examined two hypotheses: (i) leaf traits vary genotypically among Populus fremontii populations from different thermal environments; and (ii) leaf traits are related to growth rate of these P. fremontii populations. We used a common garden at the warm edge of P. fremontii distribution that included individuals transplanted from 11 provenances. Provenances varied in mean annual maximum temperature by 5·9 °C, reflecting a range of expected increases in temperature over the next 80 years.
  3. Conservative leaf traits (e.g. low specific leaf area, N content, stomatal conductance, net photosynthetic rate and high leaf water-use efficiency) were positively related to growth rates of genotypes and populations, a pattern opposite of that widely reported among species in other studies.
  4. Provenance temperature explained 75% of the variation in multivariate leaf traits with the warmest provenances having the most conservative traits and highest growth rates. Clinal genetic variation suggests that P. fremontii may be adapted to thermal environments.
  5. Leaf area-to-sapwood area ratio was positively associated with growth rate, while leaf area-based net photosynthetic rate was negatively associated with growth rate; these results suggest that hydraulic architecture was more important than leaf-level photosynthetic rate in determining growth rate.
  6. Synthesis. Our results suggest that conservative leaf traits promote rapid growth of P. fremontii genotypes in extremely hot environments. Thus, relationships between leaf economic traits among species do not necessarily apply to the range of variation among genotypes within species. The generality of this pattern should be examined for other species that will be exposed to climate warming. Moreover, our research shows that common garden provenance trials are useful for identifying genotypes best suited to a predicted warmer climate and for improving understanding of the physiological basis for adaptation to warm environments.


Strong correlations among plant traits across a diverse array of plant species have been shown repeatedly (Reich, Walters & Ellsworth 1997; Reich & Oleksyn 2004; Wright et al. 2004). Species from resource-rich environments tend to have fast growth and leaf traits suited for fast or ‘acquisitive’ resource capture [high specific leaf area (SLA), mass-based leaf nitrogen concentration and net photosynthetic rate], whereas species from resource-poor environments are characterized by slow growth and ‘conservative’ resource capture traits (low SLA, leaf nitrogen concentration and net photosynthetic rate; Chapin 1980; Reich, Walters & Ellsworth 1992; Poorter & Bongers 2006). While variation in the this so-called leaf economics spectrum (sensu Wright et al. 2004) across species and environments suggests the presence of a broad evolutionary filter controlled chiefly by macroclimate, finer-scale adaptations may occur on more local scales, such as within an individual species' geographic range. At such local scales, the relationship between leaf traits and environmental variation likely is more complex because of adaptation to multiple interacting factors (Chapin et al. 1987; James, Tiller & Richards 2005). Interestingly, local-scale intraspecific genotypic variation in inter-correlated plant traits is poorly understood (although see Burns & Beaumont 2009; Bilton et al. 2010; Seiwa & Kikuzawa 2010) despite its potential to enhance understanding of the evolutionary importance of plant functional traits.

A useful approach for examining the relationship between leaf traits and growth rates as well as assessing the potential for plant traits to be under selection is to use a common garden provenance trial (Rehfeldt et al. 1999; St Clair & Howe 2007). This approach includes transferring multiple genotypes from different provenances to a common garden. Genotypes that are ‘superior’ performers in terms of higher growth rates, survival rates or reproductive success may be locally adapted to the common garden environment (Mimura & Aitken 2007; Savolainen, Pyhäjärvi & Knürr 2007; O'Neill, Hamann & Wang 2008). The plant traits characteristic of these superior genotypes may be under selection and underlie physiological adaptations to the common garden environment (Howe et al. 2003). Evidence supporting selection of a plant trait is provided by demonstrating genetic variation/heritability for that trait (Lynch & Walsh 1998).

Species adapted to high stress environments typically have conservative leaf traits (Chapin et al. 1987; Westoby 1998; Reich & Oleksyn 2004) and inherently slow growth rates (Poorter & Remkes 1990; Poorter & Bongers 2006). However, whether this interspecific strong relationship between conservative leaf traits and slow growth also commonly occurs among genotypes at the intraspecific level is unknown. To address this question, we tested two hypotheses: (i) leaf economic traits vary genotypically among Populus fremontii (Fremont cottonwood) populations from different thermal environments; and (ii) leaf traits are related to growth rate across these same P. fremontii populations in an extremely hot environment.

We studied the relationship between a suite of leaf economic traits and growth of 67 genotypes from 11 provenances (i.e. source populations) of P. fremontii at an extremely warm [31·0 °C mean annual air maximum temperature (MAMT)] common garden in south-eastern California. Using a test site at the warm edge of a species geographic distribution allowed us to simultaneously address the potential of trait variation to signify adaptation to a hot environment and to identify which genotypes were likely to be best suited to a predicted warmer climate. In our study, the range of provenance MAMT was 5·9 °C, which is commensurate with predicted increases in temperature over the next 80 years in the south-western United States (Karl, Melillo & Peterson 2009).

