Mining-induced subsidence affects cropland use and land management in coal–cropland overlapping areas. Leveling provides high-precision subsidence time series but limited spatial sampling, whereas multi-temporal unmanned aerial vehicle light detection and ranging (UAV LiDAR) provides spatially continuous terrain-change information, although the uncertainty of digital elevation model (DEM) differencing must be assessed. Three UAV LiDAR surveys (30 May 2025, 25 December 2025, and 14 June 2026) and leveling measurements (March 2025–April 2026) were combined to analyze subsidence at the 32311 working face of Xutuan Coal Mine, Anhui Province, China, during working-face advance and after mining cessation. The level of detection at the 95% confidence level (LoD95) was calculated from elevation residuals in stable areas, and LiDAR-derived subsidence was validated against leveling measurements at 12 points. The LoD95 values of the two cross-sensor differencing periods were 131.9 and 148.8 mm, indicating that changes smaller than approximately 0.13–0.15 m could not be reliably distinguished from measurement error in this study. Leveling-derived cumulative subsidence at these points ranged from 19.76 to 392.95 mm during 30 May–19 December 2025, comprising 19.76–69.75 mm at the six weak-subsidence points and 185.13–392.95 mm at the six strong-subsidence points. LiDAR-derived subsidence showed a strong relationship with leveling measurements at the six strong-subsidence points (R2 = 0.87, RMSE = 29.9 mm), whereas the relationship was weak at the six weak-subsidence points (R2 = 0.15). At seven representative leveling points, post-cessation residual subsidence accounted for 15.9–31.2% (mean 20.8%) of the final cumulative subsidence recorded during the leveling period. Relative to 30 May 2025, the area with subsidence exceeding 150 mm increased from 15.31 ha on 25 December 2025 to 16.73 ha on 14 June 2026. The two techniques thus provide complementary temporal and spatial information: leveling measurements reveal the persistence of post-cessation subsidence, while UAV LiDAR characterizes changes in the spatial distribution of cumulative subsidence. Continued monitoring after mining cessation is therefore necessary to capture ongoing surface movement and its spatial characteristics.
SAM2 has great potential for remote sensing water body extraction due to its strong feature representation capability. However, when directly applied to the water body segmentation task, SAM2 still faces several challenges, including domain discrepancy, spectral confusion between water bodies and shadow-like objects, significant intra-class variations, and dependence on manual prompts. To address these challenges, this paper proposes the PMA-SAM2, which integrates domain prior injection, prototype memory calibration, and automatic prompt guidance for robust water body extraction. First, a Multi-level Prior Adaptation Module is designed to inject spatial geometric information and texture features into the SAM2 encoder, enabling effective alignment between general visual representations and remote sensing domain features. Then, a Prototype Memory Calibration Module is introduced to dynamically accumulate water feature prototypes under different geographical environments and imaging conditions, improving the model’s robustness to intra-class variations and inter-class spectral ambiguity. Furthermore, an Automatic Prompt Decoder is proposed to transform prototype-retrieved similarity maps into dense prompts, replacing manual prompts and enabling fully automatic water body segmentation. Experimental results on four datasets demonstrate that the proposed method achieves superior water body extraction accuracy compared with mainstream segmentation methods, verifying its strong generalization capability and scene adaptability.
