基于Landsat TM数据的黄土高原区水体识别方法研究

IDENTIFICATION OF WATER BODIES IN THE LOESS PLATEAU BASED ON LANDSAT TM DATASET

基于2007年7月9日Landsat TM数据以固原市周边约500km2的范围为研究对象,分别应用单波段阈值法、多波段谱间关系法、水体指数法等多种方法对研究区的水体进行识别提取.研究结果表明:各种水体识别方法均不能提取满意的水体信息,反映了黄土高原地区水体识别的复杂性和下垫面的特殊性,给水体识别带来了很大的不确定性.TM5单波段阈值法和NDWI与实际水体最为接近,提取了绝大部分水体,TM7单波段阈值法、CIWI、MNDWI次之,但是单波段法中混杂了其他信息.NDWI方法在研究区识别的水体信息更为真实准确,能够反映水库和大型淤地坝位置和面积信息,能够用于监测这类淤地坝水体的变化特征.

Regarding the intensive human interference to the hydrological processes in the Loess Plateau, the key to understand the reciprocal effects is how to quantify the evolution of large quantities of water bodies, such as farm dams, small reservoirs, check dams, etc. On the basis of the image of Landsat dataset on July 9, 2007, a region around Guyuan City （approx. 500 km2） was chosen as the study area and several methods were applied to investigate a reasonable way to identify water bodies in the Loess Plateau, including single band classification, multiple bands classification, water body indices, etc. The results revealed that the particular underlying surface condition in Loess Plateau and the complexity of water bodies. In comparison, water body mapping based on 5th band and NDWI provide most reasonable results. While useful information can also be derived by 7th band, CIWI and MNDWI classification. Generally, all methods cannot identify the precise water body information based on Landsat dataset. However, information derived in this study can give one a primary knowledge of small reservoir and key check dams, and can also be used to monitor and investigate the evolution of these water bodies.