粮食单产遥感预测方法

Estimation of grain yield with remote sensing method

本文介绍了通过时序遥感植被指数进行粮食单产预测的方法。以NOAA-AVHRR数据为基础数据源,提取区域作物生长过程曲线,并对曲线进行时序分析,提取曲线特征参数,采用参数-产量拟合法、产量分解法和产量差值法三种建模方法,分别对各参数与粮食单产、波动单产和差值单产进行相关分析,得到区域粮食单产敏感因子,并采用线性拟合的方法建立各地区不同类型的单产预测模型,从三种模型中选取置信度最高的模型作为最终单产预测模型。依据此方法对全国26个秋粮主产省份2004年单产进行预测,并将最终的预测结果与国家统计局数据进行对比分析,两者之间的误差在-4.9至11.59之间,相关系数为0.947。

In this paper, we make a study on the grain-yield forecasting model based on time series Normalized Difference Vegetation Index （NDVI） derived from NOAA-AVHRR. We build the grain yield-forecasting model by three different modeling patterns, including Parameter-yield mode,Decomposed yield mode and Difference-yield mode.By comparing the significances of three models,we pick out the final forecasting model for different province.The models were used to predict the autumn grain yield of different province in China and the differences are -4.9% to 11.59% by comparing with the yield data of the National Statistic Bureau,the R Square is 0.947.