Improving the Spatial Representation of Reservoir Evaporation Using SAR-Based Wind Fields

Evaporation losses from reservoirs can be substantial, yet spatial variability in evaporation rates is difficult to quantify, in part due to the difficulty of estimating high-resolution wind fields at the water-air interface. Synthetic aperture radar (SAR) offers a promising approach to estimate high-resolution wind fields over lakes. To test the utility of SAR-based winds to improve spatial representation of reservoir evaporation, we developed a framework fusing Sentinel-1 based wind fields with gridded meteorological estimates of air temperature, humidity, and solar radiation to estimate distributed evaporation rates at Lake Mead, USA using the Penman open water evaporation equation. We validated outputs at over-water buoys and benchmarked the results against wind and evaporation estimates derived solely from the Real-Time Mesoscale Analysis (RTMA) data product. Sentinel-1 wind field accuracy matched or outperformed RTMA, with a wind direction mean absolute error (MAE) of 41.3° compared to 40.3° and a wind speed MAE of 1.3 m/s compared to 1.8 m/s. This framework yielded a 25x increase in spatial resolution of wind fields (100 m) and evaporation (100 m) compared to RTMA-alone (2500 m), enabling identification of greater spatial variability across Lake Mead than was previously possible. This framework could be used globally for high-resolution evaporation mapping to support water resources monitoring, modeling, and management.