High-speed railways in the Beijing–Tianjin–Hebei (BTH) Plain are gradually becoming more widespread, covering a greater area. The operational safety of high-speed railways is influenced by the continuous development of land subsidence. It is necessary to predict the subsidence along the high-speed railways; thus, this work is of critical importance to the safety of high-speed railway operation. In this study, we processed Sentinel-1A data using the Persistent Scatterer Interferometric Synthetic Aperture Radar (PS-InSAR) technique to acquire the land subsidence in the typical BTH area. Then, we combined the Empirical Mode Decomposition (EMD) and Gradient Boosting Decision Tree (GBDT) methods (EMD-GBDT) to forecast land subsidence along high-speed railways. The results revealed that some parts of the high-speed railways in the BTH plain had passed through or approached the land subsidence area; the maximum cumulative subsidence of the Beijing–Shanghai, Tianjin–Baoding and Shijiazhuang–Jinan high-speed railways reached 326 mm, 384 mm and 350 mm, respectively. The forecasting accuracy for land subsidence along high-speed railways was enhanced by the EMD-GBDT model. The Root Mean Square Error (RMSE) and Mean Absolute Error (MAE) were 0.38 mm to 0.56 mm and 0.23 mm to 0.38 mm, respectively.

Cangzhou is located in the northeast part of the North China Plain; here, groundwater is the main water source for production and living. Due to the serious regional land subsidence caused by long-term overexploitation of groundwater, the monitoring of land subsidence in this area is significant. In this paper, we used the Small Baseline Subsets Interferometric Point Target Analysis (SBAS-IPTA) technique to process the Envisat-ASAR, Radarsat-2, and Sentinel-1A data and obtained the land subsidence of Cangzhou from 2004 to 2020. Additionally, we obtained winter wheat distribution information in Cangzhou using the Pixel Information Expert Engine (PIE-Engine) remote sensing cloud platform. On this basis, we analyzed the relationship between ground water level, winter wheat planting area, and the response of land subsidence according to the land use type and groundwater level monitoring data near the winter wheat growing area. The results show that during 2004–2020, the average annual subsidence rate of many places in Cangzhou was higher than 30 mm/year, and the maximum subsidence rate was 115 mm/year in 2012. From 2004 to 2020, the area of the subsidence funnel showed a trend of first increasing and then decreasing. In 2020, the subsidence funnel area reached 6.9 × 10³ km². The winter wheat planting area in the urban area showed a trend of first decreasing, then increasing and then decreasing, and it accounted for a large proportion in the funnel area. At the same time, we studied the relationship between the land subsidence rate and the water level at different burial depths and the response of winter wheat planting area. The results showed that the change of confined water level had a stronger response with the other two variables.

Land subsidence is a widespread geological hazard in urban areas, yet accurate prediction remains challenging due to the complex, non-linear spatio-temporal dynamics governing subsurface deformation. In this work, the PEGNet (Peridynamics-Emergent-GAT-GRU Network) method integrates Graph Attention Networks (GAT) with Gated Recurrent Units (GRU) under a three-level embedding strategy to predict land subsidence at monitoring points. Peridynamics gradient and emergent velocity features are fused with multi-source data at the feature level; multi-head graph attention captures spatial dependencies among monitoring points at the structure level; and a Peridynamics-based loss regularizer penalizes unrealistic spatial gradients while an attention entropy regularizer encourages focused attention patterns at the constraint level. Evaluated in Tongzhou District, Beijing over 60 months, the model achieves an RMSE of 5.16 mm, outperforming the GAT-GRU baseline by 25.1% and reducing RMSE by 51.7% relative to GCN-LSTM. Comprehensive ablation studies and baseline comparisons validate each component, showing that the dual-constraint framework effectively suppresses large-magnitude outliers. This work demonstrates that coupling macro-micro priors with spatio-temporal deep learning provides an effective paradigm for integrating physical and geoscience principles into data-driven subsidence prediction.

Land subsidence prediction remains challenging. Conventional grid-based or sequence-only neural networks struggle to represent these spatial dependencies and often lack structured mechanisms for incorporating region-level deformation priors and local physical consistency. This study develops PEGNet, a Peridynamics-inspired and emergent-feature-conditioned spatio-temporal graph neural network for land subsidence prediction. PEGNet uses a per-node GRU for temporal encoding and an EF-conditioned GAT for spatial aggregation. At the node level, historical InSAR deformation and auxiliary hydroclimatic observations are augmented with a PD-derived state and an EF neighborhood context, allowing a per-node GRU to learn temporal evolution and produce a base estimate of future incremental deformation. At the edge level, a fixed spatial graph supports EF-conditioned graph attention, where edge representations combine spatial distance, PD-derived bond-strain information, and the EF same/cross-region gate to generate a spatial residual prediction. At the training and prediction level, the temporal and spatial branches are fused through a learned gate, cumulative deformation is reconstructed from the last observed value, and training combines a data-fitting loss with a PD-inspired local consistency term weighted by distance and EF relations. EF therefore conditions both the network and the coupled regularizer, although it introduces no separate loss term, and the PD component remains a local consistency mechanism rather than a complete Peridynamic solver. PEGNet is evaluated using 60 months of PS-InSAR observations in Tongzhou District, Beijing, four overlapping purged test windows covering nine unique target months and five random seeds. It achieves an incremental-deformation RMSE of 2.153 ± 0.007 mm and a reconstructed cumulative-deformation RMSE of 2.631 ± 0.072 mm. These results outperform the deterministic baselines and remain comparable to the unconstrained GRU-GAT model. Mechanism diagnostics indicate that EF mainly redistributes graph attention, whereas the PD components improve local consistency without producing a substantial global accuracy gain. Overall, the comparable aggregate performance and improved local consistency support the use of PEGNet for conditional subsidence monitoring and scenario analysis.