Search Results (2,696)

X-band rainfall radar systems offer advantages such as high spatial resolution, flexible deployment options, and strong near-surface detection capabilities. However, due to the short electromagnetic wavelength in this band, the radar base data are highly susceptible to multiple coupled factors—including clutter from mountain vegetation, tall buildings, and other terrain features; electromagnetic interference from surrounding radio-frequency equipment; beam obstruction caused by complex topography; and attenuation of rainfall intensity along the precipitation path—resulting in pronounced distortion of raw echoes. This distortion significantly hinders the accuracy of quantitative precipitation estimation in mountainous regions and makes it difficult to meet the operational requirements for precise mountain torrent early warning systems. To address the problem, this study utilizes real-time observational data from field deployments of X-band rainfall radars in mountainous regions to construct a comprehensive, progressive quality control system comprising: refined removal of terrain clutter; electromagnetic interference pre-suppression; dynamic beam obstruction correction; and adaptive rainfall intensity attenuation correction. The terrain clutter suppression algorithm is optimized in this study, and an adaptive beam-blockage compensation algorithm based on the terrain-blockage fraction is further adopted. However, beam-blockage correction exhibits limited improvement in this case, which is likely attributed to the complementary observational coverage provided by higher-elevation radar scans. The echo-missing regions are dynamically corrected according to the terrain-blockage coverage ratio at different elevation angles and azimuths to realize differentiated compensation corresponding to blockage severity so as to effectively restore the true echo intensity obscured by terrain. Furthermore, considering the prominent rain-induced attenuation of radar electromagnetic waves along the propagation path, a dynamic attenuation correction scheme based on path-integrated reflectivity is introduced. The precipitation attenuation coefficient is dynamically calculated point by point from the variation characteristics of real-time echo intensity along the beam path to compensate echo loss, which addresses the limitation that fixed correction parameters cannot adapt to attenuation differences under variable rainfall intensities. Based on the Z-R power-law relationship, the radar echo-derived rainfall is inverted, using hourly measurements from dense ground-based rain gauges as the reference values, and precision is verified using the MB and RMSE—industry-standard meteorological evaluation metrics. Experimental results demonstrate that clutter suppression dominates RMSE reduction (7.37 to 1.91 mm/h). Beam blockage correction shows negligible impacts on RMSE and MB. Attenuation correction delivers marginal MB improvement (−0.52 to −0.51 mm/h) with no measurable RMSE response. After full-chain quality-control optimization, the aggregate RMSE between radar-derived rainfall and gauge observations is 1.91 mm h⁻¹, representing an approximately 74% reduction compared with the raw dataset (from 7.37 to 1.91 mm h⁻¹). Elevated RMSE values of 4.2–5.0 mm h⁻¹ are observed for rainfall intensities between 5 and 10 mm h⁻¹. In addition, the discrepancy between radar and gauge estimates grows as rainfall intensity increases. As the distance between radar and rain gauge increases, radar QPE systematically underestimates light precipitation, while persistent underestimation occurs across all ranges for heavy rainfall. This study enhances the differentiated quality control framework for X-band radar systems in complex mountainous regions, effectively improves the observation quality of radar-derived base data, and provides data support and technical references for dynamic monitoring of flash floods in mountainous river basins. Full article

