基于多生育期MODIS-NDVI的区域冬小麦遥感估产研究

以河南省开封市为研究区,采用2005—2013年种植区域冬小麦生育期内16 d合成的250 m空间分辨率的归一化植被指数(NDVI)之和,与冬小麦产量数据进行相关分析,筛选得到最佳遥感估产时相,建立单变量回归模型;同时,基于冬小麦生长的各个关键时期NDVI构建多元回归模型,再通过主成分分析方法对多元回归模型参数进行改进,得到新的估产模型;最后使用开封市2014年的产量数据对估产模型进行验证,旨在构建具有较高精度的估产模型,从而更好地指导小麦生产。结果显示,3种估产模型的估产误差均控制在10.55%内,根据3个模型得到研究区内冬小麦最佳产量预测时段为3月下旬,即拔节期;3个模型中,主成分回归估产模型的产量拟合精度最高,达93.12%,具有一定的实用价值。

Regional Scale Yield Estimation of Winter Wheat by Remote Sensing Using MODIS-NDVI Data in Different Growth Stages

This study took Kaifeng city of Henan province as study area.The correlation between the sum of 16 d synthetic 250 m normalized difference vegetation index(NDVI) in winter wheat growth stages from 2005 to 2013 and winter wheat yield was analyzed to obtain the best phase of yield estimation by remote sensing to construct single variable regression model.Meanwhile,the multiple regression model based on NDVI of winter wheat in different growth stages was constructed.Then,the principal component analysis method was used to improve the parameters of multiple regression model.Finally,the yield data of Kaifeng city in 2014 was used to verify the yield estimation model,aiming to build the yield estimation model with higher precision,so as to better guide the wheat production.The results showed that the estimation errors of three models were controlled within 10.55%,according to the three models,the optimal period of yield estimation of winter wheat was in late March,ie,jointing stage,the principal component regression estimation model had the highest fitting precision of 93.12%,and had some practical value.