We selected P. fremontii as a study organism because it is a target species for restoration in riparian ecosystems of the south-western United States (Follstad Shah et al. 2007) and because prior research suggests that P. fremontii populations vary in adaptation to temperature (Grady et al. 2011). Populus fremontii also is considered a foundation tree species because its presence drives ecological diversity and modulates ecosystem processes within riparian habitats (Whitham et al. 2006).

Materials and methods

Site Description

An experimental common garden was established in March 2007 at the Palo Verde Ecological Reserve (PVER; Fig. 1), located on the historic floodplain (prior to dam construction) of the Lower Colorado River (c. 5 km from Blythe, California and 0·25 km from the Lower Colorado River; N 33·71391, −114·49600, elevation 87 m). Prior to garden establishment, this floodplain was levelled and converted to agricultural use, formerly supporting cotton (Gossypium hirsutum) and alfalfa (Medicago sativa). The PVER garden is a component of the Lower Colorado River Multi-Species Conservation Program (LCR MSCP), a $626-million ecosystem restoration project fund managed by the Bureau of Reclamation (Follstad Shah et al. 2007; LCR MSCP 2010) and covering c. 1030 km of river length. Our common garden design consisted of propagated plantings from 16 provenances of P. fremontii and Salix gooddingii and seven provenances of S. exigua. Individual cuttings were grown in pots (6 cm diameter × 25 cm depth) and then transplanted to 66 blocks (16 × 16 rows at 2 m spacing per block) within the common garden. All cuttings were collected randomly from within 10 m of the active flood channel of perennially flowing rivers in Arizona and California from the Basin and Range Hydrogeologic Province. In this study, we present data and results for P. fremontii only. For more details regarding garden establishment, see Grady et al. (2011).

Figure 1.

Locations of provenance collection sites for Populus fremontii and the Palo Verde Ecological Reserve (PVER) common garden. Also shown are the mean annual maximum air temperatures (MAMT; averaged between 1950 and 2000; WorldClim (2011)) across the range of P. fremontii (Little, 1971) in the south-western United States and northern Mexico.

The garden was flood irrigated with c. 300 L m−2 (c. 1 acre-foot) of reclaimed water from the city of Blythe every 2–3 weeks in May through September, and once every month from October through April. Soil volumetric water content at 10 cm soil depth at PVER throughout the growing seasons in 2009 and 2010 (March to December) ranged from 0·04 to 0·32 m3 m−3 (mean of 0·24 m3 m−3; range of 0·25–0·32 during leaf gas-exchange measurement periods) as determined by continuous (10-min intervals) data logging (Onset Computer Corporation, Bourne, MA, USA) using a ECH2O Dielectric Aquameter probe (Decagon Devices, Inc., Pullman, Washington, USA) permanently installed at the approximate centre of the garden. In May 2009, depth to groundwater was c. 3·3 m based on presence of free water at this depth in a hole after auguring. Soils at the garden were composed chiefly of the United States Department of Agriculture (USDA) Soil Taxonomic Family of coarse-loamy, mixed, superactive, calcareous, hyperthermic Typic Torrifluvents (USDA NRCS 2010).

Provenance Characteristics and Climate

For this study, we focused on 11 randomly selected provenances of P. fremontii that varied in provenance mean annual maximum air temperature (MAMT) transfer distance (MAMT of the common garden minus MAMT of provenance origin), as this was previously shown to be strongly correlated with an index of above-ground net primary productivity (ANPP; Grady et al. 2011). The annual mean of maximum daily temperature (MAMT) ranged from 24·4 to 30·9 °C across provenances (Table 1). This range represents c. 50% of the total MAMT temperature range of P. fremontii based on climatic data for the western United States (PRISM Climate Group 2010) and Mexico (WorldClim 2011). Provenances used in our study represent the warmest edge of the geographical distributional range of P. fremontii (Fig. 1; MAMT), as well as the hottest recorded temperatures (>50 °C) at which this species occurs (Table 1). The range of mean annual temperature (MAT) transfer distances across provenances (0–5·9 °C for P. fremontii; Table 1) encompasses the range of predicted temperature increases via climate change over the next century in the south-western United States (Karl, Melillo & Peterson 2009). For instance, according to the International Panel on Climate Change (IPCC), a low or high carbon emission scenario is predicted to yield an MAT increase over the historic baseline (mean MAT from 1960 to 1979) of 1·7–3·6 °C or 3·9–5·6 °C, respectively, by 2090 (Karl, Melillo & Peterson 2009).

Table 1. Characteristics for 11 provenances of Populus fremontii located throughout Arizona and California and grown at the Palo Verde Ecological Reserve (PVER, indicated with *) common garden. Characteristics include number of clonal genotypes per provenance (#Geno); elevation; transfer distance in mean annual maximum air temperature (MAMT of the garden minus MAMT of the provenance); MAMT; mean annual air temperature (MAT); mean annual precipitation (MAP); record high temperature recorded since weather station became operational (c. 20th century record) and USDA Soil Taxonomic subgroup (USDA NRCS 2010). For all characteristics that include a mean value, Western Regional Climate Data (WRCC 2010) were averaged between 1971 and 2000
Provenance#GenoElevation (m)MAMT transfer (°C)MAMT (°C)MAT (°C)MAP (cm)Record high (°C)Soil subgroup
Palo Verde*487030·9229·450Typic Torripsamments
Cibola1070030·9229·450Typic Torrifluvents
Bill Williams53570·630·321·522·951·1Typic Torripsamments
Hassayampa45951·829·218·83149·4Typic Torriorthents
Verde96203·127·819·639·647·8Aquic Ustifluvents
Virgin65703·427·518·819·848·9Typic Torrifluvents
Salt56603·727·219·242·946·7Aquic Ustifluvents
Gila III610484·126·918·634·346·7Typic Torrifluvents
Sonoita412344·726·316·145·543·3Typic Torrifluvents
San Pedro812685·225·716·54643·9Aquic Torrifluvents
Agua Fria69636·524·416·341·446·1Typic Torrifluvents

