A deep learning-based hybrid model of global terrestrial evaporation

Abstract

Terrestrial evaporation (E) is a key climatic variable that is controlled by a plethora of environmental factors. The constraints that modulate the evaporation from plant leaves (or transpiration, Et) are particularly complex, yet are often assumed to interact linearly in global models due to our limited knowledge based on local studies. Here, we train deep learning algorithms using eddy covariance and sap flow data together with satellite observations, aiming to model transpiration stress (St), i.e., the reduction of Et from its theoretical maximum. Then, we embed the new St formulation within a process-based model of E to yield a global hybrid E model. In this hybrid model, the St formulation is bidirectionally coupled to the host model at daily timescales. Comparisons against in situ data and satellite-based proxies demonstrate an enhanced ability to estimate St and E globally. The proposed framework may be extended to improve the estimation of E in Earth System Models and enhance our understanding of this crucial climatic variable.

Recent global decline in rainfall interception loss due to altered rainfall regimes

“…Our model setup of NN was consistent with what used in Zhao et al 28 . The hybrid model has proved to outperform pure NN in ET prediction particularly under climate extremes and for out-of-sample extrapolation 15 , 28 . This improved behavior is the rationale for the use of this approach such that a hybrid model fitted outside of rainy conditions (HM dry ) can be compared to the hybrid model fitted during and right after rainy conditions (HM wet ).…”

“…EC measurements do offer the opportunity to measure latent heat flux (LE, ET in the form of energy) at the ecosystem level as opposed to the tree level. Such ecosystem-level measurements have great capacity to extrapolate to large spatial scales by leveraging machine learning (ML) algorithms and Earth observations 12 – 15 . Nevertheless, EC towers do not directly measure E i , or indirectly partition the E i part of this water flux.…”

A Pareto‐Based Sensitivity Analysis and Multiobjective Calibration Approach for Integrating Streamflow and Evaporation Data

“…The intensification of the global water cycle driven by an already changing climate represents a challenge for future water security (Lehner et al., 2019), and is likely to affect the occurrence of extreme events such as droughts and floods (Blöschl et al., 2020; Miralles et al., 2019; Peterson et al., 2021). Moreover, evaporation is projected to increase in the context of global warming (IPCC, 2021), and its changes can pose an important threat for water resources availability and the biosphere (Konapala et al., 2020; Koppa et al., 2022).…”

“…The growing availability of evaporation products provides an unprecedented opportunity to improve the representation of the land‐atmosphere interactions within the hydrologic system, effectively overcoming the inherent limitations of the streamflow‐only calibration (Dembélé, Ceperley, et al., 2020; Dembélé, Hrachowitz, et al., 2020; Koppa et al., 2019). While streamflow provides valuable information about the dynamics of the hydrologic system, evaporation data can help understand other relevant spatiotemporal processes, particularly in data‐scarce areas (Dembélé, Hrachowitz, et al., 2020; López López et al., 2017) and in arid and semiarid regions where evaporative fluxes are dominant (Dembélé, Ceperley, et al., 2020; Koppa et al., 2022).…”

“…The intensification of the global water cycle driven by an already changing climate represents a challenge for future water security (Lehner et al, 2019), and is likely to affect the occurrence of extreme events such as droughts and floods (Blöschl et al, 2020;Miralles et al, 2019;Peterson et al, 2021). Moreover, evaporation is projected to increase in the context of global warming (IPCC, 2021), and its changes can pose an important threat for water resources availability and the biosphere (Konapala et al, 2020;Koppa et al, 2022).The scientific underpinning for understanding and monitoring the water cycle relies upon simulations of water stores and fluxes (Fersch et al, 2020;Pokhrel et al, 2021). Our ability to reproduce existing hydroclimatic conditions and anticipate their future changes depends on an accurate representation of hydrologic processes within both climate and hydrologic model structures (Clark et al, 2015;Koppa et al, 2022).…”