山东省冬小麦单产监测与预报方法研究

针对传统农业估产方法效率低、成本高的现状,以山东省为研究区,基于山东省10年表面反射率8 d合成产品MOD09A1、全球陆地蒸发蒸腾8 d合成产品MOD16A2数据和历史产量数据,以增强型植被指数(EVI)、作物水分胁迫指数(CWSI)和经过历史产量分解得到的技术产量为输入,利用最小二乘法构建了山东省级和市级尺度的冬小麦单产估算模型,并在监测和预报两种模式下进行了模型的应用和精度验证。结果表明,在监测模式下,省级估产精度为96.91%,各市监测精度均不小于89.41%,其中菏泽市监测精度最高,为99.31%,济宁市监测精度最低,为89.64%;在预报模式下,返青期结束(第89天)、拔节期结束(第121天)和乳熟期结束(第145天)时的省级小麦预报精度均不低于96.44%,各市预报精度均不小于89.41%,其中青岛市预报精度最高,3次预报的平均精度为99.07%,济宁市预报精度最低,3次预报的平均精度为89.81%。本文建立的估产模型对市级和省级作物单产估算均有较高的适用性,可以实现动态产量预报。本研究对及时了解冬小麦的生长状况、制定科学有效的农业生产决策具有参考价值。

Monitoring and Forecasting Method of Winter Wheat Yield in Shandong Province

In view of the low efficiency and high cost of traditional agricultural yield estimation methods, taking winter wheat in Shandong Province as an example, and the cumulative enhanced vegetation index (EVI), cumulative crop water stress indicator (CWSI) and trend yield were used to build a statistical yield estimation model in Shandong Province with least square method. Cumulative EVI was calculated from 8-day surface reflectance products (MOD09A1), cumulative CWSI was calculated from 8-day global terrestrial evapotranspiration products (MOD16A2), and trend yield was calculated using historical yield data calculated by method of time trend analysis. The yield estimation model was operated and verified in monitoring mode and forecasting mode respectively. In the monitoring mode, the provincial yield estimation accuracy was 96.91%, and the estimation accuracy of each city was above 89.64%. Heze City had the highest monitoring accuracy, which was 99.31%, and Jining City had the lowest value, which was 89.64%. In the forecasting mode, the model was operated and verified in three time points of growth period: the end of the rejuvenated period (the 89th day), the end of the jointing period (the 121st day) and the end of the milk ripening period (the 145th day). The prediction accuracy of wheat was over 96.44% at provincial level and over 89.41% at municipal level during three-time points. The forecast accuracy of Qingdao City was the highest, with an average of 99.07%, and that of Jining City was the lowest, with an average of 89.81%. The yield estimation model had a high applicability to the estimation of crop yield at municipal and provincial levels, which can realize the constantly yield prediction. The method of monitoring and forecasting yield was conducive to timely understanding the growth condition and changes of winter wheat, and it had a certain reference value for the government departments to make scientific and effective agricultural production decisions.