不同产量分离方法在水稻产量预报上的对比分析OA

Utilizing meteorological observation data and rice yield and growth stage data of Xinyang City from 1993 to 2021,this study compared and analyzed the applicability of five-year sliding average,three-year sliding average,HP filtering,double exponential smoothing,ARIMA model,and grey prediction model GM(1,1)in forecasting rice yield in Xinyang City.The correlation analysis was performed to determine the key meteorological factors affecting rice yield in Xinyang City.A rice yield forecasting model was established based on the stepwise regression method.The forecasting models were then used to conduct back testing on rice yields from 1993 to 2021 and forecasting tests on rice yields from 2022 to 2023.Two testing methods,namely accuracy rate and Taylor plot,were employed to evaluate and analyze the effects of forecasting equations based on different yield separation methods in rice yield forecasting in Xinyang City.The results showed that the key meteorological factors affecting rice yield in Xinyang City were sunshine duration in late July,average temperature in late July,and precipitation in late July.The multi-year average back-testing accuracy rates of six models ranged from 90.2％to 94.5％.Among them,HP filtering and five-year sliding average demonstrated higher accuracy rates of 94.5％and 94.4％,respectively.The multi-year average forecasting accuracy rates of six models ranged from 89.2％to 98.8％,and the two methods with higher forecasting accuracy rates were double exponential smoothing(98.8％)and ARIMA model(96.5％).