Regional earthquake forecast model for M $$\ge $$ 6 using strain rate and seismicity for Western Himalayas

Earthquake prediction plays a crucial role in hazard preparedness and offers an edge to the concerned authorities to minimize the subsequent risk to lives and the economy. However, predicting earthquakes is challenging, considering the uncertainties involved in the earth tectonics process and complicated interactions. In a general sense, the continuous movement of tectonic plates results in the gradual accumulation of strain, which leads to rupture and subsequent occurrence of earthquakes. Further, with the enhancement in the density of earthquake recording stations and technical advancement in the navigation satellite systems, both earthquake and geodetic data are readily available, but its application in successfully forecasting future earthquakes with reasonable spatiotemporal accuracy remains a challenging endeavour. This paper aims to develop a grid-based, time-independent regional earthquake likelihood model for the Western Himalayan region (RELM-WH). Here, horizontal strain rates are utilized to calculate the geodetic moment rate, which is balanced by the empirical correlation coefficient with the historical seismicity. After that, the Poisson process and Truncated Gutenberg Richter processes were employed to calculate the probability of M $$\ge $$ 6 in a period of 30 years at $$0.1^{\circ }\times 0.1^{\circ }$$ grid. The present study has developed the RELM models by utilizing two strain rate datasets. The first one is the Global Strain Rate Model version 2.1 (GSRM v.2.1) (Kreemer et al. in Geochem Geophys Geosyst 15(10):3849–3889, 2014) and the second one is from the study of Stevens and Avouac (Geophys J Int 226(1):220–245, 2021) specially tailored for the Indian region. A comprehensive comparison analysis was performed for the models developed from both data sets. From each dataset, six models are developed by considering the variable and uniform seismogenic thickness and rigidity for the region named UNI and VARI, respectively. It is observed that models with the assumption of uniformity have performed better than models with variability. Furthermore, it is also observed that models developed with regional datasets outperformed global dataset models on various performance metrics. Additionally, more than half of the study region is estimated to have a probability ( $$M \ge 6$$ in 30 years) greater than 1 $$\%$$ and a concentration of high probability has been observed near the Chamoli region of Uttarakhand and Kangra of Himachal Pradesh. Both regional and global models exhibit greater forecasting capability for earthquake epicentres than seismic moments during the retrospective testing phase, while in the pseudo-prospective phase, forecasting of seismic moment outperforms that of epicentral locations. The present study provides a comprehensive spatial distribution of potential seismic hazards for the Western Himalayan region, which the authorities can employ for hazard and risk mitigation strategies.