Predicting rice yield for Bangladesh by exploiting weather conditions

Due to climate change and its impact on agriculture, accurate estimate of future crops yield is very critical for low-lying countries like Bangladesh. Rice is the most consumed crop for Bangladesh with an annual production of more than 40 million tons per year. Despite being gifted with multiple seasons suitable for cultivating variety of rice, its productivity in Bangladesh has been changing following climatic variationsin this region over the last decades. In this paper, we present an approach, called WPSRY (Weather-based Prediction System for Rice Yield), for forecasting rice yield in different regions of Bangladesh. The proposed approach (WPSRY) firstly builds a model to predict the weather parameters applying Neural Networks (NN), and then estimates the rice yields applying Support Vector Regression (SVR) that uses as inputs predicted weather from NN as well as current agricultural data. Simulation demonstrates that WPSRY approach achieves promising prediction accuracy.