ADVANCING EVAPOTRANSPIRATION MODELING AND WATER USE EFFICIENCY ASSESSMENT USING A REMOTE SENSING-BASED TWO-SOURCE ENERGY BALANCE MODELING FRAMEWORK

Abstract

Evapotranspiration (ET) is a critical component of the water cycle, and accurate estimation, monitoring, and mapping of field-scale ET are essential for assessing crop water status and supporting sustainable agricultural water management. The accuracy of existing and newly developed ET products and tools should be evaluated across a range of cropping systems and geographic regions to assess their reliability and applicability for robust crop water use assessment. This thesis examined the performance and utility of the Shuttleworth-Wallace Sentinel-2 (SW-S2) two-source energy balance model to monitor crop water use (ET) and water use efficiency (WUE) across two contrasting agricultural systems, winter wheat (Triticum aestivum L.) in Oklahoma and rice (Oryza sativa L.) in Bangladesh. First, a new SW-S2 ET model was evaluated against eddy covariance flux tower measurements collected from winter wheat sites in Oklahoma between 2018 and 2024 and compared its performance against readily available OpenET and ECOsystem Spaceborne Thermal Radiometer Experiment on Space Station (ECOSTRESS) Jet Propulsion Laboratory ET (JET) products. SW-S2 produced ET estimates that were generally comparable to OpenET products, and results suggest that these products can complement each other. The frequent revisit time of Sentinel-2 improved temporal coverage and captured daily ET variations within the growing season effectively, although model performance declined under extreme dry conditions. The SW-S2 model was then applied to rice fields in Bangladesh, where direct ET measurements were not available, to determine whether remotely sensed variables could distinguish Alternate Wetting and Drying (AWD), a water conservation practice in rice fields, from non-AWD irrigation practices. Crop water use was found to be lower in AWD plots compared to non-AWD plots. Among the different remotely sensed variables examined, the ratio of evaporation to evapotranspiration (E/ET) and the ratio of evapotranspiration to reference evapotranspiration (ET/ETo), both derived from the SW-S2 model, showed the strongest ability to separate the two irrigation practices. The random forest classification, using E/ET and ET/ETo along with other variables, achieved high accuracy for identifying AWD plots but showed comparatively weaker ability to detect non-AWD fields. Overall, these findings show both the opportunities and challenges associated with the use of remotely sensed ET products in crop water use monitoring and provide valuable insights into sustainable irrigation management across different agricultural systems across the globe.