基于时序MODIS/NDVI影像的鄂东南低山丘陵区植被覆盖度季节变化特征

SEASONAL VARIATION CHARACTERISTICS OF VEGETATION FRACTION IN THE HILLY SOUTHEASTERN HUBEI PROVINCE,USING TIME-SERIES MODIS250 M NDVI DATA

遥感估算植被覆盖度的精度受云噪声的影响较大。最大值合成法(MVC)法能够较好消除时序NDVI影像上云污染的影响,但对于长时间多云或连阴雨天气常见的地区来说,该方法难以彻底去除云覆盖。以鄂东南低山丘陵区为研究对象,应用BISE算法对其2006年时序MODIS/NDVI影像进行噪声检测和处理。在此基础上,应用改进后的像元二分模型对这些NDVI影像进行植被覆盖度的估算和分级统计与分析。研究结果表明:根据改进后的像元二分模型和NDVI指数估算植被覆盖度的平均绝对误差和平均相对误差分别为0.149和27.7%;对时序植被覆盖度影像的分级统计结果显示:鄂东南低山丘陵区植被覆盖整体较高,全年高覆盖度(>80%)、中高覆盖度(60%～80%)和中等覆盖度(40%～60%)植被区所占的面积比之和约为75%～84%,尤其是7～9月份80%左右的区域为高覆盖度或中高覆盖度植被所覆盖;除了低覆盖度植被以外,其他等级的植被区季节变化明显,尤其是高覆盖度和中高覆盖度植被的年内变化极为显著,并且二者之间存在着明显的此消彼长的转化关系。

Cloud contamination has a significant effect on the accuracy in estimating vegetation fraction based on remote sensing.MVC method can apparently eliminate the influence of cloud elimination on the time-series NDVI images.But in the regions with long-term cloudy or rainy weather,it is difficult to completely eliminate the cloud coverage with this method.In this paper,BISE algorithm was adopted to identify and analyze noise from the time-series MODIS/NDVI data in the hilly southeastern Hubei province.Based on it,vegetation fraction was calculated according to the developed dimidiate pixel model.Compared with the vegetation fraction data derived from forest resource inventory,Mean Absolute Error（MAE）was 0.149and Mean Relative Error（MRE）27.7%.Moreover,the obtained time-series vegetation fraction data was calculated based on division into five degrees（i.e.〈20%,20%～40%,40%～60%,60%～80%and〉80%）.The results show that about 75%to 84%of the study region was covered by vegetation fraction more than 40% and the area ratio of vegetation fraction more than 60% was almost 80% during July to September.Except the low degree of vegetation fraction（less than 20%）,other types had a remarkable seasonal change,and especially the high degree（more than 80%）and the intermediate-tohigh degree（60%～80%）interchanged.