Hydrologic linkages between a climate oscillation, river flows, growth, and wood Δ13C of male and female cottonwood trees


Correspondence: S. Rood. E-mail: rood@uleth.ca


To investigate climatic influence on floodplain trees, we analysed interannual correspondences between the Pacific Decadal Oscillation (PDO), river and groundwater hydrology, and growth and wood 13C discrimination (Δ13C) of narrowleaf cottonwoods (Populus angustifolia) in a semi-arid prairie region. From the Rocky Mountain headwaters, river discharge (Q) was coordinated with the PDO (1910–2008: r2 = 0.46); this pattern extended to the prairie and was amplified by water withdrawal for irrigation. Floodplain groundwater depth was correlated with river stage (r2 = 0.96), and the cottonwood trunk basal area growth was coordinated with current- and prior-year Q (1992–2008: r2 = 0.51), increasing in the mid-1990s, and decreasing in 2000 and 2001. Annual Δ13C decreased during low-flow years, especially in trees that were higher or further from the river, suggesting drought stress and stomatal closure, and male trees were more responsive than females (−0.86 versus −0.43‰). With subsequently increased flows, Δ13C increased and growth recovered. This demonstrated the linkages between hydroclimatic variation and cottonwood ecophysiology, and we conclude that cottonwoods will be vulnerable to drought from declining river flows due to water withdrawal and climate change. Trees further from the river could be especially affected, leading to narrowing of floodplain forests along some rivers.


With fixed positions and multi-century lifespans, trees act as sentinels of climatic variation and change (Briffa 2000). Dendrochronological (tree ring) studies are central to analyses of the ecological consequences of climate change, and these studies have especially involved trees in locations with specific physical limitations on growth, such as arid zones where water is limiting and precipitation provides the dominant environmental influence (Axelson, Sauchyn & Barichivich 2009). Climatic variation includes the directional change associated with global warming, as well as oscillations, reversing alterations of temperature, precipitation and other climatic factors. The Pacific Decadal Oscillation (PDO) is the predominant oscillation affecting temperature and precipitation in the arid and semi-arid regions of western North America (Mantua & Hare 2002), and historic river flows from the ‘Crown of the Continent’, the central Rocky Mountains region straddling the United States–Canada border, are strongly coordinated with the PDO over the 20th century (Rood et al. 2005, 2008; St. Jacques, Sauchyn & Zhao 2010).

In dry regions, riparian (streamside) woodlands provide marked contrast to the non-forested uplands (Naiman, Décamps & McClain 2005). In these regions, the riparian trees are reliant upon groundwater that originates from the adjacent stream and infiltrates into the alluvial aquifer (Scott, Shafroth & Auble 1999; Horton, Kolb & Hart 2001; Amlin & Rood 2003). In many arid and semi-arid regions around the Northern Hemisphere, cottonwoods, or riparian poplars (Populus species), are the predominant streamside trees and represent the keystone species for the floodplain forests. With their widespread occurrence and dependence upon stream flow (Rood, Braatne & Hughes 2003), cottonwoods could be useful for investigating aspects of climatic variation that influence river flows. Sufficient river flows are not only essential for riparian and aquatic ecosystems but also provide primary surface water resources for agricultural irrigation, municipal and industrial needs, and hydroelectric power generation.

We have been studying riparian cottonwoods, with a focus on ecophysiological differences across species, hybrids and sexes (Rood et al. 2003; Letts et al. 2008; Nielsen et al. 2010). In the present study, we sought to coordinate two research themes to investigate whether the growth and water relations of riparian cottonwoods were coordinated with river flow patterns, and consequently with the climatic variation associated with the PDO. Our study design incorporated a two-decade interval with very high-flow years, followed by very low flows, and then a return to intermediate flows. A dendrochronological approach enabled retrospective analysis to investigate the following hypotheses:

  • H1: Temporal association Our first hypothesis was that there would be positive correspondence between river flows and cottonwood growth. We particularly anticipated growth reduction during a dry, low-flow interval due to water stress that would be reflected in decreased carbon stable isotopic discrimination (Δ13C) of the stem wood. We further expected carry-over influences from river flows of the prior year(s).
  • H2: Spatial influence Our second hypothesis was that the environmental conditions of the individual trees would influence their water status and vulnerability to river flow decline. We particularly expected increased susceptibility of trees situated higher above, and further from, the river.
  • H3: Sex differentiation Based on patterns from prior studies of riparian cottonwoods, willows (Salix species) and box elder (Acer negundo) (Dawson & Ehleringer 1993; Hultine et al. 2007), our third hypothesis was that males would be better adapted to drought stress than females, and that this would involve less effect on Δ13C and growth with a low-flow interval.


