Adaptive irrigation management for subsistence farming

Irrigation management authorities in Sri Lanka have often failed to incorporate historical information in the decision‐making process in a systematic manner. The objective of this study is to develop a water model to combine prior information systematically. The decision rule developed is based on a single‐season linear programming model, using real‐time forecasting model inflows. The forecasts are based on the most current information (via Kalman filtering) available to the forecaster. The forecasting method used is vector autoregression with Bayesian priors, allowing the reservoir management to include subjective information into the decision matrix. This adaptive technique was examined for various seasonal inflow scenarios and found to be superior to the existing water allocation method.