Modeling of earthquake-induced landslide distributions based on the active fault parameters

Earthquake-induced landslides can cause enormous damage and numerous casualties in mountainous areas worldwide; consequently, accurate assessment of the potential distribution of earthquake-induced landslides is an important issue. For a national disaster prevention agency, it is desirable to be prepared to rapidly evaluate disasters and risks from information that can be scarce or easily available. The present study examined landslides triggered by reverse and strike-slip fault earthquakes, focusing on recent cases in Japan with relatively complete landslide inventories and detailed active-fault information. The landslide distributions and parameters of active faults were compared, and it was found that the skewness and kurtosis of the landslide distribution for each earthquake event was close to zero, indicating that the distributions followed the normal distribution. The mean distance of the landslide distribution from the active fault is negatively correlated with the inclination of the active fault, while the standard deviation is negatively correlated with the fault-top depth of the active fault and positively correlated with the magnitude of the earthquake. Based on our findings, we conducted a regression analysis of the relationship between the mean distance from the active fault and the inclination of the active fault. The mean distance of the landslide from the active fault increased when the active fault had a lower inclination, and was regarded as zero when the inclination was greater than 60 degrees. In addition, a relationship between the standard deviation, depth of fault top, and earthquake magnitude was established using multiple regression analysis. We thus confirmed the accuracy of the methodology using the actual landslide distributions of recent earthquakes in Japan. The length of the active fault can be used as input for the methodology, to estimate the maximum magnitude of the respective earthquake. Furthermore, the 95% confidence interval appears to cover almost all the large landslides, which enables us to limit the extent of areas that should be considered as the most exposed to co-seismic landslide hazards. The applicability of the proposed methodology was successfully tested using the case of the 1999 Chi-Chi earthquake in Taiwan with a seismic mechanism similar to the Japanese earthquake. Consequently, the proposed methodology can be applied to estimate the potential distribution of landslides caused by reverse and strike-slip fault earthquakes, based on the parameters of the active faults.