Volume 126, Issue 5 e2020JG005873
Research Article
Free Access

Contrasting Growth Response of Jack Pine and Trembling Aspen to Climate Warming in Quebec Mixedwoods Forests of Eastern Canada Since the Early Twentieth Century

Emmanuel A. Boakye

Emmanuel A. Boakye

Chaire Industrielle CRSNG-UQAT-UQAM en Aménagement Forestier Durable, Institut de Recherche Sur Les Forêts, Université du Québec en Abitibi-Témiscamingue (UQAT), Rouyn-Noranda, QC, Canada

Contribution: Conceptualization, Formal analysis, Writing - original draft

Search for more papers by this author
Yves Bergeron

Yves Bergeron

Chaire Industrielle CRSNG-UQAT-UQAM en Aménagement Forestier Durable, Institut de Recherche Sur Les Forêts, Université du Québec en Abitibi-Témiscamingue (UQAT), Rouyn-Noranda, QC, Canada

Forest Research Centre, Université du Québec à Montréal, Montréal, QC, Canada

Contribution: Conceptualization, Supervision

Search for more papers by this author
Martin P. Girardin

Martin P. Girardin

Natural Resources Canada, Canadian Forest Service, Laurentian Forestry Centre, Québec City, QC, Canada

Contribution: Conceptualization, Supervision

Search for more papers by this author
Igor Drobyshev

Corresponding Author

Igor Drobyshev

Chaire Industrielle CRSNG-UQAT-UQAM en Aménagement Forestier Durable, Institut de Recherche Sur Les Forêts, Université du Québec en Abitibi-Témiscamingue (UQAT), Rouyn-Noranda, QC, Canada

Southern Swedish Forest Research Centre, Swedish University of Agricultural Sciences, Alnarp, Sweden

Correspondence to:

I. Drobyshev,

[email protected]

Contribution: Conceptualization, Supervision

Search for more papers by this author
First published: 30 April 2021
Citations: 4


Forest monitoring studies show contrasting trends in tree growth rates since the mid-twentieth century. However, due to their focus on annual and decadal dynamics, they provide limited insight into the effects of long-term climatic variability on tree growth. Here, we relied on a large tree-ring dataset (∼2,700 trees) of two common North American shade-intolerant tree species, trembling aspen (Populus tremuloides Michx.) and jack pine (Pinus banksiana Lambert), to assess their lifespan-long growth dynamics in the mixedwood forests of Québec. We also determined how the environmental conditions of the stands influenced tree growth. We observed a significant increase in the radial growth rate of trembling aspen during the study period, while the jack pine decline was not significant. Over the whole study region, the trees growing in sites with lower competition, and those at the lower sections of the terrain slope experienced more of the positive effects of temperature on growth rates. Our study suggests that the tree growth response to climate warming may be species-specific and will vary across the boreal mixedwoods.

Key Points

  • Climate warming impacts long-term tree growth in the boreal mixedwoods

  • Growth response to climate warming is species-specific

  • Stand conditions modulate the tree growth responses to climate

1 Introduction

In Canada, the annual temperature has increased at the rate of 0.3°C–1.0°C/decade since the 1970s (Buermann et al., 20132014; X. Zhang et al., 2019). The changes in the length of the growing season, the atmospheric demand for water, and the precipitation that accompany warming have had undeniable effects on the growth rate of trees and forest productivity (Buermann et al., 20132014; Girardin, Bouriaud, et al., 2016; Price et al., 2013). Changes in tree growth affect the net uptake of carbon, biochemical cycles, and ecosystem services (Payne et al., 2019; Silva et al., 2010). Knowledge of tree growth responses to climate warming supports forestry planning, specifically—the management of the forest as a potential carbon sink (Boucher et al., 2018; Payne et al., 2019; Silva et al., 2010).

Repeated measurements on permanent forest plots and analyses of dendrochronological and satellite records have shown contrasting trends in tree growth and forest productivity in the boreal and temperate biomes (Marchand, Girardin, Gauthier, et al., 2018). Increases in growth have been attributed to the increasing photosynthetic rates due to elevated atmospheric CO2 (Fang et al., 2014; Hember et al., 2017; Marty et al., 2017; Pretzsch et al., 2014; Stinziano et al., 2014) and increased temperature and precipitation (Boulanger et al., 2017; Hember et al., 2017; Luo et al., 2019). However, a decline in growth rates, reported for the Canadian boreal shield, Western Canada, and Alaska, has questioned the positive effects of the recent climate trends and increase in atmospheric CO2 on growth (Cahoon et al., 2018; Chen et al., 2016; Girardin et al., 20142016; Silva et al., 2010). Declining growth rates have been related to a higher incidence of periods with extreme temperatures and low precipitation (Cahoon et al., 2018; Hogg et al., 2008). These studies, however, covered a limited period and focused mainly on annual and decadal growth dynamics, which provide limited information on the long-term effects of climatic variability on tree growth.