Hydrology and weather

We analysed the sequence of hydrologic linkages extending from the PDO, through headwater and then downstream river flows, and to the alluvial groundwater that phreatophytic riparian cottonwoods are dependent upon. Monthly PDO values were obtained from the University of Washington (http://jisao.washington.edu/pdo/PDO.latest) and averaged by year. Historic river discharges (Q) were obtained from the HYDAT database of the Water Survey of Canada (http://www.wsc.ec.gc.ca/applications/H2O/index-eng.cfm), for the Waterton River at Waterton Park (#05AD025; free flowing, relatively pristine watershed), the St. Mary River at the International Boundary (#05AE027; pristine watershed but somewhat regulated flows) and the Oldman River near Lethbridge (#05AD007; extensively regulated). For the Oldman River, ‘naturalized flows,’ estimates of flows that would have occurred without regulation, were obtained from Alberta Environment's Weekly Natural Flow Database (Version 3.02, October 2004, Edmonton, AB) and these were available only to 2001. Analyses included yearly (Qyear), monthly (Qmonth) and multiple-monthly mean discharges, especially for the cottonwood growth (in leaf) season from May to October (QMay–October).

Groundwater depths (Zgw) were determined at 1 h intervals through the summer of 2009 with Solinst 3001 LT Leveloggers (Solinst Canada Ltd, Ontario), with barometric correction with measurements from a Solinst 3001 LT Barologger. The transducers were positioned in piezometers that consisted of vertical 4-m-long, 3-cm-diameter galvanized steel pipes attached to Solinst Model 615-screened drive points driven into the alluvia that consisted primarily of sands and gravels. Two piezometers were installed, with the riverside piezometer near the upstream end of the cottonwood grove, about 75 m from the river channel, and the floodplain piezometer 225 m from the channel. For comparison, hourly river stage (elevation) data were obtained for the hydrometric station 05AD007, 1.4 km downstream from the riverside piezometer. Weather data from the Lethbridge Airport, 10 km from the study site, were obtained from the Environment Canada Archive (http://climate.weatheroffice.gc.ca/climateData). These were used to calculate monthly means for precipitation (P) and temperature (T) from May to October.

Cottonwood trees

We analysed narrowleaf cottonwoods (Populus angustifolia James) in the Helen Schuler Nature Reserve, a natural woodland within the Oldman River valley at Lethbridge, Alberta, Canada (49.702°N, 112.863°W). This woodland also contains prairie cottonwoods (Populus deltoides), black cottonwoods (Populus trichocarpa) and interspecific hybrids (Gom & Rood 1999). Distinct branching patterns, leaf shapes and phenology enabled the recognition of different species and clones, and we sought apparently pure narrowleaf cottonwoods. Trees were otherwise haphazardly sampled within an 800-m-long band of mature cottonwoods that paralleled the river channel. To avoid clonal ramets, selected trees were more than 50 m apart, or were of different sexes. Reproductively mature and apparently healthy trees of relatively uniform size [diameters at breast height (DBH): males 30.1 ± 2.0 cm, n = 13; females 27.6 ± 1.4 cm, n = 18] were selected in April and May 2009 and sexes were identified from catkins. All of the trees were included for tree ring analysis, but only 17 females were sampled for 13C analyses.

Each tree was photographed and identified with a numbered metal tag, and its position was determined with wide angle augmentation system GPS (Garmin eTrex, Olathe, KS, USA). The elevation of each tree relative to the adjacent river surface was determined to ±1 cm with a transit and staff gauge on 11 May 2011, at a river discharge of 206 m3 s−1. The corresponding river stage would have been about 117 cm above the base stage, the level associated with the base flow of 20 m3 s−1, which is the minimum summer flow following the 1993 implementation of the Oldman River Dam (Rood, Kalischuk & Mahoney 1998). The extent of canopy shading from adjacent trees was estimated for each tree as the average of estimates on a scale of 1 (shaded) to 3 (exposed) at each of three aspects [north (N), southeast (SE), southwest (SW)] around each tree.

Three increment cores were extracted from each tree with a 5.15-mm-diameter Haglöf borer, with a core from the N-, SE- and SW-facing sectors. The different sectors were exposed to different environmental conditions, especially solar radiation and wind, and this provides a progressively increasing sequence of drying index of N, SE and SW (Samuelson & Rood 2011). The different environmental conditions might be reflected in differences in growth among the radial sectors, because in poplars there is partial vertical vascular alignment (Kozlowski & Winget 1963).

