366 Tailings dams often operate under more variable conditions than conventional water-retaining dams, which can increase vulnerability. Key risk factors include the use of tailings or waste rock as construction material, progressive raising (especially upstream), historical gaps in design standards and regulation, and the need for long-term surveillance during operation and after closure when access and resources may be constrained (Canadian Dam Association, 2013). Although failures are relatively infrequent, they can be catastrophic, releasing large volumes of contaminated tailings with severe downstream impacts, including loss of life and environmental damage (Rico et al., 2008; Silva Rotta et al., 2020). These events highlight the need for lifecycle risk management and monitoring systems capable of detecting early signs of instability under changing operational and environmental conditions. Monitoring typically relies on visual inspections and in-situ instrumentation (e.g., piezometers, inclinometers, survey benchmarks). Instrumentation provides quantitative data that supports performance evaluation and stability assessment (Lumbroso et al., 2019), but measurements are usually sparse and require installation, maintenance, and safe site access. For large or remote TSFs, dense sensor deployment can be impractical due to cost and logistical risk (Zare et al., 2024), motivating complementary techniques that provide wide-area deformation measurements without direct human exposure. Over the last two decades, InSAR has become a standard remote-sensing technique for deformation monitoring. By exploiting phase differences between repeat SAR acquisitions, it measures displacement along the radar line of sight with centimetre- to millimetre-scale sensitivity over wide areas (Hanssen, 2001; Massonnet & Feigl, 1998; Rosen et al., 2000). Multi-temporal approaches, particularly Persistent Scatterer Interferometry (PSI) and Small Baseline Subset (SBAS), use long image stacks to reduce noise and improve detection of slow deformation. This capability is well suited to TSFs, where changes can develop gradually before visible distress. InSAR also provides dense spatial coverage across the dam and surrounding areas, enabling detection of localised deformation that sparse instrumentation may miss (Crosetto et al., 2016; Rana et al., 2024). Sentinel-1 time series have revealed accelerating deformation prior to several failures, including Brumadinho (2019) and Cadia Valley (2018), supporting its value for early detection when interpreted with site knowledge (Das et al., 2024; F. Gama et al., 2020). Consequently, InSAR is increasingly used as a complementary monitoring layer in TSF management (Sánchez et al., 2026). However, observation alone is insufficient for proactive risk management; forecasting is needed to anticipate near-future deformation under changing operational and environmental conditions. TSF deformation reflects interacting internal processes and external forcing. Rainfall is often a key driver because could raise pore pressures and the phreatic level, reduce effective stress, and accelerate settlement or trigger instability (Zhu et al., 2022). Temperature can modulate longer-term behaviour through seasonal drying–wetting cycles and thermo-mechanical effects. Traditional approaches commonly use regression models linking displacement to cumulative rainfall, water-level indicators, or time. Multiple linear regression is widely applied due to simplicity and interpretability (Liu et al., 2023), but it can struggle with nonlinear, time-dependent behaviour. ARIMA models can reproduce short-term trends under near-linear and stationary conditions, yet performance often degrades for longer horizons in nonstationary systems with multiple interacting drivers, as is typical for tailings dams (Hill et al., 2021). Machine Learning (ML) offers an alternative by learning nonlinear relationships directly from data (LeCun et al., 2015). Long Short-Term Memory (LSTM) models, have shown strong performance in dam and landslide forecasting by capturing temporal dependencies (Nava et al., 2023; Yang et al., 2020; Zhu et al., 2022), but they typically require large, consistent datasets for robust training, which are not always available in TSF monitoring (Bansal et al., 2022). Ensemble tree methods, particularly gradient boosting, are therefore attractive for moderate sample sizes and heterogeneous predictors. LightGBM is a computationally efficient implementation that has achieved strong deformation forecasting performance using rainfall histories, hydrological indicators, and displacement “memory” features (Chen et al., 2025). In this study, we present a LightGBM framework for short-term TSF vertical deformation forecasting using Sentinel-1 MT-DInSAR, rainfall, and temperature. Zone-specific models predict the next deformation increment at the native 12-day acquisition interval, preserving operational relevance. Benchmarking against baseline approaches shows that gradient boosting can provide useful near-term forecasts to complement TSF monitoring and support risk-informed decision-making.
RkJQdWJsaXNoZXIy MTM0Mzk2