Abstract:

In order to explore the impact of different separation methods on yield forecast accuracy, summer maize yield data and meteorological data from Xuchang city, the main grain producing area in central Henan province from 1985 to 2024 were used. The 3−year moving average, 5−year moving average, HP filtering, five point quadratic smoothing, quadratic exponential smoothing and ARIMA model were used to separate meteorological yield and construct a summer maize yield forecast model. The simulation effect was evaluated by calculating indicators such as trend forecast accuracy and yield forecast accuracy. The results showed that the 3−year moving average, 5−year moving average, 5−point quadratic smoothing and quadratic exponential smoothing had better separation effects on yield, while there were differences in the positive and negative relationship of separating meteorological yield among different methods. Among the key meteorological factors selected, precipitation during the jointing-silking stage, temperature and sunshine hours during the grain-filling stage, showed a significant positive correlation with meteorological yield(P<0.05). The temperature during the seedling stage and precipitation during the grain filling stage showed a significant negative correlation with meteorological yield (P<0.05), which was consistent with the growth and development characteristics of maize. The yield trend forecast accuracy of 6 models in backtesting was over 78.1%, the yield forecast accuracy exceeded 94.5%, and RMSE was less than 410.5kg×ha⁻¹. The combination weighted forecast model performed the best in forecast testing, with an accuracy of 97.2% for yield forecast, which was better than the performance of a single method model. It can provide reliable data reference for accurate grain yield forecasting and scientific decision−making in agricultural production.

Key words: Summer maize, Yield separation method, Meteorological yield, Key meteorological factors, Yield forecasting