Cores were mounted and shaved transversely and annual radial increments (RIs) were measured to 0.002 mm precision with a dissecting microscope, a Velmex sliding stage and Acu-Rite encoder (Velmex Inc., Bloomfield, NY, USA), and MeasureJ2X software (VoorTech Consulting, Holderness, NH, USA). We used RI for comparisons of sector growth rates, but for most comparisons we compared annual basal area increments (BAIs), calculated from the average of the three RIs, and DBH. In contrast to RIs, which display an age-related pattern, the BAIs of regional cottonwoods become fairly constant within cottonwood groves with closed canopies, providing a more constant growth baseline (Willms, Pearce & Rood 2006; Berg et al. 2007).

After ring measurements, the annual wood segments were excised from the cores for three 3 year intervals: 1994–1996; 1999–2001; and 2006–2008. These represented high-flow, low-flow and intermediate-flow intervals. For each year and from each tree, the segments from the three cores from different sectors were combined, dried and finely ground for analysis of carbon isotopic composition (δ13C). The samples were sent to the University of California, Davis, Stable Isotope Facility, for analysis with a PDZ Europa ANCA-GSL elemental analyser interfaced to a PDZ Europa 20–20 isotope ratio mass spectrometer (Sercon Ltd, Crewe, UK). Annual values of carbon isotope composition, relative to the Vienna PeeDee Belemnite standard (δ13C,‰), were used to calculate annual values of carbon isotope discrimination (Δ13C) as

inline image (Farquhar, O'Leary & Berry 1982)

where δp is wood δ13C (‰); and air δ13C (δa) was assumed to be −7.82‰ in 1994, and to decrease regularly at 0.028‰ per year (Francey et al. 1999; McCarroll & Loader 2004) to yield estimates for the other years.

Δ13C in bulk wood, such as these samples, is expected to provide a similar or stronger relationship to climatic parameters as Δ13C in cellulose or lignin (Loader, Robertson & McCarroll 2003). Δ13C in lignin is higher than that in cellulose, however, and so Δ13C in bulk wood of poplars is expected to be higher than that in cellulose (Loader et al. 2003; Rasheed et al. 2011). The difference may be constant from year to year (Loader et al. 2003), or may increase gradually with age (Rasheed et al. 2011). Our interannual comparison of bulk wood Δ13C might consequently be slightly influenced by tree aging.


SPSS 18 (IBM Corp., Somers, NY, USA) was used for statistical analyses and we first investigated Pearson product correlations (r) across the weather, hydrology and growth variables. Coefficients of determination (r2) are reported and referred to as correspondence or association (%). For historic time series, moving 3- or 5-point averages were sometimes calculated for data smoothing. We undertook regression analyses to investigate associations between growth and Δ13C of the individual genotypes, with the individual environmental factors of elevation (m above the base stage), distance (log of m from the typical summer riverbank position) or the extent of canopy shading from adjacent trees. For regressions, we favoured the simplest function that provided a near-maximal r2 with the sequence: linear, log and quadratic.

For the interannual comparisons, multifactor analyses for RI, BAI or Δ13C involved univariate linear mixed-model (LMM) analyses of variance (anova), with sex and year and their interaction as fixed effects, the annual values from each tree treated as repeated measures, and the trees included as a random effect on the intercept. Based on the Bayesian information criterion, the first-order autoregressive covariance structure was chosen for the repeated measures analysis. The Bonferroni adjustment was used in multiple comparisons.

We then analysed Δ13C values across the individual trees for particular years, and especially for the extreme low-flow year 2001, and for the change in Δ13C over the low-flow interval from 1999 to 2001. These involved general linear model anovas with elevation, distance and/or canopy exposure as continuous covariates, and sex as a fixed factor. We chose the simplest (fewest factors) model that provided a near-maximal fit (r2).


Climatic variation and hydrology

Over the past century, patterns of river discharges (Q) have been closely coordinated for the different headwater streams within the Oldman River Basin (Rood et al. 2005), and these were also closely coordinated with the PDO (Fig. 1). There were some apparent lags in the association, and these were somewhat variable (e.g. 1920s and around 2000).

Figure 1.

Values (5 year moving averages) of the Pacific Decadal Oscillation (PDO, inverted axis) index and discharge (Q) of the St. Mary River, the largest headwater tributary of the Oldman River, displaying about one-half correspondence (n = 98, r2 = 0.457, P < 0.001).

Downstream, each of the major tributaries within the Oldman River Basin has been dammed to permit water storage and offstream diversion, largely for irrigation. The extent of diversion increased over the 20th century, and withdrawal increased to more than 50% of the naturalized flow during low-flow years after 1970 (not shown).

