Abstract

The impacts of climate change on future soil temperature Ts and soil moisture Ms of northern forests are uncertain. In this study, the authors first calibrated Ts and Ms models [Forest Soil Temperature Model (ForSTeM) and Forest Hydrology Model (ForHyM), respectively] using long-term observations of Ts and Ms at different depths measured at three forest sites in eastern Canada. The two models were then used to project Ts and Ms for the period 1971–2100 using historical and future climate scenarios generated by one regional and five global climate models. Results indicate good model performance by ForSTeM and ForHyM in predicting observed Ts and Ms values at various depths for the three sites. Projected annual-mean Ts at these sites increased between 1.1° and 1.9°C and between 1.9° and 3.3°C from the present 30-yr averages (1971–2000) to the periods 2040–69 and 2070–99, respectively. Increases as high as 5.0°C were projected at the black spruce site during the growing season (June) for the period 2070–99. Changes in annual-mean Ms were relatively small; however, seasonally Ms is projected to increase in April, because of earlier snowmelt, and to decrease during the growing season, mainly because of higher evapotranspiration rates. Soil moisture in the growing season could be reduced by 20%–40% for the period 2070–99 compared to the reference period. The projected warmer and drier soil conditions in the growing season could have significant impacts on forests growth and biogeochemical cycles.

1. Introduction

Global-mean surface temperature has risen at a rate of 0.07°C decade−1 over the last century while precipitation has also generally increased for regions of the Northern Hemisphere (Solomon et al. 2007). Change in climate will continue to be observed as warming of approximately 3°C and increases in precipitation are projected over the next century. Regionally, an increase in annual-mean temperature of almost 4°C and a 7% increase in annual precipitation are projected for 2080–99 for eastern North America (Solomon et al. 2007). For the southern Quebec region, an ensemble of simulations project a warming between 3° and 5°C for 2080–99 and an increase in precipitation, mostly during winter and spring, between 5% and 17% (Logan et al. 2011; DesJarlais et al. 2010).

Projected changes are likely to affect soil moisture and temperature conditions, variables that have important impacts on many important processes such as seed germination and fine root development of plants. Also, low soil moisture is associated with increased forest fire probability, and warm summer temperatures combined with low soil water content may create water stresses (Rennenberg et al. 2006). Decreases in soil water content could also reduce water and CO2 exchanges between the forest and the atmosphere because of stomata closure in order to limit water losses (Gessler et al. 2004; Hernandez-Santana et al. 2009; Rennenberg et al. 2006). In addition, warm and dry soil conditions generally impair nitrate assimilation by trees (Gessler et al. 2004). Furthermore, over extended time periods, changes in soil water content could also affect tree species distributions and forest composition.

Changes in soil temperature and water content may also have indirect impacts on forest dynamics by controlling and affecting many chemical and biological processes of ecological significance in forested ecosystems. For example, changes in soil temperature may alter temperature-dependent biochemical reactions such as soil organic matter decomposition and microbial respiration (Davidson et al. 1998). Conversely, the relationship between temperature and decomposition or respiration is also dependent on soil moisture. If soil water content falls under a certain threshold, soil microorganisms could suffer from desiccation stress, thus limiting their activity (Davidson et al. 1998). Higher soil temperatures may also influence other processes such as the rate of nutrient release from soil organic matter due to microorganism activity (Melillo et al. 2002) or the weathering of soil minerals that controls the release rate of base cations to soil solution (White and Blum 1995; Gislason et al. 2009; Houle et al. 2010). An increase in soil temperature could also potentially produce a fertilizing effect and stimulate forest productivity (Luxmoore et al. 1993; Melillo et al. 2002).

Given the importance of soil temperature and moisture on forest dynamics, there is an obvious need to understand how climate change will affect future soil conditions. However, to date, future climate scenarios specific to soil conditions are limited in number compared to those available for air temperature and precipitation. Changes in soil temperature or moisture cannot be directly inferred from expected changes in air temperature or in precipitation, especially on a seasonal or monthly basis. For instance, greater air temperature increases are projected in winter than in summer, but during winter forest soils are insulated from air temperature by the snowpack. Also, increases in precipitation are projected for the winter season, whereas few changes are expected during the growing season. It is therefore conceivable that, during the growing season, higher air temperatures, combined with only small changes in precipitation, could result in lower soil moisture due to higher evapotranspiration. Despite being plausible, robust modeling procedures are needed in order to test this hypothesis. Although global or regional climate models have land surface schemes from which soil temperature and soil moisture can be diagnosed, their usefulness for directly obtaining predictions for specific forest sites is limited as their coarse spatial resolution typically cannot adequately reproduce crucial site-specific characteristics (presence of organic soils, organic soil depth, B horizons depths, depth of rooting zone, clay content, soil porosity, vegetation type, etc.). Another reason for the scarcity of studies reporting model projections of soil conditions is the rarity of measured observations taken in situ over relatively long periods that could allow reliable model calibration or validation.

In the present study, process-oriented models developed to simulate water [Forest Hydrology Model (ForHyM)] and thermal [Forest Soil Temperature Model (ForSTeM)] fluxes in forested ecosystems (Yin and Arp 1993) were used. These models were selected because they have been built specifically for forest ecosystems, and they have been validated for several temperate and boreal forests (Arp and Yin 1992; Balland et al. 2006; Bhatti et al. 2000; Houle et al. 2002; Meng et al. 1995; Yin and Arp 1993).