Tree-ring analysis is a powerful tool that is used to assess the response of growth rates to climate variability across a range of temporal frequencies (Béland et al., 1992; Gillis et al., 2005; Johnson & Abrams, 2009). Tree-ring sampling procedures, however, can introduce various biases in such estimations, since the sampled trees are not always representative of the general tree population in terms of their age or size. Sampled individuals from older age classes typically represent slow-growing trees, whereas younger classes commonly represent fast growing trees (Duchesne et al., 2019; Hember et al., 2019). A particularly strong bias may arise from the variation in recruitment histories: trees representing a particular age class may predominantly recruit into the canopy following a stay in the understory (so-called advanced regeneration) while other age classes may reflect a wave of post-disturbance establishment followed by initial growth under the conditions of reduced competition. Such variability in regeneration histories could potentially mask the growth response to changes in climate (Duchesne et al., 2019; Hember, Kurz, & Girardin, 2019; Ung et al., 2001). Competition among trees can further alter tree response to climate variability, by modifying the availability of light and soil resources (Marchand et al., 2019). Untangling climatic versus nonclimatic effects in tree growth are, therefore, a nontrivial but important element in the analyses that help to correctly quantify the effect of climate on growth.

Site conditions modify the response of tree growth to regional climate. For example, differences in soil drainage can affect the sensitivity of tree growth to variation in precipitation. In the boreal forests of Ontario, where soil moisture is generally abundant, temperature rise during the 1950–2007 period has enhanced tree growth rates (Luo et al., 2019; Silva et al., 2010). However, trees on sites with limited available soil moisture have experienced reduced growth. The large spatial and temporal variability in growing conditions, together with the confounding effects of competition, complicates partitioning between regional climatic and stand level growth responses (Marchand et al., 20182019).

In this study, we examined the century-long trends in the dynamics of growth rates of trembling aspen (Populus tremuloides Michx.) and jack pine (Pinus banksiana Lambert) that had regenerated between the years, 1900 and 2000. Both species are shade-intolerant, typically regenerate following canopy disturbances, and experience limited variability in light conditions during their lifespans (Béland & Bergeron, 1996; Strimbu et al., 2017; Ung et al., 2001). The species' shade-intolerance made them good models to study the effects of long-term climate variability on growth since the variation in the growing conditions at the beginning of their lifespans was low. In particular, the trees, which reached the forest canopy and were eventually sampled during the inventories, likely experienced the limited variability in light environments. This facilitated the comparison of tree growth rates among cohorts. To further enhance the climate signal in tree growth, we focused on the Eastern North American Temperate and boreal mixedwoods, an ecosystem that is known to be sensitive to changes in regional climate regimes (Bergeron et al., 2014; Payne et al., 2019). We tested two hypotheses: (H1) the growth rates of jack pine and trembling aspen increased from the nineteenth century until the late twentieth century; and (H2) stand conditions (temperature, climate moisture index [CMI], terrain slope, and inter-tree competition) modified the long-term trends in the growth of jack pine and trembling aspen.

2 Material and Methods

2.1 Study Area

The study was conducted in the mixedwood region of Québec, Eastern Canada (Figure 1), extending from 47°N to 49°N and from 62°W to 79°W. The western portion of the study area has a continental climate, while the eastern portion experiences a maritime climate (Reyes et al., 2013). The continental climate is drier (the average total annual precipitation 866.6 mm; the annual mean temperature ∼2°C), whereas the maritime climate is wetter (average total annual precipitation 1,050.4 mm; annual mean temperature ∼3°C) (Aussenac et al., 2017; Bergeron et al., 2014; Reyes et al., 2013). The topography in the west is relatively flat to rolling and becomes increasingly hilly and montane toward the east. The main soil superficial deposit is till (Bergeron et al., 2014). The soil drainage regimes vary from xeric to hydric (Duchesne et al., 20162017).

Details are in the caption following the image

The study area in the mixed forest zone of Québec, bioclimatic domains, and the locations of sampled plots. The names of bioclimatic domains reflect the dominant vegetation during the late seral stages.

