1. Introduction
In many parts of the world, soil water can be separated into two hydrological layers: 1) the vadose zone (unsaturated soil), which is located between the surface and the water table and which holds soil moisture; and 2) the saturated zone, lying underneath, which is composed of aquifers with three-dimensional groundwater flows. Most global climate models and Earth system models do not represent groundwater flows. They only simulate the vertical transport of soil moisture in the vadose zone. In doing so, they may be neglecting an additional source of water for the atmosphere.
Groundwater was recently estimated to provide 23% of the water transpired by plants at the global scale (Evaristo and McDonnell 2017). This figure accounts for all the regions/seasons where groundwater is accessible to the vegetation root system. But this does not mean that adding groundwater in a global climate model should increase by as much the simulated global transpiration. Where and whenever the model simulates enough soil moisture to meet the evaporative demand in the absence of groundwater, adding groundwater will not increase transpiration—even when, in reality, part of the water transpired does indeed come from the underlying aquifer. In other words, the lack of groundwater representation in a model is likely to induce an underestimation of transpiration only where and when groundwater is available to plants whose transpiration is limited by the lack of soil moisture in the vadose zone. These situations correspond to “water-limited regimes” of evapotranspiration (Seneviratne et al. 2010) where the variations of soil moisture control those of evapotranspiration. These regimes typically occur in semiarid environments or during the evaporation season in regions of transition between wet and arid climates (Koster et al. 2006; Dirmeyer et al. 2009; Dirmeyer 2011). Under such conditions, a soil moisture–controlled increase of evapotranpiration leads to a humidification and a cooling of the near-surface atmosphere (Koster et al. 2006; Seneviratne et al. 2010) that can affect temperature and precipitation mean values (Koster et al. 2006; Dirmeyer et al. 2009; Seneviratne et al. 2010) and extremes (Fischer et al. 2007; Hirschi et al. 2011; Miralles et al. 2012). Therefore, to the extent that some aquifers are shallow enough to contribute to evapotranspiration where it is water-limited, groundwater may have a significant influence on the climate system.
Shallow aquifers are generally found under wet climates where soil moisture tends to be plentiful. However, drier environments can also sustain relatively shallow water table depths. This can happen in areas of complex terrain, with a convergence of the lateral groundwater flows in valleys, or in regions characterized by a pronounced seasonal cycle of precipitation, where the groundwater recharge occurring in the rainy season maintains the aquifer at a relatively high level during the dry season (Fan 2015) allowing plants to access groundwater in the capillary fringe of the vadose zone (Fan et al. 2019).
Over the last two decades, models simulating groundwater have been coupled to land surface and atmospheric models over limited-area domains, ranging in size from the watershed to the regional scale. This body of literature has shown that taking groundwater into account can indeed increase soil moisture and evapotranspiration, which can affect the boundary layer height and stability (Maxwell et al. 2007; Larsen et al. 2016; Forrester and Maxwell 2020), as well as mean precipitation (Anyah et al. 2008; Jiang et al. 2009; Leung et al. 2011; Larsen et al. 2016) and temperature (Anyah et al. 2008; Jiang et al. 2009), and possibly heat waves (Keune et al. 2016; Mu et al. 2022; Furusho-Percot et al. 2022). However, relatively few of these studies (Leung et al. 2011; Larsen et al. 2016; Keune et al. 2016; Furusho-Percot et al. 2022) used simulations that were long enough to provide climate-relevant results. And all of them were conducted with limited-area models over domains located in the United States, Europe, or Australia, thus failing to provide a global picture.
The possible effects of groundwater on climate have recently started to be studied at the global scale with global climate models using idealized configurations (Wang et al. 2018) or schematized representations of groundwater flows (Arboleda Obando et al. 2022). These first results indicate that even at the relatively low resolutions of global climate models, the inclusion of a groundwater scheme can indeed affect the simulated climate conditions and also modulate the regional patterns of the climate change signal (Arboleda Obando et al. 2022) as previous studies had indirectly suggested (Maxwell and Kollet 2008; Ferguson and Maxwell 2010).
In the present study, we pursue this effort of globally assessing groundwater–climate feedbacks in a changing climate. To this end, we use the CNRM-CM6-1 global climate model (Voldoire et al. 2019; Roehrig et al. 2020) and its process-based hydrogeological parameterization of unconfined aquifers (Vergnes et al. 2012, 2014; Decharme et al. 2019). We compare simulations performed with and without groundwater, under preindustrial levels of atmospheric CO2 and after climate stabilization following an abrupt quadrupling of these preindustrial levels of CO2 (4xCO2).
The model and experimental setup are described in detail in section 2. Results are presented in section 3. First, we analyze the effect groundwater on soil moisture and evaluate the realism of the groundwater contribution to evapotranspiration in CNRM-CM6-1 under preindustrial conditions. Then, we show the regional impacts of groundwater on climate change, starting with the impacts of groundwater in the stationary climate of the preindustrial world. We also explain the physical processes involved in the groundwater–climate feedbacks and their evolution with climate change. Finally, in section 4, we give the main conclusions and discuss the possible underestimation of groundwater–climate feedbacks in our simulations.
2. Methods
a. Model
CNRM-CM6-1 (Voldoire et al. 2019) is the global climate model (http://www.umr-cnrm.fr/cmip6/spip.php?article11) developed in our institute [Centre National de Recherches Météorologiques (CNRM)]. The simulations used in the present study were run in an atmosphere-only mode (i.e., not coupled to the ocean model).
The configuration we used is based on the ARPEGE-Climat v6.3 atmospheric general circulation model (Roehrig et al. 2020) and the SURFEX v8.0 surface modeling platform which includes the land surface model ISBA (Interaction Soil–Biosphere–Atmosphere) coupled to the CTRIP (CNRM version of the Total Runoff Integrating Pathways) river routing model (Decharme et al. 2019) (http://www.umr-cnrm.fr/spip.php?article1092&lang=en). A complete description and validation of the surface and atmospheric models can be found in the cited reference papers. Here, we only review the main features.
The horizontal resolution is about 1.4° at the equator for ARPEGE-Climat and ISBA, and 0.5° for CTRIP. There are 91 vertical levels up to 0.01 hPa in the atmosphere, 14 soil levels down to 12 m and 12 snow levels. At the land surface, a plant–climate interactive scheme (Delire et al. 2020) controls the vegetation transpiration and growth [prognostic leaf area index (LAI)]. There are 16 different vegetation types and 3 nonvegetation surface types in ISBA, clustered in 12 different surface tiles in the version used in CNRM-CM6-1, each with a different set of parameters, among which are the rooting depth and the vertical root density profiles.
