Application of Machine Learning Approaches for Cotton Seasonal Yield Estimation

Estimating crop yield can help farmers plan for equipment, labor, and other crop production input requirements. Forecasting crop yield is also useful for analyzing weather-related variability to guide decisions such as irrigation water and fertilizer management. This work discusses the application of Gated Recurrent Unit (GRU) and Long Short-Term Memory (LSTM) machine learning algorithms for seasonal cotton yield prediction. Simulation results from the crop model AquaCrop, consisting of irrigation depth, soil moisture content, and crop growth stage data from 2003 to 2021 were used to train the algorithms. The two developed yield-prediction models were tested against data collected from an irrigated cotton field located at Clemson University Edisto Research and Education Centre (EREC), near Blackville, South Carolina, USA during the 2023 growing season. The values of hidden layers, hidden units, dropout, learning rate and batch size hyperparameters were set to respectively, 3, 64, 0.2, 10E-3 and 64 for the GRU model and 3, 128, 0.4, 10E-3 and 64 for the LSTM model. Analysis suggested that the tested algorithms resulted in very good to excellent performance. We concluded that machine learning algorithms are useful tools that can provide insights into how much yield to expect in an upcoming season and help farmers optimize energy, water, and fertilizers applications.