Impact of interactive vegetation phenology on the Canadian RCM simulated climate over North America
- 1.2k Downloads
Biosphere–atmosphere interactions play a very important role in modulating regional climate. To capture these bi-directional interactions, a dynamic vegetation model, the Canadian Terrestrial Ecosystem Model (CTEM), has been implemented in the fifth generation of the Canadian Regional Climate Model (CRCM5). CTEM can grow vegetation from bare ground and includes processes of photosynthesis, autotrophic and heterotrophic respiration, phenology, turnover, mortality and allocation. This study focuses on assessing the impact of interactive vegetation phenology, i.e. CTEM, on the CRCM5 simulated climate over North America. This is achieved by comparing two CRCM5 simulations—one with interactive phenology and the other with prescribed vegetation, driven by ECMWF reanalysis data (ERA40 and ERA-Interim) at the lateral boundaries, for the 1971–2010 period. Comparison of simulated vegetation attributes, temperature and precipitation in both simulations to those observed indicates that introduction of interactive phenology improves the performance of CRCM5 in some regions, although it introduces new biases in other regions, which are related to the underestimation of leaf area index (LAI). Interactive phenology enhances biosphere–atmosphere interactions, which are reflected in the higher values of correlation between atmosphere and biosphere variables. Interactive phenology also introduces long-term memory in CRCM5, estimated via lagged correlations between precipitation/temperature and LAI. Improved biosphere–atmosphere interactions and long-term memory in the CRCM5 simulation with interactive phenology leads to better interannual variability, particularly noticeable in the biosphere and atmosphere states during anomalously wet and dry years. This study thus provides useful insights related to the added value of interactive phenology in CRCM5 as well as the nature and variability of biosphere–atmosphere interactions over North America.
KeywordsDynamic vegetation model Interactive phenology Regional modelling Biosphere–atmosphere interactions
It is well recognized that climate has a strong influence on the distribution and characteristics of terrestrial ecosystems (Woodward 1987; Stephenson 1990; Prentice et al. 1992; Zhao et al. 2011). In turn, vegetation forcings on climate, through surface characteristics such as albedo and roughness length, have been proven to be important by previous studies based on observations (Liu et al. 2006; Notaro et al. 2006; Wang et al. 2014) and by climate model simulations using static vegetation, i.e. vegetation with prescribed characteristics (Snyder et al. 2004; Dubreuil et al. 2012). However, the study of vegetation impacts on climate using climate models with static vegetation is limited by the fact that these models cannot be expected to adequately capture long-term evolution of vegetation attributes and variability, particularly in the context of changing climate. Pielke et al. (1998) suggest that terrestrial ecosystems could significantly influence atmospheric processes on short-term (biophysical pathway) as well as on long-term (biogeochemical and biogeographical pathways) timescales, confirming the important role of the biosphere in climate simulations.
Dynamic vegetation models (DVMs) are increasingly being used for applications in climate models to improve the representation of biosphere–atmosphere interactions (Peng 2000; Cox 2001; Quillet et al. 2010; Van den Hoof et al. 2011). A DVM takes into account different terrestrial carbon pools (stems, roots, leaves) and the changes in the terrestrial ecosystem, like vegetation structure and composition, which affect these pools. In these models, vegetation is represented as patches of plant functional types (PFTs) such as crops, grass, deciduous trees, and evergreen trees, with prognostic properties: leaf area index, stem area index, vegetation height, rooting depth, etc. The vegetation attributes change in response to changes in climate in a DVM. These changes include those affecting the biogeochemical, biogeophysical and hydrological cycles, and more specifically alterations in the biomass, productivity and energy fluxes. Consequently, climate models with DVMs are able to simulate vegetation-atmosphere interactions more realistically, particularly in the long term.
Wramneby et al. (2010) studied vegetation-climate feedbacks over Europe under future greenhouse warming using the Rossby Center’s Regional Climate Model (RCM) RCA-GUESS, which includes the dynamic vegetation model LPJ-GUESS (Smith et al. 2001), based on two simulations with and without feedbacks of vegetation dynamics. They found that vegetation feedbacks on the climate are small compared to the radiative forcing of increased global CO2 concentrations but may alter warming projections locally, regionally and seasonally compared to the simulation lacking a dynamic vegetation module. Similarly, using the University of Wisconsin’s fully coupled global atmosphere–ocean–land Fast Ocean Atmosphere Model (FOAM) with dynamic vegetation model LPJ, Notaro et al. (2007) showed that, although the majority of the projected future warming is associated with the radiative forcing of rising CO2, the vegetation physiological forcing augments the warming by weakening the hydrological cycle due to reduced evapotranspiration, particularly for tropical forests. Using the Hadley Center’s HadCM3LC with the DVM TRIFFID, Pinto et al. (2009) demonstrated that the lifting condensation level over the Amazon in future conditions could be 1,000 m higher compared to current climate due to reduced vegetation cover, and therefore decreased evapotranspiration leading to low atmospheric humidity resulting in increased sensible heat flux and therefore warmer temperatures.
