Item

Abstract

The comprehensive landslide hazard mapping of any region involves the development and integration of spatial and temporal probability of landslides. The spatial probability of landslides is governed by landslide susceptibility mapping (LSM) through weighted aggregation of its causative factors. The exceedance probability of rainfall is often considered as the proxy measure as an estimation of temporal probability of landslides. Under changing climate, the rainfall characteristics are subjected to substantial changes and the computation of temporal probability of landslide with rainfall magnitude alone is inaccurate. Also developing the estimation based on single rainfall event brings uncertainty in the estimation of landslide probability (LP). This study proposes a novel Copula based framework accounting for intensity and duration from multi-site rainfall information to develop LP map. The proposed method is applied for Idukki District of Kerala, India considering multiple extreme rainfall events of 2018. Firstly, the LSM of the district was developed using a robust Random Forest (RF) model. Then LP map was developed considering Generalized Extreme Value (GEV) formulations. Accounting for the intensity and duration of rainfall events, best fitting Frank Copula based estimation of joint probability is made as a surrogate of LP. Integration of the LS map with LP map provided the LH map of Idukki district for different return periods ranging from 1 to 200, in both cases. A comparison showed an under estimation of landslide hazards by GEV formulations over the Copula formulations, indicting necessity of bi- or multi-variate modeling of LP in the regional landslide hazard studies.