Forecasting irrigation water requirement driven by meteorological forecast error correction in the Yellow River irrigation area of Shandong Province

Irrigation water requirement (IWR) is a key parameter in agro-hydrological research and water resource optimization. Its forecast accuracy determines irrigation scheduling reliability, but it is constrained by meteorological errors, which propagate through the modeling chain. An optimal correction scheme was developed for short-term (1–3 days), medium-term (4–15 days), and long-term (16–40 days) forecasts of maximum temperature (T′ max ), minimum temperature (T′ min ), and precipitation (P′) based on 1–40-day public weather forecasts from 2019 to 2023. A regionally calibrated Hargreaves model was evaluated for estimating reference evapotranspiration (ET o ), and an IWR forecast method driven by meteorological error correction was established. The results demonstrate that optimal correction methods varied by lead time: the decaying-average method based on autocorrelation function (ACF-DAM) was optimal for T′ max at short lead times; genetic algorithm (GA)-optimized long short-term memory (LSTM) for T′ max at medium-to-long lead times and T′ min at short-to-medium lead times; stepwise regression (SR) for T′ min at long lead times; and Bayesian optimization (BO)-optimized random forest (RF) for P′ across all lead times. ET o forecasts achieved accuracy exceeding 80% at 86.36% of stations across different lead times. ET o predicti o ns ranged from 2.06 to 3.48 mm d −1 , with root mean square error (RMSE) values of 1.22 mm d −1 (short-term), 0.87–0.99 mm d −1 (medium-term), and 1.25–1.32 mm d −1 (long-term). Spatially, the Yellow River Delta and northern plains performed best, while the central region showed larger errors. Across different lead times, IWR forecasts remained stable, with IWR ranging from 145.13 to 293.62 mm for winter wheat and from 31.34 to 239.99 mm for summer maize; station-mean values were 207.11–249.39 mm and 78.21–196.69 mm, respectively. Spatially, IWR was higher in the north and lower in the southwest. Correcting weather forecast errors can significantly reduce the prediction errors of ET o , ET c , and IWR, providing a scientific basis for hydrological regulation, water allocation, and agricultural water demand forecasting in irrigation districts.