Several other studies have looked at the effect of vegetation dynamics on the climate variability. Hughes et al. (2006), using the Hadley Center GCM including TRIFFID, found that the variability of the vegetation structure, which is determined by environmental conditions through photosynthesis, and feedbacks can dampen or amplify atmospheric variability through a shift in the response timescale. Delire et al. (2011) using two coupled atmosphere-vegetation models, CCM3-IBIS from the National Center for Atmospheric Research and LMDz-ORCHIDEE from the Institut Pierre Simon Laplace, demonstrated that vegetation dynamics introduces a long-term memory into the climate system by slowly modifying the physical characteristics of the vegetated surface. Furthermore, they found that phenology alone could enhance the variability of the climate system.
Similarly to most of the existing state-of-the-art climate models, including RCMs (Stéfanon et al. 2012; Adam et al. 2014; Kraucunas et al. 2014), the fifth generation of the Canadian RCM (CRCM5) (Zadra et al. 2008; Martynov et al. 2013) is evolving to include a dynamic vegetation module. The land surface scheme used in CRCM5 is the latest version of the Canadian LAnd Surface Scheme (CLASS, version 3.5) (Verseghy 1991, 2011; Verseghy et al. 1993). CLASS recognizes four vegetation types: broadleaf, needleleaf, crops and grass. Vegetation phenology is prescribed in CLASS. The dynamic vegetation model CTEM (Canadian Terrestrial Ecosystem Model: Arora 2003; Arora and Boer 2003, 2005), developed at the Canadian Centre for Climate Modelling and Analysis (CCCma), has been implemented in CRCM5. In this framework, CTEM simulates the vegetation biomass as a function of climate, which is used by CLASS to compute water and energy fluxes at the land–atmosphere interface, allowing a two-way interaction between vegetation and regional climate. The CLASS/CTEM framework has been used in offline simulations and validated against site-specific biophysical and biogeochemical measurements from flux towers (Li and Arora 2011), as well as over all of North America, as detailed in Garnaud et al. (2014). The latter compared two simulations driven by different reanalysis products over North America. They found that the simulated biosphere is relatively sensitive to the driving data, as most simulated carbon pools and fluxes showed important differences, particularly over eastern North America, primarily due to the differences in precipitation and temperature in the driving reanalysis products. Nonetheless, Garnaud et al. (2014) concluded that the simulated biosphere in offline CLASS/CTEM simulations responds adequately to climate change, such as rising CO2 and temperatures.
The main objective of this study is to assess the impact of interactive vegetation phenology on the CRCM5 simulated climate over North America, particularly the role played by biosphere–atmosphere interactions in modulating the interannual climate variability. This is achieved by comparing two experiments with CRCM5—one with prescribed vegetation phenology (i.e. CLASS only) and the other with interactive phenology (i.e. CLASS/CTEM), for the 1971–2010 period. The paper is organized as follows. Section 2 gives a brief overview of the model along with a description of the experimental set-up and methods used. Section 3 presents analysis of the RCM simulations, with and without interactive phenology, focusing on the mean state of the simulated climate, biosphere–atmosphere interactions quantified using correlations, long-term memory of the atmosphere and biosphere through lagged correlations, and interannual variability including the state of the biosphere and atmosphere during anomalously dry and wet years. A brief summary and conclusions are given in Sect. 4.
2 Model, experiments and methods
2.1 The Canadian Regional Climate Model
The regional climate model used in this study, CRCM5, is based on a limited-area version of the Global Environment Multiscale (GEM) model used for numerical weather prediction at Environment Canada (Côté et al. 1998). GEM employs semi-Lagrangian transport and (quasi) fully implicit time marching scheme. In its fully elastic nonhydrostatic formulation (Yeh et al. 2002), GEM uses a vertical coordinate based on hydrostatic pressure (Laprise 1992). The following GEM parameterizations are used in CRCM5: deep convection following Kain and Fritsch (1990), shallow convection based on a transient version of the Kuo (1965) scheme (Bélair et al. 2005), large-scale condensation (Sundqvist et al. 1989), correlated-K solar and terrestrial radiation (Li and Barker 2005), subgrid-scale orographic gravity-wave drag (McFarlane 1987), low-level orographic blocking (Zadra et al. 2003), and turbulent kinetic energy closure in the planetary boundary layer and vertical diffusion (Benoit et al. 1989; Delage and Girard 1992; Delage 1997).
