Mine water inrush disaster seriously threatens the production of coal mine. Rapid and accurate identification of mine water inrush sources is a key premise for mine water disaster prevention. The conventional research on the identification of water inrush source has focused on a single source, and the identification of mixed water samples from multi-source aquifers in deep coal mining environment is not yet fully explored. In this study, absorption spectrum technology was introduced into the identification of water inrush sources. The absorption spectra of the water samples with different mixing ratios were prepared using the ultraviolet and visible spectrophotometry (UV–Vis) spectrophotometer. In addition, spectral data preprocessing such as scattering correction, baseline correction, smoothing and denoising, and data enhancement were conducted to reduce the influence of experimental error, environment, radiation, molecular interaction, and other factors on the spectral data. Furthermore, a genetic algorithm (GA) was used to improve the seven parameters of the extreme gradient boosting (XGBoost) algorithm, such as learning rate, base model selection, tree parameters, regularization parameters, and iteration times. The deep-learning classifier of mine mixed water sources based on GA-XGBoost was established and used to identify 66 groups of mixed water sources in the Huangyuchuan Mine. The simulation results show that spectral preprocessing and normalization enhancement effectively improved the accuracy of the discriminant model. After 100 cross-validations, the average recognition accuracy of the GA-XGBoost model was 94%, and the results were accurate and reliable. This study provides a new direction and method for the identification of water inrush sources, particularly for mixed water inrush sources. It may also serve as a technical reference for decision-makers to formulate effective coal mine water inrush prevention and control programs and for mine water disaster prevention in similar coalfields in North China.

Mining activities have significantly altered the land use patterns of mining areas, exacerbated the degree of landscape fragmentation, and thereby led to the loss of biodiversity. Ecological networks have been recognized as an essential component for enhancing habitat connectivity and protecting biodiversity. However, existing studies lack dynamic analysis at the landscape scale under multiple future scenarios for mining areas, which is adverse to the identification of ecological conservation regions. This study used the MOP-PLUS (multi-objective optimization problem and patch-level land use simulation) model to simulate the land use patterns in the balance of ecology and economy (EEB) scenario and ecological development priority (EDP) scenario for the Shendong coal base. Then, climate change and land use patterns were integrated into ecosystem models to analyze the dynamic changes in the ecological networks. Finally, the conservation priorities were constructed, and dynamic conservation hotspots were identified using landscape mapping methods. The following results were obtained: (1) From 2000 to 2020, large grassland areas were replaced by mining areas, while cultivated land was replenished. By 2030, the forest and grassland areas (967.00 km², 8989.70 km²) will reach their peaks and the coal mine area (356.15 km²) will reach its nadir in the EDP scenario. (2) The fragmentation of ecological sources intensified (MPS decreased from 19.81 km² to 18.68 km²) and ecological connectivity declined (in particular, $\alpha$ decreased by 6.58%) from 2000 to 2020. In 2030, the connectivity in the EDP scenario will increase, while the connectivity in the EEB scenario will be close to that of 2020. (3) The central and southeastern parts of the Shendong coal base have higher conservation priorities, which urgently need to be strengthened. This study offers guidance on addressing the challenges of habitat and biodiversity conservation in mining areas.

Destructive earthquakes frequently sever transportation lifelines, significantly impeding the progress of emergency rescue and post-disaster reconstruction efforts. The automated identification of road damage utilizing high-resolution remote sensing imagery is strictly constrained by the scarcity of post-disaster labeled samples and the morphological complexity of road networks. Consequently, model segmentation results frequently suffer from discontinuities in topological connectivity and confusion between background features and damaged roads. To address these challenges, this study proposes a road damage detection framework that integrates generative artificial intelligence with vector prior knowledge. A data simulation pipeline utilizing a stable diffusion model was constructed, employing topologically constrained masking to generate high-fidelity synthetic damage samples based on the DeepGlobe dataset, thereby mitigating the data deficit. The proposed Vector-Guided Damaged Road Segmentation Network (VRD-U2Net) employs wavelet convolutions (WTConv) to decouple high-frequency noise from low-frequency structural components and utilizes a Multi-Scale Residual Attention (MSRA) module to align visual features with vector priors. Furthermore, a vector-prior-driven dynamic upsampling mechanism is introduced to enforce geometric constraints on model predictions. Experimental results demonstrate that the method achieves an mIoU of 0.884 on the synthetic dataset. In validation using real-world imagery from the 2023 Turkey earthquake, the model attained an F1-score of 65.3% and recall of 72.3% without fine-tuning, exhibiting robust generalization capabilities to support manual damage assessment in data-scarce emergency scenarios.