Provenance origins ranged in elevation from 49 to 1268 m above sea level, in MAMT from 24·4 to 30·9 °C, in MAT from 16·1 to 22·0 °C, and in mean annual precipitation (MAP) from 9·4 to 46·0 cm (Table 1). Provenance climatic data were compiled for years 1971–2000 from weather stations affiliated with the Western Regional Climate Center (WRCC 2010), located within a 10-km radius and 80 m of elevation from each provenance. Two provenance origins (Verde and San Pedro) did not have weather stations that fit these criteria, so we used modelled estimates of temperature and precipitation from PRISM data (PRISM Climate Group 2010). Soils at provenance origins were similar to those at PVER; these soils were also members of the Typic Torrifluvents Soil Taxonomic subgroup or closely related subgroups and were dominated by coarse textural classes and derived primarily from calcareous parent materials (Table 1; USDA NRCS 2010).

Leaf Traits

We measured SLA, internal leaf carbon dioxide concentration (Ci), stomatal conductance (Gs), net photosynthetic rate (Pn), transpiration rate (E) and leaf nitrogen (N), carbon (C) and chlorophyll (Chl) on both a leaf areal and mass basis. We then used these measurements to calculate leaf C-to-N mass ratio (C/N), leaf-level intrinsic water-use efficiency (iWUE = Pn/Gs) and leaf-level nitrogen-use efficiency (NUE = Pn/N). We used iWUE rather than Pn/E to minimize the influence of variation in vapour pressure deficit during measurements and to provide a surrogate of the trade-off between C uptake and water balance. Measurements were taken on between 3 and 10 ramets from 4 to 10 genotypes from each of 11 provenances (Table 1) in May 2009 and on the same plants in four provenances in May 2010 when plants were 2·2 and 3·2 years old, respectively. In 2010, we remeasured all variables listed previously on the two provenances with the highest and the two provenances with the lowest mean population Pn in 2009 to evaluate interannual stability in gas-exchange measurements among provenances. Furthermore, all trees flowered in 2010 but not in 2009 thus results of our study may be pertinent to both juvenile and mature trees. The number of genets and the number of ramets per genet that were collected at provenance sources varied among provenances due to differences in the availability of plant materials at each site. Due to land-use alteration, hydrological modifications and exotic species encroachment, many provenances contained only a few trees resulting in some imbalances in replication at the garden. Genotypic identification was initially based on spatial discreteness of genets and a minimum distance of 10 m between individuals and later confirmed by genetic-based microsatellite technology analyses (Bangert et al. 2012).

Leaf gas-exchange was measured using a Li-Cor 6400 infrared gas analyzer (IRGA; Li-Cor, Inc., Lincoln, NE, USA) on fully expanded leaves in the upper canopy of trees exposed to full sunlight. Measurements were taken between 09:00 and 13:00 h when environmental conditions were conducive to the highest rates, as determined by preliminary diel measurements (data not shown). We measured gas exchange on one randomly selected leaf from each of three separate branches exposed to full sunlight per tree. The branches were either attached or detached from trees using pole pruners when leaves were inaccessible from the ground (i.e. the majority of trees were taller than 3 m in height by the time they were 2 years old). Leaves on detached branches were placed in a leaf cuvette within 30 s of detachment to minimize effects of desiccation on gas exchange. Gas-exchange values were recorded three times at 10-s intervals once rates stabilized, and the mean of these three measurements was used for statistical analyses. We did not find significant differences in leaf gas exchange between leaves attached to the tree and detached from the tree after 30 s, agreeing with other studies of riparian trees (Kolb, Hart & Amundson 1997; Horton, Kolb & Hart 2001). Using the same protocol as mentioned previously, we used ten different trees in which three leaves from different branches were measured while attached to the tree, and three leaves from detached branches were measured after being positioned in the cuvette <30 s after severing. Paired Students t-tests showed that all gas-exchange measurements were similar for leaves from attached and detached branches (P > 0·91 for Ci, Gs, Pn, and E, n = 10). Throughout measurements, photosynthetically active radiation (PAR) was maintained at 1500 μmol photons m−2 s−1 using the chamber lamp; IRGA block temperature was measured at ambient condition (mean = 36·9 °C, standard deviation = 1·3, n = 776 trees); cuvette carbon dioxide concentration was maintained at ambient concentration (mean = 381 μmol CO2/mol air, standard deviation = 3·8, n = 776 trees); and relative humidity (RH) was maintained near ambient condition (mean = 15·2%, standard deviation = 1·3, n = 776 trees; weather station range of RH during measurement periods ranged from 12% to 25%). Relatively similar environmental conditions across measurement periods combined with random sampling of genotypes assured that sampling was not biased and measurements were representative of site conditions. Concomitant with gas-exchange measurements, we measured specific chlorophyll content (SPAD) using a Minolta SPAD-502Plus meter (Konica Minolta Sensing, Inc., Tokyo, Japan). Measured leaves were subsequently refrigerated until laboratory analyses for SLA and C and N concentrations.