The land surface scheme CLASS in CRCM5 models three soil layers, 0.1, 0.25 and 3.75 m thick in its standard formulation, corresponding approximately to the depth influenced by the diurnal cycle, the rooting zone and the annual variations of temperature, respectively. CLASS includes prognostic equations for energy and water conservation for the three soil layers and a thermally and hydrologically distinct snowpack where applicable (treated as a fourth variable-depth layer). The thermal budget is performed over the three soil layers, but the hydrological budget is done only for layers above the bedrock. In order to simply mimic subgrid-scale variability, CLASS adopts a “pseudo-mosaic” approach and divides the land fraction of each grid cell into a maximum of four sub-areas: bare soil, vegetation, snow over bare soil and snow with vegetation. The energy and water budget equations are first solved for each sub-area separately and then averaged over the grid cell, using spatially varying structural attributes and physiological properties of the four CLASS PFTs in CLASS (needleleaf trees, broadleaf trees, crops and grasses). These structural attributes include albedo, leaf area index (LAI), roughness length, canopy mass and rooting depth.
The dynamic vegetation model CTEM that has recently been implemented in CRCM5 is a process-based ecosystem model (Arora 2003; Arora and Boer 2003, 2005, 2006; Li and Arora 2011) designed to simulate the terrestrial carbon cycle. It is able to grow vegetation from bare ground and to simulate several vegetation structural attributes such as leaf area index, vegetation height, root distribution and canopy mass. It includes processes such as photosynthesis, autotrophic respiration, heterotrophic respiration, phenology, turnover, allocation, fire and land-use change. CTEM simulates two dead carbon pools—litter and soil organic carbon—and three live vegetation pools—stems, leaves and roots. Terrestrial ecosystem processes in CTEM are modeled for nine different plant functional types (PFTs): evergreen and deciduous needleleaf trees, broadleaf evergreen and cold and drought deciduous trees, and C3 and C4 crops and grasses. The manner in which CLASS and CTEM interact is explained in the following section.
As discussed earlier, this study investigates the effects of interactive phenology on the regional climate and its variability over North America. Two simulations are performed for the 1958–2010 period at 0.5° resolution; these simulations are forced by the European Center for Medium range Weather Forecast’s (ECMWF) ERA40 reanalysis data (Uppala et al. 2005) for the 1958–1978 period and by ERA-Interim reanalysis data (Dee et al. 2011) for the 1979–2010 period at the lateral boundaries. The ERA40 reanalysis (Uppala et al. 2005) is available for the period 1957–2001 at 2.5° (~250 km) resolution. The ERA-Interim reanalysis (Dee et al. 2011) is available from 1979 to the present day at a resolution of 1.5°.
2.2.1 CRCM5_STD versus CRCM5_CTEM
The most important differences between CRCM5_STD and CRCM5_CTEM, i.e. simulations with prescribed and interactive vegetation phenology respectively, relate to the canopy resistance and photosynthesis, phenology, root distribution, canopy mass and vegetation height.
Photosynthesis and canopy conductance are of crucial importance in vegetation modelling. In CRCM5_STD, the canopy conductance (g c = 1/r c) formulation handled in CLASS is similar to that of Schulze et al. (1995), where the canopy resistance (r c) is expressed as a function of minimum stomatal resistance (r min) and a series of environmental dependences, such as incoming solar radiation, air temperature, vapour pressure deficit and soil moisture suction. Thus, the effect of atmospheric CO2 concentration on canopy resistance is not explicitly modeled in this formulation.
In CRCM5_CTEM, CTEM models photosynthesis, as described in detail in Arora (2003). In addition to the environmental variables cited above, the CTEM formulation includes the effect of atmospheric CO2 concentration on stomatal conductance. This is essential to simulate the physiological effects of increasing CO2 amounts on stomatal conductance, and thus canopy resistance, in climate simulations. The photosynthesis sub-module used in CTEM is based on the biochemical approach (Farquhar et al. 1980; Collatz et al. 1991, 1992) with some minor differences. The photosynthesis rate, A, is co-limited by assimilation rates based on the photosynthetic enzyme Rubisco, the amount of available light, and the photosynthetic product transport capacity. CTEM can also simulate the effect of soil moisture stress on photosynthesis in order to account for soil moisture stress via stomatal closure. Net canopy photosynthesis rate, A n, is then used to estimate canopy conductance. A n is obtained by subtracting canopy leaf maintenance respiration from the canopy photosynthesis rate (Arora 2003). Thus, when CTEM is coupled to CRCM5/CLASS, the canopy conductance is estimated by CTEM’s photosynthesis sub-module, which uses canopy temperature, aerodynamic conductance, soil moisture and other variables simulated by CLASS. CLASS and CTEM’s photosynthesis sub-module, simulating the fast biophysical processes, such as photosynthesis, canopy conductance and leaf respiration, operate at the model timestep (20 min) while other biogeochemical processes are modeled at a daily timestep. The canopy conductance is then passed back to CLASS, where it is used in its energy and water balance calculations.