To evaluate the potential effects of climate change on forested soils, we first evaluated the ability of ForHyM and ForSTeM to adequately simulate soil moisture and temperature for three midlatitude boreal forests by comparing field measurements with model simulations. Second, climate model projections of the evolution of air temperature and precipitation for the period 1961–2100 were used as driving variables for ForHyM and ForSTeM in order to simulate future soil conditions and assess potential climate change effects on soil moisture and temperature.

2. Methods

a. Site description and data collection

The study focused on three forested sites located within three calibrated watersheds (Lake Clair, Lake Laflamme, and Lake Tirasse) in southern Quebec (Fig. 1 and Table 1). Latitudinal differences between sites provide a gradient in climate conditions, as well as in soil and forest types. The dominate vegetation type for Lake Clair, Lake Laflamme, and Lake Tirasse is sugar maple (SM), balsam fir (BF) and black spruce (BS), respectively. Sites will hereafter be referred to by their dominant tree species. Each site is equipped with a complete meteorological station. Site weather data for the required climate variables of daily minimum and maximum air temperature and weekly cumulative precipitation are available for a minimum of 10 yr. Moreover, daily soil volumetric liquid water content was measured using time-domain reflectometry (Campbell Scientific CS615) and soil temperature was measured using thermistor probes (Thermometric DC95F232V and YSI 401). Both variables were measured at three different depths corresponding to the forest floor (FF; 2 cm above the boundary between organic and mineral horizons), mineral soil (34 cm at SM and 22 cm at BF and BS) and subsoil layers (70 cm at SM and 81 cm at BF and BS). To account for in-site spatial variability, average soil moisture and temperature values for each depth were measured at multiple soil profiles. For the BF and BS sites, variables were measured from four and three profiles, respectively. At the SM site, soil moisture and temperature values were averaged from two and three profiles, respectively (except subsoil moisture, for which only one probe yielded reliable measurements). The average standard error (expressed as percentage of the mean) calculated from individual monthly averages values ranged from 4% to 22% among the soil and subsoil soil moisture measurements. Spatial variability of soil temperature measurements had a mean standard error of ±1°C, calculated on individual average monthly values for all soil depths. However, it is important to note that our objective is to present projected changes in average soil moisture and soil temperature in the form of a difference (delta) as compared to the actual average soil conditions (see below), and thus spatial variability is not a crucial issue in our work and will not be further discussed.

Fig. 1.

Site locations of the three study sites: Lake Clair watershed (LCW; SM site), Lake Laflamme watershed (LLW; BF site), and Lake Tirasse watershed (LTW; BS site).

Fig. 1.

Site locations of the three study sites: Lake Clair watershed (LCW; SM site), Lake Laflamme watershed (LLW; BF site), and Lake Tirasse watershed (LTW; BS site).

Table 1.

Site characteristics: data were recorded during the periods 1987–2008, 1981–2008, and 1997–2008 for Clair (SM), Laflamme (BF) and Tirasse (BS) sites, respectively. The soil type is according to the Canadian system of soil classification (Soil Classification Working Group 1998).

Site characteristics: data were recorded during the periods 1987–2008, 1981–2008, and 1997–2008 for Clair (SM), Laflamme (BF) and Tirasse (BS) sites, respectively. The soil type is according to the Canadian system of soil classification (Soil Classification Working Group 1998).
Site characteristics: data were recorded during the periods 1987–2008, 1981–2008, and 1997–2008 for Clair (SM), Laflamme (BF) and Tirasse (BS) sites, respectively. The soil type is according to the Canadian system of soil classification (Soil Classification Working Group 1998).

b. Soil humidity and soil temperature models

1) Model descriptions

ForHyM (hydrological conditions and water flow simulation) and ForSTeM (temperature changes and heat transfer simulation) have been described in Arp and Yin (1992) and Yin and Arp (1993), respectively, and only a brief description is given here. Data requirements for ForHyM and ForSTeM include monthly air temperature, total precipitation, and snowfall fraction of precipitation. Other site-specific descriptors such as latitude, forest stand composition, soil rooting zone, soil layer depths, and soil characteristics (porosity and texture) are also considered in the models. Both models are one-dimensional process-based models that also include some general empirical relationships. Forest canopy, snowpack, forest floor, soil (i.e., above the rooting zone), and subsoil (subsoil depth is fixed to 1 m) are represented by the model layers. Water and heat fluxes between model layers are simulated over a period of time, with flux intensity being dependent on soil layer physical, hydrological, and thermal properties. Once determined for a given site these properties permit the models to predict all major water and heat fluxes through upland forests. It is important to note that soil temperature is closely related to soil moisture, especially in northern regions because of phase changes of soil water and snow dynamics. As such, in ForSTeM, thermal conductivity of each layer depends on, among other characteristics, soil water content. For this reason, the outputs from ForHyM are used to feed ForSTeM to take into account the impact of the changes in soil water content on soil temperature changes.

2) Model calibration

For each site, the models were driven with the observed in situ meteorological data and measured site characteristics. The ForHyM calibration of water content values was performed only for the soil and subsoil layers, while the ForSTeM calibration of temperature was based on the comparison of simulated and measured values for the humus, soil, and subsoil layers.

To calibrate the models, parameters were adjusted iteratively to optimize (highest correlation coefficient) the fit between simulated and observed soil moisture and temperature values. Care was taken to ensure that the calibration yielded a good model performance on monthly values for the entire observation period as well as on the average seasonal patterns.