In data-sparse Central Asian watersheds, machine learning and Geographic Information Systems (GIS) are increasingly applied to hydrological hazard classification. This article compares Random Forest (RF), XGBoost, and Long Short-Term Memory (LSTM) models for daily elevated-discharge classification in the Koksu River basin, Zhetysu Region, Kazakhstan (1614 km², semi-arid, snowmelt- and glacier-influenced climate; 2005–2023), using verified discharge, precipitation, and temperature records from four monitoring stations. The elevated-discharge threshold (80th percentile, ~90.5 m³/s) was computed using only the training period (2005–2017) to avoid temporal data leakage into validation and test periods. Class imbalance, addressed via class weighting, was moderate under this threshold. All models were benchmarked against a naive persistence baseline. Contrary to expectation, the persistence baseline achieved the highest Critical Success Index (CSI = 0.821, 95% block-bootstrap CI [0.725, 0.891]), narrowly ahead of XGBoost (CSI = 0.813, [0.703, 0.897]), Random Forest (CSI = 0.809, [0.691, 0.897]), and LSTM (CSI = 0.763, [0.630, 0.873]); the confidence intervals overlap substantially, indicating no statistically distinguishable advantage of any trained model over simple persistence in this basin. Feature-importance and ablation analyses confirmed that lagged discharge, not precipitation, drove nearly all predictive skill: a discharge-lags-only model performed as well as or better than the complete pipeline, while a meteorology-and-seasonality-only model performed markedly worse (CSI ≈ 0.57–0.59). As a preliminary, exploratory illustration rather than a core result, Gumbel-derived return-period discharges (Q10, Q25, and Q100) were separately translated into illustrative, uncalibrated, DEM-proximity-based inundation extents, a candidate direction for future work rather than an engineering-grade flood-mapping contribution of this study. These results indicate that, in this basin and period, machine learning classifiers provide no clearly demonstrated advantage over simple discharge persistence, underscoring the necessity of routine persistence benchmarking before adopting machine learning approaches for early warning in Central Asian watersheds facing increasing flood risk under climate change. Full article

Constructed wetlands (CWs) offer a low-cost and sustainable option for wastewater treatment, but their performance is commonly assessed through infrequent, periodic field sampling that yields temporally sparse records. This study evaluated whether five-day biochemical oxygen demand (BOD₅) removal efficiency at the Gložan horizontal subsurface-flow constructed wetland (Serbia) could be predicted from freely available satellite-derived and meteorological variables alone, using an interpretable machine learning workflow. Sixteen candidate predictors—the mean and spatial standard deviation of Landsat-derived indices (land surface temperature [LST], Normalized Difference Vegetation Index [NDVI], Normalized Difference Water Index [NDWI], Normalized Difference Suspended Sediment Index [NDSSI], Modified NDWI [MNDWI], and chlorophyll index) together with 7- and 14-day air temperature and precipitation—were screened against 38 BOD₅ removal-efficiency observations (n = 38; 2005–2026) using Pearson correlation and three Random Forest importance measures, refined through variance-inflation-factor analysis and regularization-guided elimination, and evaluated across twelve regression algorithms under nested leave-one-out cross-validation (LOOCV). A one-component Partial Least Squares (PLS) regression using four predictors—14-day mean air temperature, spatial variability of land surface temperature (LST), and the mean and spatial variability of the Normalized Difference Water Index (NDWI)—achieved the best performance in the external (outer-loop LOOCV) evaluation (RMSE = 5.73 percentage points, MAE = 4.37 percentage points, R² = 0.41). Linear-family models consistently outperformed tree-ensemble and kernel-based methods, and the explicit removal of multicollinearity during predictor selection was key to this advantage, allowing the linear models to surpass their nonlinear counterparts. These results indicate that a compact set of satellite-derived and meteorological variables can provide meaningful and interpretable information on CW treatment performance without in situ operational data, offering a complement to—rather than a replacement for—traditional physicochemical analyses in the monitoring of constructed wetlands. Full article

