Abstract: Based on the comparison of enhanced vegetation indices(EVIs) from moderate resolution imaging spectroradiometer(MODIS), linear, quadratic nonlinear and stepwise regression models were constructed with the 16day and 250m resolution vegetative indices of Terra and Aqua(MOD13Q1, MYD13Q1) combined EVIs multiply by township rice planting area and total rice grain yield in Liling City, Hunan Province, China. The optimal fitting models were selected by error analysis, and then the total rice yield in the next year was forecasted. More than 50% absolute values of errors between Terra and Aqua MODIS EVIs were less than 0.03, and 93.22% and 99.50% of them were less than 0.08 and 0.10, respectively. The relative errors of total rice grain yield for all optimal fitting models were less than 0.10%, although forecasting errors were larger than that of fitting, the relative errors of forecasting total rice grain yield were less than 5%.

Key words: moderate resolution imaging spectroradiometer, enhanced vegetation index, rice, yield estimation, remote sensing

摘要： 以湖南省醴陵市为研究区域，在比较分析Terra卫星与Aqua 卫星中分辨率成像光谱仪增强型植被指数的基础上，结合这两个卫星提供的空间分辨率为250 m、时间分辨率为16 d的植被指数产品，建立水稻各主要生育期增强型植被指数平均值乘以水稻面积的结果与乡镇级水稻总产的一次线性、二次非线性及逐步回归模型。通过误差分析，选择最优遥感拟合模型，并在此基础上，预测下一年水稻总产。结果显示，水稻种植区Terra卫星与Aqua 卫星中分辨率成像光谱仪增强型植被指数有超过50%的偏差绝对值小于0.03；93.22%、99.50%的偏差绝对值分别小于0.08、0.10。水稻遥感拟合模型模拟结果相对误差小于0.1%，预测模型的估产结果比遥感拟合模型的拟合结果的误差大，但相对误差仍然小于5%。

关键词: 中分辨率成像光谱仪, 增强型植被指数