To reach the optimal calibration, the ForHyM parameters were adjusted as shown in Table 2. For ForSTeM calibration, thermal conductivity between soil layers generally needed to be reduced. Thermal conductivity coefficients of forest floor were adjusted with correction factors of 0.76, 0.90, and 0.62 for SM, BF, and BS, respectively. Similarly, correction factors of 0.86, 0.95, and 0.97 were applied on soil thermal conductivity. For the subsoil layer, the correction factors of SM, BF, and BS were 0.95, 1.03, and 0.95, respectively.

Table 2.

Parameter values for the three sites (SM, BS, and BF) and for the five scenario calibrations for one simulation of the CRCM (2041–70) at the BF site (Lake Laflamme). The results of these scenario calibrations on soil moisture and soil temperature projections are shown in Fig. 4 (see the text).

Parameter values for the three sites (SM, BS, and BF) and for the five scenario calibrations for one simulation of the CRCM (2041–70) at the BF site (Lake Laflamme). The results of these scenario calibrations on soil moisture and soil temperature projections are shown in Fig. 4 (see the text).
Parameter values for the three sites (SM, BS, and BF) and for the five scenario calibrations for one simulation of the CRCM (2041–70) at the BF site (Lake Laflamme). The results of these scenario calibrations on soil moisture and soil temperature projections are shown in Fig. 4 (see the text).

The impact of model calibration (five different calibrations) on projected changes of soil moisture using ForHyM was also tested for the BF site with a scenario (2041–70) of the Canadian Regional Climate Model (CRCM; see the results below in section 3c). The set and range of parameters used for this sensitivity analysis are shown in Table 2.

c. Future projections

1) Climate scenarios

To obtain a range of variability in future projections of soil temperature and moisture made with ForHyM and ForSTeM, six different climate scenarios were used as drivers. First, a future climate scenario was produced using simulated output from the CRCM 4.2.3 (Caya and Laprise 1999; Music and Caya 2007; hereafter referred to as CRCM). The simulation was carried out for a domain covering all of North America with a horizontal grid-size mesh of 45 km (true at 60°N) for the period 1961–2100, driven at its boundaries by atmospheric fields taken from a simulation output of the fourth run of the Canadian Centre for Climate Modelling and Analysis (CCCma) Coupled General Circulation Model, version 3 (CGCM3). Both global and regional simulations were performed using the Intergovernmental Panel on Climate Change (IPCC) Special Report on Emissions Scenarios (SRES) A2 greenhouse gas (GHG) and aerosol projected evolution for the period 2001–2100 and with the IPCC 20C3M scenario for years prior to 2001.

Climate projections from global climate models (GCMs) were also used to simulate future soil conditions. The use of an ensemble of climate simulations provides a measure of the robustness, as well as a quantitative assessment of uncertainty in the projected future conditions (Duchesne et al. 2009). The selection of the GCMs was based on a cluster analysis of monthly average change values of air temperature and precipitation for each simulation calculated as the difference or ratio between the average future conditions for the period 2071–99 and the reference period 1971–2000 (Logan et al. 2011). To determine an appropriate number of future scenarios a series of k-means cluster analyses were performed and a value of five groups was determined that offered both good group separation and a manageable number of future climate scenarios. As such, the 86 climate simulations made available from phase 3 of the Coupled Model Intercomparison Project (CMIP3; Meehl et al. 2007) for the studied regions and periods, were classified into five subgroups and a single simulation from each subgroup was retained for the subsequent analyses. The selected global simulations that were used to complement the CRCM were as follows: Model for Interdisciplinary Research on Climate 3.2, medium-resolution version [MIROC3.2(medres); run 1 of SRES A1B], ECHAM and the global Hamburg Ocean Primitive Equation (ECHO-G; run 1 of SRES A1B), ECHAM5 (run 1 of SRES A1B), Meteorological Research Institute Coupled General Circulation Model, version 2.3.2a (MRI CGCM 2.3.2a; run 3 of SRES A2), and MRI-CGCM 2.3.2a (run 3 SRES B1).

2) Individual sites

Outputs of monthly air temperature, precipitation, and snowfall fraction from the different climate scenarios were used as inputs in the ForHyM and ForSTeM simulations. Time series of simulated data for the input climate variables of interest were extracted for the selected GCM simulations and for the CRCM simulation by finding the corresponding data series for the nearest-neighbor model grid cell between each climate model and the three study sites.

Furthermore, a bias correction between simulated temperature and precipitation values and the observed values at the three sites was performed before importing the various climate projections into ForHyM and ForSTeM. For the temperature, monthly biases were calculated between average observed and simulated temperature for the reference period. Bias correction was then performed by adding the corresponding monthly bias to all years of simulated time series for the corresponding month (Salzmann et al. 2007). For the precipitation, bias correction factors were obtained from the fraction of the corresponding percentiles of the reference period data simulations and the observations. This relationship was then applied on the corresponding percentiles of the simulated data (Mpelasoka and Chien 2009). Differences in spatial resolution between the various global and regional climate models make direct comparison of absolute modeled soil temperature and humidity values somewhat problematic, and it is therefore important to note that the results presented here are restricted to a presentation of relative changes between simulated future and present periods rather than absolute values.

To examine the results of ForHyM and ForSTeM for the presence of statistically significant changes in simulated future soil moisture and temperature conditions by the CRCM, t tests were performed comparing the two future horizons (2040–69) and (2070–99) with respect to a reference period (1971–2000).