Existing unmanned aerial vehicle fire monitoring methods primarily recognize visible smoke or flame cues but cannot reliably determine whether an observed anomaly is an uncontrolled fire requiring intervention. This study reformulates UAV fire monitoring from a conventional visual-recognition problem into an event-level reasoning and decision task. We propose Fire-Verify-Expert (FVE), a compact vision–language model that jointly performs smoke and fire cue confirmation, real-fire verification, event-type attribution, and ordered response-priority assessment. FVE incorporates fire-management logic into multimodal preference alignment through a multidimensional Group Relative Policy Optimization objective. Its reward mechanism constrains determination correctness, event-type consistency, rationale–decision consistency, and reasoning quality, while imposing asymmetric penalties on operationally costly errors, particularly missed real fires and severe underestimation of response priority. To support this task, we establish a UAV-oriented dataset of 600 images spanning varied environmental contexts and viewing perspectives, including real fires and visually similar controlled or currently non-actionable smoke/fire events. The dataset provides structured reasoning annotations and ordered response-priority labels. FVE achieves an accuracy/F1-score of 0.917/0.915 for real-fire verification, with a recall of 1.000 for real fires, and an accuracy/F1-score of 0.833/0.706 for response-priority assessment. These results demonstrate the effectiveness of FVE in aligning multimodal vision–language reasoning with fire-management requirements, providing a feasible approach for automated fire-event verification and response prioritization in frontline UAV fire-monitoring workflows.
Mesoscale eddies and internal solitary waves (ISWs) influence upper-ocean chlorophyll variability, yet their roles in the spatial variations in chlorophyll during their interaction remain poorly revealed. Using satellite ocean-color and sunglint imagery, we observed different chlorophyll variabilities in three cases: an ISW-only case, an eddy-only case, and an eddy–ISW interaction case. In the ISW-only case, chlorophyll exhibited coherent, banded, and phase-dependent anomalies, characterized by reduced concentrations near the wave trough but enhanced ones in the trailing region, with a chlorophyll anomaly reaching approximately 0.02 mg m−3. In the eddy-only case, chlorophyll exhibited a distinct low-core but high-periphery structure, with elevated filamentary features concentrated around the eddy boundary and spatially consistent with enhanced finite-size Lyapunov exponent (FSLE). During the eddy–ISW interaction, the ISW crests became strongly deformed near the eddy periphery, consistent with modulation by the mesoscale eddy. Within the eddy interior, the ISW-related banded chlorophyll structure remained evident, whereas near the eddy periphery these bands became curved and spatially overlapped with filamentary chlorophyll structures. The difference in chlorophyll between the eddy–ISW and eddy-only cases increased toward the eddy periphery, reaching approximately 50% at ~1.2R. These observations reveal a multiscale spatial modulation of chlorophyll variability, in which the mesoscale eddy establishes a broader spatial pattern, while ISWs induce propagating fine-scale perturbations. The eddy–ISW interaction primarily regulates and enhances chlorophyll variability near the eddy periphery, rather than producing a uniform response throughout the eddy.
Cross-platform land surface temperature (LST) harmonization is often evaluated by agreement of individual observations, although urban analyses commonly rely on spatial differences and temporal changes. We investigated whether improving agreement in urban-core and surrounding-ring LST also improves their core–ring contrast and pre-to-post-sunset transition. Using Advanced Baseline Imager observations from Geostationary Operational Environmental Satellites (GOES)-16, -17, -18, and -19 over western North American cities during 2019–2026, we evaluated five cohorts using established mean-squared-error (MSE) decomposition and cross-validated model selection at each measurand. Component root-mean-squared error (RMSE) decreased by 29.3–59.7%, whereas core–ring contrast RMSE ranged from an 11.5% reduction to a 13.1% increase. Loss of residual covariance offset 94.6–114.0% of gross variance-plus-bias gains. In the post hoc 2026 temporal evaluation of the same 75 cities used for 2025 calibration, core and ring RMSE declined by 59.7% and 56.1%, but contrast RMSE reduction was −0.14% (95% crossed city–interval bootstrap interval, −6.75% to +6.48%, with cities and intervals resampled independently). The historical 2021 evaluation likewise showed only a small, uncertain contrast improvement (2.9%; 95% interval, −4.3% to +9.8%). Calibration-based selection retained raw contrasts and transitions, while a residual interval rule withheld uncertain transition direction claims. Independent three-station radiometric comparisons assessed raw satellite products at station scale rather than the complete urban workflow. These findings demonstrate why harmonization performance should be evaluated at the derived scientific quantity, with processing decisions based on its observed cross-platform consistency.