In the soil, the evolution of the temperature and the water content are computed with an explicit diffusion scheme using the one-dimensional Fourier and Darcy laws and accounting for the hydraulic and thermal properties of the soil organic carbon. The use of a multilayer snow model of intermediate complexity allows us to separate the water and energy budgets in the soil and the snowpack. CTRIP simulates the river flow, inundation dynamics, and groundwater flow.
The CNRM-CM6-1 configuration we used is almost identical to the one used for the experiments from phase 6 of the Coupled Models Intercomparison Project (CMIP6; Eyring et al. 2016), except for the activation of the interactive LAI scheme (which was turned off in the CNRM-CM6-1 CMIP6 experiments) and a slight modification in the groundwater scheme (see next paragraph).
b. Groundwater representation
CNRM-CM6-1 represents unconfined aquifer processes in the world’s major groundwater basins at a 0.5° resolution. The hydrogeological modeling of groundwater dynamics is based on the well-known MODCOU hydrogeological model (Ledoux et al. 1989). It consists of a one-layer diffusive 2D scheme (embedded in CTRIP) that computes the piezometric head as a function of the lateral groundwater flow, the two-way water exchange with the river (also computed in CTRIP), and the two-way vertical water exchange with the unsaturated soil column of the vadose zone (represented in ISBA), as detailed in Vergnes et al. (2014) and Decharme et al. (2019).
Groundwater basins boundaries were defined using the following global maps: the Worldwide Hydrogeological Mapping and Assessment Programme (WHYMAP), the hydrogeological map over the United States from the U.S. Geological Survey (USGS), and the global map of lithology (Dürr et al. 2005). The latter was also used to determine the transmissivity and the effective porosity coefficient in each basin. In each grid cell, the water table depth (WTD) accounts for the 1-km-resolution topography, which is extracted from the Global Multiresolution Terrain Elevation Data 2010 (Danielson and Gesch 2011). The WTD is computed in relation to the mean elevation of the 1-km subgrid points located below the first decile of the subgrid topography (instead of the mean elevation of the grid cell). That way, WTD is representative of the “lowland” part of the grid cell, whose elevation is close to that of the river. Consequently, the upward capillary flux into the ISBA soil column is allowed only over a fraction fwtd that corresponds to the area over which the water table head is close to the surface. fwtd is computed dynamically as a function of the riverbed elevation and the “subgrid” water table depth (computed as the depth of the water table head at a resolution of 1-km, using the subgrid topography) (Vergnes et al. 2014). Over this fwtd fraction of the grid cell, the water flux between the aquifer and the soil column is bidirectional (downward recharge from the soil to the aquifer and upward capillary rise from the aquifer to the soil column). Over the rest of the grid cell (1 − fwtd), this water flux only represents the downward recharge of groundwater (see appendix A for further details). Given that the water table depth is representative of the “lowland” part of the grid cell, it would be unrealistic to simulate capillary rise over the whole grid cell. This feature can also be seen as a way to account for the subgrid hillslope groundwater flow.
The modeling of groundwater and other hydrological processes in ISBA-CTRIP has been thoroughly validated in previous publications, both at the regional and global scales, in offline (Decharme et al. 2019; Vergnes et al. 2012; Vergnes and Decharme 2012; Vergnes et al. 2014; Munier and Decharme 2022) and inline configurations (Voldoire et al. 2019; Roehrig et al. 2020). Model results were compared to in situ data of piezometric head, large datasets of river discharge observations, and GRACE terrestrial water storage estimates. The water table depths were also compared to the global dataset of Fan et al. (2013) derived from a high-resolution groundwater model constrained with observations (Decharme et al. 2019). The ISBA-CTRIP land surface system was then used in a number of studies dealing with global hydrology and/or climate change (Ardilouze et al. 2019; Giffard et al. 2019; Douville et al. 2020; Padrón et al. 2020; Pellet et al. 2020; Saint-Martin et al. 2021).
In the version of the groundwater parameterization we used here, the coupling between the saturated zone (groundwater in CTRIP) and the vadose zone (soil column in ISBA) was slightly improved compared to the formulation described in the reference papers (Vergnes et al. 2014; Decharme et al. 2019). In the latter, the water table is always considered to be below the vadose zone for the coupling, even when the water table depth computed by CTRIP is shallower than the vadose zone depth in ISBA. The coupling formulation was improved to allow the water table to actually penetrate the vadose zone. The corresponding equations are detailed in appendix A. This improvement has a minor impact on water table depths, which was evaluated both in offline and inline configurations (not shown).
In ISBA-CTRIP, plant rooting depth only depends on the vegetation type, regardless of the typical range of water table depth in a given environment. Studies have shown that in reality, plants adapt their rooting depth to the local profile of soil water availability. If the water table remains shallow throughout the year, roots also stay shallow to avoid anoxia in the saturated zone. In drier environments, plants can send deep roots in the capillary fringe to sustain their water demand (Fan et al. 2017, 2019). Therefore, having a fixed rooting depth for each vegetation type is somewhat unrealistic. The way it may affect our results regarding the impact of groundwater in our model will be discussed in sections 3a and 4.
c. Experimental setup
We performed two pairs of simulations, with and without aquifers: one was carried out with preindustrial (PI) levels of atmospheric CO2 concentration, and the other with a quadrupling of the preindustrial CO2 concentration (4xCO2). The simulations with aquifers were performed using the groundwater parameterization described in the previous paragraph (i.e., with 2D groundwater flows, two-way water exchanges with the river and the unsaturated soil column). The simulations without aquifers have no representation whatsoever of groundwater; the water drained at the bottom of the unsaturated soil column is directly transported to the river. In the following, PIa (C4a) refers to the simulation with aquifers under preindustrial (4xCO2) conditions, and PIr (C4r) is the reference simulation without aquifers.
All simulations were run in a stand-alone configuration (i.e., not coupled to the ocean). The model was forced with monthly climatologies of sea surface temperature and sea ice cover derived from the corresponding fully coupled simulations performed with CNRM-CM6-1 for CMIP6 (namely, the piControl and the abrupt-4xCO2 simulations). The sea surface temperature and sea ice cover climatologies were built with the same procedure as the one used to run all atmosphere stand-alone simulations in CMIP6 (Taylor et al. 2000).
The initialization was done using restart files extracted from the piControl and abrupt-4xCO2 CMIP6 simulations run with CNRM-CM6-1. We used the restart files of the year 1850 of the piControl simulation for PIr and PIa, and those of the 150th year of the abrupt-4xCO2 simulation for C4r and C4a so that the model has reached its equilibrium state in both cases.
As the interactive plant–climate scheme was not activated in CNRM-CM6-1, we extracted the LAI and plant carbon variables values from the restart files of the CNRM-ESM2-1 (Séférian et al. 2019) CMIP6 simulations. CNRM-ESM2-1 is the Earth system model version of CNRM-CM6-1 (http://www.umr-cnrm.fr/cmip6/spip.php?article10). In addition to the processes represented by CNRM-CM6-1, it simulates the global carbon cycle, which requires the use of additional components and parameterizations, such as the interactive plant–climate scheme.