As a mining country, China faces enormous challenges in the context of the global commitment to achieve carbon neutrality. In order to achieve this goal, the Chinese government is actively promoting the green and low-carbon transformation of the energy system. Consequently, an increasing number of mines with poor production capacity and depleted resources are being closed down or eliminated, leading to a large quantity of stranded land resources that are now idle. However, in the process of rapid economic development, China is facing serious problems, such as land shortage and land use conflicts. Abandoned mining land (AML), as a kind of reserve land resource, has an important regulating role in solving the dilemma of land resource tension faced by national land spatial planning. In order to realize the rational planning and utilization of AML, this study proposes a high-precision AML planning model and simulates the planning of AML in multiple policy scenarios, using Ningbo City as an example. The results show that AML has great economic and ecological potential; the economic development scenario (EDS) enhanced the economic benefits of the mine region by 396%, and the ecological protection scenario (EPS) enhanced the ecological benefits of the mine region by 74.61%, when compared with the baseline scenario (BAU). The overall level of optimization is as follows: EDS > EPS > BAU. In addition, the optimal utilization of AML in all three scenarios significantly enhanced the ecological quality of the mining region, and the enhancement effect was EPS > BAU > EDS. Therefore, AML, as a kind of free land resource, has an important supporting effect for the spatial planning of the national territory. Furthermore, it is of great significance to scientifically and reasonably guide the optimal utilization of AML, according to the policy planning for future development, in order to achieve efficient economic development and improve the quality of the ecological environment.

Accurate mapping of building damage after destructive natural disasters is essential for emergency response and recovery planning. However, annotated post-disaster samples are often scarce, and conventional supervised models usually generalize poorly across scenes and hazard types. These constraints limit the operational use of remote sensing-based damage assessment during emergency response. To address this problem, this study proposes SCTSC-VLM, a training-free framework for damage grading of pre-localized building instances, implemented using a locally deployed Q4_K_M 4-bit quantized version of the Qwen3-VL-30B vision–language model. The framework has three main components. First, an instance-level visual input construction strategy integrates independent assessment of building instances localized by the supplied footprint data without a dedicated instance segmentation model. Second, a hierarchical reasoning mechanism decomposes the assessment into four progressive stages: multi-view structured feature extraction, independent visual description, cross-validation, and rule-constrained classification. This design is intended to reduce hallucination and visually ungrounded reasoning. Third, a multi-sampling ensemble decision mechanism mitigates the randomness of autoregressive generation. The framework was evaluated using post-disaster optical imagery from the 2025 Dingri earthquake and the 2025 Eaton Fire in California. On the full Dingri evaluation dataset comprising Areas A–C (n = 398), SCTSC-VLM achieved a Macro-F1 score of 88.2% and a Kappa coefficient of 0.806. On the separate full California evaluation dataset (n = 386), it achieved a Macro-F1 score of 83.8% and a Kappa coefficient of 0.800. Separate ablation experiments were conducted on explicitly defined analysis populations to examine the contributions of the framework components. In both full-dataset evaluations, the fully damaged and undamaged classes obtained higher F1-scores than the partially damaged class. These results provide case-level evidence that structured reasoning can support training-free damage grading of pre-localized buildings, although broader generalization and large-area operational efficiency remain to be established.