Optimisation of deficit irrigation strategies under climate change conditions based on artificial intelligence

Abstract

This paper presents a novel modelling framework that integrates artificial intelligence (AI) tools with deficit irrigation (DI) management. Different levels of DI were applied to summer and winter crops, with six scenarios assessed for each season. Four metaheuristic algorithms – LSHADE (Success-History based Adaptive Differential Evolution with Linear Population), PSO (Particle Swarm Optimisation), GWO (Grey Wolf Optimiser), and NRO (Nuclear Reaction Optimisation) – were implemented and automated using Python-based models to identify optimal solutions. The optimisation model aims to maximise the project’s total net profit by determining the most efficient allocation of land and water resources under DI conditions. For the summer season, the findings showed that the best crop pattern was obtained in scenario S2, the balanced traditional scenario, using the LSHADE algorithm. This scenario increased financial profit by 29%, used 89% of the land, and applied 12.5% deficit irrigation. The winning crop pattern consisted of 25% maize, 5% tomato, 5% cucumber, 4.23% aubergine, 10% watermelon, and 40% maize forage. For the winter season, the optimal pattern was obtained in scenario S6 (water-stressed) using the LSHADE algorithm. Financial profit increased by 44%, land use reached 100%, and DI was 12.5%. The most suitable agricultural pattern included 45% wheat, 30% barley, 5% potatoes, 3% spinach, and 7% alfalfa.

The findings confirm that AI-driven optimisation effectively supports DI management. Advanced metaheuristic algorithms should be integrated into irrigation planning, and their adoption should be supported through simple decision-support tools for farmers.