In situ nondestructive continuous monitoring of plant water status coupled with online data analysis is a core technical prerequisite for developing intelligent irrigation decision-making systems. The LPCP (leaf patch clamp pressure probe, ZIM-probe) sensor can detect leaf turgor pressure and shows promising application prospects. In this paper, the LPCP was used to monitor the leaf turgor pressure of maize during the silky growth stage cultivated in the North China Plain. The experiment was arranged in a randomized block design with two treatments: full irrigation (CK) and natural drought (ND). The results showed that the probe output pressure (P_p) underwent three stages: P_{p min} increasing stage, P_{p max} early-occurring stage, and P_p curve inversion stage, along with soil water decreasing, which correspond to mild, moderate, and severe water stress, respectively, and the ranges of soil and leaf water content of each stage were identified. As water stress intensified, the peak times of transpiration rate and stomatal conductance occurred earlier than normal, and the values decreased; meanwhile, the relationships between P_p, sap flow (SF), and leaf physiology indicators were quadratic parabolic, but the parabola opening directions, determination coefficients of regression equations, and model significance differed under different water stress stages. Under mild and moderate water stress, P_p positively correlated with SF, vapor pressure deficit (VPD), photosynthetically active radiation (PAR), and air temperature (T), while negatively correlated with relative humidity (RH), and path analysis results revealed that PAR and T exerted direct effects on P_p variations, whereas SF and VPD influenced P_p indirectly through other variables, and RH exhibited a negative effect on P_p changes. However, these correlations reversed under severe water stress. A regression model including PAR, T and VPD was established to simulate the P_p values of maize under full irrigation conditions, and by comparing the variation trend of P_p curves between the predicted and the measured values, whether maize was under water stress could be determined. Full article

Sudden water pollution events in Chaohu Lake pose serious public environmental risks, making reliable short-term prediction of water quality grades essential for effective water quality early warning systems and scientific lake management. Existing approaches inadequately characterize pollutant transmission between monitoring stations, fail to reasonably utilize meteorological information, and pay insufficient attention to scarce high-pollution-grade samples. Moreover, most studies merely evaluate overall model performance, lacking targeted applicability for early warning of pollution outbreaks in Chaohu Lake. To address these issues, this study proposes a hybrid model that integrates both water quality and meteorological data, leveraging the pollution transmission relationships between different monitoring stations, which can improve the prediction performance for Grade V and below-Grade V water quality. The proposed model employs a condition-aware fusion strategy to dynamically combine both water quality and meteorological data, constructs a spatial hydrological transport graph, encoding pollution transmission patterns between stations, and designs a hierarchical classifier to address the severe imbalance in water quality data. Four-hour water quality observations from seven monitoring stations between June 2022 and May 2025 were paired with previous-day meteorological data. Data from 1 June 2022 to 29 February 2024 were used for training, data from 1 March to 31 May 2024 were used for validation, and data from 1 June 2024 to 31 May 2025 were used for independent testing. Across five independent runs, the full model achieved an overall accuracy of 0.7256 ± 0.0022 and a polluted-class recall of 0.8044 ± 0.0069. Relative to the Water-GRU baseline, polluted-class F1 increased from 0.5814 to 0.6554 (12.7%), while sudden-pollution recall increased from 0.4310 to 0.4879 (13.2%). Multi-step evaluation further showed that polluted-class F1 remained 0.6001 at a 12 h forecasting horizon. These results indicate that condition-aware meteorological information and hydrologically informed spatial aggregation can improve warning-oriented water quality grade prediction in Chaohu Lake, particularly by reducing missed pollution events. Full article

The vegetation cover in the basins of small and medium-sized rivers of the southwestern Caspian region performs crucial ecosystem functions; however, arid and semi-arid conditions, dissected topography, and intensive anthropogenic pressure make these landscapes highly vulnerable. Conventional ground-based monitoring is hampered by limited accessibility and a sparse meteorological network, while comprehensive predictive studies using a basin-wide approach have not been previously conducted. This study assesses long-term dynamics (2001–2024) and forecasts vegetation-cover changes in nine river basins (Sunzha, Sulak, Shuraozen, Ulluchay, Samur, Karachay, Atachay, Kheraz, and Gorgan) using NDVI, EVI, and NDMI derived from MODIS data (Google Earth Engine, MOD09A1 and MOD13Q1). Statistical characteristics, Theil–Sen trends with Mann–Kendall testing, and Hurst exponent-based persistence forecasts were calculated for each basin. Spatial distribution of all indices was highly heterogeneous. Statistically significant positive NDVI and EVI trends (indicating increasing green phytomass) and a significant negative NDMI trend (declining moisture availability) were found across the entire territory. The lowest multi-year mean NDVI was recorded for the Kheraz basin (0.18), and the highest for the Ulluchay basin (0.44). The positive NDVI trend was significant (p < 0.05) for all basins except Shuraozen, Karachay, and Atachay. Predictive mapping indicates continuous degradation in the estuarine and most intensively developed parts of the Sunzha, Sulak, and Shuraozen basins, particularly reflected by NDMI moisture deficit, whereas sustained improvement is projected for the lower Samur basin. For most of the territory, the change tendency remains uncertain. Full article