3. Results

a. ForHyM and ForSTeM performance

Simulated values of soil and subsoil water content were generally in good agreement with observed values for the three forest types studied (Fig. 2). High correlation coefficients were obtained between measured and simulated monthly values for the soil (0.91, 0.85, and 0.84) and the subsoil (0.61, 0.80, and 0.94) at SM, BF, and BS. The seasonal variation was also well reproduced at the three sites where correlation coefficients for monthly averages were all above 0.94 (except for SM subsoil, which was at 0.82). At SM, the major discrepancies between observed and simulated values were caused by the difficulty of ForHyM to adequately reproduce the low soil water content values observed during the dry summers of 2002, 2005, and 2006. For this site, soil and subsoil water contents during dry summers may be overestimated (this point is discussed below). However, for the BF and BS sites, observed and simulated soil water contents during summer droughts and spring snowmelt were in good agreement in both timing and magnitude.

Fig. 2.

Observed (solid blue) and simulated (dashed red) monthly values of soil and subsoil water content and monthly averages for the complete period of calibration. Standard errors of monthly averages are shown by vertical bars.

Fig. 2.

Observed (solid blue) and simulated (dashed red) monthly values of soil and subsoil water content and monthly averages for the complete period of calibration. Standard errors of monthly averages are shown by vertical bars.

Observed soil temperature values were well simulated by ForSTeM (Fig. 3), where there was good agreement between observed and simulated temperatures for the forest floor, soil, and subsoil at the three sites (correlation coefficients ≥ 0.96), for both monthly values and seasonal variations.

Fig. 3.

Observed (solid blue) and simulated (dashed red) monthly values of FF, soil, and subsoil temperature and monthly averages for the complete period of calibration. Standard errors of monthly averages are shown by vertical bars.

Fig. 3.

Observed (solid blue) and simulated (dashed red) monthly values of FF, soil, and subsoil temperature and monthly averages for the complete period of calibration. Standard errors of monthly averages are shown by vertical bars.

b. Climate change scenarios

At the three sites, annual-mean air temperature is projected to increase between 1.5° and 4.4°C for 2050, depending on the climate model and emissions scenario (Table 3). For 2080, projected changes in air temperature ranged from 2.3° to 6.0°C. A median increase of 7% (0%–12%) of annual precipitation is projected for SM and BF in 2050 while it should reach 10% (7%–15%) in 2080 (Table 3). At BS, precipitation is projected to increase by 13% (8%–18%) and 18% (15%–24%) in 2050 and 2080, respectively (Table 3).

Table 3.

Intersimulation variations in the projected change in precipitation and air temperature for the future horizons 2050 and 2080 compared to the reference period for three study sites. Values represent 30-yr monthly averages.

Intersimulation variations in the projected change in precipitation and air temperature for the future horizons 2050 and 2080 compared to the reference period for three study sites. Values represent 30-yr monthly averages.
Intersimulation variations in the projected change in precipitation and air temperature for the future horizons 2050 and 2080 compared to the reference period for three study sites. Values represent 30-yr monthly averages.

c. Effect of model calibration on future projections of soil moisture and temperature

A sensitivity analysis was conducted in order to evaluate the potential effects of model calibration on future projections. ForHyM and ForSTeM were calibrated to obtain the best possible fit with observed values with a relatively small number of changes as well as small changes to the absolute parameter values (see above). However, it is possible that a different set of model parameters could produce relatively similar values of soil moisture or temperature but could respond differently in future climatic conditions. Projected changes in soil water contents at BF for 2050 obtained from five different calibrations of ForHyM were thus compared (Fig. 4). The following parameters were changed in different combinations: forest floor (2.8–10) and subsoil (0.5–2.2) percolation coefficients, forest floor (20–40) and soil (70–83) field capacity, and potential evapotranspiration (0.87–1.3). Correlation coefficients between measured and simulated soil moisture values ranged from 0.75 to 0.94 with the five different calibrations. Future projected outputs using the various calibrations displayed similar seasonal patterns with nearly identical absolute values of projected changes relative to the reference period. A maximum standard error of 2.6% for monthly soil moisture values was observed in May while a maximum standard error of 0.07°C was observed for July soil temperatures between calibrations.

Fig. 4.

Comparison of the simulated changes in soil water content and temperature for future horizon 2050 at the BF site, projected by five calibrations of ForHyM.

Fig. 4.

Comparison of the simulated changes in soil water content and temperature for future horizon 2050 at the BF site, projected by five calibrations of ForHyM.

d. Future projections of annual and average monthly soil moisture

For all the sites, simulations of annual soil moisture based on the CRCM forcing data show only little change in water content values of the forest floor and soil layers between the future and reference periods (Table 4). At SM, the only significant change in average annual water content was seen for the soil layer in 2080 where the ratio of the future versus the present water content is 0.95. All other statistical tests indicated no significant change for this site. At BF, the only significant changes are projected for the soil and subsoil layers in 2080 with ratios of 0.94 and 0.99, respectively. No change in annual water content could be observed for the two future horizons and in the two soil layers at BS.

Table 4.

Projected change in annual soil water content and temperature at SM, BF, and BS based on climate scenarios from CRCM 4.2.3. “NS” means no significant change.

Projected change in annual soil water content and temperature at SM, BF, and BS based on climate scenarios from CRCM 4.2.3. “NS” means no significant change.
Projected change in annual soil water content and temperature at SM, BF, and BS based on climate scenarios from CRCM 4.2.3. “NS” means no significant change.

