Objective Affected by multiple factors such as extreme hazard events and climatic forcing, the Joshimath town on the southern slope of the Himalayas exhibits complex deformation characteristics. Joshimath is located in Chamoli district, Uttarakhand, northern India, between two major active faults—the Main Central Thrust and the Munsiari Thrust—and the town is largely built on paleo-landslide deposits. This study aims to analyze the spatiotemporal evolution patterns of deformation of the Joshimath landslide body and its associations with climatic factors and extreme hazard events.

Methods This study integrated time-series SBAS-InSAR, change-point detection, and signal decomposition methods to monitor ground deformation across nine areas in Joshimath town and to derive subsidence and uplift displacement time series. The decomposed displacement components (trend and periodic terms) were independently related to monthly air temperature and precipitation to identify deformation mutations and quantify their climatic associations. Finally, SARIMAX and LSTM models were compared to predict future deformation trajectories and assess the robustness of nonlinear trend extrapolation.

Results 1) Deformation showed strong spatial heterogeneity within the paleo-landslide body. Subsidence areas exhibited greater deformation intensity than uplift areas. Areas III and VII were strong-subsidence, high-risk areas, with sustained downward motion and clear dominance of the long-term trend component; by contrast, other areas showed weaker subsidence or alternating uplift–subsidence behavior. 2) Abrupt changes in displacement rate were detected in multiple areas. These change-point signals provided evidence that the 2021 Ronti ice-rock avalanche acted as a key external perturbation capable of altering deformation behavior. The clearest change point signatures and post-event trajectory shifts were observed in areas VI and VIII, where the displacement evolution showed a distinct change relative to the pre-event stage, demonstrating a stronger sensitivity to extreme-event forcing than in other areas. 3) The deformation time series showed a significant lagged response of 1−2 months to climatic factors. Air temperature exerted a more pronounced influence on deformation than precipitation, with partial correlation coefficients of 0.5−0.8, highlighting its dominant role in modulating deformation at the monthly scale. 4) LSTM outperformed SARIMAX in capturing nonlinear deformation and periodic climate-driven responses. The LSTM-based risk heat map indicated that the subsidence areas III and VII were future high-risk areas, while the uplift-dominant area VIII should also be considered as a potentially unstable area.

Conclusions This study reveals the joint control of extreme events and lagged climatic effects on deformation of a reactivated paleo-landslide, and predicts deformation trends using an SSA–LSTM framework. The results provide a scientific basis for stability assessment and hazard prevention for paleo-landslides on the southern Himalayan slope under ongoing climate change.