The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences
Publications Copernicus
Articles | Volume XXXIX-B8
27 Jul 2012
 | 27 Jul 2012


C. Liu, W. Li, P. Lu, K. Sang, Y. Hong, and R. Li

Keywords: landslide, empirical model, ANN learning, susceptibility, hazard mapping

Abstract. Under the circumstances of global climate change, nowadays landslide occurs in China more frequently than ever before. The landslide hazard and risk assessment remains an international focus on disaster prevention and mitigation. It is also an important approach for compiling and quantitatively characterizing landslide damages. By integrating empirical models for landslide disasters, and through multi-temporal ground data and remote sensing data, this paper will perform a landslide susceptibility assessment throughout China. A landslide susceptibility (LS) map will then be produced, which can be used for disaster evaluation, and provide basis for analyzing China's major landslide-affected regions. Firstly, based on previous research of landslide susceptibility assessment, this paper collects and analyzes the historical landslide event data (location, quantity and distribution) of past sixty years in China as a reference for late-stage studies. Secondly, this paper will make use of regional GIS data of the whole country provided by the National Geomatics Centre and China Meteorological Administration, including regional precipitation data, and satellite remote sensing data such as from TRMM and MODIS. By referring to historical landslide data of past sixty years, it is possible to develop models for assessing LS, including producing empirical models for prediction, and discovering both static and dynamic key factors, such as topography and landforms (elevation, curvature and slope), geologic conditions (lithology of the strata), soil type, vegetation cover, hydrological conditions (flow distribution). In addition, by analyzing historical data and combining empirical models, it is possible to synthesize a regional statistical model and perform a LS assessment. Finally, based on the 1km×1km grid, the LS map is then produced by ANN learning and multiplying the weighted factor layers. The validation is performed with reference to the frequency and distribution of historical data.

This research reveals the spatiotemporal distribution of landslide disasters in China. The study develops a complete algorithm of data collecting, processing, modelling and synthesizing, which fulfils the assessment of landslide susceptibility, and provides theoretical basis for prediction and forecast of landslide disasters throughout China.