Abstract:

To identify the response characteristics of winter wheat yield in southeastern Gansu Province to meteorological factors and assess the trend of county-level yield changes in the near future, this study selected Qinzhou District, Maiji District, Qin'an County, Gangu County, and Wushan County in Tianshui City as the research areas. Based on the winter wheat yield data from 2015 to 2024, AgERA5 agrometeorological reanalysis data, and CMIP6 future climate data, a meteorological index system for different growth stages was constructed. Correlation analysis, partial correlation, grey correlation, multiple linear regression, random forest, and KNN methods were comprehensively adopted for analysis. The results show that the winter wheat yield in the study area is highly sensitive to heat conditions. The maximum average temperature during the entire growth period and the filling and maturation period has a stable negative correlation with the yield. Indicators such as solar radiation during the reviving-jointing period, extreme low temperature, and precipitation during the filling and maturation period also have significant impacts. The historical fitting effect of the random forest was better than that of multiple linear regressions, and the KNN residual model performed well in the independent validation in 2024. The prediction results from 2025 to 2035 indicate that the winter wheat yield in the five counties and districts shows fluctuating changes and does not show a continuous upward trend, and there are certain spatial differences among the counties. The study suggests that high-temperature heat stress is an important meteorological risk limiting the stable production of winter wheat in the region, and the reviving-jointing period and the filling and maturation period are the key stages of climate affecting wheat yield.

Key words: winter wheat, meteorological factors, AgERA5, random forest, KNN, CMIP6