Hybrid Machine-Learning Model Adds Uncertainty Estimates to Irrigation Scheduling
A new arXiv paper presents a hybrid mathematical and machine-learning approach for irrigation decision support that also quantifies confidence in its soil-moisture forecasts. The authors note that irrigation is typically scheduled reactively, even though agriculture uses about 70% of global freshwater withdrawals. The model aims to give growers both a forward-looking moisture prediction and a measure of how reliable that prediction is.