With respect to phenology, in CRCM5_STD, CLASS adopts an approach where the air temperature and the temperature of the top soil layer determine leaf onset and offset, as described in Verseghy (2011). A threshold air temperature of 2 °C is used and when this threshold is exceeded the LAI increases linearly from a specified minimum to specified maximum value in certain number of days: 2 months for needleleaf trees and 1 month for broadleaf trees. With respect to crops, the Earth is divided into 10° latitude bands, and as specified in Verseghy (2011), the beginning of crop growth and the end of harvest are specified for each band as occurring on certain days of the year. It is assumed that crops take 2 months to reach maturity, and that 1 month elapses between the time that senescence begins and the time that harvest is over. Since the annual variations in height and leaf area index of grass can be considered as negligible, its height and LAI are assigned to a maximum value throughout the year. Thus, seasonality is modeled but not long-term variations in canopy cover or vegetation structure.
In CRCM5_CTEM, CTEM simulates the leaf onset through a carbon benefit approach, and the leaf offset is initiated by unfavorable environmental conditions that stress the plant and imply carbon loss. These unfavorable conditions include shorter day length, cooler temperatures, and drier soil moisture conditions. These processes are fully described in Arora and Boer (2005). In CTEM, four plant growth states determine the plant behavior and allocation patterns, as summarized in Table 2 of Arora and Boer (2005). These are maximum growth, normal growth, leaf-fall/crop-harvest and dormancy/no-leaves. During maximum growth, for trees and crops, all net primary productivity is allocated to leaves. For grass, the allocation is to leaves and roots for structural stability reasons. During normal growth, the allocation is shared between leaves, stem (if applicable) and roots. During leaf fall and crop harvest, the allocation to leaves ceases but continues for stem and roots. During dormancy/no-leaf state, no allocation occurs in the absence of CO2 uptake. The set of conditions used to trigger transition from one plant state to another for each PFT in CTEM are described in Arora and Boer (2005).
In CRCM5_STD, CLASS calculates the vegetation height and the canopy mass as a function of a PFT-dependent roughness length for momentum at vegetation maturity and maximum value of canopy mass, respectively. For trees, these values are invariant; for crops and grass, the vegetation effective height and the canopy mass vary annually with snow depth and growth stage. The rooting depth remains at its PFT-dependent maximum value for trees and grass; for crops, it is further corrected for growth stage.
In CRCM5_CTEM, CTEM allocates positive net primary production (NPP) between leaf, stem, and root components, which increases their biomass, while negative NPP results in the decrease of component biomass because of respiration. As a result of these allocation processes, the vegetation biomass may vary diurnally. CTEM then uses the simulated leaf biomass to calculate the LAI, which is passed on to CLASS and used in energy and water balance calculations over the vegetated fraction of the grid cell. The root biomass is converted to a rooting depth and root distribution profile through a variable root distribution parameterization (Arora and Boer 2005), and they are then used to estimate the fraction of roots in each soil layer required for calculating transpiration in CLASS. The aboveground canopy mass from CTEM is used to estimate the canopy heat capacity.
In CRCM5_CTEM, the vegetation structural attributes of CTEM’s nine PFTs are clustered to four PFTs (needleleaf trees, broadleaf trees, crops and grass) when they are passed on to CLASS (see Fig. 1). CLASS and CTEM’s photosynthesis sub-module that simulates the fast biophysical processes, such as photosynthesis, canopy conductance and leaf respiration, operates at a 20 min timestep while the rest of CTEM runs at a daily timestep.
2.3 Methods of model output evaluation and analysis
Prior to studying the impact of interactive phenology on the simulated climate over North America, the climate and biosphere simulated by CRCM5_STD are compared to observations in order to assess the model performance and to help identify biases. CRCM5_STD simulated temperature and precipitation are compared to gridded observational datasets available from the Climate Research Unit (CRU) (Mitchell and Jones 2005) and from the University of Delaware (UDEL) (Willmott and Matsuura 1995). The CRU TS 2.1 dataset covers the period 1901–2002 at a monthly temporal resolution and has a spatial resolution of 0.5°. The UDEL V3.01 dataset is composed of monthly values from 1901 to 2010 with a 0.5° global coverage. The simulated biosphere is evaluated against the observation-based green leaf area index product from the International Land Surface Climatology Project Initiative (ISLSCP II) FASIR-adjusted NDVI Biophysical Parameter Fields measured by the satellite mounted AVHRR sensor (Los et al. 2000; Hall et al. 2006; Sietse 2010). This monthly global dataset is available for the 1982–1998 period at 0.5° × 0.5° resolution. To limit the impact of initial conditions on results, the analysis presented in this study focuses on the 1971–2010 period. The impact of interactive vegetation phenology on the simulated climate over North America is then investigated by comparing CRCM5_CTEM and CRCM5_STD, particularly with respect to biosphere–atmosphere interactions.