The main tree species in the boreal mixedwoods are white birch (Betula papyrifera Marsh.), trembling aspen (P. tremuloides Michx.), balsam fir (Abies balsamea [L.] Mill.), white spruce (Picea glauca [Moench] Voss) and eastern white cedar (Thuja occidentalis L.). In contrast, the temperate mixedwoods are dominated by balsam fir and yellow birch (Betula alleghaniensis Britt.) (Girard et al., 2004). Aspen and jack pine (P. banksiana Lambert) often mark areas subjected to fires in the mixedwood zone (Bergeron, 2000; Bergeron et al., 2014). Fire and insect outbreaks are the dominant disturbance factors that control species distribution and successional pathways in this biome (Bergeron, 2000; Brandt et al., 2013). Spruce budworm (Choristoneura fumiferana Clemens, SBW) is the common factor of canopy tree mortality for black spruce (De Grandpré et al., 2019) whereas, outbreaks of forest tent caterpillar (Malacosoma disstria Hübner; FTC) can cause the mortality of aspen (Brandt et al., 2013).

2.2 Data Collection

The data originated from permanent and temporary circular plots managed by Québec's Ministère des Forêts, de la Faune et des Parcs (MFFP). The plots had the radii of 11.28 m, and the sampling is based on a stratified randomized design with the proportional allocation of samples, according to the surface area of each stratum (MFFP, 2016a2016b). The different strata were determined from aerial photographs based on stand characteristics (composition, density, height, and age), edaphic properties (slope, drainage, and deposit), and the disturbance history (MFFP, 2016a2016b).

For each plot, the following information was available: diameter at breast height (DBH) of every tree, the composition of the understory, slope, soil, and the type of superficial deposits (Table 1; MFFP, 2009; MRN, 2001). We used data collected between 1970 and 2012.

Table 1. Summary of Generalized Additive Mixed Model of the Effect of Tree Size, Age and Stand Environmental Variables on the Growth Rates of Jack Pine and Trembling Aspen
Characteristic Jack pine Trembling aspen
Leaf type Needleaf Broadleaf
Number of trees 1,454 1,246
Max tree age (years) 188 130
Min tree age (years) 11 12
Mean values (±SE)
BAI (cm2 year−1) 4.60 ± 0.01 5.84 ± 0.02
Basal area (cm2) 331.09 ± 0.88 479.23 ± 1.44
CI (m2ha−1) 7.52 ± 0.02 9.54 ± 0.03
Slope (%) 10.37 ± 0.12 10.25 ± 0.04
TMean (°C) 13.08 ± 0.00 13.24 ± 0.00
TMax (°C) 18.31 ± 0.00 18.55 ± 0.00
CMI (mm) 8.61 ± 0.03 5.05 ± 0.03
Model fit
Deviation explained (%) 51.30 58.20
RMSE 0.27 0.25
Parametric coefficients (±SE, P-value)
log(BA) 0.60 ± 0.03, <0.001* 0.31 ± 0.03, <0.001*
Significance of smoothing terms (edf, F, P-value)
Cambial age 2.00, 643.32, <0.001* 2.00, 1598.72, <0.001*
TMax* CI 4.24, 3.14, 0.063ns 6.59, 2.18, 0.018*
CMI* CI 3.69, 2.70, 0.018* 3.24, 0.56, 0.404ns
TMax* Slope 2.87, 6.58, 0.000* 1.00, 1.83, 0.176ns
CMI*Slope 3.02, 5.13, 0.002* 3.00, 3.76, 0.010*
Subject tree 1344.80, 13.23, <0.001* 1126.66, 22.95, <0.001*
  • Note. CI, competition index; TMean, growing season (May–September) mean temperature of site; TMax–growing season maximum temperature of site, CMI, growing season climate moisture index; BA, tree basal area; edf, estimated degrees of freedom; SE, standard error. P-values indicate the significance of the effect, with the star indicating values of P lower than 0.05, “ns” indicates nonsignificant effects. BAI, basal area growth; CMI, climate moisture index; RMSE, root mean square error.

In each permanent plot, up to nine trees were cored, five of which were selected randomly. Two trees were selected randomly among the four biggest trees (in terms of diameter at breast height, DBH) of the dominant species. One additional tree was selected to have a diameter closest to the mean diameter of the dominant tree species, and one more tree was selected to have a basal area at breast height closest to the 30th percentile of the distribution of stem basal area for the dominant species. In each temporary plot, three trees (DBH > 90 mm) were selected: one tree was selected randomly among all trees within the plot, another was selected randomly among the four biggest trees (in terms of DBH) of the dominant species, and the third had a diameter closest to the mean diameter of the dominant tree species (Duchesne et al., 2017). These trees were cored at the height of one meter. In 2012, the database included tree-ring measurements of close to 365,000 trees from 42 tree species covering the entire boreal and temperate regions of Québec (Duchesne et al., 2019). In this study, we analyzed a total of 1,454 jack pine trees and 1,246 trembling aspen trees, representing the mixedwood forest zone (Figure 1).