All laboratory analyses were conducted within 2 weeks of gas-exchange measurements. Leaves were scanned, following petiole abscission, with a Hewlett–Packard scanner (Hewlett–Packard Inc., Palo Alto, CA, USA); leaf area was measured on the digital image using WinFolia software (Regent Instruments Inc., Quebec City, QC, Canada). Leaves were then oven-dried for 48 h at 70 °C and weighed. Specific leaf area was calculated by dividing projected (i.e. one-sided) leaf area by dry mass. Dry leaves were ground through a 40-mesh screen with a Thomas-Wiley Mill Model ED-5 (Thomas Scientific, Swedesboro, NJ, USA) and analysed for C and N concentration following combustion using a FLASH EA 1112 Elemental Analyzer (CE Elantech Inc., Lakewood, NJ, USA).

For each tree used in leaf analyses, we also measured trunk diameter at breast height (DBH; c. 1·4 m) within 2 weeks of leaf gas-exchange measurements. We converted DBH to an index of above-ground growth rate using allometric equations to estimate total above-ground biomass [growth rate (kg biomass tree−1 year−1) = wood biomass/tree age + foliage mass], and assuming that initial mass of the small cuttings used to establish the common garden was approximately constant among trees. We used two separate equations: one for estimating the woody components of above-ground biomass (modified from Lojewski et al. 2009), and an equation derived from trees growing at the common garden in the present study for estimating foliar biomass (detailed later).

Sapwood Area, Leaf Area and Total Tree Leaf Area Estimates

We estimated the ratio of leaf area to sapwood area (AL/AS) and mean individual-leaf area on a subsample of four populations: the two populations with the lowest and the two populations with the highest mean population Gs, as determined from gas-exchange measurements conducted in 2009. We focused on extremes of Gs as it has been widely reported that Gs is strongly related to AL/AS (Whitehead & Jarvis 1981; Whitehead, Edwards & Jarvis 1984). From each population, we selected one ramet from each of five separate genotypes that spanned a range of DBH from c. 10 to 20 cm. We felled trees at the end of the growing season (November 2010) immediately prior to leaf senescence. After felling, we measured tree height, DBH and diameter at stem base (DSB; 0·3 m height) and cut a 4-cm thick cross-section at DSB (we cut our cross-sectional sample at DSB as opposed to DBH due to the presence of large forking branches between 0·3 and 1·4 m). We scanned the cross-sectional area using a Hewlett–Packard flatbed scanner and estimated the sapwood area (heartwood subtracted from xylem cross-sectional area) using WinFolia software (Regent Instruments, Inc.).

We developed allometric equations relating AL to both AS and DSB for each of the four populations separately to test for variation among populations in AL/AS, as well as to create a site-specific allometric equation to predict leaf biomass from DBH. We divided each felled tree into three canopy height zones (zone 1: 0–2·99 m; zone 2: 3–5·99 m; and zone 3: >6 m) and collected four branches representing a range of branch diameters (0·13–5·00 cm) from each zone. For smaller branches (0·13–0·25 cm diameter), we collected all leaves from each branch. For larger branches (>0·26 cm diameter), we counted all leaves on the branch and then collected a random sample of 100 leaves, which were stored at 4 °C for up to 1 week until measurement of their area (using WinFolia software) and mass (after drying for 48 h at 70 °C). For these branches, total leaf area and leaf mass were estimated by multiplying the average leaf area and mass, respectively, by the total number of leaves. For smaller branches (i.e. <0·26 cm diameter), we directly measured total branch leaf area by scanning all collected leaves.

We measured the diameter and recorded the height zone of every branch with a diameter >5·0 cm for each felled tree. For each population (using branches from five genotypes), we developed logarithmic regressions for each height zone relating branch diameter to total branch leaf area and biomass. We then applied population and height-specific equations to all branches to estimate total tree leaf area of the 20 sampled trees. We then developed allometric equations relating DBH to leaf area and calculated AL/AS (m2 cm−2) for each felled tree. We divided the total tree leaf area by the total number of leaves to estimate the mean individual-leaf area.

Statistical Analysis

Variation in each trait across genotypes and populations was tested using restricted maximum likelihood (REML; Conner & Hartl 2004). Data were log-transformed when the distribution of data violated assumptions of normality or homogeneity of variance. The REML model included population and genotype (nested within population) as random effects because populations were originally selected randomly and not according to temperature variation. Significance of variation in traits across genotype and population was tested using a log-likelihood ratio Chi-squared test (Shaw 1987). The proportion of phenotypic variation in each trait that was genetically inherited was estimated across populations using the REML model described previously (Lynch & Walsh 1998). Broad-sense heritability (H2) was estimated using the following equation, where σ2 denotes variance:

display math

Significance of H2 estimates was tested by log-likelihood ratio Chi-squared tests.