Atmospheric water vapor plays a critical role in the climate system, governing energy balance, climate variability and precipitation processes. Accurate monitoring of precipitable water vapor (PWV) is therefore essential for both meteorological and climate-related studies. The Global Navigation Satellite System (GNSS) provides an effective and continuous tool for PWV estimation through the calculation of zenith tropospheric wet delay (ZWD). Given the scarcity of continuous ground-based observations in Antarctica, the TUR1 and TUR2 permanent GNSS stations, established on Horseshoe Island during the Turkish Antarctic Expedition-4 (TAE-4) under the TÜBİTAK Polar Research Project (No. 118Y322), provide a valuable infrastructure for continuous atmospheric water vapor monitoring, which constitutes the primary contribution of this study. The conversion of ZWD to PWV requires an accurate estimation of the weighted mean temperature (Tm), which is typically derived from empirical models. However, globally applied Tm models may not adequately represent the regional atmospheric variability of high-latitude environments, where the vertical atmospheric structure and water vapor distribution differ substantially from mid-latitude conditions. This limitation is especially pronounced in Antarctica, which serves as a natural laboratory for climate change research while remaining one of the most observationally constrained regions on Earth due to sparse meteorological infrastructure and logistical challenges. In this study, a locally derived Tm parameterization (HRS) was obtained from radiosonde profiles at near sea-level stations within the 66°S–70°S latitude belt and evaluated against both the existing regional Antarctic Tm model (ANT) and the globally applied GPT2, GPT3 and Bevis models, using radiosonde observations as an independent reference. The HRS parameterization reproduced the radiosonde-derived Tm values with an RMSE of 2.90 K, clearly outperforming the global models (4.20–4.51 K) and showing close agreement with the existing regional Antarctic Tm model (r = 0.987), independently confirming the transferability of the regional approach to the Horseshoe Island region. The GNSS-derived PWV based on the regional Tm parameterization was then evaluated against ERA5 reanalysis data as an independent reference. The estimates showed strong agreement, with a correlation of 0.95, an RMSE of 1.65 mm and a mean bias of +1.22 mm, the seasonal agreement being strongest in austral summer (RMSE ≈ 1.24 mm) and weakest in austral winter (RMSE ≈ 1.92 mm), consistent with the lower water vapor content and stronger surface temperature inversions that characterize the cold season. The close agreement between the two independent TUR1 and TUR2 stations further supports the repeatability of GNSS-based PWV retrieval under coastal Antarctic conditions, highlighting the observational value of these stations for atmospheric water vapor monitoring in Antarctica, where continuous observations remain limited. Full article