After the initialization, a 40-yr spinup was run for each simulation to ensure that all variables have adjusted to each other in the newly defined settings (i.e., without aquifers, with forced SST and SIC and with the plant–climate interactive scheme). Then, all four simulations (PIr, PIa, C4r, and C4a) were run for 90 years.
d. Statistical significance computations
For all significant tests performed on field differences, we used a false detection rate (FDR) test described by Wilks (2016). It is based on local t tests for the computation of P values. To determine the statistical significance of the differences over each grid points, P values are compared to a threshold which depends on the P values of the other grid points, the number of grid points, and the “level of confidence” of the test (in our case, 95%). This method allows to reduce the rate of false significance, which can be quite high in the case of autocorrelated fields when P values are directly compared to a fixed threshold corresponding to the level of confidence of the test (Wilks 2016).
For the tests performed on the 2-m temperature fields, the ocean grid points were excluded from the computation of the threshold. As the simulations are forced by the same sea surface temperatures whether or not the groundwater scheme is activated, there is very little difference in the 2-m temperatures over the ocean in PIa (C4a) compared to PIr (C4r). Therefore, when testing the significance of the temperature differences, the P values over ocean grid points are very close to zero, and this falsifies the computation of the significance threshold.
3. Results
a. Contribution of groundwater to soil moisture and transpiration under preindustrial conditions
As mentioned in section 2, the CNRM-CM6-1 groundwater scheme has been thoroughly validated in previous publications. Our purpose here is not to go over this validation again. In this subsection, we analyze the effects of groundwater on soil moisture and evaluate the order of magnitude of the groundwater contribution to the global transpiration flux under preindustrial conditions (PI simulations).
Figure 1a shows that aquifers present a rather shallow water table over a large portion of the land surface PIa (see Fig. B1 in appendix B for the seasonal variations of Figs. 1a, 1c, and 1d). As explained in section 2b, this water table depth is only representative of the “lowland” part of each grid cell, whose fraction is given by fwtd (Figs. 1b and B1); fwtd is larger over flat regions and when the WTD is shallow. The presence of groundwater significantly affects the root zone water content only if the water table is not much deeper than the plants rooting depth (less than ∼1.5 m) (Figs. 1c,d and B1).
(a) Mean annual water table depth (WTD) in PIa; (b) mean annual fraction of the grid cells over which capillary rise is allowed fwtd; (c) difference between the mean annual WTD and the vegetation rooting depth; and (d) mean annual root zone water content difference with and without groundwater (PIa − PIr). The mean WTD in Pia shown in (a) is representative of the fwtd fraction of the cell, shown in (b). These fields present very limited seasonal variations (see Fig. B1).
Citation: Journal of Climate 36, 21; 10.1175/JCLI-D-22-0767.1
In some regions, the mean water table is shallower than the rooting depth, whereas in reality roots do not grow in the saturated zone. On average over these regions, we find that 34% of the total root zone liquid water content is located below WTD. But deep roots have a low density; layers located below WTD only contribute to 2.2% of the total amount of water available to transpiration over these regions (the water available to transpiration is computed as the liquid water content weighted by the vertical profile of root density). Figure B2 shows this ratio of water availability below WTD for each grid cell. On average over the regions where WTD lies above the rooting depth, this ratio is equal to 1%. So the unrealistic presence of roots located below WTD in our model does not lead to a notable overestimation of transpiration. It also has a very limited impact on the increase of transpiration due to the presence of groundwater. As shown in Fig B2, most of the increase of vegetation water availability in PIa, compared to PIr, involves soil layers located above the water table (98% on global average and 96.5% on average over the regions where the WTD is shallower than the rooting depth). This means that the presence of shallow aquifers increases soil moisture mostly through the combined effect of capillary rise and a reduction of drainage efficiency.
We now consider the realism of the groundwater contribution to global transpiration simulated by CNRM-CM-6-1 under preindustrial conditions. The validations presented in previous publications offered an indirect validation of the evapotranspiration simulated in the presence of groundwater, as adding groundwater improved river discharge and terrestrial water storage annual cycles (Vergnes et al. 2012; Vergnes and Decharme 2012). However, the increase of evapotranspiration induced by the presence of groundwater falls within the range of uncertainties of these gridded estimates (Decharme et al. 2019), which makes it difficult to assess the realism of the groundwater impact on evapotranspiration at the global scale.
In a recent meta-analysis study of in situ data using water isotopes, groundwater was estimated to represent 23% of the global transpiration flux (Evaristo and McDonnell 2017). But as mentioned in section 1, this figure cannot be compared to the global increase of transpiration obtained with the activation of a groundwater scheme in a global climate model (+2% over the whole land surface and +8% above the large groundwater basins represented in CNRM-CM6-1). Indeed, when transpiration is not limited by soil moisture, shallow aquifers may still provide water to the vegetation and thus contribute to the transpiration fluxes in the observed data, but this situation will not result in an increase of transpiration when comparing simulations run with and without groundwater. If there is enough soil moisture to meet the evaporative demand in the first place, the addition of groundwater will not lead to an increase of transpiration. We can however derive an upper estimate of the proportion of groundwater transpired by groundwater-dependent ecosystems, Tgw, in PIa, the preindustrial simulation with aquifers (where the simulated climate is fairly close to the present-day one) and compare it with the results of another meta-analysis study of in situ data (Barbeta and Peñuelas 2017) which concluded that groundwater accounts for 38% of the global transpiration flux of groundwater-dependent ecosystems (instead of all ecosystems).