The relationship between growth rate and plant traits was evaluated using Pearson's correlation coefficients (r) among populations (n = 11) and among genotypes (n = 67). An exception to this approach was the correlation between growth rate and AL/AS, where we used a subset of four populations of the 11 total populations and 20 genotypes of the total 67 genotypes. Estimates of plant trait values from the REML model were used to calculate genotype and population means. Population and genotype means of growth rates and each trait were used to compute correlations across populations. We tested for population differentiation in AL/AS and mean individual-leaf area using anova followed by Tukey's HSD.

We used principal components analysis to reduce the dimensionality of the inter-correlated leaf traits. We included mass-based values of Gs, Pn, E, leaf N, Chl, SLA, NUE, iWUE and growth rates. We excluded from this analysis the following traits: areal-based measures because they strongly covaried with mass-based estimates (data not shown); leaf C/N ratio because it strongly covaried with leaf N and Ci and leaf C because they were weakly related to other traits in preliminary investigations of bivariate plots (data not shown). We used regression analysis to evaluate relationships between the principal components (PC) and MAMT transfer distance. We evaluated differences in trait values among populations using two complimentary techniques. First, we used analysis of variance (anova) to determine whether at least one population had significantly different PC1 axis scores than the rest. We followed significant anova results with the Tukey's HSD test to determine pairwise differences in PC1 between each population pair. Second, given that PC1 accounts for only the first dominant functional gradient, we also used permutational multivariate analysis of variance (permanova; Anderson 2001) to assess differences among populations in all leaf traits simultaneously. We used Euclidean distance in the permanova analysis because traits were linearly correlated with each other; 9999 permutations were used to assess significance of the differences. We used permanova in the ‘vegan’ library of r (R Development Core Team 2010) to test for pairwise differences; a Bonferroni correction was not made to the alpha value because permanova produces an exact P value (Anderson 2001). All REML and anova analyses were performed using jmp 9.0 statistical software (SAS Institute Inc., Cary, NC, USA) using the alpha = 0·05 significance level.


Because 2009 gas-exchange data were strongly correlated with 2010 data across genotypes, we focus our interpretation on the more extensive 2009 data. For instance, 2009 and 2010 measurements of Gs, Pn and SLA were correlated across genotypes and populations, indicating similar patterns for both years (for Gs: r = 0·92, P < 0·0001, n = 26 clonal genotypes; r = 0·94, < 0·02, n = 4 populations; for Pn: r = 0·89, < 0·0001, n = 26 clonal genotypes; r = 0·89, < 0·05, n = 4 populations; for SLA: r = 0·88, < 0·0001, n = 26 clonal genotypes; r = 0·94, < 0·02, n = 4 populations).

Consistent with our hypothesis that leaf traits were heritable and varied by population, we found that broad-sense heritability ranged from 0·13 to 0·46 across traits and was statistically different from 0·00 for all traits (Table 2). Traits varied significantly among both genotypes and populations (Table 2). Variation in leaf traits among genotypes aligned primarily along a single dimension representing leaf economic traits. The first two principal component eigenvalues were >1 (Table 3). The first principal component (PC1) accounted for 61·5% of the overall variation in traits among genotypes (Table 3). Pn, E, Gs, N, Chl, SLA and NUE loaded positively on PC1, whereas growth rate and leaf iWUE loaded negatively on PC1. The second principal component accounted for 19·8% of the total variance, as well as additional variation in Chl, iWUE and NUE.

Table 2. Range and broad-sense heritability (H2), genotype differentiation and population differentiation (including χ2 test results) for several Populus fremontii traits. Measurements were conducted on replicated clonal genotypes from 11 provenances collected throughout Arizona and California and grown in a common garden at the Palo Verde Ecological Reserve. Plant traits include specific leaf area (SLA), leaf internal carbon dioxide concentration (Ci), leaf carbon to nitrogen (C/N) mass ratio, leaf-level water-use efficiency (iWUE = Pn/Gs), leaf-level N-use efficiency (NUE = Pn/N) and above-ground annual growth rate (GR). Additionally, the following traits are presented on both an areal and mass basis: stomatal conductance (Gs), net photosynthetic rate (Pn), transpiration rate (E), leaf nitrogen (N), leaf carbon (C) and leaf chlorophyll (Chl)
MaxMin X 2 P value χ 2 P value
AreaGs (mol m2 s−1)0·5470·1990·3342·430·000169·240·0001
Pn (μmol m2 s−1)18·811·40·2221·750·000144·920·0001
E (mol m2 s−1)11·26·60·2329·880·000147·260·0001
N (g m−2)2·031·440·318·130·002216·870·0001
C (g m−2)51·832·40·3456·830·000137·590·0001
Chl (g m−2)0·4220·2910·46108·920·000181·260·0001
MassGs (mol kg−1 s−1)6·411·730·3231·990·000190·710·0001
Pn (μmol kg−1 s−1)2241020·2722·400·0001107·350·0001
E (mol kg−1 s−1)138560·2625·760·000183·480·0001
N (g kg−1)22·614·70·2635·640·0001132·320·0001
C (g kg−1)4394030·1311·510·000328·370·0001
Chl (g kg−1)4·923·210·2731·910·000164·900·0001
SLA (m2 kg−1)13·68·20·3046·830·000172·160·0001
Ci (μmol mol−1)2632030·1312·880·000265·090·0001
C : N (kg kg−1)29·419·10·2643·370·0001107·330·0001
iWUE (μmol mol−1)75·634·50·2121·560·000160·520·0001
NUE (μmol g−1)10·96·80·2325·620·000163·890·0001
GR (kg−1 tree−1 year−1)6·682·040·1418·210·000112·470·0002
Table 3. Eigenvalues, per cent of variation explained, cumulative per cent variation explained and eigenvectors from the principal component analysis of 67 genotypes from 11 populations of Populus fremontii throughout California and Arizona grown at the Palo Verde Ecological Reserve common garden. Per cent values are the per cent of total variance (i.e. the sum of the diagonal elements in the correlation matrix) accounted for by the first three principle components (PC1, PC2 and PC3). See Table 2 caption for abbreviations
Cumulative per cent61·581·389·4

