Synergizing Open Geospatial Data from Cloud Platform and Machine Learning (ML) Algorithms in Open Source Environment for Crop Yield Estimation

Climate change has a significant influence on the production and maintenance of agricultural crops. Predicting or forecasting the crop yield well in advance assists farmers, agriculture companies and governmental bodies in risk management and the development of appropriate crop insurance policies. The traditional method of estimating yield is a ground survey, which is subjective, costly and time-consuming. Empirical and crop growth methods using weather data have been developed to resolve issues of the traditional method, but these approaches also have issues owing to the complexity involved due to the spatial distribution of weather stations and high data demand, which causes the production of information to be delayed. Moreover, complete estimates from these methods are made available only after the actual harvesting of the crops. Machine Learning techniques make the prediction process inexpensive and efficient and also increase the accuracy of yield prediction by using data from the previous year. The approach attempted in this study uses fourteen years (2000–2014) of weather, satellite data derived products (Normalised Difference Vegetation Index) and yield data to train and compare machine learning regressor models such as Support Vector Machine (SVM), Gradient Booster Tree (GBT), Random Forest (RF) and Artificial Neural Network (ANN). Parameters of each algorithm were optimised to ensure that the best configuration was used to evaluate the performance of each model. For the present study, Mean Squared Error (MSE), Root Mean Squared Error (RMSE) and coefficient of determination (R-squared) were calculated. A comprehensive and objective comparison was conducted between all the models and the result indicates that SVM, with R2 of 0.93, RMSE of 0.160 and MSE of 0.025, proved to be a better performance in the crop yield prediction when compared to other models. The results can be used by researchers or policymakers for food security.