Any long-term memory introduced by interactive phenology is studied through lag correlations. Also, the interannual variability of selected biosphere and climate characteristics in both simulations is investigated using the coefficient of variation (i.e. the ratio of the standard deviation to mean). In connection with the interannual variability, modeled atmosphere and biosphere states for observed anomalously dry/hot and wet years are also studied. To this end, the year 1988, which was a very dry and warm year over large regions of the United States, covering the central and eastern parts, is selected. The year 1993, which was an anomalously wet year over approximately the same region, is also considered. For the dry and warm year of 1988, the simulated number of hot days (NHD) is compared to that observed. Using a similar approach to Fischer et al. (2007), NHD is defined as the number of days with maximum temperature exceeding the long-term (1981–2010) 90th percentile daily maximum temperatures, calculated for each month of the summer (JJA) season. Since no single high-resolution dataset of daily temperature is available over the study domain, two observation datasets, covering Canada and the USA, were used. The gridded observational dataset over Canada (Hopkinson et al. 2011) is generated from daily observations at Environment Canada climate stations using a thin plate smoothing spline surface fitting method. The gridded meteorological data from the University of Washington (UW; Maurer et al. 2002) is used over the USA. The response of the biosphere in CRCM5_STD and CRCM5_CTEM to the anomalous amounts of precipitation and temperatures for the studied dry and wet years is also explored in detail.
3 Results and discussion
3.1 CRCM5_STD evaluation
3.2 Mean climate: CRCM5_STD versus CRCM5_CTEM
The underestimation of LAI in CRCM5_CTEM over the western boreal forest is clearly visible for summer. This could be due to a negative LAI-temperature feedback in spring. Comparison of CRCM5_CTEM and CRCM5_STD in Fig. 5b shows that CRCM5_CTEM simulates much cooler temperatures in spring over the boreal region, which may lead to a lower annual and summer LAI by delaying the annual increase in LAI of evergreen needleleaf trees. It is worth noting that, in Garnaud et al. (2014), ERA40- and NCEP-driven CLASS/CTEM offline simulations showed a similar underestimation of LAI in the western boreal forest. Also, results from Peng et al. (2014) show that, while the needleleaf evergreen PFT of the model performs well for the coast of British Columbia, it yields lower than observed GPP in the interior of the province. This is due to the colder and drier climate in the interior of British Columbia, to which CTEM’s needleleaf evergreen trees are not adapted. These results indicate that while the broad classification of PFTs in CTEM is sufficient to capture terrestrial ecosystem process at the global scale, it is inadequate for representing regional scale processes and another needleleaf evergreen PFT is probably required, as suggested by Peng et al. (2014). However, it must be noted that there are large uncertainties in the observation datasets. For example, Gibelin et al. (2006) showed that the ISLSCP II LAI is higher than other satellite based estimates, especially for the boreal forest. Garrigues et al. (2008) also showed that LAI datasets derived from remote sensing data all have their weaknesses, especially over forests. The underestimation of LAI over central and eastern North America is mainly due to the weak representation of C3 crops in CTEM, and this bias was also reported in Garnaud et al. (2014), but it is probably amplified by the lower precipitation in CRCM5_CTEM over central North America (Fig. 5a). On the contrary, the LAI of crops is overestimated in CRCM5_STD compared to observations, thus creating large differences between the two simulations.
The effect of the biosphere on precipitation is more complex as the precipitation source could be local recycling or convergence of moisture advected into the region. As shown in Figs. 5a and 6b, similar to the temperatures, the largest differences between the two simulations occur in spring and summer, when the biosphere–atmosphere interactions are strong. CRCM5_CTEM only slightly improves the dry bias along the west coast compared to CRCM5_STD, despite the difference in LAI (Fig. 7a), as the main source of precipitation in this region is the moisture advected into the region. In spring, the lower LAI in CRCM5_CTEM (Fig. 7a) across western and central US implies reduced LHF (Fig. 7d), which could lead to reduced contribution of local moisture to the total atmospheric westerly transport of transpired water. In summer, the differences in LAI are strongest between the two simulations (Fig. 7a). This is the season with maximum convective activity and therefore the biosphere has a more direct effect on the local precipitation. The model is able to reproduce these biosphere–atmosphere interactions: in summer, the spatial patterns of the differences in LHF (Fig. 7d), a variable that is greatly affected by LAI, and precipitation (Fig. 5a) are very similar.
However, the main patterns of biases to CRU and UDel in both temperature and precipitation remain after dynamic coupling of vegetation. This is an indication that the biases could be due to deficiencies in model physics (not related to vegetation) or due to errors in the driving data and model resolution.
Thus, although the impact of interactive phenology on the 1971–2010 mean climate can be significant depending on the region and the season, it does not always improve the model when compared to observations. However, the possible improvement from CTEM is limited by its own internal biases, along with the biases from the simulated climate, greatly influencing the simulated vegetation. Indeed, Garnaud et al. (2014) have shown that the vegetation simulated by CLASS/CTEM is sensitive to the driving climate. Despite these limitations, the implementation of CTEM in CRCM5 enables better representation of certain aspects of biosphere–atmosphere interactions, as discussed in the following section.
