Assessing wildfire susceptibility requires integrating environmental and anthropogenic factors to quantify the probability and vulnerability of fires in a given area. Many existing machine-learning models offer high predictive power but limited interpretability, restricting their utility for operational decision-making. This study is the first to apply the intrinsically interpretable deep network TabNet to wildfire susceptibility modeling. By fusing multi-source data and leveraging TabNet’s feature-mask matrix, we achieve accurate prediction and built-in explanation without relying on auxiliary tools. On a dataset of 133,811 samples, the proposed model achieves an Area Under the Curve (AUC) of 0.760, recall of 0.883, precision of 0.395, and an F1.5 score of 0.640, outperforming XGBoost (version 1.5.0) and other baseline models. The importance rankings derived from the feature-mask matrix align with the Shapley Additive Explanations (SHAP) results, confirming the reliability of the explanations. This approach combines predictive accuracy with transparency, providing a deployable framework for wildfire early warning, risk management, and ecosystem conservation.

This study presents a case study of the Ms5.5 Dezhou Earthquake to document the spatiotemporal characteristics of Outgoing Longwave Radiation (OLR) anomalies and their concurrent patterns with tidal force cycles. Based on NOAA satellite OLR data, synchronous monitoring and comparative analysis were conducted with tidal force variation cycles. The results show that pronounced OLR anomalies were concentrated exclusively in the co-seismic tidal cycle (Cycle C: 23 July–5 August 2023), while no significant anomalies were detected in pre-seismic Cycles A/B and post-seismic Cycle D. Temporally, the OLR anomalies in Cycle C exhibited a distinct six-stage evolutionary pattern: initial warming (31 July) → rapid intensification (1–3 August) → peak (4 August) → abrupt decline (5 August) → post-seismic pulse (6 August) → exponential decay (7–9 August). Spatially, the anomalies were closely distributed along the Liaocheng–Lankao Fault, showing a NE-trending (N35°E) distribution that matches the structural characteristics of the fault zone. Additionally, the spatial extent of OLR anomalies (within 400 km of the epicenter) is consistent with the effective detection range of co-seismic electromagnetic signals reported in existing studies. This study provides a typical observational case of OLR anomaly characteristics associated with medium-magnitude earthquakes, offering a reference for understanding the spatiotemporal evolution of seismic thermal anomalies.