Despite small projected changes in the overall mean annual soil moisture content values, significant changes in the seasonal variation of forest floor, soil, and subsoil water contents (detected as a significant change for at least one monthly average) are projected at the three sites for both the 2050 and 2080 horizons (Fig. 5). The ForHyM simulations forced with the CRCM, data projected an increase in soil moisture in early spring and a decrease during late spring that is caused by earlier snowmelt. The CRCM forced simulation is generally within the 95% confidence interval of the five GCM simulations (gray area in Fig. 5). Reduced soil moisture is projected over the following summer months for every climate scenario at the three sites and was maintained over the entire growing season. At SM, the maximum increase is observed in March where water contents for the 2080 horizon are 1.5 times higher than the reference period for the soil layer (Fig. 5). The largest decrease in soil moisture, corresponding to a ratio of future (2080) versus present values of approximately 0.8, is projected for the April–September period in the soil layer (Fig. 5). Similarly, at BF, the largest increase is projected in March for the forest floor in 2080 (Fig. 5). The largest decrease in soil moisture at BF is projected in August for 2080 for the forest floor and both soil layers (ratio of 0.6 as compared to the reference period; Fig. 4). At BS, the greatest increase in soil water content occurs in April for the horizon 2080 and was 1.6 times higher than reference conditions (Fig. 6). At this site, the largest decrease in water content is projected to occur in 2080 for the soil layer with ratio values around 0.7–0.8 times lower than the reference conditions during the entire growing season (Fig. 5).

Fig. 5.

Simulated changes in FF and soil water contents for future horizons 2050 and 2080 based on CRCM climate projections (thick line), with the 95% confidence interval (gray) based on the variability observed in the five GCMs selected. Presented values are 30-yr monthly averages.

Fig. 5.

Simulated changes in FF and soil water contents for future horizons 2050 and 2080 based on CRCM climate projections (thick line), with the 95% confidence interval (gray) based on the variability observed in the five GCMs selected. Presented values are 30-yr monthly averages.

Fig. 6.

Simulated changes in FF and soil temperatures for future horizons 2050 and 2080 based on CRCM climate projections (thick line), with the 95% confidence interval (gray) based on the range of variability observed in the five GCMs. Presented values are 30-yr monthly averages.

Fig. 6.

Simulated changes in FF and soil temperatures for future horizons 2050 and 2080 based on CRCM climate projections (thick line), with the 95% confidence interval (gray) based on the range of variability observed in the five GCMs. Presented values are 30-yr monthly averages.

e. Future projections of annual and average monthly soil temperature

Significant changes in annual average soil temperatures based on the CRCM forcing data were projected at the three sites for all soil layers (Table 4). The projected increases at SM ranged from 1.2° to 1.3°C and from 1.9° to 2.1°C for the different soil horizons in 2050 and 2080, respectively. At the BF site, the changes ranged from 1.6° to 1.9°C and from 2.7° to 3.3°C for the different soil horizons in 2050 and 2080, respectively. At BS, increases between 1.1° and 1.4°C and between 2.0° and 2.4°C are projected for the three soil layers in 2050 and 2080, respectively. A similar pattern of seasonal changes is projected for the three sites. The changes projected by ForSTeM when forced with the CRCM data are within the 95% confidence interval of the four GCMs (Fig. 6). The projected increase is not constant throughout the year since few changes are projected from December to March while maximum increases occur in June and could reach values as high as 4.8°, 5.2°, and 3.8°C in the soil layer in 2080 for SM, BF, and BS, respectively.

f. Projection variability between climate models

Simulated soil moisture and temperature based on the various CRCM and GCM scenarios differed in the magnitude of projected changes, yet there is a consensus between simulations in terms of seasonal variations (Figs. 4, 5). For instance, all scenarios project an increase in spring forest floor and soil water content similar to what is projected using the CRCM scenario, where there is a maximum increase observed in either March or April for the different sites. Projected change of spring water content based on the CRCM scenario is typically in the middle to upper range of the other projections for future horizons 2050 and 2080. Furthermore, there is a consensus between the vast majority of simulations (and across sites) for the occurrence of a summer drying of the forest floor and the soil. The timing of the greatest reduction in soil moisture content is variable between sites, ranging from the months of April and May seen at the SM for the forest floor and soil, respectively, to the month of August for BF. The summer CRCM-based results are in the middle range of these projections for the future horizon 2050, while they are in the lower range for 2080.

Similarly, all CRCM- and GCM-based simulations project increases in forest floor and soil temperature. Projected increases in soil temperature are higher during summer and fall, with a maximum reached around June. Projected increases in forest floor and soil temperature, based on the CRCM scenario, are typically in the middle range of the other projections for the horizon 2050. In the case of the horizon 2080, projected increases in forest floor and soil temperature based on the CRCM scenario are usually in the upper range of the other projections.

