关键词: 产量预测, 产量分离, ARIMA模型, 灰色模型, 二次指数平滑, 枇杷

Abstract: Four yield separation methods were used in this study to analyze the applicability on meteorological yield separation of regional cash crops, which are the ARIMA model, GM model, linear trend and quadratic exponential smoothing method. With the use of these four methods, the yield forecast models based on meteorological factor are built for analyzing the prediction accuracy. The results showed that the meteorological yield of loquat separated by GM model and linear trend method were well matched with the agrometeorological disaster records. The correlation coefficients between the meteorological yield of loquat separated by four methods and nine meteorological factors were high. The correlation coefficients were 0.95(separated by ARIMA), 0.94(separated by GM model and linear trend method) and 0.93(separated by quadratic exponential smoothing method). The root mean square error(RMSE) and the mean absolute percentage error(MAPE) between the loquat yield predicted by GM model and the actual yield were the largest, which was reduced by 50.1% after taking the meteorological factors into account. RMSE between the loquat yield predicted by linear trends, ARIMA model and secondary exponential smoothing methods and the actual yield was 49.3%, 16.7% and 14.4%. Comparing the four yield separation methods, the GM model significantly outperformed other methods, followed by the linear trend method. Moreover, the ARIMA model was the worst. The RMSE and the MAPE between the loquat yield predicted by GM model and the actual yield were 3.0kg·ha⁻¹ and 15.2%. Overall, loquat meteorological yields separated by GM model and the linear trend method matched well with the agrometeorological disaster records. The GM model performed the best on loquat yield prediction, which shows that the GM model is more suitable for yield separation and prediction of regional cash crops.

Key words: Loquat production forecast, ARIMA model, GM model, Meteorological output