For each of the N grid points where the presence of groundwater affects the mean annual transpiration flux, [TPIa(i) − TPIr(i)] represents the increase of transpiration due to the inclusion of groundwater; it is always positive. The term fwtd(i)TPIa(i) corresponds to the transpiration flux over the fraction of the grid cell over which the vegetation can be considered groundwater-dependent. We find Tgw to be equal to 34%. If the increase of transpiration in PIa solely stemmed from groundwater, Tgw would represent the relative contribution of groundwater to transpiration flux for groundwater-dependent ecosystems in our model. But as previously mentioned, the increase of water available to transpiration in PIa is not only due to capillary rise from the aquifer but also to a less efficient drainage of soil moisture above shallow aquifers. Our modeling framework does not allow us to disentangle these two effects in order to quantify the actual contribution of groundwater in the increase of water availability. We can only state that groundwater contributes to part of the additional transpiration in PIa, making Tgw an upper estimate of the contribution of groundwater to transpiration. The value of Tgw being slightly lower than the 38% found by Barbeta and Peñuelas (2017), the contribution of groundwater to transpiration in groundwater-dependent environments is thus likely to be underestimated in our model. This may be due to the lack of dynamical rooting depth in ISBA, which could limit the uptake.
b. Groundwater impacts on climate under preindustrial conditions
Before we assess the impacts of the presence groundwater on the climate change signal between the preindustrial and 4xCO2 simulations, we explore the impact of groundwater on a stationary climate by comparing the PIa and PIr simulations (with and without aquifers). As explained in the introduction, a shallow water table depth is not sufficient to enhance evapotranspiration. For this to happen, a number of conditions have to be met. As shown in the previous subsection, the water table depth must not be much deeper than the plant rooting depth to affect the root zone water content. Then, for the increase of soil water content to translate into an increase of vegetation water availability, the soil must neither be frozen (as in northwestern Russia in winter and spring) nor already close to the field capacity (as in Indonesia) and the increase of water content must not affect only deep soil layers with a low root density (as in Amazonia) (Fig. 2). Finally, the increase of vegetation water availability must occur in water-limited regimes of evapotranspiration, which can be characterized by strong values of the cross-correlation between evapotranspiration and vegetation water availability in PIr (Fig. 2).
(left) Mean seasonal differences (PIa − PIr) of the vegetation water availability index (computed as the Soil Wetness Index, with a weighting of soil water content with root density along vertical layers). (center) Mean seasonal cross-correlation of water stress index with evapotranspiration. (right) Mean seasonal differences (PIa − PIr) of evapotranspiration.
Citation: Journal of Climate 36, 21; 10.1175/JCLI-D-22-0767.1
In our simulations, all these conditions are predominantly met during boreal summer and fall (JJA and SON) in western and eastern United States of America, northwestern Europe, northern Australia, the western part of the Brazilian Nordeste, the plateaus of Angola, and a wider area in the eastern part of the geographical Europe we will refer to as “eastern Europe.” Results indicate that this increase of evapotranspiration reduces the daily maximum temperatures by 0.5°–2°C in summer (JJA) over the three later regions (Brazil, Angola, and eastern Europe) but has no statistically significant effect on precipitation (Fig. 3).
Impact of groundwater under preindustrial conditions (PIa − PIr) on mean seasonal daily maximum 2-m (left) temperature and (right) precipitation.
Citation: Journal of Climate 36, 21; 10.1175/JCLI-D-22-0767.1
c. Regional impacts of groundwater on climate change
Climate change impacts on shallow aquifers (which are the ones susceptible to impact climate in return) are significant almost everywhere (Fig. 4). They are mostly driven by precipitation changes, as the recharge rates are mainly controlled by precipitation.
Climate change impact (C4a − PIa) on (left) mean seasonal precipitation, (center) precipitation minus evapotranspiration, and (right) water table depth shallower than 100 m. All differences are statistically significant at the 95% confidence level.
Citation: Journal of Climate 36, 21; 10.1175/JCLI-D-22-0767.1
In the southwestern United States, Brazil Nordeste, and Angola plateau regions, the water table is deeper under 4xCO2 conditions and thus groundwater has a smaller effect on evapotranspiration. However, this does affect the 2-m temperature and precipitation (Fig. 5).
Impact of groundwater under 4xCO2 conditions (C4a − C4r) on (left) seasonal evapotranspiration, (center) daily maximum 2-m temperature, and (right) precipitation. The stippling indicates statistically significant differences at the 95% confidence level.
Citation: Journal of Climate 36, 21; 10.1175/JCLI-D-22-0767.1
In eastern Europe, the situation is reversed with a higher impact of groundwater on evapotranspiration under 4xCO2 conditions. Figures 6 and 7 offer a closer look at the effects of groundwater on temperature and precipitation during the extended boreal summer season (June to September) in eastern Europe.
Impact of groundwater on the mean extended summer (JJAS) 2-m daily maximum air temperature (°C) under preindustrial and 4xCO2 conditions over Europe. (a) Climate change signal without groundwater (C4r − PIr); (b) climate change signal with groundwater (C4a − PIa); (c) impact of groundwater in the preindustrial simulations (PIa − PIr); (d) impact of groundwater in the 4xCO2 simulations (C4a − C4r); (e) impact of groundwater on the climate change signal [(d) − (c)]; (f) relative impact of groundwater on the climate change signal [100 × (e)/(a)]. In (c) and (d), the stippling shows significantly statistical differences at the 95% confidence level. In (a), (b), (e), and (f), the stippling is as in (d). The rectangle in (e) shows the “eastern Europe” box over which variables are spatially averaged for Figs. 8, 10, and B3 and Table 1.
Citation: Journal of Climate 36, 21; 10.1175/JCLI-D-22-0767.1
As in Fig. 6, but for the impact of groundwater on mean precipitation (mm day−1).
Citation: Journal of Climate 36, 21; 10.1175/JCLI-D-22-0767.1
In this region, the differences on the mean summer daily maximum temperatures remain below 1°C in PIa compared to PIr, but they reach 2°C in C4a compared to C4r, with a cooling zone spreading farther south (Fig. 6). In other words, there is a significant differential impact of groundwater ([C4r − C4a] − [PIa − PIr]) on maximum daily temperatures, which locally amounts to 20% of the climate change signal. If we consider the spatially averaged percentiles of daily minimum and maximum temperatures over eastern Europe (40°–62°N, 19°–60°E; Fig. B3), we find that the cooling induced by the presence of groundwater is stronger for the warmer values of temperatures.
The impact of groundwater on precipitation is not significant in the preindustrial simulations, but we find an increase of summer precipitation over eastern Europe in the warmer climate under 4xCO2 conditions (Fig. 7). In C4a, the mean precipitation is 0.4 mm day−1 larger (i.e., 30%) than it is in C4r, with a maximum relative increase centered around the median values of daily precipitation (Fig. B3). In this region, the climate change signal corresponds to a drying in the south and a wetting in the north, so the presence of aquifers leads to a southward shift of the drying/wetting limit.
To further understand the processes involved in this differential impact of groundwater, we now consider the annual cycles of the water exchanges between the atmosphere, the vadose zone, and the deep saturated zone (aquifers) in all four simulations over the eastern Europe box (Figs. 8 and 9).
Semi-conceptual sketch of the main mechanisms involved in the groundwater–climate feedbacks in a changing climate over eastern Europe.
Citation: Journal of Climate 36, 21; 10.1175/JCLI-D-22-0767.1
Mean annual cycles spatially averaged over the eastern Europe box drawn in Fig. 6e (40°–62°N, 19°–60°E) of (a) precipitation (mm day−1), (b) infiltration divided by precipitation, (c) water table depth (m), (d) vegetation water availability index [computed as the Soil Wetness Index (SWI) for liquid water in the root zone], (e) leaf area index, (f) evapotranspiration (mm day−1), and (g) transpiration divided by evapotranspiration, for PIa (blue solid line), PIr (blue dotted line), C4a (red solid line), and C4r (red dotted line).
