基于MODIS数据的河南省冬小麦产量遥感估算模型

小麦是世界上最重要的粮食作物，小麦生产对中国的粮食保障起着十分重要的作用，及时、准确、大范围对小麦产量进行监测预报，对于农学经济发展和粮食政策制定具有极为重要的现实意义。对作物产量进行遥感监测的原理是建立在其遥感特征基础之上的，通过建立作物长势指标与遥感信息的定量关系，可实现对作物产量的监测预报。文章基于2009年MODIS遥感数据和气象数据，利用Arcgis和ENVI提取纯小麦像元，并提取纯小麦像元对应的NDVI、NPP和LAI，获取分县NDVI、NPP和LAI均值，利用统计软件对产量数据和分县遥感参数均值进行数据整理和分析，建立了河南省冬小麦产量估算模型。以往研究多采用遥感图像上某像元和地面调查点进行研究，具有很大的不确定性，文章以县为单位，对冬小麦平均单产和县域内冬小麦种植像元遥感参数的均值进行相关研究，提高了模型模拟精度。同时文章选用多种遥感参数和多项气象因子建立估产模型，避免了针对一个参数进行估产的局限性。在最佳时相的选择上，根据冯美辰（2010）以往的研究结果，从4月以后，5月8日和4月20Et植被指数和产量相关性最大，4月份之前冬小麦处于返青到拔节期，对产量来说还有很多不确定闪素，因此文章选用5月8El和4月20日进行冬小麦估产研究。结果表明，5月8日的估产模型优于4月20日，加入气象冈子的遥感气象估产模型优于只采用遥感参数进行估产的遥感模型。利用2010年产量数据对模型精度进行检验，遥感气象模型预测精度在70．2％N99．7％之间，平均精度为90．7％；遥感模型预测精度在68．1％到95．5％之间，平均精度为83．9％。表明遥感气象模型模拟精度更高，其精度可以满足大面积估产要求，可以对产量预报提供科学参考。

Remote sensing estimation model of Henan province winter wheat yield based on MODIS data

Wheat is the world＇s most important food crops, its production plays a very important role in China＇s food security, and it makes extremely great practical significance to timely, accurately and wide range monitor and forecast wheat yield for agricultural economic development and food policy making. The principle of crop yield remote sensing monitoring is based on the spectral characteristics, and the crop yield forecast can be realized by establishing a quantitative relation between the crop growing index and the spectral information. In this paper, based on the 2009 MODIS remote sensing and meteorological data, we use Arcgis and ENVI to extract pure wheat pixels and the corresponding NDVI, NPP, LAI; further, we calculate the county mean value ofNDVI, NPP, LAI By using statistical software to process and analyze the above crop yield data and county mean value of remote sensing parameters, we establish an estimation model of Henan province winter wheat yield. Previous studies mainly expanded by using a pixel on a remote sensing image and ground survey point, which had great uncertainties. Here, by taking county as the unit, the correlation between the winter wheat average yield and the mean value of county plant pixel remote sensing parameters has been studied, which improves the model simulation accuracy. At the same time, various remote sensing parameters and meteorological factors are selected to build yield estimation model, which avoids the limitations of yield estimation by considering only one parameter. As to the optimum temporal selection, according to the previous studies of Feng Meichen （2010）, vegetation index and yield have the maximum correlation between May 8th and April 20th after April; winter wheat stays in the returning to the jointing stage before April, and there are many uncertain factors for yield. Thus, May 8th and April 20th are chosen here for the investigation of winter wheat yield estimation. The results show that yield estimation model of May 8 is better than the one of April 20; and the remote sensing meteorological yield model is superior to the remote sensing yield model. The accuracies of the models have also been tested by using 2010 annual yield data, the prediction accuracy of remote sensing meteorological yield model is between 70.2% and 99.7%, and the average is 90.7%; the precision of remote sensing yield model is between 68.1% and 95.5%, and the average is 83.9%. The study shows that the simulation accuracy of remote sensing meteorological yield model is better, which can meet the requirements of large area yield estimation, and can be applied to provide scientifically reference for yield forecasting.