The study period for cottonwood growth rings extended from 1992 to 2008 and stem wood Δ13C was assessed from three shorter intervals. The first interval (1994–1996) included 1995, with the highest peak flow of the century-long record (Rood et al. 1998). The middle interval (1999–2001) represented an extremely low-flow period due to low snowpacks and rainfall, and substantial water withdrawal for irrigation (Rood & Vandersteen 2010). The annual hydrographs for the low-flow years 2000 and 2001 (Fig. 2) reveal the nearly complete trapping of the spring peak that would have occurred in May and June, the primary interval of cottonwood leaf production and branch elongation (Willms et al. 1998).

Figure 2.

Measured discharge of the Oldman River for 2000 and 2001, and the naturalized discharge that would have occurred without damming and diversion.

Depth to the groundwater table (Zgw) was analysed in 2009, a moderate-flow year. The two piezometers, which flanked the sampled cottonwoods, indicated strong correspondence between Zgw and river discharge. At the riverside piezometer, the Zgw pattern was tightly associated with the adjacent river stage (r2 = 0.96, P < 0.001, Fig. 3). The Zgw pattern at the floodplain piezometer was also coordinated with the river discharge, but the pattern was attenuated, reflecting alluvial infiltration and drainage. The observed depth of Zgw fluctuation is likely to incorporate the primary zone of coarse roots, as revealed in a cut-bank exposure 2 km upstream (Rood, Bigelow & Hall 2011). Interestingly, there was diurnal fluctuation of Zgw at the riverside piezometer, and more prominent fluctuation at the floodplain piezometer (Fig. 3). That would quantitatively reflect the groundwater drawdown from the daily cottonwood forest transpiration (Butler et al. 2007; Lautz 2008).

Figure 3.

Elevations of the water table for two piezometers (wells), near the study grove of cottonwoods, at 1 h intervals, and of the adjacent Oldman River for the 2009 growth season. The plots are vertically offset for clarity. The inset expands the boxed floodplain piezometer plot for 10–16 September.

For the study interval from 1992 to 2008, and particularly for the cottonwood growth period from May to October, flows of the Oldman River were positively correlated with the local precipitation at the Lethbridge Airport and tended to be negatively associated with temperature (Table 1). There was negative correspondence between precipitation and temperature at Lethbridge, and thus while the observed higher river flows largely reflected precipitation and snowmelt regimes in the Rocky Mountain region, these were also associated with wet and cool conditions in the prairie region.

Table 1. Correspondence (r2) for the interval 1992–2008 between narrowleaf cottonwood yearly BAIs, mean May–October discharge of the Oldman River at Lethbridge (Q, growth year, and prior-year weighted one-half), mean May–October weather at Lethbridge, Alberta, and the PDO (3 year running mean)
 QQyr-1Q and Qyr-1PrecipitationTemperaturePDO
  1. t (trend) = P < 0.1; * = P < 0.05; ** = P < 0.01.
  2. BAI, basal area increment; PDO, Pacific Decadal Oscillation.
Q 0.0730.776**0.500**0.202t0.117
Qyr-1  0.276**0.189t0.0860.153
Q and Qyr-1   0.2440.0300.323*
Precipitation    0.504**0.170
Temperature     0.122

Cottonwood growth

There was substantial variation in the annual RIs over the 17 year interval (Fig. 4, Table 2). Growth rate increased after 1992 to a peak in 1995, fell to the minimum around 2000, and subsequently, gradually increased. There was no significant effect of sex, or year × sex interaction in the anova (Table 2), but for the 17 year average there was a trend for increased RI in the females (Fig. 4; and t = 1.702, P = 0.101).

Figure 4.

Annual radial increments (a) and basal area increments (b) for 1992–2008 for female (n = 18) and male (n = 13) cottonwoods along the Oldman River. Mean ± SE.

Table 2. Linear mixed-model analyses of variance for growth of male and female cottonwoods
  1. t = P < 0.1; ** = P < 0.01; *** = P < 0.001.
  2. BA, basal area; BAI, basal area increment; RI, radial increment.
RIs – yearly values averaged from 3 cores per tree
Year (17 years)6.740.000***
Sex (17 female: 13 male)1.6840.204
Sex × year0.6010.882
BAIs – yearly values
Sex × year1.1190.337
BAI for the low-flow interval – 1999 to 2001, with initial BA as covariate
Year (3 years)12.5410.000***
Sex × year0.0880.916
Initial BA14.3290.001**
RI – comparison across the three sectors and two sexes, with initial BA as covariate; non-significant interactions removed
Year (17 years)11.5870.000***
Initial BA8.3190.005**

The BAI also showed highly significant interannual variation, but the female and male BAIs were more similar than the RIs (Fig. 4, Table 2). There was no significant year × sex interaction for BAI over the 17 year interval but there was apparent differentiation during the low-flow interval from 1999 to 2001, with higher BAI in the female trees, suggesting less growth inhibition than for the males.