In agreement with our hypothesis that traits were related to differences in thermal environments among populations, we found that the relationship between mean PC1 scores for each population, and MAMT transfer distance was significant and positive (R2 = 0·75, P = 0·0006; Fig. 2). In contrast, the relationship between PC2 and MAMT transfer distance was not significant (R2 = 0·01; P = 0·38). Given that PC1 was clearly the most important axis and that PC2 was not related to the temperature gradient, we subsequently focused our analyses on PC1. Analysis of variance on PC1 indicated significant variation among populations (F10,56 = 15·7, P < 0·0001). Tukey's HSD tests on PC1 indicated that the five coldest provenances (i.e. greatest transfer distances) had significantly higher PC1 values than the three warmest provenances (Fig. 2). An assessment of provenance differences in all leaf traits simultaneously using permanova produced similar overall results (F10,56 = 15·1, = 0·0001; Fig. 2). Post hoc permanova tests indicated that the five coldest provenances had significantly different leaf trait values than the four warmest provenances. In further support of our hypothesis, we also found that provenances with high MAMT transfer distance had significantly lower AL/AS than provenances with low transfer distance (< 0·0001; Fig. 3b).

Figure 2.

Relationship between the first principal component (PC1), representing variation in the leaf economics spectrum among populations (Table 3), and provenance transfer distance (mean annual maximum air temperature (MAMT) of common garden minus MAMT of provenance) for Populus fremontii. The linear regression is fit through the population means (two populations occurred at a MAMT transfer distance of zero). Vertical bars represent standard errors (i.e. variation of genotypes within populations). Populations with different lowercase letters below the bars indicate significant pairwise differences in PC1 scores using Tukey's HSD test. Populations with different capital letters above the bars indicate significant pairwise differences using permanova on all traits simultaneously.

Figure 3.

Relationship of leaf area to sapwood ratio (AL/AS) to growth rate and air temperature for 20 genotypes from four populations of Populus fremontii collected from throughout Arizona and California and grown in a common garden at the Palo Verde Ecological Reserve. (a) The relationship at both the population- and genotype-level between (AL/AS) and above-ground annual growth rate. (b) Mean population plus standard error of leaf area to sapwood area and mean individual-leaf area by provenance transfer distance (mean annual maximum air temperature (MAMT) of common garden minus MAMT of provenance). Different lower case letters denote significant differences between populations (P < 0·05).

We addressed the sensitivity of these results relative to sample size as we were only able to evaluate four genotypes in three of the provenances (Table 1). We therefore excluded these provenances in a separate test of the relationship between PC1 scores and MAMT transfer distance (Fig. 2) to test whether this relationship was sensitive to including data from these provenances. We found that the relationship in Fig. 2 was similar with (R2 = 0·75, = 0·0006) and without these provenances (R2 = 0·66, = 0·0144).

In agreement with our hypothesis that leaf traits would be correlated with growth rate, we found that most traits were significantly correlated with growth rate among genotypes and populations and that trees with faster growth rates produced more leaf area per amount of sapwood than trees with slower growth. Among populations, we found negative correlations between growth rate and both leaf mass- and areal-based measures of gas-exchange, leaf N and SLA (Table 4). We also found positive correlations with growth rate for leaf C content (areal-based) and C-to-N mass ratio. We found similar correlations among genotypes except that SLA and leaf C content (areal-based) were not correlated with growth rate, and iWUE was positively correlated with growth rate (Table 4). We also found that AL/AS was strongly and positively correlated with growth rate among both genotypes and populations (Fig. 3a). Trees with high AL/AS also had larger leaves on an individual-leaf basis (Fig. 3b).

