Mapping Evapotranspiration and Diffuse Recharge via Variational Assimilation of GOES LST and SMAP Soil Moisture

Evapotranspiration (ET) and groundwater recharge are fundamental components of the terrestrial water cycle, critical for understanding land–atmosphere interactions, ecosystem dynamics, and climate feedbacks. However, large‐scale mapping of these fluxes is hindered by sparse in situ observations and uncertainties in land surface models. We present a scalable variational data assimilation (VDA) framework—the Land Integrated Data Assimilation system (LIDA‐2)—for mapping ET and recharge by jointly assimilating hourly GOES land surface temperature (LST; 5 km) and SMAP surface soil moisture (SSM; 9 km, every 2–3 days) into a coupled water balance and dual‐source energy model. LIDA‐2 estimates key parameters—soil and canopy evaporative fractions (EF s , EF c ), bulk heat transfer coefficients (C HN ), and the Brooks–Corey parameter (B)—while quantifying uncertainty via second‐order Hessian analysis. Implemented over the Oklahoma Panhandle and Southern Great Plains (2016–2018) at 5 km resolution, LIDA‐2 produces high‐resolution maps of ET and recharge. Model outputs were evaluated against in situ soil moisture, latent heat flux, ERA5‐Land data, and groundwater well anomalies. Results show substantial improvement in ET and soil moisture estimates relative to open‐loop runs, particularly during dry‐down periods. Annual recharge patterns match hydrogeologic gradients and prior regional studies, while daily soil moisture and EF outputs reveal seasonally varying land–atmosphere coupling: strongest in spring, greater in sandy soils, and higher in grasslands than croplands. The framework preserves physical realism, supports ecohydrological analysis, and offers a robust, observation‐informed approach—explicitly designed for GOES LST and SMAP SSM—for mapping water and energy fluxes at scale, with applications to drought monitoring, groundwater sustainability, and climate feedback research.