Within-tree comparisons – growth by radial sector

We anticipated that the different sectors would experience different sunshine and wind patterns that could produce differing growth responses during the low-flow, drought interval. However, the growth patterns were very similar across the three sectors, with common patterns of growth reduction around 2000 and 2001, and there was no sex × sector interaction (Table 2, Supporting Information Fig. S1).

Correspondence between hydrology and tree growth

There was strong correspondence between the interannual BAI and the mean annual river discharge (Q) and even closer correspondence with the mean Q for the cottonwood growth season from May to October (Figs 4,5, Table 1). Growth closely tracked Q from 1992 to 2001 but there was subsequently more deviation with the irregular Q pattern from 2002 to 2008. In particular, Q fell substantially in 2004 while the BAI remained constant. This attenuated growth pattern indicates that the growth rate in a particular year responded not only to conditions of that current year but also reflected the effects of the prior-year water availability. This principle is supported by the increased association between BAI and Q of the current and prior years (Table 1). The further consideration of growth season precipitation or temperature did not substantially increase the observed correspondence over that between BAI and the two-season Q.

Yearly wood Δ13C

There was considerable variation of stem wood Δ13C among the trees, but the relative values were quite consistent, producing parallel patterns over the time series (Supporting Information Fig. S2). Individual measurements of Δ13C (i.e. in wood of individual years from individual trees) ranged from 18.0 to 24.1‰ and the mean values for individual trees across all years ranged from 19.4 to 23.1‰. For individual trees, Δ13C was fairly similar for the initial, high-flow interval of 1994–1996, and for the final, intermediate-flow interval of 2006–2008. In contrast, Δ13C decreased during the low-flow interval from 1999 to 2001 (Fig. 5). The LMM anova confirmed the variation across years (Table 3), and in paired comparisons, the Δ13C in 2001 was lower than in any other year (P < 0.01 for every comparison), and lower in 2000 than in 1995 or 1999 (P < 0.05).

Figure 5.

Interannual patterns of (a) Oldman River discharge [Q, mean of current and one-half weighted prior-year growth season (May–October)], (b) wood Δ13C, and (c) basal area increment (BAI) of annual rings of male and female cottonwoods. To provide standardized measures for Δ13C and BAI, the interannual mean for each tree was determined and the yearly difference from that mean was calculated; these differences were then averaged across the trees.

Table 3. Linear mixed-model analyses of variance for Δ13C content in wood of male and female narrowleaf cottonwoods growing along the Oldman River
  1. * = P < 0.05; *** = P < 0.001.
Δ13C for sexes across years, with distance as a covariate
Year (9 years)13.176< 0.001***
Sex (17 female: 13 male)0.2520.619
Sex × year1.5220.155
Distance (log)4.4760.044*
Δ13C for sexes across the low-flow interval from 1999 to 2001
Year (3 years)33.97< 0.001***
Sex × year4.1750.021*
Distance (log)7.4920.011*

For the full data set, there was no overall effect associated with sex, or for the sex × year interaction (Table 3). We subsequently analysed the shorter interval of 1999–2001 to exclude the relatively invariant early and late sampling intervals. This analysis revealed a significant sex × year interaction for the low-flow interval (P = 0.02; Table 3). Thus, during the interval of low river flow and relative drought, Δ13C decreased more in the male trees than in the female trees and this difference was about twofold.

Much of the difference in Δ13C across the different trees in this study could be explained by the effects of environmental conditions within the grove, and particularly by each tree's distance from the river, elevation above the river and the extent of shading from adjacent trees. Across the trees, the Δ13C values were significantly (P < 0.05) negatively correlated with (log) distance from the river in each year from 2000 to 2008. The correlation was strongest in the low-flow years of 2000 and 2001 (P < 0.01; Fig. 6), consistent with increased stomatal closure in response to greater water stress. There was no correlation between Δ13C and distance in the high-flow years of 1994–1996.

Figure 6.

Variation in (a) Δ13C of the 2001 (very low-flow year) growth ring with the (log) distances of male and female cottonwood trees from the Oldman River; and (b) the difference in Δ13C from 1999 to 2001 versus the trees’ elevations above base flow river stage. Correspondences for combined sexes were (a) r2 = 0.250, P = 0.001; and (b) r2 = 0.344, P = 0.001.