4. Discussion

a. Sites and climate variability

The three sites studied vary in many respects. Not only do they represent the three dominant forest types in eastern Canada, but they also show a range in mean annual precipitation (from 924 to 1414 mm) and mean annual temperature (from 1.0° to 4.2°C). Annual average soil temperature and average soil water content also ranged from 3.4° to 4.7°C and from 14.7% to 33.3%, respectively, between the three sites. The onset of snowmelt is also different, with peak soil water content being observed during the month of April at the SM site and in May at the BF and the BS sites. The timing of snowmelt also influences the monthly pattern of soil temperature, which shows a sharp increase a month sooner at the SM site, due to an earlier disappearance of ground snow cover. The period of observation also offers a wide range in mean annual air temperature and precipitation with particularly dry summers in 2002, 2005, and 2006. Also, because of mild winter and spring temperatures, snowmelt occurred a month earlier at BS in 1998. At the three sites, January air temperatures in 2003–05 were 2°–6°C colder than what is usually observed for this month.

b. Model calibration and sensitivity of future simulations to calibration

The ForHyM and ForSTeM performed well in predicting monthly soil water content and soil temperature through the range of variations encountered at the three sites and between years (Figs. 2, 3). Obtained model performance in this study was generally comparable or better than the fits reported in other studies for both temperature and soil water content (Balland et al. 2006; Bhatti et al. 2000; Houle et al. 2002; Yin and Arp 1993). Multiple model calibrations using different sets of parameters can sometimes provide comparable quality fits between observed and simulated values. In our case, the number of adjustable parameters was relatively low, but we nevertheless tested the effect of different types of calibrations by applying them to the projected soil temperature and water content for 2050 at the BF site. The results showed that the projected soil temperature and soil water content values were very similar (Fig. 4), illustrating the small dependency of future projections on model calibration. This, along with the good performance of model versus observed values of soil water content and soil temperature, provide robust conditions to predict future forest soil conditions.

c. Projections

At the three sites and for all soil horizons, substantial increases in average annual temperatures, varying from 1.9° (soil, SM) to 3.3°C (forest floor, BF) are to be expected for 2080 (Table 4). Largest changes are projected during spring and most particularly in June (Fig. 6), reaching increases as high as 4°–5°C. The higher temperatures persisted throughout summer and early fall. The seasonal variations are comparable at the three sites and for each of the climate scenarios used, despite differences in absolute temperature increases. Similar to these results, a recent study (Qian et al. 2011) reported significant increases in spring (March–May) soil temperatures at depths of 5–150 cm for soils of various sites in Canada between 1958 and 2008. They also reported significant trends for soil temperature in summer but not for winter or for annual means. As in the projected future results in the present study, the role of earlier snowmelt and reduced snow depth was shown to play a key role in these observed trends. Relatively similar observations and projections were made for Scots pine stands in boreal forests of Sweden (Mellander et al. 2005, 2007), where changes in the date of snowpack disappearance had a large impact on spring soil temperature as well as on the spatial variability of spring soil temperature.

The average annual soil water content values are not projected to change substantially in the future, although small but statistically significant decreases in the ratios from 0.99 (subsoil, BF) to 0.94 (soil, BF) are predicted for 2080. However, as observed for soil temperature, seasonal patterns of projected soil water content show pronounced changes for the 2050 and 2080 periods. An earlier snowmelt generally increases the soil water content a month sooner than the reference period. Similar results were obtained for future discharge data at small forested catchments of the Hubbard Brook Experimental Forest in New Hampshire (Campbell et al. 2011). Overall, higher values of soil moisture are observed in winter whereas lower values are observed when the ground is snow free. Of concern is the marked decrease in soil moisture projected over most of the growing season at the three sites for the forest floor and the soil layers where the vast majority of roots are located. The selected climate scenarios (which cover the variability observed within a larger set of 86 scenarios) offer a wide range of projected climate and, within this projected variability, none of the scenarios predicts a general increase in soil water content during the growing season. Overall, the causes for lower soil moisture in summer include an earlier snowmelt that decreases the amount of soil water available at the beginning of the growing season and higher plant transpiration during the following months.

5. Conclusions

Although there is a growing number of climate models and climate scenarios made available to the scientific community, the impact of global changes on future forest soil temperature and soil moisture has received relatively little attention despite their critical importance on forest dynamics. In this study, we used a soil temperature model (ForSTeM) and a soil moisture model (ForHyM) to simulate future soil conditions for three forest sites based on one climate scenario from the Canadian Regional Climate Model (CRCM) and five climate scenarios issued from global climate models chosen to represent the range of variability observed in an ensemble of 86 climate scenarios. Average annual soil temperatures are projected to increase between 1.1° and 1.9°C in 2050 and between 1.9 and 3.3°C in 2080, depending on sites and horizon depths, while no significant changes in annual soil water content are projected. Although the results may vary in terms of absolute predicted changes, similar changes in seasonal patterns of both soil temperature and soil water contents are projected at the three sites. Soils are projected to be slightly moister and warmer in winter, but large increases in soil temperature and large decreases in soil water content are projected during the growing season. Warm summer conditions can produce an increased evapotranspiration demand from plants that can result in a drought stress, particularly when soil moisture level is low (Rennenberg et al. 2006). Increased vulnerability due to water depletion could result in reduced photosynthesis, transpiration, and growth caused by reduced water and CO2 fluxes within the plant mainly due to stomata closure to regulate water losses (Gessler et al. 2004; Hernandez-Santana et al. 2009; Rennenberg et al. 2006).

