Statistical analyses of landslide size and spatial distribution triggered by 1990 Rudbar-Manjil (Mw 7.3) earthquake, northern Iran: revised inventory, and controlling factors

Abstract

This study is a revised inventory of landslides triggered by the Rudbar-Manjil earthquake (Mw = 7.3), 1990, obtained through the study of Komak Panah and Hafezi Moghadas (1993), interpretations of the stereoscopic aerial photos, and field data. We mapped the 223 coherent landslides with a total area of 11.3 km2 and a calculated volume of 3.8 × 108 m3 using a digital elevation model (DEM, 12.5 m) within the geographical information system (GIS) software. The correlation between landslide occurrence (size and distribution) and the controlling factors was analyzed using the bivariate model and the linear automated modeling (LINEAR) procedure in the SPSS software. Both LINEAR models of landslide volume (LV, m3) and landslide area (LA, m2) indicate the distance from the fault surface rupture is the most influencing factor on size prediction of coseismic landslides. The LA is more dependent on seismic factors, whereas the LV is on the lithology and topographic factors. The result of principal component analysis (PCA) model, using the GIS, showed that most landslides occurred in the red band (including 95% of the variance of controlling factors). Thus, the selected factors to investigate the landslide occurrence are valid and the PCA can be used to identify regions prone to coseismic landslides. The curves of landslide number density (LND, landslides/km2), landslide area percentage (LAP, landslide area/total area%), LA, and LV have a multimodal distribution (several patterns of response) by the topographic and seismic factors. It is recommended that other models be applied to better investigate the local effects of factors on the coseismic landslides.