Benchmarking Water User Associations for Improved Irrigation Efficiency: A DEA Approach with SHAP and Machine Learning Integration
摘要
Water user associations (WUA) have played a crucial role in enhancing the efficiency of irrigation water distribution, managing minor irrigation systems seamlessly, and resolving conflicts among farmers and other stakeholders. This paper aims to develop an analytical framework to benchmark and objectively evaluate the efficacy of WUAs. Additionally, the framework provides recommendations for improving the efficiency metrics of these associations. The study uses data from the Rohini Canal System in northern Uttar Pradesh, India, and employs data envelopment analysis (DEA) with an input-oriented model to perform the analysis. The results demonstrate the significant impact of WUAs in increasing efficiency and improving irrigation water management. Notably, there were marked improvements at the tail ends of the minors, which previously experienced irregular water distribution. To further understand the dependency of various parameters on canal efficiency, an explainable artificial intelligence (SHAP) analysis was conducted. The parameters with the highest SHAP values were then used to develop and test two AI/ML models, LSTM and random forest. The random forest model proved to be the most effective and was subsequently used to forecast the canal system’s efficiency in the near future.
Graphical Abstract