Time-Dependent Probabilistic Forecasting of Moderate-to-Large Earthquakes in Northeast India Using the Das Magnitude Scale Mwg: Considering the Misuse of the M or Mw Scale Below 7.5

Abstract

The adjoining parts of Northeast India represent one of the most seismically active regions of the Indian subcontinent. This region is classified as Zone V by the Indian Seismic Code, indicating maximum seismic hazard potential. This region has a history of several great earthquakes, including the 1897 Shillong and 1950 Assam events that resulted in widespread ground shaking, liquefaction, landslides and flooding damage. Hence, accurate probabilistic forecasts are essential for regional seismic hazard mitigation. An updated earthquake catalog was compiled for the area between 87–98° E and 20–30° N, with a focus on events of Mwg ≥ 5. A unified catalog was prepared for the analysis covering the period from 1897 to 2025. The region was divided into nine seismogenic zones to account for the spatial variability in recurrence patterns. The inter-event times between successive earthquakes were calculated for each zone, and three renewal models, lognormal, Weibull and gamma, were fitted by maximum likelihood estimation. The logarithmic probability of the likelihood function (ln L) was used to assess the model performance for a range of elapsed times (t) and forecast intervals (τ). The results indicate that the lognormal distribution at zero elapsed time (t = 0) provided the best fit and is, therefore, the most robust for short-term earthquake forecasting. Meanwhile, the Weibull model always gave the maximum of the conditional probabilities for longer elapsed times, especially for τ values between 13 and 25 years, and would, therefore, be preferable for long-term forecasts. Most of the zones had a gamma model that lay between the two, with moderate fits and probabilities. The zone analysis showed strong variability, with some zones with short recurrence times and high probabilities for the short term and others with long recurrence intervals. The Indo-Burma Range and the Eastern Himalaya zones had conditional probabilities of more than 90% for Mwg ≥ 5 earthquakes during 2044–2050, implying a higher seismic risk in these areas. The results suggest that there is no single renewal model that is optimal for all seismogenic zones and that the choice of model should be adapted to the elapsed time and the seismic characteristics of each zone. The present study offers a robust time-dependent probabilistic approach for the prediction of moderate-to-large earthquakes in Northeast India by combining an updated catalog and comparative renewal modeling. The findings are anticipated to contribute to seismic hazard assessment, infrastructure resilience planning and disaster preparedness strategies in one of the country’s most vulnerable regions.