Improving groundwater evapotranspiration modeling in saline areas by integrating remote sensing data

Groundwater evapotranspiration (ETg) is a crucial upward water flux in the water budget, especially in arid and semi-arid saline areas. Modeling ETg is challenging as it involves complex biogeophysical processes in both soil and vegetation dynamics. However, these processes are vastly oversimplified in commonly used process-based models like MODFLOW, where the ETg modeling relies solely on groundwater table depth. To disentangle this issue, this study presents the Evapotranspiration Package with Multi-factor (ETM), an enhancement to MODFLOW which additionally incorporates soil properties, vegetation information, and salinity levels to simulate spatiotemporal ETg. Compared to the original MODFLOW-EVT package, the proposed ETM package mitigates structural uncertainty by involving external soil and vegetation information based on optical remote sensing data. We conducted intensive experiments in Hetao, a one-thousand-year irrigation district in China. Daily groundwater table depth time-series for 108 observation wells were collected and used for calculating ground truth ETg based on the groundwater level fluctuation method. We evaluate the proposed ETM package in both well-level and regional-level experiments. In the well-level experiments, the ETM outperformed the EVT package with the coefficient of determination increasing from -1.698 to 0.449 and the RMSE reducing from 1.906 mm to 0.861 mm. Additionally, we employed the ETM package to model regional ETg for a 3,000-ha experimental area. Compared to the original EVT package which primarily considers groundwater level and results in more homogeneous outputs, the proposed ETM package demonstrated diverse ETg estimates in which the spatial pattern aligns with the prior knowledge. This improved approach addresses the shortcomings of previous models and contributes to more informed agricultural water resource management and planning through a deeper understanding of groundwater dynamics.