3.3 Biosphere–atmosphere interactions: CRCM5_STD versus CRCM5_CTEM
In CRCM5_CTEM, the correlations between the LAI and precipitation are very strong with mainly positive values (Fig. 8a). Over southern North America for example, where water is a growth-limiting factor, vegetation is largely dependent on precipitation. This leads to a positive LAI-precipitation feedback loop, with a positive precipitation anomaly boosting plant productivity and LAI, which leads to increased evapotranspiration, as shown in Fig. 8b, and possibly the amount of local rainfall. In the northern latitudes, the strong positive correlations between LAI and precipitation may be the result of both the increased LHF contributing to local rainfall and the reduced heat stress due to increased cloud cover. The correlations between the LAI and temperature (Fig. 8c) are negative over southern North America, since plants suffer from heat stress at higher temperatures in these regions, consequently decreasing the vegetation productivity. The correlations are positive in the higher latitudes, where vegetation will benefit from warmer temperatures due to the lengthening of the growing season. Over most regions, where correlations between LAI and SHF are negative, correlations between LAI and LHF are positive. However, in CRCM5_CTEM, the region of eastern Canada, shown in Fig. 8a, stands out. Indeed, in this region, though the LAI-LHF correlations are positive, the LAI-precipitation correlations are mostly non-significant. This leads to the conclusion that vegetation does not suffer from moisture limitation in this region.
For the same region, the correlations between LAI and temperature are positive, though the correlations between LAI and SHF are mostly non-significant. To better understand the interactions, a path analysis is performed to decompose the effect of (1) LAI and SHF on temperature, and (2) LAI and temperature on SHF, as explained in Sect. 2.3 and shown in Fig. 8e, f. The magnitudes of the direct effect of SHF and the indirect effect of LAI (through SHF) on temperature–SHF correlations (Fig. 8e) confirm that an increase in SHF leads to an increase in temperature. This suggests that the missing link in the LAI–SHF–temperature–LAI interactive loop resides in the correlation between LAI and SHF. Figure 8f shows that the LAI–SHF correlation is non-significant due to the counteracting direct effect of LAI on SHF (−0.25) and indirect effect of temperature on SHF through its effect on LAI (0.22).
3.3.1 Long-term memory
The results presented above for instantaneous correlations and lead-lag correlations show great resemblance in patterns to those of Notaro et al. (2006) who used observed data in order to study the vegetation-atmosphere feedbacks across the United States. Thus, the model simulation CRCM5_CTEM is consistent with observations and other studies. It was discussed in Sect. 3.2 that CRCM5_CTEM does not clearly improve the model with respect to the mean climate. However, since it introduces biosphere–atmosphere feedbacks and long-term memory in the model, it could be hypothesized that CRCM5_CTEM captures better the variability of biosphere and climate, as it is discussed in the following section.
3.4 Interannual variability and extremes
The interannual variability, quantified in terms of CV, of maximum LAI is clearly higher for CRCM5_CTEM, compared to CRCM5_STD (Figs. 10a, 11a) for both seasons, and is closer to that observed (figure not shown). In spring, the interannual variability of the maximum LAI in CRCM5_STD (Fig. 10a) can be high as it depends on the date of budburst. In summer however (Fig. 11a), the majority of the vegetation has reached its prescribed maximum LAI, except in the high latitudes, where the length of the growing season has a great impact on the maximum LAI. In these high latitude regions, a link can be made with Fig. 8c where the spatial pattern of the correlation between the maximum LAI and mean spring/summer temperature is very similar to the interannual variability of the maximum LAI in summer. As explained previously, this is due to the fact that in CRCM5_STD the phenology of the plants is driven by the 2-m air and the soil temperatures. However, the interannual variability of LAI does not seem to be linked to the interannual variability of the energy fluxes and the climate in CRCM5_STD, which is expected since the maximum LAI is prescribed in CRCM5_STD.
In CRCM5_CTEM, the region of maximum LAI variability shifts from the high-latitude regions in spring to a southerly location, dominated by crops, in summer. For spring, amplification of interannual variability can be noticed for SHF in CRCM5_CTEM, for the same high-latitude regions where higher interannual variability in maximum LAI is noted, which is also reflected in the interannual variability of mean spring temperature (Fig. 10e). The amplification in LHF variability for the same region is much smaller compared to SHF and is expected since the impact of LAI on albedo and therefore on SHF is dominant compared to LHF in these regions.
