Assessing the probability of rainfall-induced landslides along the vulnerable slope-aligned transportation corridor in Darjeeling-Sikkim Himalayas

Shallow debris slides triggered by rainfall are widely distributed along the NH-10 in the Darjeeling–Sikkim Himalayan region, and most of them are reactive in nature. The Himalayan region is vulnerable to reactivated landslides due to extreme rainfall during the monsoon season. The occurrence of landslides largely depends on rainfall patterns. Therefore, the present study focuses on establishing the probability of landslide events P(A|B) with a certain amount of rainfall intensity (0 ≤ I <15, 15 ≤ I < 35, 35 ≤ I < 55, and I ≥ 55 mm/d) and predicting the expected number of landslide events and the number of landslides in different return periods. To achieve these objectives, a Bayes’ probabilistic approach and the Gumbel method have been employed. The pre-requisite data sets, such as rainfall information has been collected from India Meteorological Department (IMD) and tea garden adjacent to the highway, while landslides data have been gathered from Public Works Department (PWD) of National Highway Division-IX, previous research studies daily newspaper and field surveys for the last 15 years (from January, 2011 to December, 2025). During 8 years (2011–2018), the highway witnessed 666 landslides from 278 landslide events were used to calibrate the models. In that period, total 911 rainfall events have been identified from historical rainfall datasets. Whereas, 33 field-identified landslides that occurred between 2019 to 2025 were used to validate the proposed model. The conditional probability of landslide events for a given rainfall intensity and landslide return periods has been estimated using the Gumbel model along NH-10. In addition, similar estimates have been made for four major lithostratigraphic units that span overs the highway. The results indicate significant variations in landslide probability among lithostratigraphic units. The Siwalik Group and Gondwana Supergroup exhibit lower conditional probabilities across all rainfall-intensity classes than the slate and chlorite sericite schist rocks. The highest conditional probability (0.56) was observed for rainfall intensity of 35–55 mm/d on slate, and the lowest (0.18) for rainfall intensity of 15–35 mm/d in Gondwana Supergroup. In addition, the maximum numbers of landslide events (86 and 101) and landslides (270 and 326) are expected to occur during T 3 - and T 5 -year return periods for rainfall intensities of 15–35 mm/d on slate. The validation showed that nearly 88% of landslides observed when rainfall intensity is 15–35 mm in a day, demonstrating the reliability of the proposed model. Such efficient model may perform well in landslide forecasting and associated disaster risk reduction in complex mountainous terrain of Darjeeling–Sikkim Himalayas.