基于MODIS时序数据提取河南省水稻种植分布

以河南省为研究区,利用2009年多时相8d合成MODIS地表反射率产品提取水稻种植分布。根据稻田含水量变化特征及水稻生长规律,构建水稻种植分布提取流程。为减少云等噪声的影响,对地表水含量指数（ILSW,land surface water content index）和增强型植被指数（IEV,enhanced vegetation index）的时序数据进行平滑重建。然后,依据豫北和豫南稻区水稻物候期差异,分别建立标准水稻IEV生长线,以计算像元尺度的水稻相似性指数作为影像分类的特征波段。同时,对重建的ILSW和IEV时序数据分别进行主成份分析,选择各自的前3个成份作为特征波段。在此基础上,采用支持向量机分类算法对组建的特征波段进行分类,提取影像中水稻的种植分布。结果显示,提取的河南省水稻种植分布与实际情况吻合较好,豫北稻区水稻分布呈现集中连片的特征,多分布在沿黄河两岸,而豫南稻区水稻种植广泛,多在大型水库灌区周边及沿淮和低洼易涝地区。与各地区水稻统计面积相比,MODIS提取的水稻面积平均相对误差为6.56%,根均方误差为5.63khm2。受到混合像元影响,以及个别地区水稻种植分散且面积相对较小,使该地区水稻面积相对误差超过±60%。

Mapping rice paddy distribution in Henan Province based on multi-temporal MODIS imagery

Multi-temporal 8-day composite MODIS Surface Reflectance Product of 2009 were used to map the distribution of rice paddy in Henan Province.By taking into consideration the characteristic of rice cultivation and the rice growth patterns,a scheme for rice mapping has been proposed.In this scheme,to reduce the influence of factors like cloud,the time series of land surface water content index（ILSW） and enhanced vegetation index（IEV） were calculated from MODIS imagery and two different filters were applied to rebuild the time series data.Then,the principal component analysis method was employed to reduce the dimensionality of the times series data,with the first three components as ILSW and IEV respectively reserved as feature bands.Another feature band is the rice similarity index,which was obtained by calculating the similarity index between a standard rice growth curve of IEV and the temporal curve of IEV for each pixel.However,there were obvious differences in rice phenology between the north and south rice planting regions in Henan Province.Based on the obtained rice GPS samples,a standard rice growth curve of IEV was established for each region.Finally,support vector machine（SVM） classification method was used to retrieve rice distribution from the feature bands.The results showed that the obtained rice distribution in Henan Province is well consistent with the real situation.Rice mainly distributed on both sides of the Yellow river in the north of Henan Province,while it distributed along the Huaihe River or around large reservoirs in the south of Henan Province.Compared with the statistical data,the rice area obtained from MODIS data bears a mean relative error of 6.56%,and a Root Mean Square Error of 5.63 khm2.Because of mixed pixels and dispersed rice distribution in some regions,the relative error can be larger than ±60%.However,the proposed rice mapping scheme based on the temporal MODIS data still shows its advantages in rice distribution mapping for large-scale areas.