<p>Optimizing the autumn irrigation is an effective way for the Inner Mongolia Hetao Irrigation District (HID) in China to achieve high crop yields, reduce resource consumption, and alleviate secondary salinization of the soil. However, relatively few studies on a regional scale optimize and evaluate the autumn irrigation patterns based on coupled genetic algorithm (<i>GA</i>) and distributed SWAP-WOFOST model. In this study, a distributed SWAP-WOFOST model that couples crop growth to soil water, heat and salt transport was constructed, calibrated and validated on a regional scale. The calibrated and validated model was then used to analyze the effects of various autumn irrigation patterns on sunflower yield and Water Productivity (<i>WP</i>= yield/evapotranspiration) in the HID from 2000 to 2017 under different precipitation year types. To improve the <i>WP</i> of sunflower, a region-scale autumn irrigation pattern was optimized and evaluated based on <i>GA</i>. The findings indicated that the distributed SWAP-WOFOST model could better model the dynamics of crop growth, soil water, salinity and temperature transport in HID. The average annual sunflower yield was 3423 ~ 4012&#xa0;kg/ha and the average annual <i>WP</i> was 1.1 ~ 1.2&#xa0;kg/m³ under different autumn irrigation patterns from 2000 to 2017 in HID. The results of optimizing the autumn irrigation pattern suggested that sunflower maintained stable yields, with average <i>WP</i> increased ranging from 1.1% to 2.1%. Additionally, based on retrospective scenario analysis using long-term precipitation statistics, the optimized autumn irrigation patterns exhibited a potential to reduce irrigation water diversion from the Yellow River by approximately 410&#xa0;million m³/year and 158&#xa0;million m³/year under wet and normal climatic conditions, respectively.</p>