Improving Skill of Rainfall Thresholds for Moisture-Driven Landslides by Integrating Root-Zone Soil Moisture at the Northeastern Himalayas

Moisture-driven landslides (MDLs) are a recurrent natural hazard in the Northeastern Himalayas (NEH) during the southwest monsoon season, where steep terrain and prolonged wetness frequently trigger catastrophic slope failures, underscoring the need for a credible early warning systems. In our recent work (Monga & Ganguli, 2025), we propose the compounding role of triggering and antecedent moisture content at an optimal d-day time lag to derive regional and local scale Event–duration (E–D) threshold model for northeastern Himalayas (NEH); however, we have not explicitly quantify the role of subsurface soil saturation in modulating the landslide likelihood. Here, we present at-site analysis of over the 21 landslide-prone sites across the NEH, considering the compound interaction of d-day time lag antecedent moisture, triggering rainfall and sub-surface root-zone saturation (up to 200 cm depth), and develop a moisture-preconditioned ED threshold model for landslides in a Bayesian probabilistic framework coupled with non-crossing quantile regression. To this end, we analyze 764 rainfall-induced landslides over 13-year (2007–2019) across the NEH and consider at-site rainfall time series from gauge-based high-frequency daily observations. The site-specific antecedent moisture content shows a mid-to-long-term memory, spanning from 2–3-week, prior to slope failure, reflecting the need to consider preceding antecedent accumulated moisture content in developing the ED threshold model. The derived 3-d E-D thresholds, computed at the modest (20th percentile) hazard level, demonstrate significant spatial variability: approximately 30% (6/21) of the sites show the robust control of antecedent moisture content over triggering rainfall, with varying optimal time lags that range from 3 to 60 days in triggering landslides. Conversely, ~25% (5/21) of the sites are more responsive to intense, short-duration rainfall in triggering slope failure. Within the Bayesian probabilistic framework, incorporating root-zone saturation, alongside the compounding role of triggering rainfall and antecedent moisture content, systematically elevates the landslide likelihood. At Kalimpong, accounting for effective soil saturation (S) of 0.85, we find an increase in the skill score by a factor of two in derived E–D thresholds, indicating the new model outperforms our earlier model as well as the one proposed in the literature.  Regionally, landslide likelihood peaks when high rainfall co-occurs with elevated sub-surface soil saturation, confirming a strong nexus between accumulated antecedent moisture content, subsurface soil saturation and short-duration record rainfall, in triggering slope failure. The derived insights aid in operational early warning systems, offering improved landslide forecast credibility in the NEH region with predominant space-time rainfall seasonality.