Tree cores were dried, glued to wooden holders, and sanded successively with 120, 220, and 320 grit sandpaper to obtain a smooth surface. Annual rings were measured with the WinDendro Image Analysis System for tree-ring measurement (Regent Instruments Inc, Canada) at a resolution of 1,000 dots per inch (Duchesne et al., 2017). A calendar year was attributed to each ring. Since these data have been developed for gaining information on forest productivity, they have not gone through a rigorous standard cross-dating process that is usually carried out in dendrochronological studies (MRN, 2001; MFFP, 2009). In order to test if the noncross-dated tree ring data has an impact on the estimation of long-term growth trends, we compared the results of a sample of the tree ring data to that of the geographically nearest dated reference chronologies in Western Québec (Bergeron, 2000; Hofgaard et al., 1999). We observed that in the same period both the cross-dated and noncross-dated tree ring data produced a similar growth trend (Figure S1). We would like to note that the use of noncross-dated tree ring series for the assessment of decadal to century-long growth trends has similarly been used in tropical (Fichtler et al., 2003; Groenendijk et al., 2014) and boreal regions (Duchesne et al., 2019).

2.3 Acquisition of Site Level Variables

We obtained historical monthly weather data for all plots, using an interpolation of data from Environment Canada's weather stations network within the BioSIM 10 software (Régnière et al., 2014) over the period between 1950 and 2012. This period was common across trees' lifespans and was believed to feature the data of the highest quality. To obtain the climatology of each plot, we averaged the following climate variables for each plot and growing season (May–September) over 1950–2012: mean maximum temperature (°C), mean total precipitation (mm), and CMI. CMI characterizes the available moisture and is the balance of monthly potential evapotranspiration (PET) and monthly precipitation (Preci) over a time period i in mm of water, CMI = Precii-PET i (Berner et al., 2017; Hogg, 1997). The CMI is a hydrologic index and is typically well-correlated with tree growth in boreal and temperate forest ecosystems (Berner et al., 2017; Hogg et al., 2013). PET was estimated with the R package SPEI using the Penman–Monteith algorithm with inputs of monthly average daily minimum and maximum temperature, latitude, incoming solar radiation, the temperature at the dew point, and altitude (Vicente-Serrano et al., 2010).

Terrain slope for each plot was originally recorded in classes. We converted them to quantitative values using the lower range value (e.g., 9%–16% was replaced with 9%).

To account for the effect of competition for resources from neighboring trees on the growth of sampled trees, we calculated a plot-level competition index (CI) as the sum of all individual basal areas for trees with DBH > 1 cm in the plot and scaled it to a hectare (m2 ha−1). The choice of CI was based on the results of an earlier study (Huang et al., 2013), which has indicated that the sum of basal areas could act as a simple yet robust representation of competition effects in the studied forests. We calculated the CI (CI, m2ha−1) using the average of the estimates of the basal areas from 1970 to 2012 (permanent plots), and 1980–1993, 1992–2003, and 2004–2012 (temporary plots) inventory data. The distribution of environmental and stand variables for both species are shown in Figure S2.

2.4 Statistical Analyses

We converted ring-width data into chronologies of basal area increments (BAIs) for each tree using the formula for the area of a circle: BAI = πR2t- πR2t-1 where R is the tree radius and t is the year when the ring was formed. Yearly growth rates were expressed as cm2 year−1. We used bai.out function in R package dplR to compute BAI (Bunn, 2008; R Core Team, 2018). Rings that formed during the first 10 years were removed, since at that age the climatic signal in growth tends to be weak due to the strong effect on growth exercised by the local conditions, such as snow loads, browsing, light/drought conditions, and proximity to taller trees. Removing rings representing the first decade in a tree's lifespan was also expected to remove sections of chronologies with the highest probability of dating mistakes (Marchand et al., 2019). As the initial exploratory step, we analyzed basal area growth rates as a function of the calendar year and cambial age using the interp function in the R package akima (Akima & Gebhardt, 2016). This bivariate function constructed a smooth interpolated surface of the median values of logarithmically scaled BAI (logBAI). This step allowed us to get a graphical representation of variation in BAI along the lifespan of the trees.

At the second step, we quantified the drivers of tree growth, using the Generalized Additive Mixed Model (GAMM) (Wood, 2017). We used GAMM to model the basal growth rate of tree j in a site k at specific year t as a function of tree basal area (variable BA) and cambial age (Age). We computed BA as the inner-bark basal area of tree i at specific year t. Age was the cambial age (1-m height ring count) of tree i at year t. The inclusion of BA and Age in the model helped us to retain low-frequency variation in chronologies (Dietrich & Anand, 2019).

