动态端元组合混合像元分解法在植被覆盖度动态监测中的应用——以长汀县为例

传统的像元信息分解法在计算端元丰度时端元组合是固定不变的，而在土地利用复杂地区，固定端元组合并不符合实际情况。文章针对研究区用地类型特点，预先设定数种端元组合，通过最小二乘法选择误差最小的端元组合进行端元丰度反演，进而计算出植被覆盖度；并以实测的地表反射率作为端元反射率，避免了大气差异造成的多时相误差，有利于进行动态监测；预先进行水陆分离，仅对陆地像元进行计算，可以避免水体像元的干扰并提高模型效率。辅以同期成像的高分影像对模型精度进行验证，相关系数达到0.98。将模型应用于长汀县近18 年4 个时相的Landsat 卫星数据，认为近年来长汀县植被覆盖度总体呈现变好的趋势，尤其是汀江沿岸；而城市化、火烧坡、大型工程建设是长汀县植被覆盖度局部变差的主要原因。

Application of Mixed Pixel Decomposition Method Using Dynamic Endmember Combination in Dynamic Monitoring of Vegetation Abundance: A Case Study of Changting County

The combination of endmembers is fixed in the calculation of the abundance of endmembers using traditional pixel information decomposition method. However, this situation may change for a region withcomplex land use. In this paper, several endmember combinations are predetermined regarding the characteristics of land use of the study area. Endmember combination with minimum error, which is identified by using the least square method, is used to retrieve endmember abundance as well as vegetation abundance. Measured actual surface reflectance is adopted instead of endmember reflectance to avoid multi-temporal errors introduced by atmosphere differences and to benefit dynamic monitoring. Meanwhile, water and land are separated in advance, and only land endmembers are calculated. So the effect of water endmember is avoided and the simulation efficiency is improved. The validation with synchronized high-resolution images shows that the correlation coefficient is more than 90%. The model is applied to Landsat data acquired in Changting County in four different periods of recent 18 years. The results indicate an improving trend of the vegetation abundance in Changting County, especially in the bank regions of the Ting River. The declined vegetation abundance in parts of Changting County is mainly due to urbanization, slope burning, and massive constructions.