Citation: Journal of Climate 36, 21; 10.1175/JCLI-D-22-0767.1
Figure 8 is a semi-conceptual sketch showing the interactions between the annual cycle changes shown in Fig. 9. In Fig. 8, the y axis represents the depth (in meters) below the surface, and the x axis the time of the year (in months). The seasonal variations of the soil moisture (SM) are shown by the colored shading that represents the mean annual cycle of the liquid water content (m3 m−3) in the vadose zone, averaged over the eastern Europe box in the C4a simulation. Below are the mean annual cycle of the water table depth (WTD) in the preindustrial (PI) and 4xCO2 simulations (C4). The arrows and inequalities illustrate the physical processes detailed in the main text. Here T refers to the air temperature, Pr the precipitation, ET the evapotranspiration, SMfroz/liq the ratio of frozen and liquid water contents in the vadose zone, Inflitr the infiltration of liquid water, and LAI the leaf area index.
We find that during the groundwater recharge season (from October to April/May), the precipitation rates are much larger in C4a than in PIa (+35%). Additionally, in the warmer climate under 4xCO2 conditions, there is less frozen water in the vadose zone, which allows for a better infiltration of the precipitation. Indeed, in the preindustrial climate, a larger fraction of the winter precipitation ends up in surface runoff, either as rain falls on a frozen ground or later on, during the spring thaw. These two features are part of climate change and their amplitude does not significantly differ whether or not groundwater is represented in the model. Combined, these features result in a larger groundwater recharge and a shallower water table with the 4xCO2 climate forcing. As the summer progresses, the water table deepens in both C4a and PIa, and the differences between the two are reduced as more water is transferred to the vadose zone in C4a. The presence of groundwater thus causes a larger gain of summer soil moisture with the 4xCO2 climate forcing. The induced increase of evapotranspiration is subsequently amplified, leading to stronger cooling and wetting effects of groundwater under the 4xCO2 conditions. Finally, the increase of precipitation in C4a creates a positive feedback on evapotranspiration. This feedback can explain why the plant transpiration does not increase more than the bare soil evaporation does, contrary to what could be expected and has been verified in offline settings where the land surface was not coupled to the atmosphere (Maxwell and Condon 2016).
d. Groundwater impacts on heat waves in eastern Europe
As groundwater has a larger impact on the warmer maximum daily temperatures over eastern Europe (Fig. B3), summer heat waves are likely to be affected as well, and possibly to a different extent with the preindustrial and 4xCO2 climate forcings.
Heat waves are here defined as events characterized by a duration, a spatial extent, and an intensity. The selection of days and grid points experiencing a heat wave is based on the exceedance of a percentile threshold computed in PIr (C4r) for the preindustrial (4xCO2) simulations: a grid cell is considered to experience a hot day when both the daily maximum and minimum temperatures exceed the 95th percentile of their reference distributions. For PIr and PIa (C4r and C4a), reference distributions are empirically estimated from Tmax and Tmin values of all JJAS days of the PIr (C4r) simulation. Then, a heat wave event is defined when at least 5% of the spatial domain (here the eastern Europe box previously defined) experiences a hot day for at least 3 consecutive days. This minimum extent has been defined in order to get a reasonable sample of heat waves in the PIr and C4r simulations. Heat waves separated by less than 3 days are concatenated. The heat wave mean intensity is defined as the maximum exceedance of the Tmax or Tmin criteria, averaged over the heat wave duration and all the grid points affected by the event. The heat wave severity is then defined as the product of duration, mean extent, and mean intensity. The average of the severity across several heat waves is performed through the geometric mean, which is less sensitive to very high departures than the arithmetic mean is. This procedure is a slightly adapted version of the procedure used in previous studies (Schoetter et al. 2015; Douville et al. 2016).
The statistical significance of changes in heat waves characteristics (duration, extent, intensity) is assessed with a bootstrap procedure: for each simulation we generate 1000 ensembles of N events randomly resampled among the N events of the simulation (with replacement), and then empirically estimate the 95% level confidence interval associated with each characteristic.
Figure 10 offers a two-dimensional view of the heat waves simulated over the eastern Europe box, giving the number of heat waves for each duration and range of extent. It shows that the effect of aquifers on heat waves is stronger in the 4xCO2 simulations. Overall, there are 57% fewer heat waves in C4a compared to C4r, with a decrease in the number of heat waves for almost every duration and extent. The mean duration and extent are respectively reduced by 18% and 12% while the mean intensity remains the same, and the mean severity (defined as the product of duration, extent, and intensity) is 39% weaker in C4a (Table 1). In PIa, the total number of heat waves is reduced by 15% compared to PIr, but the signal along the spectrum of durations and extents is somewhat unclear (Fig. 10); in fact, none of the mean or maximum features of heat waves are significantly reduced (Table 1).
(a)–(d) Number of heat waves (HW) events per duration (in days, x axis) and mean spatial extent [in % of the eastern Europe box domain (40°–62°N, 19°–60°E), y-axis] for the (a) PIr, (b) C4r, (c) PIa, and (d) C4a simulations. (e) Impact of groundwater in the preindustrial simulations [(c) minus (a)]. (f) Impact of groundwater in the 4xCO2 simulations [(d) minus (b)]. (g) Impact of groundwater on heat wave changes [(f) minus (e)]. For (a) and (d), the total number of heat waves events (N) is indicated in the top-right corner, as well as the mean severity normalized by the PIr value (S).
Citation: Journal of Climate 36, 21; 10.1175/JCLI-D-22-0767.1
Heat wave mean and maximum characteristics in all four simulations.
It is not possible to directly assess the impact of climate change on heat waves in our simulations because nearly every summer day in the 4xCO2 simulations meets the criteria defining a heat wave in the preindustrial climate. However, since the effects of groundwater on heat waves are larger under the 4xCO2 conditions, compared to the preindustrial climate, one can say that groundwater has a dampening effect on the climate change–induced worsening of heat waves in eastern Europe.
4. Discussion and conclusions
In this study, we carried out a set of four global climate simulations to assess the impact of groundwater on a stabilized climate, under preindustrial and 4xCO2 climate forcings.
Under preindustrial conditions, we found that the inclusion of groundwater has a limited yet significant impact on daily maximum 2-m temperatures in a number of regions (eastern Europe, parts of Brazil, and southern Africa) where the presence of shallow unconfined aquifers has a cooling effect in summer, due to an increase of evapotranspiration. Then we showed that in eastern Europe, this cooling effect of groundwater is stronger in the 4xCO2 simulations, thus reducing the intensity of the climate change–induced warming by 5%–20%. This differential impact of groundwater on summer temperatures translates into a reduced worsening of heat waves with climate change over this region.
