Climate variability as a driver of slope stability: integrating satellite data and hydro-geotechnical modeling for tropical railway corridors.

Climate variability exerts a fundamental control on the timing and recurrence of rainfall-induced landslides, particularly in tropical regions characterized by deeply weathered soils, pronounced wet–dry seasonality, and sparse ground-based monitoring networks. In this context, climate variability primarily acts as a preparatory factor by regulating antecedent moisture conditions, soil suction, and seasonal hydrological states, while also modulating the frequency and intensity of rainfall events that act as triggers. Although advances have been achieved in climate science, remote sensing, and slope stability modeling, these developments remain only partially incorporated into engineering geological assessments of infrastructure slopes. This study addresses this gap by presenting a climate-informed framework that links large-scale climate variability to local hydro-mechanical slope response in tropical railway environments.

The proposed framework integrates multi-source satellite data with probabilistic and physically based analyses to assess rainfall-induced slope instability. Precipitation data were obtained from CHIRPS (0.05° spatial resolution; 1981–2023), while soil moisture was derived from SMAP products. Topography was represented by the ALOS PALSAR Digital Elevation Model (12.5 m; JAXA, 2021), and vegetation conditions were characterized using NDVI from CBERS-4A imagery acquired on 4 August 2020 (12.5 m). Landslide susceptibility along the railway corridor was mapped using a probabilistic Random Forest model and independently validated with ground deformation data derived from descending-orbit Sentinel-1 SAR images (22 May 2022–26 September 2023) processed using the SqueeSAR InSAR technique. The framework also incorporates hydro-geotechnical characterization, transient numerical modeling, and UAV-based LiDAR surveys.

At the slope scale, the framework emphasizes unsaturated soil behavior, recognizing rainfall infiltration and suction loss as dominant triggering mechanisms in tropical soils. Field and laboratory investigations define soil–water retention characteristics and hydraulic conductivity functions, enabling representation of seasonal moisture dynamics. These parameters are incorporated into coupled transient seepage and slope stability simulations driven by long-term satellite-based rainfall time series. Furthermore, the simulations account for soil–climate interactions by explicitly considering evapotranspiration effects and antecedent moisture conditions, capturing the interactions between climate variability, infiltration processes, and mechanical response.

The susceptibility analysis demonstrates the effectiveness of the Random Forest model in identifying zones prone to shallow landsliding along the railway, with strong agreement between predicted high-susceptibility classes and observed slope instabilities. These results support the selection of critical slopes for detailed numerical investigation. Subsequent coupled seepage and slope stability simulations reveal strong sensitivity of slope stability to rainfall intensity and antecedent moisture conditions, with distinct responses to daily extreme rainfall events and multi-day cumulative rainfall. Seasonal and interannual variability associated with ENSO phases modulates pore-pressure evolution and safety margins, producing periods of increased vulnerability even in the absence of significant long-term precipitation trends.

By coupling climate signals, hydrological processes, and mechanical behavior, the proposed framework provides a practical pathway for integrating climate information into engineering geological assessments. The approach is particularly suited to data-scarce regions such as the Amazon, where satellite observations can partially compensate for limited in situ monitoring, supporting improved slope susceptibility evaluation and climate-informed decision-making.