The expected impact on biogeochemical cycling of nutrients is difficult to ascertain because of the complex interactions or antagonistic effects that higher temperatures and lower soil moisture may have on microorganism activity and abiotic soil chemical reactions such as weathering. The positive relationship between microorganism activity and temperature becomes invalid when soil moisture falls under a certain threshold (Almagro et al. 2009; Curiel et al. 2007; Davidson et al. 1998). Reduced soil moisture may thus limit soil biotic processes and offset the potential effects of increased soil temperature. Finally, recent studies have reported significant microorganism activity and nutrient cycling under the snow cover (Monson et al. 2006; Schmidt and Lipson 2004), highlighting the importance of winter processes for soil biogeochemistry in winter (Campbell et al. 2005). Since the projections point toward warmer and wetter soils below the snowpack in winter, the contribution of winter nutrient cycling to annual nutrient fluxes could potentially increase.

Acknowledgments

We thank the staff of the Ministry of Natural Resources and Wildlife of Quebec for field data collection and Isabelle Charron for her help in editing the manuscript. The costs associated with this research were covered by the Ministère des Ressources Naturelles et de la Faune du Québec and Le Fond Vert du Ministère du Développement Durable, Environnement, et Parc du Québec within the framework of the Action Plan 2006-2012 on climate change in association with Natural Resources Canada. We thank John Scinocca and Charles Curry from the Canadian Centre for Climate Modelling and Analysis for their comments on an earlier version of the manuscript. We also thank the two anonymous reviewers that provided many valuable comments and advices on the original submitted version. CRCM data were provided by Ouranos. For GCM data, we acknowledge the modeling groups, the Program for Climate Model Diagnosis and Intercomparison (PCMDI), and the WCRP Working Group on Coupled Modelling (WGCM) for their roles in making available the WCRP CMIP3 multimodel dataset. Support of this dataset is provided by the Office of Science, U.S. Department of Energy.

