Remote sensing and GIS based comparison of land-use classification methods and data mining for degree of land degradation: a case study in Leilongwan Area, Hengshan County, Shannxi Province
Abstract:This study employed the maximum likelihood classifier and spectral angle mapping (SAM) method to classify land-use types in the Lilongwan area, Hengshan County, Shannxi province, using Landsat ETM+ remote sensing data. The SAM adopted the minimum noise fraction rotation (MNF) and pixel purity index (PPI) to extract terminal elements of land-use types and based on that to develop the land-use map. The maximum likelihood classifier achieved higher accuracy on water and arable land than the other land-use categories with omission errors occurred on the sandy land category. As the SAM achieved better classification results for sandy land, urban and built-up land and water categories, certain level of confusion existed between woodland and grassland categories. We then established a model of land degradation degree index (DDI) using Albedo, NDVI and wetness from remote sensing data and selected parameters of the model to obtain an improved classification result. We employed GIS technology to produce the map for degree of land degradation. The comparison analysis between SAM classification and PPI samples showed a good correlation between DDI and land-use types. This study indicates that the methodologies developed in this study can be used to reveal the degree of land degradation information of the study area.