IN AND END OF SEASON SOYBEAN YIELD PREDICTION WITH HISTOGRAM BASED DEEP LEARNING

Abstract

Abstract. One sector that feels the effects of global warming and climate change on all levels is agriculture. In order to prepare for possible yield loss, as well as market, storage, and import planning challenges brought on by climate change, businesses can utilise agricultural decision support applications. Within the scope of this study, a crop yield prediction module has been developed that can provide in and end of season estimation of crop yields to be obtained from the determined regions. The Python programming language was used in the creation of the module as a QGIS plugin. The area for which crop yield predictions are to be made is covered by retrieving MODIS SR, MODIS LST, and Daymet data from the Google Earth Engine data catalogue. Histograms obtained from remotely sensed images are used as input data to two deep learning methods (CNN-LSTM and HistCNN). As a result, the HistCNN model outperformed CNN-LSTM for in season soybean yield prediction, with an R2 of 0.72, while the CNN-LSTM model outperformed it for in end of season soybean yield prediction, with an R2 of 0.67.