REFERENCES

REFERENCES
Almagro
,
M.
,
J.
Lopez
,
J. I.
Querejeta
, and
M.
Martinez-Mena
,
2009
:
Temperature dependence of soil CO2 efflux is strongly modulated by seasonal patterns of moisture availability in a Mediterranean ecosystem
.
Soil Biol. Biochem.
,
41
,
594
605
.
Arp
,
P. A.
, and
X. W.
Yin
,
1992
:
Predicting water fluxes through forests from monthly precipitation and mean monthly air temperature records
.
Can. J. For. Res.
,
22
,
864
877
.
Balland
,
V.
,
J.
Bhatti
,
R.
Errington
,
M.
Castonguay
, and
P. A.
Arp
,
2006
:
Modeling snowpack and soil temperature and moisture conditions in a jack pine, black spruce and aspen forest stand in central Saskatchewan (BOREAS SSA)
.
Can. J. Soil Sci.
,
86
,
203
217
.
Bhatti
,
J. S.
,
R.
Fleming
,
N. W.
Foster
,
F.-R.
Meng
,
C. P. A.
Bourque
, and
P. A.
Arp
,
2000
:
Simulations of pre- and post-harvest soil temperature, soil moisture, and snowpack for jack pine: Comparison with field observations
.
For. Ecol. Manage.
,
138
,
413
426
.
Campbell
,
J. L.
,
M. J.
Mitchell
,
P. M.
Groffman
,
L. M.
Christenson
, and
J. P.
Hardy
,
2005
:
Winter in northeastern North America: A critical period for ecological processes
.
Front. Ecol. Environ.
,
3
,
314
322
.
Campbell
,
J. L.
,
C. T.
Driscoll
,
A.
Pourmokhtarian
, and
K.
Hayhoe
,
2011
:
Streamflow responses to past and projected future changes in climate at the Hubbard Brook Experimental Forest, New Hampshire, United States
.
Water Resour. Res.
,
47
,
W02514
,
doi:10.1029/2010WR009438
.
Caya
,
D.
, and
R.
Laprise
,
1999
:
A semi-implicit semi-Lagrangian regional climate model
.
Mon. Wea. Rev.
,
127
,
341
362
.
Curiel
,
Y. J.
,
D. D.
Baldocchi
,
A.
Gershenson
,
A.
Goldstein
,
L.
Misson
, and
S.
Wong
,
2007
:
Microbial soil respiration and its dependency on carbon inputs, soil temperature and moisture
.
Global Change Biol.
,
13
,
2018
2035
.
Davidson
,
E. A.
,
E.
Belk
, and
R. D.
Boone
,
1998
:
Soil water content and temperature as independent or confounded factors controlling soil respiration in a temperate mixed hardwood forest
.
Global Change Biol.
,
4
,
217
227
.
DesJarlais
,
M.
, and
Coauthors
,
2010
:
S’avoir S’adapter aux Changements Climatiques. Ouranos, 128 pp
.
Duchesne
,
L.
,
D.
Houle
,
M.-A.
Côté
, and
T.
Logan
,
2009
:
Modelling the effect of climate on maple syrup production in Québec, Canada
.
For. Ecol. Manage.
,
258
,
2683
2689
.
Gessler
,
A.
,
C.
Keitel
,
M.
Nahm
, and
H.
Rennenberg
,
2004
:
Water shortage affects the water and nitrogen balance in central European beech forests
.
Plant Biol.
,
6
,
289
298
.
Gislason
,
S. R.
, and
Coauthors
,
2009
:
Direct evidence of the feedback between climate and weathering
.
Earth Planet. Sci. Lett.
,
277
,
213
222
.
Hernandez-Santana
,
V.
,
J.
Martinez-Vilalta
,
J.
Martinez-Fernandez
, and
M.
Williams
,
2009
:
Evaluating the effect of drier and warmer conditions on water use by Quercus pyrenaica
.
For. Ecol. Manage.
,
258
,
1719
1730
.
Houle
,
D.
,
L.
Duchesne
,
R.
Ouimet
,
R.
Paquin
,
F. R.
Meng
, and
P. A.
Arp
,
2002
:
Evaluation of the FORHYM2 model for prediction of hydrologic fluxes and soil temperature at the Lake Clair watershed (Duchesnay, Quebec)
.
For. Ecol. Manage.
,
159
,
249
260
.
Houle
,
D.
,
S.
Couture
, and
C.
Gagnon
,
2010
:
Relative role of decreasing precipitation sulfate and climate on recent lake recovery
.
Global Biogeochem. Cycles
,
24
,
GB2029
,
doi:10.1029/2009GB003757
.
Logan
,
T.
,
I.
Charron
,
D.
Chaumont
, and
D.
Houle
,
2011
:
Atlas de Scénarios Climatiques pour la Forêt Québécoise. Ouranos et MRNF, 55 pp
.
Luxmoore
,
R. J.
,
S. D.
Wullschleger
, and
P. J.
Hanson
,
1993
:
Forest responses to CO2 enrichment and climate warming
.
Water Air Soil Pollut.
,
70
,
309
323
.
Meehl
,
G. A.
,
C.
Covey
,
K. E.
Taylor
,
T.
Delworth
,
R. J.
Stouffer
,
M.
Latif
,
B.
McAvaney
, and
J. F. B.
Mitchell
,
2007
:
THE WCRP CMIP3 multimodel dataset: A new era in climate change research
.
Bull. Amer. Meteor. Soc.
,
88
,
1383
1394
.
Melillo
,
J. M.
, and
Coauthors
,
2002
:
Soil warming and carbon-cycle feedbacks to the climate system
.
Science
,
298
,
2173
2176
.
Mellander
,
P. E.
,
H.
Laudon
, and
K.
Bishop
,
2005
:
Modelling variability of snow depths and soil temperatures in Scots pine stands
.
Agric. For. Meteor.
,
133
,
109
118
.
Mellander
,
P. E.
,
M. O.
Lofvenius
, and
H.
Laudon
,
2007
:
Climate change impact on snow and soil temperature in boreal Scots pine stands
.
Climatic Change
,
85
,
179
193
.
Meng
,
F. R.
,
C. P. A.
Bourque
,
K.
Jewett
,
D.
Daugharty
, and
P. A.
Arp
,
1995
:
The Nashwaak experimental watershed project: Analyzing effects of clearcutting on soil temperature, soil moisture, snowmelt and streamflow
.
Water Air Soil Pollut.
,
82
,
363
374
.
Monson
,
R. K.
,
D. A.
Lipson
,
S. P.
Burns
,
A. A.
Turnipseed
,
A. C.
Delany
,
M. W.
Williams
, and
S. K.
Schmidt
,
2006
:
Winter forest soil respiration controlled by climate and microbial community composition
.
Nature
,
439
,
711
714
.
Mpelasoka
,
F. S.
, and
F. H. S.
Chien
,
2009
:
Influence of rainfall scenario construction methods on runoff projections
.
J. Hydrometeor.
,
10
,
1168
1183
.
Music
,
B.
, and
D.
Caya
,
2007
:
Evaluation of the hydrological cycle over the Mississippi River basin as simulated by the Canadian Regional Climate Model (CRCM)
.
J. Hydrometeor.
,
8
,
969
988
.
Qian
,
B.
,
E. G.
Gregorich
,
S.
Gameda
,
D. W.
Hopkins
, and
X. L.
Wang
,
2011
:
Observed soil temperature trends associated with climate change in Canada
.
J. Geophys. Res.
,
116
,
D02106
,
doi:10.1029/2010JD015012
.
Rennenberg
,
H.
,
F.
Loreto
,
A.
Polle
,
F.
Brilli
,
S.
Fares
,
R. S.
Beniwal
, and
A.
Gessler
,
2006
:
Physiological responses of forest trees to heat and drought
.
Plant Biol.
,
8
,
556
571
.
Salzmann
,
N.
,
C.
Frei
,
P.-L.
Vidal
, and
M.
Hoelzle
,
2007
:
The application of Regional Climate Model output for the simulation of high-mountain permafrost scenarios
.
Global Planet. Change
,
56
,
188
202
.
Schmidt
,
S. K.
, and
D. A.
Lipson
,
2004
:
Microbial growth under the snow: Implications for nutrient and allelochemical availability in temperate soils
.
Plant Soil
,
259
,
1
7
.
Soil Classification Working Group
,
1998
:
The Canadian system of soil classification. 3rd ed. Agriculture and Agri-Food Canada Publication 1646, 202 pp
.
Solomon
,
S.
,
D.
Qin
,
M.
Manning
,
M.
Marquis
,
K.
Averyt
,
M. M. B.
Tignor
,
H. L.
Miller
Jr.
, and
Z.
Chen
, Eds.,
2007
:
Climate Change 2007: The Physical Science Basis. Cambridge University Press, 996 pp
.
White
,
A. F.
, and
A. E.
Blum
,
1995
:
Effects of climate on chemical weathering rates in watersheds
.
Geochim. Cosmochim. Acta
,
59
,
1729
1747
.
Yin
,
X. W.
, and
P. A.
Arp
,
1993
:
Predicting forest soil temperatures from monthly air temperature and precipitation records
.
Can. J. For. Res.
,
23
,
2521
2536
.