Precise irrigation of dryland cotton under canal irrigation system constraints based on the CERES-CROPGRO-Cotton model

Abstract

Xinjiang agriculture faces significant challenges due to water resource scarcity and uneven distribution, making accurate predictions of irrigation's impact on cotton yield crucial for decision-making. Existing studies primarily examine the relationship between irrigation and soil but often overlook the combined effects of irrigation networks, time constraints, and the interactions between crop growth patterns, weather, soil, and irrigation strategies. This study integrates the Decision Support System for Agrotechnology Transfer (DSSAT) model with machine learning models, accounting for weather, soil, crop conditions, and irrigation time constraints, to propose an intelligent irrigation decision-making framework. Meteorological data from 1980 to 2024, 13 sets of soil data, and field experiments conducted in 2023 and 2024 were used to calibrate the DSSAT model (with a calibration rate of 0.856). Additionally, water channel quota time constraints were established to determine the optimal irrigation timing. The framework analyzes the interactions between weather, soil, irrigation strategies, and other factors, enhancing cotton yield prediction. The results indicate that the intelligent decision-making algorithm outperforms traditional methods under data limitations, reducing the irrigation water consumption-yield ratio ($U_i$) by 3.99 %, while increasing yield by 8.5 % to 9724 kg/ha, thus achieving both water-saving and yield-enhancing objectives. This research offers a refined solution for intelligent irrigation decision-making in cotton cultivation in arid regions and paves the way for the application of intelligent agricultural decision systems.