Estimation of wheat yields by using remotely sensed and modeled agro-meteorological data-driven statistical and crop growth models

Download
2021-6
Bulut, Burak
Estimation of wheat yield is essential not only for agricultural sectors but also for making economic and strategic decisions at the national level. In this thesis, wheat yield estimation was carried out on the cities and districts where the highest wheat production is made with rainfed agriculture in Turkey and on TİGEM research farms. Two different modeling approaches were evaluated within the scope of the thesis; a statistical multiple linear regression (MLR) model based on analysis of possible agro-meteorological variables and periods that affects wheat yield and a crop growth model (AquaCrop) adapted to regional operation. Wheat yields were estimated on the study areas using grid-based agro-meteorological data obtained from remote sensing and reanalysis sources. The performance of both models has been validated using independent validation methods. The AquaCrop adapted for regional wheat estimation validation statistics were calculated as 40.6 kg/da RMSE on city-based, 47.3 kg/da on district-based, and 79.2 kg/da on farm-based models. In addition, the r2 values were calculated as 0.78, 0.65, and 0.69 for the city, district, and farm-based models. The MLR model statistics for the prediction year 2019 were calculated over cities, districts, and farms as 28.5 kg/da, 52.5 kg/da, and 74.6 kg/da RMSE, and the r2 values were calculated as 0.90, 0.82, and 0.59. The results obtained from the study show that wheat yields are predicted consistently in both model approaches. The results obtained from the study show that wheat yields are predicted consistently in both model approaches.
Citation Formats
B. Bulut, “Estimation of wheat yields by using remotely sensed and modeled agro-meteorological data-driven statistical and crop growth models,” Ph.D. - Doctoral Program, Middle East Technical University, 2021.