Table 4. Pearson correlation coefficients (r) and their statistical significance (P value) between Populus fremontii annual growth rate and leaf-level traits among genotypes (n = 67) and among populations (n = 11). Genotypes were collected from 11 provenances in Arizona and California and grown in a common garden at the Palo Verde Ecological Reserve (PVER). See Table 2 caption for abbreviations
 TraitAcross populationsWithin populations
r P-value r P-value
C : N0·900·00020·440·0016


Climate-Driven Variation in Leaf Traits

Using common gardens to select favourable genotypes for use in restoration (McKay et al. 2005; O'Brien, Mazanec & Krauss 2007) and assisted migration (Ledig & Kitzmiller 1992; Rehfeldt et al. 2003; Hulme 2005; Aitken et al. 2008) is becoming an increasingly useful strategy of climate change mitigation (Grady et al. 2011). Incorporating plant functional traits into genotype selection trials may improve the effectiveness of selecting genotypes most suitable to a changing climate. For example, we found that a conservative suite of leaf economic traits (i.e. low mass-based Pn, Gs, N and SLA; Wright et al. 2004) was positively correlated with growth rate of P. fremontii genotypes in an extremely warm common garden (Table 4). The leaf economics spectrum reduced to a single principal component (PC), reflecting conservative traits with negative PC scores and acquisitive traits with positive PC scores, showed a similar pattern among populations. Provenance genotypes with conservative traits grew faster than those with acquisitive traits (Fig. 2). These patterns strongly suggest that a conservative strategy of leaf traits for resource acquisition in a hot desert–riparian environment is associated with faster growth. Our results imply that P. fremontii genotypes with conservative leaf traits will be the most productive in a warmer climate. Thus, linking leaf traits with growth rates among genotypes in a common garden located at the warm edge of a species distribution may be a useful approach for screening genotypes best suited to a future warmer climate.

The premise that local genotypes are locally adapted to hot conditions of the common garden (MAMT of > 31 °C), and that inter-correlated leaf traits are potentially adaptive, is supported by significant heritability of all measured plant traits and strong correlations between traits and growth rate. Heritability of plant physiological and morphological traits is not surprising given their prominent functional role in C acquisition and growth. Our results are supported by a wide range of studies that have reported heritability of similar plant traits for many species across diverse environments (Lambers, Chapin & Pons 2008). However, we acknowledge that relationships between heritable plant traits and growth alone do not necessarily indicate genotypic fitness, which is best measured by long-term survival and reproductive success.

The Leaf Economics Spectrum

Our finding of a positive relationship between conservative traits on the leaf economics spectrum (i.e. low mass-based Pn, Gs and N and low SLA) and plant growth rate differs from the widely reported negative correlation between conservative traits and growth rate across thousands of plant species (Poorter & Remkes 1990; Reich, Walters & Ellsworth 1992; Shipley 1995; Grubb 1998; Poorter & Bongers 2006). Why would interspecific correlations between plant functional traits and growth across a broad range of species differ from intraspecific correlations in P. fremontii?

In one of the few studies to investigate intraspecific correlations between leaf traits and growth rate of trees in a common garden, and the only study to have this common garden located at the warm edge of a species geographic distribution, we found that genotypes from warmer provenances (i.e. with the lowest MAMT transfer distances) had more conservative leaf traits and grew faster than genotypes from cooler provenances with less conservative, or more acquisitive, traits. In contrast, a common garden experiment in a more temperate environment in Belgium showed that hybrid Populus clones with acquisitive leaf traits had higher growth rates (Pellis, Laureysens & Ceulemans 2004). The difference between these studies suggests that the relationship between leaf traits and growth rate can vary, even between closely related species, depending on site-specific environmental conditions.

Although many studies across species have shown a positive relationship between acquisitive traits and growth rate (Reich, Walters & Ellsworth 1992; Shipley 1995; Poorter & Bongers 2006), results of intraspecific studies in common gardens are idiosyncratic. For instance, another intraspecific study in a common garden trial of two perennial grasses in England showed that acquisitive traits were not related to total above- and below-ground biomass production and were negatively related to shoot biomass production among genotypes (Bilton et al. 2010). However, they found that C allocation to reproduction was positively correlated with acquisitive traits, indicating a trade-off between growth and reproduction. Taken together, these examples at the intraspecific level illustrate how fine-scale environmental filters and trade-offs in C allocation can lead to relationships between leaf traits and growth rates that would not be predicted from interspecific patterns.

At both global and regional scales, environmental conditions have been shown to affect plant traits in predictable patterns across species (Díaz et al. 2004; Wright et al. 2004; Cornwell & Ackerly 2009). The interspecific literature on leaf traits demonstrates that plants from resource-rich environments are characterized by a fast return on investment in leaf tissue (fast growth rates and high tissue turnover), while plants from resource-poor environments are characterized by slow growth rates and high tissue longevity to conserve limiting resources (Chapin 1980; Reich, Walters & Ellsworth 1997; Wright et al. 2004). However, our data suggest that relationships in leaf economic traits among species may not accurately describe localized adaptive variation at the intraspecific level.