We also found that while the presence of groundwater has no significant effect on precipitation in the preindustrial simulations, it leads to an increase of summer precipitation in eastern Europe in the 4xCO2 simulations, thus affecting the climate change signal with a northward shift of the drying/wetting limit in this region.
There are good reasons to assume that if anything, the groundwater–climate feedbacks could be underestimated in our simulations. In section 3a, we showed that the proportion of groundwater in the water transpired by plants that actually rely on aquifers was probably a little underestimated. Moreover, the fraction of this groundwater-dependent vegetation is likely to be underestimated because of the model’s resolution (0.5° for groundwater) and the lack of dynamical rooting depth.
The 0.5° resolution allows for a good representation of groundwater in relatively flat regions. But when the subgrid topography is more complex, the extent of shallow water table depths is underestimated, partly by construction (see section 2b) and partly because the lateral groundwater fluxes are weaker than they would be at a higher resolution (Krakauer et al. 2014). In CNRM-CM6-1, depending on the regions, the mean lateral groundwater flux is approximately 5–20 times smaller than the mean recharge flux (not shown)—locally, this ratio can drop below 1 or exceed 100. Therefore, most of the groundwater-induced increase of evapotranspiration is due to the use of the groundwater stored during the rainy season, and not to the spatial convergence of groundwater in valleys. However, there are regions such as the U.S. Rocky Mountains where the lateral flow was proven to dominate the groundwater influence on evapotranspiration (Forrester and Maxwell 2020). So the effects of groundwater may be underestimated in the regions we identified. And with a higher resolution, other regions could also turn out to be affected.
Another possible source of underestimation of the effects of groundwater lies in the fact that plant rooting depths are fixed in CNRM-CM6-1. As mentioned in sections 2 and 3a, studies have shown that to a certain extent, plants can grow deeper roots in drier environments to access an underlying groundwater resource (Fan et al. 2017, 2019). In regions where groundwater already affects climate in our simulations, the inclusion of a dynamical plant rooting depth could accentuate the increase of transpiration and the subsequent effects on air temperature and/or precipitation. The dynamical deepening of roots could also foster a groundwater–climate coupling in some of the regions where the simulated water table is currently too deep for groundwater to impact the atmosphere.
However, it is difficult to foresee how an increased resolution or a dynamical representation of plant rooting depth would affect the regional impact of groundwater on the climate change signal.
Ultimately, our study shows that even at the current resolution of global climate models and Earth system models, where the effects of groundwater may not be fully accounted for, it is worth representing aquifers, given that failing do so can regionally bias the model’s response to climate change. This conclusion supports the recommendations issued by other authors in the groundwater and climate modeling communities (Clark et al. 2015; Fan et al. 2019; Gleeson et al. 2021; Arboleda Obando et al. 2022) also calling for the inclusion of groundwater processes in Earth system models. And although the intensity and location of the groundwater impacts could vary from one model to another, the mechanism we unraveled should remain the same: wherever shallow water table depths coincide with water-limited regimes of evapotranspiration, groundwater may have a cooling and/or a wetting effect, and these effects are likely to grow stronger (weaker) in the future if mean precipitation rates increase (decrease) with climate change.
Acknowledgments.
This work was funded by Météo-France and the CNRS. The authors thank the entire CNRM-CM team for their support, in particular S. Sénési for his technical assistance, and Thomas Fiolleau who created the graphic elements of Fig. 3. We also thank the anonymous reviewers for their useful comments which helped improve the manuscript. We acknowledge the participants of the I-GEM project (ANR-14-CE01-0018) and the associated workshops for the fruitful discussions which helped put our work in perspective.
Data availability statement.
All the simulations outputs analyzed in this study are available in the Zenodo repository at https://zenodo.org/record/7137879.
APPENDIX A
Groundwater–Soil Coupling Formulation
APPENDIX B
Supplementary Figures
Figures B1–B3 provide supplementary informations regarding 1) the seasonal variations of WTD, fwtd, and root zone water content (Fig. B1), 2) the contribution of soil water content in layers deeper than WTD to water availability in PIa and the increase of water availability in PIa compared to PIr (Fig. B2) and 3) the changes variability in daily minimum and maximum 2-m temperatures and daily precipitation over the eastern Europe box (Fig. B3).
(top) Seasonal mean water table depth (WTD) in PIa; (center) seasonal mean fraction of the grid cells over which capillary rise are allowed fwtd; and (bottom) seasonal mean root zone water content difference with and without groundwater (PIa − PIr).
Citation: Journal of Climate 36, 21; 10.1175/JCLI-D-22-0767.1
Contribution of soil layers located below the water table depth to water availability (liquid water available weighted with the vertical root density profile). (top) contribution to water availability in PIa. (bottom) contribution to water availability changes (PIa − PIr).
Citation: Journal of Climate 36, 21; 10.1175/JCLI-D-22-0767.1
Differences on the summer (JJAS) percentile values (y axis) of (a) the daily maximum 2-m temperature (°C) and (b) the daily minimum 2-m temperature (°C). (c) Relative differences (%) of the summer percentile values (y axis) of the daily precipitation rates greater than 0.1 mm day−1 for each percentile rank (x axis). The percentiles were estimated empirically and their values were averaged over the eastern Europe box (40°–62°N, 19°–60°E). The blue lines represent the differences for the preindustrial simulations (PIa − PIr) and the red lines represent the differences in the 4xCO2 simulations (C4a − C4r), and the dots on the lines mark the 5th, 10th, 50th 75th, 90th, and 95th percentiles from left to right, respectively.
Citation: Journal of Climate 36, 21; 10.1175/JCLI-D-22-0767.1
REFERENCES
Anyah, R. O., C. P. Weaver, G. Miguez-Macho, Y. Fan, and A. Robock, 2008: Incorporating water table dynamics in climate modeling: 3. Simulated groundwater influence on coupled land–atmosphere variability. J. Geophys. Res., 113, D07103, https://doi.org/10.1029/2007JD009087.
Arboleda Obando, P. F., A. Ducharne, F. Cheruy, A. Jost, J. Ghattas, J. Colin, and C. Nous, 2022: Influence of hillslope flow on hydroclimatic evolution under climate change. Earth’s Future, 10, e2021EF002613, https://doi.org/10.1029/2021EF002613.
Ardilouze, C., L. Batté, B. Decharme, and M. Déqué, 2019: On the link between summer dry bias over the U.S. Great Plains and seasonal temperature prediction skill in a dynamical forecast system. Wea. Forecasting, 34, 1161–1172, https://doi.org/10.1175/WAF-D-19-0023.1.