To understand the influence of stand environmental conditions on growth, we included in the GAMM model the following variables as growth predictors: competition with neighboring trees (CI), terrain slope, maximum growing season temperatures (TMax), and the CMI. The full fitted GAMM was:

In this expression, β (log(BAjkt)) is the parametric portion of the model, where β is the vector of the parameter associated with logarithmically transformed tree inner-bark basal area (BA). The nonparametric part of the expression is formed by ƒ, which is a smoothing function for Age effect resolved at tree level, stand, and environmental effects. As for smoothing functions, we used cubic regression splines. The interaction between each of the two variables (TMax and CMI) and terrain slope and CI was modeled with a tensor (te) product function to address the different units in which temperature, precipitation, and terrain slope had been measured (Wood, 2017). To avoid over-fitting, we set the number of knots for the splines at the low value of 3. The subject tree identity (Zjk) was used as a random effect in the GAMM. We included an auto-regressive (AR(1)) term to account for the temporal correlation of the data. The results from the GAMM growth model are associated with estimated degrees of freedom (EDF) that indicated whether the relationship was linear (EDF = 1) or nonlinear (EDF > 1). GAMM was realized in R package mgcv (Wood, 2017).

For each species, the full model was fit on a randomly selected subset of the database, comprising 80% of the trees of a studied species. We computed a correlation matrix among the stand and climatic variables to ascertain if multicollinearity will influence the growth model (Figure S3). We also evaluated concurvity of the smoothing variables in the growth model terms. Concurvity refers to the degree to which a smooth model term can be approximated by one or more smooth model terms (Johnston et al., 2018). Like multicollinearity in a linear modeling framework, concurvity can lead to instability of the estimated coefficients of the smoothing terms in GAMMs. The concurvity index is calculated on a scale of 0–1, with 0 indicating no concurvity and one indicating high concurvity (Figure S4; Morlini, 2006). We further tested to know how each climate variable performs separately in competing for GAMM for each species (Table S1). We observed that the full model with both CMI and TMax had the lowest Akaike's information criteria with a small sample bias adjustment (AICc).

The predictive capacity of the final model was validated using the remaining 20% of the trees to compute the root mean square error (RMSE) of predicted versus observed growth rates. The diagnostic plots of the GAMM model used to assess if the data give a reasonable description of the relationships for estimating growth for each species are in Figure S5. The GAMM model predicted BAI on a natural logarithmic scale. We back-transformed the prediction to the original scale (cm2 year−1), by following the procedure described by Girardin et al. (2016). We weighted the difference between observed (O) BAI (OBAI, cm2) and predicted (P) BAI (PBAI, cm2) by the PBAI and expressed it as a percentage:

We estimated the long-term growth trend as a slope of the linear regression of BAI over the time period (1950–2012). This period is marked by significant increases in mean annual temperatures in eastern Canada (Price et al., 2013).

2.5 Methodological Considerations

Estimating growth trends indirectly from growth patterns using Akima plots facilitates the understanding of growth dynamics over time and the identification of its driving factors. Although it is not based on a statistical model, an Akima plot is an illustration of the time-evolving plot commonly used in dendrochronology, making it possible to derive the growth rate at any given cambial age and a calendar year (Lenz et al., 2014).

The growth dynamics of a species may be attributed to the periodic oscillations in the climate system (Girardin 2011), tree ontogeny (tree age and size), and inter-tree interactions that all can distort the climatic signal in growth chronologies (Marchand et al., 2019). Nonlinear models, such as the general additive mixed effect model (GAMM) have the potential to minimize the fixed effect of tree age and size and adequately quantify effects of climate on growth (Dietrich & Anand, 2019; Marchand et al., 2019). While GAMM has more predictive power or is more flexible than typical statistical techniques employed in tree-ring science, we do caution that the amount of dating error will be reflected in the overall noise and goodness-of-fits. However, the predictor's CI, Slope, CMI, and TMax are here all time-irrelevant, which means that the impact of dating errors is not that important. For the other two time-related predictors, Age and BA, the shift of the year-recognition of the whole series also has no impact on the average growth rate.

3 Results

3.1 Growth Temporal Patterns

The akima plots (Figure 2) revealed visible differences in the pattern of variation of the basal area growth rate (logBAI) between jack pine and trembling aspen over time. For jack pine, the plot showed a highly fluctuating growth rate over time: for the cambial ages 11–111 years, the growth revealed an ∼20-year periodicity between 1840 and 1950 (Figure 2a). Higher growth rates were observed for cambial ages between 11 and ∼31 years and between ∼111 and ∼171 years in the post-1950s period. For aspen, the growth rate at the cambial ages of ∼11 and ∼31 years was lower prior to1970 than post-1970 (Figure 2b). For aspen trees with cambial ages greater than 31 years, we observed a relatively constant growth rate over the period between 1960 and 2012.

