Debris flow susceptibility zonation using statistical models in parts of Northwest Indian Himalayas—implementation, validation, and comparative evaluation

Debris flows are natural disasters with devastating consequences and frequent recurrence in changing climatic regime of the Indian Himalayas. Therefore, it is necessary to delineate the debris flow susceptible areas in higher reaches of Himalayan terrain. With changing climate and unprecedented monsoon precipitation, debris flow susceptibility zonation may prove to be an effective way of disaster risk reduction. In the present study, debris flow susceptibility zonation has been attempted for a part of Chamoli district, Garhwal Himalayas, India. Three well-established statistical models involving frequency ratio, information value and certainty factor concepts have been implemented to integrate important influencing factors to generate the debris flow susceptibility zonation maps. Pre-2013 and post-2013 debris flow inventories have been used for model implementation cum success rate verification and prediction rate validation based on area under success rate curve and area under prediction rate curve concepts, respectively. It is inferred that both frequency ratio and certainty factor models have the fair classification quality for pre-2013 events and also have fair prediction capability for post-2013 events. Debris flow susceptibility maps generated using frequency ratio model and certainty factor model are found to be acceptable and appropriate as higher percentages of debris flows are found to occur in the very high susceptibility zone and a decreasing trend has been observed toward lower susceptibility zones. In contrary, the susceptibility map produced using information value model is found to be ambiguous as almost 99.81% areas of debris flows are observed in moderate and high susceptibility zones only. Comparative evaluation of debris flow susceptibility zonation maps is also attempted using density, error matrix, as well as difference image analysis. About 78.53% of the pixels are matching in both the frequency ratio as well as certainty factor-based debris flow susceptibility maps. On the other hand, information value-based susceptibility map has maximum mismatching of pixels with other two susceptibility maps. Such an attempt of debris flow susceptibility zonation mapping in Indian Himalayas considering both verification and validation data sets of debris flow events of two distinct periods is the novelty and uniqueness of the present study. Keeping in view the higher degree of risks involved in Himalayan debris flows due to their runout effects, such debris flow susceptibility zonation studies can be replicated in other parts of the Himalayan terrain.