Entropy-Based Modelling for Flash Flood Hazard Mapping in Uttarakhand Himalaya

In light of recent hazards witnessed in Uttarakhand such as Kedarnath flash flood 2013, Chamoli flash flood 2021 and Maldevta flash flood 2022, flash flood hazard susceptibility (FFHS) mapping has become an important requirement for undertaking flood mitigation activities especially in mountainous regions. A number of approaches have been used for modelling flash flood susceptibility; however, ML-based techniques have shown significant potential. In the present study, the maximum entropy model (MAXENT model) approach was used to generate the FFHS map. Maxent is a presence of only model where the input used to train the model is the point locations of previous flash floods. About 100 field-based georeferenced points obtained from Uttarakhand State Disaster Management Authority were used to train the model whereas 29 for validating the modelled output. Terrain-derived parameters like elevation, slope, aspect, terrain wetness index, terrain ruggedness index, stream power index, sediment transport index, distance to river, distance to landslides, planform curve, profile curve and land cover were provided as input to the model. The model returned an area under the receiver operating characteristic curve value of 0.91 (91%). From the Jack-knife test, it was observed that distance to river, distance to landslide, elevation and land use/land cover are the four most critical conditioning factors for FFHS mapping. From the FFHS map, it is assessed that about five percentage of the study area falls under high and very high flash flood hazard potential classes. The very high-to-high hazard is mainly concentrated along the main stem of the Mandakini river.