We suggest three potential mechanisms to explain our findings that P. fremontii genotypes with low photosynthetic capacity per unit leaf area can have a more rapid above-ground growth rate compared with P. fremontii genotypes with high photosynthetic capacity: (i) a trade-off in C allocation between above-ground growth and reproduction or below-ground growth; (ii) variation in phenology among genotypes, with genotypes from warmer provenances having a longer growing season; and (iii) a negative relationship between photosynthetic capacity at the whole-tree level and photosynthetic capacity at the leaf area level. Regarding the first mechanism, although we did not measure reproductive output, several recent studies have shown that growth rates are either positively correlated with or independent of reproductive potential in hardwood trees (Knops, Koenig & Carmen 2007; Satake & Bjørnstad 2007). For the trees in our study that became reproductive at 3 years of age, we found the same negative relationship between growth rate and leaf area-based photosynthetic rate in both the second (prior to flowering) and third years (first year of flowering) after planting. We also found the same relationship between provenance MAMT transfer distance and growth rate prior to and after flowering (i.e. genotypes that were the least productive prior to maturity were still the least productive after maturity). These findings suggest that reproductive output had little effect on genetic variation in growth and does not explain the negative relationship between growth and photosynthetic rate among genotypes. However, genotypic variation in the trade-off in C allocation to above- versus below-ground growth has been demonstrated in Populus clones (e.g. Souch & Stephens 1998) and may have occurred in our study. Availability of stream and ground water can alter C assimilation and allocation in riparian species (Horton, Kolb & Hart 2001; Williams et al. 2006). The genotypes used in our study were transplanted from perennial streams, and the common garden was well watered during the growing season. Thus, variation among genotypes in response to varying water conditions was unlikely to drive variation in below-ground allocation. We cannot address this trade-off because we did not measure below-ground growth in our study, but this trade-off clearly is an important subject for future research.

Several studies on Populus species have shown variation in growth phenology among provenances from different latitudes, with lower-latitude provenances having longer growing seasons, primarily due to late autumn bud set (Howe et al. 2003; Keller et al. 2011). In our study, it is possible that a longer growing season contributed to greater growth of low-latitude provenances. While we did not measure phenology, genetic variation in growth phenology was unlikely to account for the almost twofold difference in ANPP (Fig. 3a) among provenances. For example, Populus nigra provenances that varied by 11° latitude showed a 27-day difference in bud set (Rohde et al. 2011). In our study, provenances varied by only 6° latitude; thus, we might expect that variation in bud set would be less than that reported for P. nigra. Growing season length at our common garden averages 300 days (Grady et al. 2011). Consequently, provenance variation in bud set <27 days would result in less than a 9% difference in annual growing season length, which is unlikely to cause a twofold difference in annual growth. Hence, we conclude that the second potential mechanism explaining the above-ground growth patterns among genotypes is also improbable.

Regarding the third mechanism, our results suggest that total tree leaf area was more important than leaf area-based photosynthesis in determining overall growth. We found a positive correlation between leaf area to sapwood area ratio and annual growth rate for a subset of genotypes that varied widely in growth rate (Fig. 3). Genotypes with the highest growth rate (those from the warmest provenances; Fig. 3) had the highest leaf area to sapwood area ratio (Fig. 3), in part, due to having larger individual leaves (Fig. 3b). Whole-tree hydraulic models indicate a negative relationship between canopy stomatal conductance and AL/AS when other hydraulic influences are constant (Whitehead & Jarvis 1981; Whitehead, Edwards & Jarvis 1984). This relationship suggests that P. fremontii genotypes with low stomatal conductance can support greater AL/AS, which could increase whole-canopy photosynthesis and compensate for low leaf area-based photosynthetic rate. Thus, we suggest that a conservative leaf economic strategy, by limiting water loss per leaf in an extremely hot desert environment via higher water-use efficiency, can result in higher growth rates by allowing more leaf area to be supported in the canopy. This hypothesis is further supported by a significant positive correlation between leaf-level water-use efficiency and growth rate among genotypes in our study (Table 4).

We found that canopy hydraulic architectural (i.e. AL/AS) was more important than acquisitive leaf-level characteristics in determining rapid growth rates of P. fremontii genotypes. This finding suggests that combining tree-level with leaf-level measurements improves our mechanistic understanding of tree growth over that gained from using single-scale measurements alone. These results also agree with a growing body of research that cautions against using measurements recorded at a single spatial scale, such as the individual-leaf scale, in attempts to understand relationships between physiology and whole-tree growth or water use (Andrade et al. 1998; Meinzer et al. 2010). More broadly, the importance of taking a multi-spatial scale approach for evaluating linkages among ecological processes has also been highlighted in investigations of species richness–productivity relationships in terrestrial plant communities (Chalcraft et al. 2004). Addressing scale dependence of global patterns of leaf trait–productivity relationships will improve our understanding of plant function; this will be especially important to consider in predicting how traits and growth rates will respond to novel environments shaped by climate change.


This research was supported by a Science Foundation of Arizona Fellowship Award, the Bureau of Reclamation Grants CESU-06FC300025, 04FC300039 and NSF FIBR grant DEB-0425908. We thank Chris Updike, Stephanie Jackson, Ben Sullivan, Faith Walker, Karla Kennedy, Laura Hagenauer, Scott Norris and Chris Pope for field and laboratory assistance. We thank the Whitham laboratory group for help in establishment and maintenance of the garden.