EARLY WARNING SYSTEM FOR RAINFALL-INDUCED LANDSLIDES ALONG NH-5, USING RAINFALL THRESHOLD AND SLOPE STABILITY ANALYSIS

Abstract

The regular occurrence of landslides along National Highway 5 during the monsoon season is a cause for concern as they lead to substantial damage and traffic congestion. An integrated empirical and numerical modelling approach is required to mitigate the damage because there has never been an empirical rainfall threshold established, particularly for this section. This study is an effort to determine the local thresholds for rainfall that cause landslides based on the amount and timing of precipitation before a landslide event occurs along the hilly section of NH5 in Himachal Pradesh, India. To determine the threshold for rainfall, statistics on about 80 landslide episodes are gathered with their dates of occurrence. However, rainfall episodes involving 68 of the 80 landslides were examined to determine an empirical intensity-duration threshold for landslide occurrences. A frequentist approach is utilized to calculate the power law curve's slope ‘β’ and intercept ‘α', which serve as indicators of the rainfall threshold. The relationship between rainfall intensity and duration that was fitted to the lower boundary of rainfall events that caused landslides is I=0.8046 D0.9111 (I=rainfall intensity in millimeters per day and D=duration in days). For the validation of the threshold, numerical modelling is carried out on a slope that had previously failed, considering the rainfall event that preceded it, which had a mean rainfall intensity of 1.8 mm/day. After that, the effect of rainfall on the slope's Factor of Safety is assessed both before and after rainfall. From the obtained threshold equation, it can be concluded that a continuous rainfall of 1.513 mm/day for the minimum duration of 2 days can trigger landslides. The slope failure and validation of the threshold equation are confirmed by the reduction of the factor of safety to a value of 0.623, which was larger than 1 before the rainfall event.