Details are in the caption following the image

The dynamics of log-transformed basal area growth rates (logBAI) in relation to the cambial age and the calendar year chronology for trembling aspen and jack pine. The color scale denotes the median values of logBAI with low values being in purple and high values—in green. The number of tree rings analyzed as a function of time is shown in the inserts. BAI, basal area increments.

3.2 Stand and Environmental Drivers of Growth

The GAMM models (Table 1) for jack pine (Figure 3) and trembling aspen (Figure 4) had an identical list of predictors, which included the inner-bark basal area (a proxy of tree size), cambial age, terrain slope, CI, mean growing season CMI and maximum temperature. There were relatively low absolute correlation (<0.6) and estimated concurvity (<0.3) among the model individual variables, suggesting weak functional relationships (i.e., low collinearity) among them. Each species-specific GAMM had a RMSE lower than 0.5, indicating a good predictive skill of the models (Table 1).

Details are in the caption following the image

The isolated effects of tree size (expressed as basal area, BA), age, and stand environmental variables on the growth rate of jack pine. Stand variables are the competition index, expressed as the sum of basal areas for all trees in the plot, scaled to a hectare (m2ha−1), terrain slope (%), maximum temperature (TMax, °C), and climate moisture index (CMI, mm) of the growing season (May–September). In the plots A and G, the OX axis represents a covariate, while the OY axis represents effect values, labeled s or f (cov, edf), where cov is the covariate name, and edf is the estimated degrees of freedom of the smooth. Plot F shows a diagnostic quantile-quantile plot of the estimated random effects versus Gaussian quantiles. Dash lines on effect plots denote the 95% confidence intervals. In the plots B–E, the axes represent the covariates. Colored areas in plots B–E refer to the regions in two-dimensional covariate space with significant effects (p < 0.05) interactions.

Details are in the caption following the image

The isolated effects of tree size (expressed as basal area, BA), age, and stand environmental variables on the growth rate of trembling aspen. Stand variables are the competition index, expressed as the sum of basal areas for all trees in the plot, scaled to a hectare (m2ha−1), terrain slope (%), maximum temperature (TMax, °C), and climate moisture index (CMI, mm) of the growing season (May–September). In the plots A and G, the OX axis represents a covariate, while the OY axis represents effect values, labeled s or f (cov, edf), where cov is the covariate name, and edf is the estimated degrees of freedom of the smooth. Plot F shows a diagnostic quantile-quantile plot of the estimated random effects versus Gaussian quantiles. Dash lines on effect plots denote the 95% confidence intervals. In the plots B–E the axes represent the covariates. Colored areas in plots B–E refer to the regions in two-dimensional covariate space with significant effects (p < 0.05) interactions.

Tree cambial age (Table 1) had a significant nonlinear effect on the growth of jack pine (Figure 3a) and trembling aspen (Figure 4a). It had a positive effect during the first ∼60 years of growth (juvenile stage), after which the age-growth relationship flattened for another ∼40-year period (peak growth). Following the peak growth, age had a negative effect on growth (i.e., a growth decrease with age after ∼100 years). For both species, the inner-bark basal area had a significant and positive effect on growth rate (Table 1; Figures 3g and 4g).

Stand environmental variables were strong predictors of the growth rate of jack pine and trembling aspen. The interaction between maximum temperature (TMax) and competition (CI) has a significant effect on the growth of trembling aspen (Figure 4b), but not jack pine (Figure 3b). The interaction between CMI and CI on the other hand has a significant effect on the growth of jack pine (Figure 3c) but not the trembling aspen (Figure 4c). Tree growth was higher in sites with higher CMI and lower competition.

The interaction between TMax and terrain slope has a significant effect on the growth of jack pine (Figure 3d) but not the trembling aspen (Figure 4d). Tree growth was lower on steep slopes and higher on flatter parts of landscapes with higher temperatures. The interaction between CMI and terrain slope also has a significant effect on the growth of both jack pine (Figure 3e) and trembling aspen (Figure 4e). For jack pine trees, the growth was relatively higher at the lower sections of the slope and at lower CMI. Trembling aspen growth was, however, higher at the lower sections of the slope, and at higher CMI.

3.3 Quantification of Growth Trends

We observed contrasting temporal growth trends between jack pine and trembling. Jack pine showed an increasing growth trend until the 1960s when growth declined up to 1990. The growth increase occurred between 1990 and 2012 (Figure 5a). Linear regression showed a nonsignificant growth rate (−0.068% year−1 ± [std] 0.010% year−1, p = 0.056, R2 = 0.105) from 1950 to 2012. Trembling aspen showed a constant increase in growth until the 1990s and a slightly decreasing growth rate afterward (Figure 5b). Between 1950 and 2012, linear regression suggested a significant growth increase (0.111% year−1 ± 0.022% year−1, p ≤ 0.000, R2 = 0.291).