To address the lack of quantitative cross-industry comparisons in safety risk management and control in power engineering construction, this study develops a cross-industry safety feature-analysis framework that integrates the DeepSeek large language model (LLM) with Python-based automated text mining. The framework is applied to 36 real-world safety cases from 18 enterprises across four major power engineering sectors (hydropower, thermal power, renewable energy, and power transmission and transformation). It uses Python to automatically traverse directories and extract text from paragraphs and tables in documents. A domain-specific lexicon is constructed using prompt engineering with the DeepSeek LLM to generate candidate professional terms. These candidate terms are subsequently verified against the original texts and manually screened before being incorporated into the final lexicon. The validated lexicon is then dynamically loaded into the Jieba tokenizer to improve the recognition of long specialized terms, thereby generating matrix-based word frequency statistics. Both raw and case-normalized word frequencies (CNWF) are calculated to support cross-sector comparison. The results reveal a common characteristic across the examined cases: “dual prevention” institutional support combined with technology-enabled safety management, including grid-based management and video surveillance. However, distinct emphases are observed across subsectors: hydropower cases show greater emphasis on microseismic monitoring and digital twins to mitigate geological risks; thermal power prioritizes institutional checklists and pre-shift safety briefings; renewable energy cases focus on meteorological warnings and remote video surveillance; and the transmission and transformation sector emphasize high-risk construction procedures and comprehensive supervision documentation. These findings indicate that safety management characteristics vary across engineering scenarios and highlight the value of differentiated and scenario-specific safety strategies. Given the relatively small and unevenly distributed sample, the findings should be interpreted as exploratory rather than as statistically representative of the entire power engineering construction industry. Full article

Satellite-derived atmospheric CO₂ observations provide important information for regional carbon monitoring, but their sparse spatial and temporal coverage limits their use in data-sparse drylands. We reconstructed seamless monthly column-averaged atmospheric CO₂ (XCO₂) over Mongolia at 0.1° resolution from 2015 to 2023 by integrating OCO-2 retrievals with vegetation, meteorological, and socioeconomic predictors using a LightGBM–Optuna method. The model achieved an R² of 0.952, RMSE of 1.510 ppm, and MAE of 1.060 ppm. Detrended seasonal XCO₂ was then related to SPEI-3-derived mean drought intensity (MDI) and drought frequency (DF) across typical steppe, forest, and Gobi Desert regions. The results showed that summer was the main season of XCO₂ sensitivity to moisture limitation, but regional responses differed substantially. The typical steppe showed the strongest summer sensitivity; forest showed weak summer MDI sensitivity but greater relevance of repeated moisture-limited months; and spring XCO₂ variability in the Gobi Desert was more closely associated with MDI. These findings indicate that seasonal moisture limitation is associated with contrasting atmospheric CO₂ variability across Mongolia’s typical steppe, forest, and Gobi Desert regions, providing a regional atmospheric perspective on carbon–water interactions in arid and semi-arid drylands. Full article

High-resolution urban Weather Research and Forecasting (WRF) simulations resolve interactions among urban form, near-surface meteorology, atmospheric dynamics, and building-energy processes, but their computational cost limits repeated scenario analysis. This study compares eight pointwise temporal surrogates (Persistence, previous-day same-hour, Ridge, Random Forest, XGBoost, multilayer perceptron (MLP), gated recurrent unit (GRU), and long short-term memory (LSTM)) for recursive 24 h prediction of T₂, Q₂, W₁₀, and WRF-BEM diagnostic air-conditioning power density (AC) from a 120 h WRF-BEP/BEM-LCZ heatwave simulation over Chongqing. Model development used three expanding-window folds, fold-specific scaling, fixed monitoring cells, and five fixed seeds for stochastic families; the h97–h120 full-domain test remained untouched until all modeling choices had been finalized. Station comparisons provided a limited check on the physical plausibility of the WRF meteorological fields; AC was not validated against metered energy use. Model performance depended strongly on the target variable. Random Forest performed best for T₂ (1.081 ± 0.002 °C), LSTM for Q₂ (0.969 ± 0.063 g/kg), and MLP for AC (0.505 ± 0.187 W/m²), whereas deterministic Ridge was best for W₁₀ (1.095 m/s). W₁₀ had weaker temporal memory (median lag-1/lag-24 ACF 0.581/0.211). Ridge outperformed the Previous-day baseline for 75.0% of grid cells, although errors increased sharply in the top wind-speed decile and during rapid changes. For this single-event, single-city benchmark, recurrent models did not improve W10 accuracy over Ridge. Model choice depended on the target and the balance between accuracy, spatial fidelity, and computational cost; transfer to other events and meteorological regimes remains untested. Full article