Barbeta, A., and J. Peñuelas, 2017: Relative contribution of groundwater to plant transpiration estimated with stable isotopes. Sci. Rep., 7, 10580, https://doi.org/10.1038/s41598-017-09643-x.
Clark, M. P., and Coauthors, 2015: Improving the representation of hydrologic processes in Earth system models. Water Resour. Res., 51, 5929–5956, https://doi.org/10.1002/2015WR017096.
Danielson, J. J., and D. B. Gesch, 2011: Global multi-resolution terrain elevation data 2010 (GMTED2010). USGS Open-File Rep. 2011-1073, 26 pp., https://doi.org/10.3133/ofr20111073.
Decharme, B., A. Boone, A. Delire, and J. Noilhan, 2011: Local evaluation of the interaction between soil biosphere atmosphere soil multilayer diffusion scheme using four pedotransfer functions. J. Geophys. Res., 116, D20126, https://doi.org/10.1029/2011JD016002.
Decharme, B., and Coauthors, 2019: Recent changes in the ISBA‐CTRIP land surface system for use in the CNRM‐CM6 climate model and in global off‐line hydrological applications. J. Adv. Model. Earth Syst., 11, 1207–1252, https://doi.org/10.1029/2018MS001545.
Delire, C., and Coauthors, 2020: The global land carbon cycle simulated with ISBA‐CTRIP: Improvements over the last decade. J. Adv. Model. Earth Syst., 12, e2019MS001886, https://doi.org/10.1029/2019MS001886.
Dirmeyer, P. A., 2011: The terrestrial segment of soil moisture–climate coupling. Geophys. Res. Lett., 38, L16702, https://doi.org/10.1029/2011GL048268.
Dirmeyer, P. A., A. C. Schlosser, and K. L. Brubaker, 2009: Precipitation, recycling, and land memory: An integrated analysis. J. Hydrometeor., 10, 278–288, https://doi.org/10.1175/2008JHM1016.1.
Douville, H., J. Colin, E. Krug, J. Cattiaux, and S. Thao, 2016: Midlatitude daily summer temperatures reshaped by soil moisture under climate change. Geophys. Res. Lett., 43, 812–818, https://doi.org/10.1002/2015GL066222.
Douville, H., and Coauthors, 2020: Drivers of the enhanced decline of land near-surface relative humidity to abrupt-4xCO2 in CNRM-CM6-1. Climate Dyn., 55, 1613–1629, https://doi.org/10.1007/s00382-020-05351-x.
Dürr, H. H., M. Meybeck, and S. H. Dürr, 2005: Lithologic composition of the Earth’s continental surfaces derived from a new digital map emphasizing riverine material transfer. Global Biogeochem. Cycles, 19, GB4S10, https://doi.org/10.1029/2005GB002515.
Evaristo, J., and J. J. McDonnell, 2017: Prevalence and magnitude of groundwater use by vegetation: A global stable isotope meta-analysis. Sci. Rep., 7, 44110, https://doi.org/10.1038/srep44110.
Eyring, V., S. Bony, G. A. Meehl, C. A. Senior, B. Stevens, R. J. Stouffer, and K. E. Taylorand, 2016: Overview of the Coupled Model Intercomparison Project Phase 6 (CMIP6) experimental design and organization. Geosci. Model Dev., 9, 1937–1958, https://doi.org/10.5194/gmd-9-1937-2016.
Fan, Y., 2015: Groundwater in the Earth’s critical zone: Relevance to large‐scale patterns and processes. Water Resour. Res., 51, 3052–3069, https://doi.org/10.1002/2015WR017037.
Fan, Y., H. Li, and G. Miguez-Macho, 2013: Global patterns of groundwater table depth. Science, 339, 940–943, https://doi.org/10.1126/science.1229881.
Fan, Y., G. Miguez-Macho, E. G. Jobbágy, R. B. Jackson, and C. Otero-Casal, 2017: Hydrologic regulation of plant rooting depth. Proc. Natl. Acad. Sci. USA, 114, 10 572–10 577, https://doi.org/10.1073/pnas.1712381114.
Fan, Y., and Coauthors, 2019: Hillslope hydrology in global change research and Earth system modeling. Water Resour. Res., 55, 1737–1772, https://doi.org/10.1029/2018WR023903.
Ferguson, I. M., and R. M. Maxwell, 2010: Role of groundwater in watershed response and land surface feedbacks under climate change. Water Resour. Res., 46, W00F02, https://doi.org/10.1029/2009WR008616.
Fischer, E. M., S. I. Seneviratne, P. L. Vidale, D. Lüthi, and C. Schär, 2007: Soil moisture–atmosphere interactions during the 2003 European summer heat wave. J. Climate, 20, 5081–5099, https://doi.org/10.1175/JCLI4288.1.
Forrester, M. M., and R. M. Maxwell, 2020: Impact of lateral groundwater flow and subsurface lower boundary conditions on atmospheric boundary layer development over complex terrain. J. Hydrometeor., 21, 1133–1160, https://doi.org/10.1175/JHM-D-19-0029.1.
Furusho-Percot, C., K. Goergen, C. Hartick, L. Poshyvailo-Strube, and S. Kollet, 2022: Groundwater model impacts multiannual simulations of heat waves. Geophys. Res. Lett., 49, e2021GL096781, https://doi.org/10.1029/2021GL096781.
Giffard, P., W. Llovel, J. Jouanno, G. Morvan, and B. Decharme, 2019: Contribution of the Amazon River discharge to regional sea level in the tropical Atlantic Ocean. Water, 11, 2348, https://doi.org/10.3390/w11112348.
Gleeson, T., and Coauthors, 2021: GMD perspective: The quest to improve the evaluation of groundwater representation in continental- to global-scale models. Geosci. Model Dev., 14, 7545–7571, https://doi.org/10.5194/gmd-14-7545-2021.
Hirschi, M., and Coauthors, 2011: Observational evidence for soil-moisture impact on hot extremes in southeastern Europe. Nat. Geosci., 4, 17–21, https://doi.org/10.1038/ngeo1032.
Jiang, X., G.-Y. Niu, and Z.-L. Yang, 2009: Impacts of vegetation and groundwater dynamics on warm season precipitation over the central United States. J. Geophys. Res., 114, D06109, https://doi.org/10.1029/2008JD010756.
Keune, J., F. Gasper, K. Goergen, A. Hense, P. Shrestha, M. Sulis, and S. Kollet, 2016: Studying the influence of groundwater representations on land surface‐atmosphere feedbacks during the European heat wave in 2003. J. Geophys. Res. Atmos., 121, 13 301–13 325, https://doi.org/10.1002/2016JD025426.
