An ensemble data assimilation approach to improve farm-scale actual evapotranspiration estimation

Abstract

Estimation of actual evapotranspiration (ET) is key to irrigation water application and basin-scale agricultural water demand assessment. While modelers and water managers rely on stand-alone ET estimation model application in their planning and management, several uncertainties including model structure, parameter set, and initial condition exist, cascading in ET calculation leading to inaccurate results. In this study, an ensemble data assimilation approach is employed to explore the benefit of remotely sensed actual ET to improve the simulations of the widely used Priestley-Taylor ET model while accounting for uncertainties. The study is conducted at farm-scale for three different crops (corn, cotton, and soybean) in the Mobile River basin in Deep South United States, which has experienced severe droughts during the cropping seasons in the past. Prior to employing data assimilation, the Priestley-Taylor model is modified for each crop to simulate actual ET instead of reference ET. Following which the model is calibrated over 320,000 farms in the river basin for identifying the optimal parameters. The calibrated model is later used for the Open-Loop simulation, as well as in the development and implementation of data assimilation. The simulated and observed actual ET is used to calculate the Kalman gain and update the model initialization every time step during the assimilation period. The findings of the study showed that assimilating the actual ET observation into the Priestley-Taylor model results in more accurate and reliable model initialization and also posterior ET estimates at farm-scale compared to open-loop simulation. These results highlight, the importance of digital agricultural tools in robust agricultural planning and management and open door for further research.