Across the tree positions, there was no correlation between distance from the river and elevation above the river (r2 = 0.004; P = 0.983), because the floodplain surface was relatively flat with localized elevated depositional bands and depressed channel remnants. Consequently, we also considered the possible negative association of Δ13C with elevation, and in 2001 this tended to occur for the males (r2 = 0.240; P = 0.089) but not for the females (r2 = 0.058; P = 0.353). We calculated a composite measure for proximity to water for each tree location, as the product of (log) distance × elevation, and this provided a stronger negative association than for distance alone for the males (r2 = 0.449, P = 0.012; females: r2 = 0.168, P = 0.102).

The final environmental factor considered was canopy exposure (the inverse of shade), and there was a statistical trend towards greater exposure in the male than the female trees [t(28) = 1.871; P = 0.072]. For all trees, exposure was negatively correlated with the mean Δ13C for each tree over all years (r2 = 0.138, P = 0.044), suggesting increased drought stress and stomatal closure with increased exposure.

Analyses of covariance were undertaken to consider multiple factors for Δ13C of individual years and especially the low-flow, drought year 2001 (Table 4). Distance was the predominant environmental factor and accounted for 25% of the variation across the trees. The addition of sex increased the model fit to 32%, or the addition of exposure similarly increased the fit to 33%. Distance, exposure and sex provided a three-factor model with 38% fit, while the optimal three-factor model included distance, exposure and elevation, and explained 40% of the variation. A four-factor model with distance, exposure, sex and elevation only further increased the fit to 41%, suggesting some overlap between sex and environmental condition, and especially exposure, which might reflect the increased clonality of females in this riparian woodland (Gom & Rood 1999).

Table 4. Univariate analyses of variance for wood Δ13C in the low-flow year 2001, and the Δ13C difference from 1999 to 2001, with sex as a fixed factor and continuous covariates: (log) distance from the river, elevation above the base river stage, and canopy exposure
  1. The one-, two- or three-factor models that provided the maximal r2 values are shown, along with two-factor pairs with sex. Corrected model significance: ** = P < 0.01; *** = p ≤ 0.001.
One factorDistance0.250**Elevation0.302**
Two factorsDistance and sex0.322**Elevation and sex0.345**
Distance and exposure0.333**Elevation and distance0.427***
Three factorsDistance, exposure and elevation0.395**Elevation, distance and sex0.485***

In contrast to its weak association with the 2001 Δ13C, elevation was strongly associated with the change in Δ13C from 1999 to 2001, accounting for 30% of the overall variation (Fig. 6, Table 4). There were significant effects from sex in a two-factor model, which increased the fit to 35%. Distance also provided a significant influence and the optimal three-factor model of elevation, distance and sex accounted for about one-half of the observed variation (49%). These analyses demonstrated that the position of the trees relative to the river and groundwater contributed to the drought response, and also that there was differentiation across the sexes, with the males being more responsive to drought, with lower wood Δ13C.

Interannual correspondences between Q, Δ13C and BAI

The influence of tree position (distance and elevation) especially on Δ13C confounded the investigation of the prospective correspondences between the river flow (Q) and the responses of the riparian cottonwoods. Consequently, there was no significant (P < 0.05) correlation between the Δ13C and the BAIs of the 17 trees for the 9 years with overlapping data. To reduce the influence from the environmental position and also to partly compensate for the different tree sizes and ages, standardized values for Δ13C and BAI were calculated. For each tree, the interannual mean was calculated and then each yearly value was expressed as the difference from this mean. These differences were then averaged across the trees of the two sexes or for all trees. These standardized values did reveal correlated patterns of Q, Δ13C and BAI. All displayed relatively higher values in the mid-1990s and lower values during the drought interval of 2000 and 2001, and then subsequent rebounded in the mid-2000s (Fig. 5). There was subsequently greater than one-half correspondence between Q and Δ13C or BAI (r2 = 0.525, P = 0.027 and r2 = 0.512, P = 0.001), and about one-third correspondence between Δ13C and BAI (r2 = 0.355, P = 0.090), for the combined male and female trees. With the greater response in the males, the association was apparently stronger (BAI and Δ13C – males: r2 = 0.400, P = 0.068; females: r2 = 0.277, P = 0.146).