In spring, despite the higher interannual variability in the southern parts of the domain for maximum LAI in CRCM5_CTEM, a dampening of LHF variability is noted for the southwestern regions of the North American landmass (Fig. 10b) and some amplification in SHF over the central Great Plains (Fig. 10c). The dampened interannual variability in LHF is not reflected in precipitation, as the source of precipitation for these regions is not only related to local recycling of moisture, but also to moist air advected into the region by winds from the adjoining ocean. The impact of the increased interannual variability in SHF for the central Great Plains on the mean temperature is less evident. As for summer, similar to spring, dampening of LHF variability can be noted in CRCM5_CTEM. Though variability in SHF is higher in high-latitude regions in CRCM5_CTEM in summer compared to CRCM5_STD, no important changes are noticed for other regions.
As for the wet summer of 1993, both simulations capture the positive precipitation anomaly, accompanied by anomalously cooler temperatures. Once again, these two climate factors have different effects on the biosphere in the two simulations. In CRCM5_STD, the cooler temperatures in 1993 and 1994 cause the mean LAI to drop, with precipitation having no direct effect on the biosphere. In CRCM5_CTEM however, vegetation benefits from the increased precipitation although this effect is dampened by the cooler temperatures. It should be noted that overly large precipitation, resulting in flooding, could lead to unfavorable conditions for the biosphere, and this is not currently represented in CTEM.
4 Summary and conclusions
The impact of interactive vegetation phenology on CRCM5 simulated climate over North America for the 1971–2010 period is studied by comparing two simulations—CRCM5_STD with prescribed vegetation phenology represented by the land surface scheme CLASS and CRCM5_CTEM which models vegetation phenology as a dynamic component through CTEM coupled with CLASS. Both simulations are driven by ERA-40/ERA-Interim reanalyses at the lateral boundaries.
Comparison of simulated and observed spatial distribution of the biospheric state variable, LAI, suggests that CRCM5_CTEM, particularly in summer, captures better the distribution, except over western Canada where it underestimates the LAI. This bias in the simulated biosphere over western Canada in CRCM5_CTEM could lead to underestimation of biosphere–atmosphere feedbacks, particularly due to the strength of the albedo effect in this region (Loranty et al. 2014), and it could possibly be rectified by creating a 10th PFT adapted to this region, i.e. a needleleaf evergreen PFT better adapted to drier summers and colder winters through longer leaf life span and lower sensitivity to drought- and cold-stress, as suggested by Peng et al. (2014). It must however be noted that there are big uncertainties in the LAI observation datasets. For example, Gibelin et al. (2006) have shown that the ISLSCP II LAI is higher than other satellite based estimates, especially in the boreal forest.
The differences in LAI, the vegetation characteristic that has the biggest impact on the atmosphere-biosphere interactions, between CRCM5_CTEM and CRCM5_STD lead to differences in surface albedo, SHF and LHF between the two simulations over various regions, which are reflected in the simulated temperature and precipitation fields. In fact, CTEM improves the model (CRCM5) in some regions, although it introduces new biases in other regions such as western USA. Thus the impact of interactive phenology on the 1971–2010 mean climate can be significant depending on the region and the season. It should be noted that biases in the simulated climate stemming from other sources could influence the simulated vegetation. Garnaud et al. (2014) have shown that the vegetation simulated by CLASS/CTEM is sensitive to the driving climate.
Despite these limitations, the implementation of CTEM in CRCM5 introduces feedbacks between the biosphere and the atmosphere. In fact, the correlations between biospheric and atmospheric variables are significantly stronger in CRCM5_CTEM, particularly with respect to precipitation and surface fluxes. Furthermore, since the biosphere is a slow integrator of short-term climate changes, the implementation of a dynamic vegetation module in a climate model introduces long-term memory in the climate system, thus influencing the climate in the long term, as suggested by Delire et al. (2011) and noted in CRCM5_CTEM. The strongest lagged correlations are between the LAI and temperature, with the average temperature of a given year having a significant impact on the LAI of the following year. Although not as pronounced, the impact of LAI on temperature the following year shows a similar behavior. Once again, this is an indication that CRCM5_CTEM is able to simulate feedback between the biosphere and the climate that could lead to significant alterations in the climate in sensitive regions in the long term. These findings are similar to those obtained by Liu et al. (2006) and Wang et al. (2014) using observations. CRCM5_CTEM simulates better the interannual variability in LAI, which is also reflected in the ability of the model to simulate more realistically the state of the atmosphere and biosphere during anomalously wet and dry years.
This study thus provides important insights on the impact of interactive phenology on the regional climate of North America, particularly related to the variability, both in space and time, in biosphere–atmosphere interactions over the region. As mentioned earlier, this study does not take into account competition between plant functional types. As shown by Smith et al. (2011), competition is important to model vegetation shifts and changing tree line, due to non-negligible effects that this can have on temperature and precipitation feedbacks, particularly in the context of a changing climate. Work is in progress to implement competition in CTEM, and thus future work will focus on the impact of competition on regional climate.
This project was funded by the Canada Research Chair in Regional Climate Modeling and the Ouranos Consortium. The authors would like to sincerely thank Ms. Katja Winger and Mr. Luis Duarte for their valuable technical assistance.