Details are in the caption following the image

Temporal variability of basal area growth (BAI) of jack pine and trembling aspen. The Y-axis is the ratio of observed BAI (cm2) to GAMM predicted BAI back-transformed to cm2, and the X-axis is the calendar time. Ratios (red curves) of jack pine (left panels) and trembling aspen (right panels). Gray shading delimits the bootstrapped 95% confidence intervals. The GAM smoothing curve is shown by blue color. The length of temporal variability has been adjusted to cover the period between 1920 and 2012. GAMM, general additive mixed model.

4 Discussion

Our study documented generally increased growth of trembling aspen and stable growth of jack pine in the mixedwood forests of Québec between 1950 and 2012, although we observed a strong variability in growth rates. The accelerated growth of trembling aspen and the lack of a long-term trend in jack pine are consistent with trend assessments from the forest inventory in the boreal region of Québec since the second half of the twentieth century (Anyomi et al., 2012; Girardin et al., 20112012). The trend estimates, however, varied across the boreal mixedwood and depended on stand conditions and local climate features.

4.1 Ontogeny Effects on Tree Growth

The observed effect of cambial age on growth rate was consistent with the classical nonlinear sigmoidal model (Table 1, Figures 3 and 4a), suggesting a higher growth rate at the early stages of tree lifespan, constant growth in mature trees, and a growth decline in trees approaching the end of their lifespan (Groover, 2017; Johnson & Abrams, 2009). The increase in growth rate is primarily related to the increase in canopy size and the photosynthetic capacity of the tree. Stabilization of the growth rate at ∼60–100 years (Figures 3 and 4a) reflects an increasing allocation of photo-assimilates toward maintenance functions and marks the onset of the period with relatively constant canopy volume (Groover, 2017; Johnson & Abrams, 2009; Weiner & Thomas, 2001). Further increasing maintenance and defense costs, higher hydraulic resistance with increase in tree height contribute to decline in growth rates in older (>100 years) trees (Figures 3 and 4a) (Weiner & Thomas, 2001).

Tree growth increased continuously with tree size in both jack pine (Figure 3g) and trembling aspen (Figure 4g). The result reflects the fact that these species establish canopy dominance early in stand development, which results from stand-replacing disturbance events benefiting from high sunlight availability (Luyssaert et al., 2008; Matsushita et al., 2015; Stephenson et al., 2014). Being well-exposed to sunlight, tree canopies can, therefore, provide high levels of photosynthesis throughout the tree lifespans (Strimbu et al., 2017; Z. Zhang et al., 2017). BAI models have typically reported a positive effect of size on growth and a negative effect of age, when both covariates have been included in the same model (D'Orangeville et al., 2018; Foster et al., 2016), as in the current study.

4.2 Stand Conditions and Climate-Growth Relationship

The stand environmental conditions altered the tree growth response to climate and often in a nonlinear manner. Trees growing on sites with lower competition, in particular, trembling aspen (Figure 4b) experienced the positive effect of temperature on growth because trees tend to increase growth in uncrowded stands where light energy is sufficiently available (Cavard et al., 2011; Huang et al., 2013). Jack pine (Figure 3c) trees growing on sites with lower competition on the other hand experienced the positive effect of CMI on growth because of the increased moisture availability in the less competitive site. Increased temperatures under no deficit in moisture availability generally favor photosynthetic activity and, subsequently, tree growth (Anderegg et al., 2012; Dietrich et al., 2016).

Jack pine trees (Figure 3d) at the flatter sections of the slope experienced the positive effect of temperature on growth because of the good availability of soil water at the lower portions of the slopes, combined with the increase temperatures, that favored the tree growing there. Nevertheless, because Jack pine typically grows on dry sites, excess moisture can constrain growth and impact negatively on the trees (Figure 3e). Trembling aspen grows in high moisture areas which could have resulted in the more positive effect of CMI at the lower portions of the slope (Figure 4e) (Anyomi et al., 2012; Marchand et al., 2019).

