ارزيابي كارايي روش هاي آماري در تعيين پتانسيل خطر زمين لغزش

Evaluation of statistical methods in landslide hazard

In this research, evaluation of statistical methods in Predication of landslide hazard has been performed using data Processed by Geographic Information System (GIS) For this purpose, Rudbar area as an appropriate example for north part of the Iran was selected. Our investigation indicates that lithology distance from faults, vegetation cover, land use, rain fall, and maximum acceleration rate are the main landslide controlling factors in the area. Each of these factors was used as a thematic layer for landslide hazard zonation mapping. Univariate and multivariate statistical analysis were used for landslide hazard analysis. Regression analysis indicates that classification of each parameter map into a number of relavant calsses such as slope, rainfall, ... reduces accuracy of prediction. Also weighting of samples based on area of each unit and landslide occurrence increase accurracy. In general, among univariate statistical analaysis, area density method represents better results. Multivariate analysis indicates appropriate results for continous data compared to discrete data. On the other hand weighting of samples based on values such as area of ground units or percentage of landslide in each unite improves the results.