- Adam JC, Stephens JC, Chung SH, Brady MP, Evans RD, Kruger CE et al (2014) BioEarth: envisioning and developing a new regional earth system model to inform natural and agricultural resource management. Clim Change 1–17. doi: 10.1007/s10584-014-1115-2
- Cox P (2001) Description of the “TRIFFID” Dynamic Global Vegetation Model. Hadley Center technical note 24Google Scholar
- Dee DP, Uppala SM, Simmons AJ, Berrisford P, Poli P, Kobayashi S, Andrae U, Balmaseda MA, Balsamo G, Bauer P, Bechtold P, Beljaars ACM, van de Berg L, Bidlot J, Bormann N, Delsol C, Dragani R, Fuentes M, Geer AJ, Haimberger L, Healy SB, Hersbach H, Hólm EV, Isaksen L, Kållbergn P, Köhler M, Matricardi M, McNally AP, Monge-Sanz BM, Morcrette J-J, Park B-K, Peubey C, de Rosnay P, Tavolato C, Thépaut J-N, Vitart F (2011) The ERA-Interim reanalysis: configuration and performance of the data assimilation system. Quart J Royal Meteorol Soc 137:553–597. doi: 10.1002/qj.828 CrossRefGoogle Scholar
- Garrigues S, Lacaze R, Baret F, Morisette JT, Weiss M, Nickeson JE, Fernandes R, Plummer S, Shabanov NV, Myneni RB, Knyazikhin Y, Yang W (2008) Validation and intercomparison of global Leaf Area Index products derived from remote sensing data. J Geophys Res 113. doi: 10.1029/2007JG000635
- Gibelin A-L, Calvet J-C, Roujean J-L, Jarlan L, Los SO (2006) Ability of the land surface model ISBA-A-gs to simulate leaf area index at the global scale: Comparison with satellites products. J Geophys Res 111. doi: 10.1029/2005JD006691
- Hall FG, Brown de Colstoun E, Collatz GJ, Landis D, Dirmeyer P, Betts A, Huffman G, Bounoua L, Meeson B (2006) The ISLSCP initiative II global data sets: surface boundary conditions and atmospheric forcings for land–atmosphere studies. J Geophys Res 111. doi: 10.1029/2006JD007366
- Kraucunas I, Clarke L, Dirks J, Hathaway J, Hejazi M, Hibbard K et al (2014) Investigating the nexus of climate, energy, water, and land at decision-relevant scales: the Platform for Regional Integrated Modeling and Analysis (PRIMA). Clim Change 1–16. doi: 10.1007/s10584-014-1064-9
- Los SO, Collatz GJ, Sellers PJ, Malmström CM, Pollack NH, DeFries RS, Bounoua L, Parris MT, Tucker CJ, Dazlich DA (2000) A global 9-year biophysical land-surface data set from NOAA AVHRR data. J Hydrometeor 1:183–199. doi: 10.1175/1525-7541(2000)001<0183:AGYBLS>2.0.CO;2
- Martynov A, Laprise R, Sushama L, Winger K, Separovic L, Dugas B (2013) Reanalysis-driven climate simulation over CORDEX North America domain using the Canadian regional Climate Model, version 5: model performance evaluation. Clim Dyn 41:2973–3005. doi: 10.1007/s00382-013-1778-9 CrossRefGoogle Scholar
- Sietse OL (2010) ISLSCP II FASIR-adjusted NDVI Biophysical Parameter Fields, 1982–1998. In: Hall FG, Collatz G, Meeson B, Los S, Brown de Colstoun E, Landis D (eds) ISLSCP Initiative II Collection. Data set. Available on-line [http://daac.ornl.gov/] from Oak Ridge National Laboratory Distributed Active Archive Center, Oak Ridge, Tennessee, U.S.A. doi: 10.3334/ORNLDAAC/972
- Taylor KE, Williamson DL, Zwiers FW (2000) The sea surface temperature and sea-ice concentration boundary conditions for AMIP II simulations. Program for Climate Model Diagnosis and Intercomparison, Lawrence Livermore National Laboratory, p 25Google Scholar
- Verseghy DL (2011) CLASS—The Canadian land surface scheme (version 3.5). Technical Documentation (version 1). Environment Canada, Climate research division, Science and Technology branchGoogle Scholar
- Webb RS, Rosenzweig CE, Levine ER (1991) A global data set of soil particle size properties. NASA Technical Memorandum, 4286Google Scholar
- Woodward FI (1987) Climate and plant distribution. Cambridge University Press, Cambridge, p 174Google Scholar
- Zadra A, Caya D, Côté J, Dugas B, Jones C et al (2008) The next Canadian Regional Climate Model. Phys Can 64(2):75–83Google Scholar
Open AccessThis article is distributed under the terms of the Creative Commons Attribution License which permits any use, distribution, and reproduction in any medium, provided the original author(s) and the source are credited.