中文摘要：

为了将遥感观测到的玉米生长期间作物冠层方向反射波谱的时间序列变化信息用于区域玉米产量估算，该文将时间序列中分辨率成像光谱仪（moderate resolution imaging spectroradiometer，MODIS）数据和高空间分辨率Landsat TM遥感观测数据相结合，以叶面积指数（LAI）作为耦合作物生长模型（crop environment resource synthesis-Maize，CERES-Maize）和植被冠层反射率模型（scattering by arbitrarily inclined leaves，SAIL）的关键参数，提出了将耦合模型与时间序列遥感观测数据同化进行区域玉米产量估算的方案。该文选择吉林省榆树市为研究区，采用MODIS和Landsat TM2种尺度数据集，利用SCE．uA（shuffled complex evolution method developed at the University of Arizona）算法分别进行玉米产量同化估产研究，得到玉米单产空间分布的估计结果，结合遥感估算的种植面积求算榆树市玉米总产量。结果表明，与玉米统计总产量相比，2007、2008和2009年遥感数据同化估算的总产量误差分别为9．15％、14．99％和8．97％；与仅利用CERES．Maize模型模拟得到的产量误差相比，3a间遥感估算总产量的误差分别减小了7．49％、1．21％和5．23％，且采用MODIS和TM遥感数据估算的玉米产量表现了其空间差异性。利用榆树市3a间玉米产量的明显差异，分析了时序遥感数据对作物长势和产量变化信息的表达能力，同年份内时序归一化差值植被指数越大，对应的玉米产量越高；年际间遥感观测反射率的差异通过数据同化方法能够反映年际间玉米产量差的变化。该文提出的玉米估产方案为将来进一步结合多源遥感数据、植被冠层反射率模型与作物生长模型进行区域玉米估产研究提供了参考。

英文摘要：

In order to apply time-series variation information of reflectance spectrum at direction of crop canopy which was observed by remote sensing in crop growing stages to estimate the regional corn yield, a practical scheme to assimilate time-series remote sensing data and the CERES-Maize （crop environment resource synthesis-Maize）growth model coupled with the canopy reflectance model SAIL（scattering by arbitrarily inclined leaves） through leaf area index for regional maize y!eld estimation was addressed based on the time-series MODIS （moderate resolution imaging spectroradiometer） and high spatial resolution TM observation data. The proposed scheme was applied to Yushu city located in Northeast China. The spatial distribution of maize per unit yield was determined by applying the SCE-UA （shuffled complex evolution method developed at the University of Arizona） optimization algorithm to the maize yield estimation, and the gross maize yield in Yushu city was estimated through per unit yield multiplied by maize planting area which was extracted using remote sensing data. The results indicated that comparing to the statistical data of maize yield, the estimation errors of the total maize yield estimation by using remote sensing data assimilation were 9.15%, 14.99% and 8.97% in 2007, 2008 and 2009 respectively. Comparing to those only running the CERES-Maize model, the errors of the total maize yield estimation by using remote sensing data assimilation decreased by 7.49%, 1.21% and 5.23% in 2007, 2008 and 2009 respectively. The maize yield estimated by MODIS and TM data revealed its spatial heterogeneity. The expression capability of maize growth condition and yield variation was analyzed using time-series remote sensing data in this paper. The greater time-series NDVI in the same year was, the higher the maize yield was. The inter-annual variations of remote sensing observations mirrored the inter-annual variations of maize yield gap by using the data assimilation method. This scheme for estimating corn