Evaluating a land surface model at a water-limited site: implications for land surface contributions to droughts and heatwaves

Abstract

Abstract. Land surface models underpin coupled climate model projections of droughts and heatwaves. However, the lack of simultaneous observations of individual components of evapotranspiration, concurrent with root-zone soil moisture, has limited previous model evaluations. Here, we use a comprehensive set of observations from a water-limited site in southeastern Australia including both evapotranspiration and soil moisture to a depth of 4.5 m to evaluate the Community Atmosphere-Biosphere Land Exchange (CABLE) land surface model. We demonstrate that alternative process representations within CABLE had the capacity to improve simulated evapotranspiration, but not necessarily soil moisture dynamics–highlighting problems of model evaluations against water fluxes alone. Our best simulation was achieved by resolving a soil evaporation bias, using a more realistic initialisation of the groundwater aquifer state and higher vertical soil resolution informed by observed soil properties, and further calibrating soil hydraulic conductivity. Despite these improvements, the role of the empirical soil moisture stress function in influencing the simulated water fluxes remained important: using a site-calibrated function reduced the soil water stress on plants by 36 % during drought and 23 % at other times. These changes in CABLE not only improve the seasonal cycle of evapotranspiration but also affect the latent and sensible heat fluxes during droughts and heatwaves. The range of parameterisations tested led to differences of ∼150 W m−2 in the simulated latent heat flux during a heatwave, implying a strong impact of parameterisations on the capacity for evaporative cooling and feedbacks to the boundary layer (when coupled). Overall, our results highlight the opportunity to advance the capability of land surface models to capture water cycle processes, particularly during meteorological extremes, when sufficient observations of both evapotranspiration fluxes and soil moisture profiles are available.

Increasing aridity will not offset CO2 fertilization in fast‐growing eucalypts with access to deep soil water

“…Beyond this particular eucalypt plantation, our findings suggest that projections from models relying on SWC observations in the upper soil layers will overestimate drought impacts on tree species with the ability to mobilize water from deep soil (e.g., Humphrey et al, 2018;Uddin et al, 2018). Therefore, including the ability of vegetation to access deep soil water reservoirs could improve dy- Thus far, stand-scale simulation models tend to better describe tree water uptake than DGVMs (e.g., Mu et al, 2021;Ukkola et al, 2016).…”

“…Additionally, six levels of precipitation were considered: +5%, +0%, −5%, −10%, −15%, and −20% of annual precipitation; no seasonal variations were considered. Although long-term deep soil water dynamics are strongly linked to precipitation patterns (Berg et al, 2017;Berg & Sheffield, 2018 and references therein;Seneviratne et al, 2010), there are still important uncertainties about the dynamics of such soil water reservoirs in SW Australia, mostly driven by the impact of vegetation (Mu et al, 2021). Therefore, for each precipitation reduction scenario, we assumed a linear decrease in S deep proportional to the reduction in precipitation.…”

“…Nevertheless, in the long term, deep soil water availability is tightly linked to precipitation patterns over land (Berg et al, 2017). In addition, increased water consumption by vegetation due to increasing evaporative demand may reduce such deep soil water reservoirs (Mu et al, 2021). In the context of increasing aridity, this may limit productivity and survival of eucalypts relying on such deep soil water reservoirs.…”

Simulating shrubs and their energy and carbon dioxide fluxes in Canada's Low Arctic with the Canadian Land Surface Scheme Including Biogeochemical Cycles (CLASSIC)

“…This study addressed a high soil evaporation bias that has also been observed in other models' simulations of grasslands, wetlands, and forests, but especially in regions with sparse vegetation (Sun and Verseghy, 2019;Decker et al, 2017;Kauwe et al, 2017;Mu et al, 2021). The empirical formulation by Merlin et al (2011) of the soil evaporation efficiency reduces simulated high rates of ground evaporation, particularly in spring, which avoided overdrying the 0-10 cm soil layer and greatly underestimating summer soil temperature, LE, and GPP, particularly in years with less summer rainfall.…”

“…The parameters η and κ are PFT-specific (Table 2). The parameter values for trees, crops, and grasses are based on Lüdeke et al (1994), while η and κ for shrubs were estimated from values of C s , C R , and C L for shrub tundra (Nobrega and Grogan, 2007;Grogan and Chapin, 2000;Murphy et al, 2009;Wang et al, 2016), resulting in the same κ values as for grasses but higher η (Table 2). (Lafleur and Humphreys, 2018) and 200 to 300 mm of precipitation on average (ECG, 2012).…”