Early Prediction of Landslide Using IoT and Deep Learning Model

Abstract: Landslides constitute a pervasive geohazard in monsoon-driven topographies, inflicting substantial socioeconomic devastation through abrupt slope failures triggered by hydrogeological stressors. This research presents an IoT-RNN/LSTM framework for early landslide prediction, fusing real-time multivariate sensor telemetry rainfall intensity, soil moisture saturation (>30%), pore pressure gradients, inclinometer tilt angles, and seismic vibrometer with deep recurrent architectures to model spatiotemporal failure precursors. ESP32 edge nodes aggregate data via MQTT, preprocessing through min-max normalization and variational mode decomposition (VMD), feeding hybrid LSTM that attain 95.2% F1-score and 24-48-hour lead times on benchmark datasets. Deployed alerts cascade through LED/buzzer/LCD/GPS interfaces, achieving