The growth of the trembling aspen (Figure 5b) increased between 1920 and 2012. Climate warming since the end of the Little Ice Age around 1850, coupled with increasing precipitation, likely favored that increase (Bergeron & Archambault, 1993; Bergeron et al., 2002; Drobyshev et al., 2017; Girardin et al., 2013; Girardin & Wotton, 2009). Although trembling aspen is found in predominantly mesic sites where soil moisture is generally considered as not limiting, a combination of a warmer and wetter climate appears beneficial for this species. In line with this observation, enhanced growth rates of trembling aspen have been documented in tundra regions in western Canada (K. Zhang et al., 2008) and in boreal and temperate forests of eastern Canada (Anyomi et al., 2012; Girardin et al., 2013). However, a decreasing trend in growth has been reported for trembling aspen along the entire stretch of southern boreal forests in western Canada from 1983 to 2005 (K. Zhang et al., 2008). These results suggest that the growth of aspen is controlled by both regional trends and variability in site-level conditions within single landscapes/watersheds. Patterns of growth variability at larger geographical scales appear to reflect interactions among regional climatology, biotic agents, and tree growth.

Jack pine did not show a region-wide growth trend over the studied period (Figure 5a), supporting the result of an earlier study done for the 1983–2005 period (Girardin et al., 2012). Jack pine dominates on mesic to xeric soils, where an increase in temperature and a subsequent increase in evapotranspiration demand are more likely to constrain photosynthesis and lead to reduced growth (Anderegg et al., 2012; Dietrich et al., 2016). However, increased growth in this species has been observed in northern boreal Québec (Girardin et al., 2011). This suggests that the growth of trees at the northern fringes of the jack pine population range in Eastern North America is temperature limited.

For both species, the lack of consistency in the study's result may arise from the differences in species demography, variability among stand conditions, analyzed time periods, and the complexity of analytical setups. Competition and site conditions appear to be the most important modifiers of the climate-growth relationships. Selection of shade-intolerant species for this study and the statistical correction of estimates for BAI trends (Dietrich & Anand, 2019; Girardin et al., 2016) both ensured that the trends in growth rate, observed in this study, are climatically driven. We realize that our methodology might not account for the influence of insect outbreaks, one of the main disturbing agents in the North American mixedwoods. This may be particularly true for aspen, which is affected by the forest tent caterpillar. Disregard for this factor could make the direct (e.g., not modified by biotic agents) climate signal appear less strong in tree ring chronologies (Boucher et al., 2018; Brandt et al., 2013; Girardin et al., 2013).

4.3 Management Implications

Our study pointed to species-specific sensitivity of growth to changes in climate, modulated by stand-level environmental variables. This implies that growth trends will likely diverge between the two species in the future, with potentially important implications for regional biogeochemical cycles and forestry. The faster turnover of trembling aspen stands may increase the carbon sequestration capacity of forests with a tangible aspen component. However, it is not clear what the overall effect of the observed trends may be, since higher growth rates will likely be associated with the shorter lifespans of the trees that, in turn, will lower carbon residence time in the living biomass (Büntgen et al., 2019), possibly even limiting the capacity for carbon sequestration in early successional and mixed forests. Shorter rotation periods of aspen stands and an increased production rate of aspen timber should favor the profitability of aspen-oriented forest management programs. The earlier maturation and shorter lifespan would likely cause a decrease in overall wood density, negatively affecting the quality of aspen wood for oriented strandboards, laminated veneer lumber, and production of pulp and paper (Balatinecz & Kretschmann, 2001; Bigler, 2016).

For the jack pine, the slowing growth could increase stand rotation age and increase the contribution of this species to the pool of carbon stored in the living biomass (Bigler, 2016). However, the management programs focusing on this species will likely suffer from lower profitability, although increased wood density, which often accompanies a decrease in growth, may partially offset the drop in the long-term economic value of jack pine stands used for pulpwood and lumber production (S. Y. Zhang & Koubaa, 2008).

The apparent lack of similarity in growth trends between trembling aspen and jack pine should make managers cautious in generalizing on the effects of climate change for long-term planning. For both species, a projected change in the frequency of natural disturbances could increase the incidence of background mortality, which may promote a higher rate of deadwood production and carbon release into the atmosphere through wood decomposition (Boucher et al., 2018; Brandt et al., 2013).


The authors thank Marie-Claude Lambert of the Ministère des Forêts, de la Faune et des Parcs for providing plot data. We thank anonymous reviewers for their insightful comments and suggestions to improve the quality of this manuscript. The authors also thank Xiao Jing Guo for statistical advice. Financial support was provided by the NSERC-UQAT-UQAM Industrial Chair in Sustainable Forest Management (Yves Bergeron) and MITACS Accelerate grant MixCanada (Igor Drobyshev) in partnership with OURANOS, the Consortium on Regional Climatology and Adaptation to Climate Change of Canada.

    Data Availability Statement

    The raw tree ring data are available at “Direction des inventaires forestiers du Ministère des Forêts, de la Faune et des Parcs (MFFP)”. MFFP can be contacted at: [email protected]. Processed data can be downloaded from data.world digital repository at https://data.world/eaaboakye/forest-inventory-data.