Level-3 (L3) Earth observation (EO) products are commonly generated through temporal compositing, smoothing, and gap-filling of Level-2 retrievals. While these procedures provide spatially and temporally continuous datasets, they rely on interpolation assumptions that cannot adequately capture high-frequency, non-linear physiological dynamics of vegetation. As a result, reconstructed time series often exhibit attenuated variability, temporal lag, and aliasing, with limited consistency with underlying ecosystem processes. This review proposes a shift from conventional gap-filling toward dynamic reconstruction, in which L3 products are interpreted as inferred trajectories of observable vegetation variables conditioned on sparse observations, complementary information, and explicit constraints. Dynamic reconstruction integrates satellite observations with complementary information, including multi-sensor EO data, meteorological drivers, spatial context, and model-based priors. The review critically synthesizes the principal methodological families for dynamic reconstruction and examines how complementary information and explicit constraints improve the reconstruction of vegetation dynamics while introducing trade-offs in uncertainty representation, physical consistency, scalability, and scale compatibility. Examples spanning physiological, environmental, and event-driven regimes illustrate how conventional L3 processing is particularly challenged by highly dynamic variables such as solar-induced chlorophyll fluorescence, evapotranspiration, land surface temperature, and stress indicators. These variables remain undersampled by satellite observations, leading to loss of short-term variability and distorted timing of rapid responses. Dynamic reconstruction provides a unifying framework for improving the representation of vegetation dynamics by explicitly accounting for observability, uncertainty, and cross-scale consistency. Overall, for highly dynamic vegetation variables, L3 products are increasingly understood as reconstructed trajectories rather than merely as gap-filled observations. This perspective may enhance the reliability and interpretability of EO-based monitoring of vegetation function and stress in an era of increasingly frequent, multi-sensor observations. Full article

Background: Asthma is a chronic respiratory disease shaped by environmental, meteorological, and behavioral factors. Few approaches combine population-level surveillance with individual-level monitoring within a single analytical framework. Methods: This work developed a dual machine learning framework. Study 1 modeled the daily count of respiratory admissions (chapter X of the ICD-10) in São Paulo, Brazil, from 2017 to 2022 as a nowcasting task, combining ElasticNet, residual CatBoost, direct CatBoost, and adaptive blending, validated by walk-forward over 30 bimonthly windows. Study 2 applied an XGBoost and Random Forest pipeline to 913 diaphragmatic-breathing sessions from 17 patients aged 9 to 16 years in the Respire Bem system, with Asthma Control Test and salivary cortisol represented using evidence-based synthetic simulation. Results: Study 1 achieved a mean MAE of 18.22, RMSE of 23.99, and R² of 0.675, exceeding the seasonal baseline by 41.5%, with a significant advantage over all three baselines (Wilcoxon and Diebold–Mariano, p ≤ 0.038). Ablation showed each single-component configuration to be significantly worse than the full hybrid, but removing the environmental block cost only 0.27 admissions per day, an effect indistinguishable from zero. SHAP rankings were stable across windows (Kendall W = 0.640), led by NO₂, PM2.5, and temperature. In Study 2 the pipeline ran end-to-end on real behavioral data, but because the outcomes were simulated, no predictive-accuracy metric is reported. Conclusions: Study 1 delivers a validated population-level nowcasting model whose accuracy rests mainly on the temporal structure of the series. Study 2 contributes a real behavioral dataset and a reproducible pipeline; the clinical validity of the digital biomarkers remains open and requires prospective work with directly measured outcomes. Full article

