Improving maize crop yield estimation by assimilating remotely sensed biophysical variables in AquaCrop agrometeorological model : a case study of Western Kenya

Details

Supervisors

Defourny, Pierre

Faculty

Faculté des bioingénieurs

Degree label

Master [120] en sciences agronomiques et industries du vivant, à finalité spécialisée: Geo-Information Science and Earth Observation for Environmental Modelling and Management

Abstract

Crop yield estimation is vital in providing insights to policymakers on the state of food security within a given country. The application of crop yield estimation modelling has been realised to provide important information for data-driven decision-making. However, the models used have some shortcomings in providing accurate yield estimations. Studies have shown that yield estimation significantly improves by assimilating remotely sensed biophysical variables into the crop estimation models to curb this challenge. This study employs AquaCrop agro-meteorological model in maize crop yield estimation, following its successful application in previous studies in arid and semi-arid areas (ASALs). The study addresses the simulation of maize yield under rain-fed conditions in Western Kenya within two counties, i.e., Uasin Gishu and Keiyo Marakwet, for 2021. The meteorological data used in the model was obtained from ERA5-land, which included precipitation, maximum temperature, minimum temperature, and potential evapotranspiration. In addition, satellite-derived FCover obtained from Sentinel-2 was assimilated into the model to improve its performance. The model simulation performance of maize yield before FCover assimilation was R2 = 0.198 at the parcel level and R2 = 0.812 at the ward level. After the FCover assimilation, the accuracy improved to R2 = 0.334 at the parcel level and R2=0.843 at the ward level. These results demonstrate that remotely sensed biophysical variables improve the performance of crop yield estimation models.