Integrating remotely sensed evapotranspiration into hydrological modelling of Mediterranean tree-grass ecosystems

Mediterranean mountain ecosystems play a key role in water supply and biodiversity conservation but are highly vulnerable to climate change. In these environments, vegetation strongly regulates key hydrological fluxes such as evapotranspiration and interception. However, the complex structure and marked seasonal dynamics of Mediterranean agroforestry systems remain poorly represented in many hydrological models, which often rely on static vegetation parameterizations or climate-driven evapotranspiration formulations. 

In this study, we investigate the integration of remotely sensed evapotranspiration and vegetation dynamics into the distributed, physically based hydrological model WiMMed (Watershed Integrated Model in Mediterranean regions) for a mountain catchment located in the Parque Natural de Cardeña y Montoro (southern Spain). Vegetation heterogeneity and dynamics are explicitly accounted for by combining MODIS-derived evapotranspiration estimates obtained from the Two-Source Energy Balance (TSEB) model with satellite-based fractional vegetation cover (FCV). Model performance and hydrological responses are evaluated against a conventional modelling approach based on a crop-modified Hargreaves formulation and static vegetation representation. 

Differences between the two approaches are analyzed in terms of evapotranspiration patterns, soil moisture dynamics, interception, and runoff generation. The inclusion of remotely sensed ET improved the simulation of water balance components and their spatial variability. Dynamic vegetation scenarios better capture seasonal and interannual variability in ET and runoff, highlighting vegetation-water interactions that are not reproduced by climate-only ET formulations.  Results show that neglecting vegetation dynamics or assuming static full cover leads to substantial biases in water flux partitioning, e.g., static vegetation scenarios “overestimate” interception (up to ~11.5% of annual precipitation), whereas incorporating dynamic vegetation reduces interception to ~3–4% and significantly alters the partitioning between infiltration and runoff, particularly during wet years. Overall, this approach provides a framework for identifying system inflection points, evaluating future climate and land-use scenarios.