For the first hydrological linkage, there was strong correspondence between the PDO and headwater river flows over the past century, a result consistent with prior analyses (Rood et al. 2005, 2008; St. Jacques et al. 2010). The pattern extended downstream and the interannual differences in river flows were amplified by water withdrawal, which was proportionally greater during low-flow intervals. The hydrological linkage was further extended to the groundwater zone of the floodplain woodland, as river stage was very closely associated with groundwater table depth. The observed diurnal fluctuation in the water table surface also confirmed that these riparian cottonwoods were phreatophytic, obtaining their moisture from the capillary fringe above the saturated phreatic zone (Butler et al. 2007; Lautz 2008; Rood et al. 2011). Our study thus supported the sequential hydrological linkages from the climatic oscillation to the floodplain groundwater.

The subsequent analyses showed clear and consistent interannual patterns in growth rate across the cottonwood trees. The results clearly supported our primary hypothesis (H1) that there would be positive correspondence between river flows and cottonwood growth. The RI and BAI patterns demonstrated increased radial growth rates during and following the high-flow years of 1995 and 1996, and subsequent reduction as the river flows decreased to the very low flows of 2000 and 2001.

Growth corresponded best with river flows of the growth year and the prior year, indicating multiple-year influence on physiology and growth. Further, the strongest association was displayed for river flows during the cottonwood growth season from May to October and this also applied to the prior year. This suggests that the influence of prior-year river flow was related to the physiological condition of the cottonwoods rather than resulting from a broader recharge of the alluvial aquifer, which would occur year-round (Shepherd, Gill & Rood 2010).

Prior studies of the association between stream flows and the growth of riparian cottonwoods in dry regions have provided variable results (reviewed in Willms et al. 1998; Rood et al. 2003). Some studies did reveal positive correlations between river flows and trunk growth (Stromberg & Patten 1990; Willms et al. 1998) while other studies indicated complacent growth, with limited, uncoordinated interannual variation (Reily & Johnson 1982; Dudek, McClenahen & Mitsch 1998; Disalvo & Hart 2002). Our new results, combined with the prior analyses, indicate that in some years, floodplain water supply is not limiting to cottonwood growth and this would be consistent with the virtually unlimited alluvial groundwater. Conversely, cottonwood growth was limited in low-flow conditions, and consistent with this, cottonwood growth declines following artificial and substantial depression of the alluvial groundwater table (Stromberg, Tiller & Richter 1996; Scott et al. 1999; Horton et al. 2001; Amlin & Rood 2003; Hultine, Bush & Ehleringer 2010).

The association between reduced river flows and cottonwood growth would indicate a physiological response due to drought stress and this was supported by lower wood Δ13C during the low-flow interval of 2000 and 2001; this may reflect stomatal closure. Prior studies with riparian cottonwoods and box elder have also consistently revealed that wood or cellulose Δ13C is lower during drier intervals or with drier environments (Leffler & Evans 1999, 2001; Ward, Dawson & Ehleringer 2002; Potts & Williams 2004). Interestingly, changes in Δ13C have been observed without substantial changes in growth patterns, suggesting high sensitivity in this integrative physiological indicator (Dawson et al. 2002; Ward et al. 2002; Rasheed et al. 2011).

Our second hypothesis (H2) proposed an association between the positions of the trees and effects on growth and Δ13C, due to accompanying differences in environmental conditions. This was confirmed, with strong associations between Δ13C and distance from the river, and elevation above the river, both of which would influence the proximity of the tree to the alluvial groundwater. It was surprising that distance and elevation were not strongly correlated in this riparian woodland. However, the floodplain consisted of a relatively flat surface with localized depressions and rises and we would expect that in many other riparian zones, there would be correspondence between distance and elevation, as the ground level would progressively rise with increasing distance from the river. Following from our present study, we would encourage future researchers to measure both distance and elevation, to better analyse proximity to alluvial groundwater.

The proximity to groundwater is an indication of water supply to the specific trees (Stromberg et al. 1996; Scott et al. 1999), while differences in evaporative demand are probably related to canopy exposure. There would be some sheltering and shading with decreasing exposure and this could reduce the water demand, consistent with the observed negative correlation between Δ13C and canopy exposure. However, it would also be likely that while increasing stand density would provide shading and shelter that would reduce transpirational water loss, associated root competition could reduce water uptake. It would be difficult and destructive to analyse cottonwood root distributions, but there could be study opportunities that benefit from natural hydraulic excavation along river cut-banks (Rood et al. 2011). In relation to the trade-offs between shoot and root interactions, we might predict a pattern whereby some canopy shading and shelter would be beneficial to a tree's water balance, but higher trunk densities would be unfavourable due to competition for water, nutrients and light (Holmgren, Scheffer & Huston 1997).

