基于混合像元分解的南方地区植被覆盖度遥感监测——以广州市为例

Remote Sensing Monitoring of Vegetation Coverage in Southern China Based on Pixel Unmixing:A Case Study of Guangzhou City

通过测量图像端元的地表反射率,对遥感图像进行精确大气校正;在对混合像元分解模型进行改进的基础上,建立了基于地表反射率的线性混合像元分解(Liner Spectral Unmixing,LSU)模型,有效地避免了因大气时间、空间差异所造成的多时相误差,实现了多时相对比;通过增加土壤湿度因子,消除了土壤湿度差异造成的误差,建立了适用于南方地区的植被覆盖度遥感监测模型。实地验证结果表明,该模型具有较高的精度。将该模型运用于广州市1998～2009年植被覆盖度时空变化监测,认为城市化、大型工程建设与陡坡开荒是造成广州市植被覆盖度变化的主要原因。

Based on the measurement of the ground spectral reflectance of basal land covers and the accurate atmospheric correction for Landsat TM data, the authors improved the linear spectral mixture model （ LSU ） and developed a vegetation coverage retrieval model suitable for southern China. The effects of the atmospheric environment and the imaging time of remote sensing data were both reduced, contributing to the multi - temporal comparison, by the utilization of the ground spectral reflectance from field survey. The soil moisture factor was considered to eliminate its remarkable spatial differentiation error in southern China. The vegetation coverage retrieval model was proved to be efficient with high precision over the in situ field verification and was applied to extract the vegetation coverage information in Guangzhou from 1998 to 2009. It is inferred that the urbanization, the large- scale architectural engineering and the reclamation activities constitute the main factors responsible for the formation of the spatiotemporal vegetation change in this area.