Koster, R. D., and Coauthors, 2006: GLACE: The Global Land–Atmosphere Coupling Experiment. Part I: Overview. J. Hydrometeor., 7, 590–610, https://doi.org/10.1175/JHM510.1.
Krakauer, N. Y., H. Li, and Y. Fan, 2014: Groundwater flow across spatial scales: Importance for climate modeling. Environ. Res. Lett., 9, 034003, https://doi.org/10.1088/1748-9326/9/3/034003.
Larsen, M., J. Christensen, M. Drews, M. B. Butts, and J. C. Refsgaard, 2016: Local control on precipitation in a fully coupled climate–hydrology model. Sci. Rep., 6, 22927, https://doi.org/10.1038/srep22927.
Ledoux, E., G. Girard, G. de Marsily, J.-P. Villeneuve, and J. Deschenes, J., 1989: Spatially distributed modeling: Conceptual approach, coupling surface water and groundwater. Unsaturated Flow in Hydrologic Modeling, H. J. Morel-Seytoux, Eds., NATO ASI Series, Vol. 275, Springer, 435–454, https://doi.org/10.1007/978-94-009-2352-2_16.
Leung, L. R., M. Huang, Y. Qian, and X. Liang, 2011: Climate–soil–vegetation control on groundwater table dynamics and its feedbacks in a climate model. Climate Dyn., 36, 57–81, https://doi.org/10.1007/s00382-010-0746-x.
Maxwell, R. M., and S. J. Kollet, 2008: Interdependence of groundwater dynamics and land-energy feedbacks under climate change. Nat. Geosci., 1, 665–669, https://doi.org/10.1038/ngeo315.
Maxwell, R. M., and L. E. Condon, 2016: Connections between groundwater flow and transpiration partitioning. Science, 353, 377–380, https://doi.org/10.1126/science.aaf7891.
Maxwell, R. M., F. K. Chow, and S. J. Kollet, 2007: The groundwater–land-surface–atmosphere connection: Soil moisture effects on the atmospheric boundary layer in fully-coupled simulations. Adv. Water Resour., 30, 2447–2466, https://doi.org/10.1016/j.advwatres.2007.05.018.
Miralles, D. G., M. J. Van Den Berg, A. J. Teuling, and R. A. M. De Jeu, 2012: Soil moisture–temperature coupling: A multiscale observational analysis Geophys. Res. Lett., 39, L21707, https://doi.org/10.1029/2012GL053703.
Mu, M., A. J. Pitman, M. G. De Kauwe, A. M. Ukkola, and J. Ge, 2022: How do groundwater dynamics influence heatwaves in southeast Australia? Wea. Climate Extremes, 37, 100479, https://doi.org/10.1016/j.wace.2022.100479.
Munier, S., and B. Decharme, 2022: River network and hydro-geomorphological parameters at 1/12° resolution for global hydrological and climate studies. Earth Syst. Sci. Data, 14, 2239–2258, https://doi.org/10.5194/essd-14-2239-2022.
Padrón, R. S., and Coauthors, 2020: Observed changes in dry-season water availability attributed to human-induced climate change. Nat. Geosci., 13, 477–481, https://doi.org/10.1038/s41561-020-0594-1.
Pellet, V., F. Aires, F. Papa, S. Munier, and B. Decharme, 2020: Long-term total water storage change from a Satellite Water Cycle reconstruction over large southern Asian basins. Hydrol. Earth Syst. Sci., 24, 3033–3055, https://doi.org/10.5194/hess-24-3033-2020.
Roehrig, R., and Coauthors, 2020: The CNRM global atmosphere model ARPEGE‐Climat 6.3: Description and evaluation. J. Adv. Model. Earth Syst., 12, e2020MS002075, https://doi.org/10.1029/2020MS002075.
Saint-Martin, D., and Coauthors, 2021: Tracking changes in climate sensitivity in CNRM climate models. J. Adv. Model. Earth Syst., 13, e2020MS002190, https://doi.org/10.1029/2020MS002190.
Schoetter, R., J. Cattiaux, and H. Douville, 2015: Changes of western European heat wave characteristics projected by the CMIP5 ensemble. Climate Dyn., 45, 1601–1616, https://doi.org/10.1007/s00382-014-2434-8.
Séférian, R., and Coauthors, 2019: Evaluation of CNRM Earth‐System model, CNRM‐ESM2‐1: Role of Earth system processes in present‐day and future climate. J. Adv. Model. Earth Syst., 11, 4182–4227, https://doi.org/10.1029/2019MS001791.
Seneviratne, S. I., T. Corti, E. L. Davin, M. Hirschi, E. B. Jaeger, I. Lehner, B. Orlowsky, and A. J. Teuling, 2010: Investigating soil moisture–climate interactions in a changing climate: A review. Earth-Sci. Rev., 99, 125–161, https://doi.org/10.1016/j.earscirev.2010.02.004.
Taylor, K. E., D. Williamson, and F. Zwiers, 2000: Report 60: 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, 28 pp., https://pcmdi.llnl.gov/report/ab60.html.
Vergnes, J.-P., and B. Decharme, 2012: A simple groundwater scheme in the TRIP river routing model: Global off-line evaluation against GRACE terrestrial water storage estimates and observed river discharges. Hydrol. Earth Syst. Sci., 16, 3889–3908, https://doi.org/10.5194/hess-16-3889-2012.
Vergnes, J.-P., B. Decharme, R. Alkama, E. Martin, F. Habets, and H. Douville, 2012: A simple groundwater scheme for hydrological and climate applications: Description and offline evaluation over France. J. Hydrometeor., 13, 1149–1171, https://doi.org/10.1175/JHM-D-11-0149.1.
Vergnes, J.-P., B. Decharme, and F. Habets, 2014: Introduction of groundwater capillary rises using subgrid spatial variability of topography into the ISBA land surface model. J. Geophys. Res. Atmos., 119, 11 065–11 086, https://doi.org/10.1002/2014JD021573.
Voldoire, A., and Coauthors, 2019: Evaluation of CMIP6 DECK experiments with CNRM‐CM6‐1. J. Adv. Model. Earth Syst., 11, 2177–2213, https://doi.org/10.1029/2019MS001683.
Wang, F., A. Ducharne, F. Cheruy, M.-H. Lo, and J.-Y. Grandpeix, 2018: Impact of a shallow groundwater table on the global water cycle in the IPSL land–atmosphere coupled model. Climate Dyn., 50, 3505–3522, https://doi.org/10.1007/s00382-017-3820-9.
Wilks, D. S., 2016: “The stippling shows statistically significant grid points”: How research results are routinely overstated and overinterpreted, and what to do about it. Bull. Amer. Meteor. Soc., 97, 2263–2273, https://doi.org/10.1175/BAMS-D-15-00267.1.