Our final hypothesis (H3) proposed differentiation between the male and female trees. Prior research into sex differentiation in ecophysiological traits of dioecious riparian shrubs and trees, including willows (Salix sp.), cottonwoods and box elder (A. negundo), has shown that physiological differences are generally subtle (Dawson & Ehleringer 1993; Letts et al. 2008; Xu et al. 2008b; Hughes et al. 2010; Rood et al. 2010). Greater physiological differentiation probably follows from cumulative differences and consequently there is some spatial segregation of the sexes, and sex ratios are unequal within some environments (reviewed in Hultine et al. 2007; Ueno, Suyama & Seiwa 2007; Hughes et al. 2010). The prior studies are fairly consistent in indicating that males are more prevalent in higher and drier positions and better adapted to drought, while females are often more abundant and display superior performance in lower and wetter environments (Hultine et al. 2007; Xu et al. 2008a; Nielsen et al. 2010).

With respect to our third hypothesis, we anticipated that due to increased drought tolerance, the males would display less growth inhibition than females during an interval with low stream flows. This was not observed, and there was actually differentiation in the opposite direction, with greater change in the males in 2000 and 2001. Thus, the response of the males was apparently more conservative, and this is consistent with the lower Δ13C. This suggests increased stomatal closure, possibly higher water use efficiency, and slower growth rates in the males during the low-flow interval. In contrast, the females displayed less responsivity to the drought interval and they might thus be characterized as more opportunistic. This sex differentiation could be important for the increased drought adaptation of the males, as their more conservative physiological responses could reduce their vulnerability to drought-induced xylem cavitation and mortality (Tyree et al. 1994; Sparks & Black 1999; Hultine et al. 2007). We thus conclude that cottonwood males may display increased drought avoidance as part of their increased drought adaptation.

Our observed higher responsivity to drought of Δ13C in the male cottonwoods in a native riparian grove is somewhat consistent with the findings of Xu et al. (2008b) following a greenhouse study of Populus cathayana originating from riparian zones. Corresponding differentiation of males and females with respect to growth and δ13C across water treatments was also reported in an irrigation study for the dioecious box elder, although in that case, the differentiation was expressed in wet rather than dry conditions (Ward et al. 2002). Thus, there have been indications that males of riparian trees display lower Δ13C in some environments, a finding consistent with our conclusion that males display a more conservative strategy relative to drought stress.

With respect to growth differentiation across the sexes, our results within the native riparian zone along the Oldman River are somewhat contrary to some findings following drought imposition on poplars in controlled environments. In some studies, male poplars have displayed greater growth than females during artificial drought (Xu et al. 2008b; Chen et al. 2010), although there have also been complicating interactions with temperature (Xu et al. 2008b). In reviewing the collective literature, there is strong evidence that for dioecious riparian trees and shrubs, males display increased drought adaptation (Hultine et al. 2007; Hughes et al. 2010), but the particular adaptive strategy – drought avoidance versus drought tolerance – remains somewhat uncertain.

River regulation as an analogue for climate change

A further aspect was important within the study, namely the amplification of the climatic oscillation with diminished river flows due to water diversion offstream for irrigation. This was prominent in southern Alberta especially in 2000 and 2001 (Rood & Vandersteen 2010), and there are compounding effects because irrigation demand and water withdrawals are increased when conditions are warm and dry. These withdrawals reduced the Oldman River flows to levels well below those that would have naturally occurred during the warm PDO phase, and this exaggerated the drought effect on the riparian cottonwoods.

Analyses of historic trends of river flows from the central Rocky Mountains have revealed gradual declines in annual flows and particular declines in late summer flows (Rood et al. 2005, 2008). The instream flow decline due to water withdrawal has provided hydrologic conditions that might be expected to occur in the future if river flows continue to decline due to climate change. Thus, the growth reduction observed in the Oldman River cottonwoods through 2000 and 2001 might become more typical in the future due to climate change, and even more severe and prolonged low-flow conditions would occur due to the combination of reduced headwater inflows due to climate change, and increased water diversion for irrigation and other human uses. Our present study revealed the strong correspondence between river flows and cottonwood growth and water relations, but this involved an acute, 2 year drought event. With more extreme, longer term or repetitive drought, chronic responses would accumulate. Following our present study, we predict that there would be increasing differentiation of response between the sexes, leading to further spatial segregation of the sexes, and unbalanced sex ratios (Braatne et al. 2007; Hultine et al. 2007; Nielsen et al. 2010).


This study was funded by the Alberta Water Research Institute and the Natural Sciences and Engineering Research Council of Canada (NSERC). We extend thanks also to Ian Ouwerkerk and Basil Jefferies for field assistance, to Alberta Environment for the provision of data and interest, and to David Ellis (City of Lethbridge) and Coreen Putman (Helen Schuler Nature Centre, Lethbridge) for interest and property access.