Hydrological drought, expressed as anomalously low water availability in rivers, aquifers, and reservoirs, poses a growing threat to water security in densely populated regions such as the Metropolitan Region of Sao Paulo (MRSP), Brazil. Characterizing how drought propagates into large-scale terrestrial water storage (TWS) deficits, and how these deficits translate into measurable surface deformation, remains challenging in complex, human-managed hydrological systems such as the Cantareira Water Supply System. This study conducts an integrated, multi-sensor geodetic assessment of drought-related hydrological dynamics in the Cantareira System, combining vertical/LOS ground displacement derived from continuous GPS observations and Sentinel-1 InSAR time series with terrestrial water storage anomalies (TWSAs) from GRACE/GRACE-FO, groundwater level records, reservoir storage, and meteorological drought indicators. Cross-correlation analysis reveals a strong and statistically significant coupling between GRACE-TWSA and GPS-derived vertical displacement, with correlation coefficients of $r=-0.8$ (Upper Tietê basin) and $r=-0.6$ (PCJ basin), consistent with an elastic crustal response to hydrological loading. Reservoir storage in the PCJ basin is similarly correlated with regional TWSA (up to $r=0.7$ for the Jaguari–Jacareí reservoir), reinforcing GRACE’s sensitivity to the main upstream storage component of the Cantareira System. Empirical Orthogonal Function (EOF) decomposition of the InSAR deformation fields, retaining the first four modes (72% of variance in Upper Tietê, 66% in PCJ), further demonstrates that deformation patterns are strongly influenced by hydrogeological controls, with distinct spatial responses between the PCJ basin, where hydroclimatic forcing is more pronounced, and the more urbanized and anthropogenically influenced Upper Tietê basin. The integrated dataset captures major drought episodes between January 2020 and Dicember 2025, demonstrating the capability of combining GPS, InSAR, and GRACE observations to quantitatively link climatic drought forcing, terrestrial water storage deficits, and surface deformation, thereby providing a robust approach for monitoring water storage changes and supporting water resource management in densely populated regions. Full article

Surface deformation triggered by coal extraction activities, together with the consequent development of unstable slopes within rugged mountainous landscapes, constitutes a critical focus for geological risk assessment and mitigation strategies. Conventional SBAS-InSAR processing pipelines suffer from inadequate tropospheric phase mitigation in topographically complex environments, while existing clustering-based recognition approaches fail to incorporate sufficient geophysical constraints. To overcome these deficiencies, the present investigation introduces a refined methodology that synergizes InSAR measurements with an enhanced clustering scheme for the automated screening of potentially unstable slope units. First, a two-stage coupled atmospheric correction framework is constructed within the SBAS-InSAR processing chain, comprising spatially varying stratified atmosphere estimation based on geographically weighted robust regression (GWRR-M) and turbulent atmosphere compensation based on structure-guided deformation-preserving interpolation (SGDPI); both stages require no external meteorological data and effectively protect deformation signals from overcorrection. Second, a spatiotemporally constrained density peak clustering algorithm (STC-DPC) is developed, which constructs a multi-dimensional feature space integrating spatial location, deformation rate, temporal evolution characteristics, and topographic-geological background, and introduces a spatiotemporally constrained distance metric together with an Unstable Slope Index (USI) to achieve automatic identification and quantitative discrimination of unstable slopes. The proposed method was evaluated using 120 ascending-track Sentinel-1A SAR images acquired from 2019 to 2023 over the coal-mining mountainous areas of Mentougou and Fangshan districts in western Beijing, China. The results show that the improved atmospheric correction reduces the phase standard deviation of a representative interferogram from 1.6 rad to 0.6 rad, with an average reduction of 42.3% across all interferograms. A total of 187 unstable slopes were identified by the STC-DPC algorithm, mainly distributed in abandoned mining areas and steep terrain with gradients of 10–35°, with a mean deformation rate of −25.3 mm/a; field investigations at representative sites confirmed significant deformation evidence (e.g., tension cracks and bulging), providing qualitative support for the identification results. Compared with the identification results obtained without atmospheric correction (79 unstable slopes), the improved method improves the detectability of weak deformation signals in areas with strong topographic relief and diverse deformation patterns. This study provides a practical technical pathway for the early screening and monitoring of geological hazards in coal-mining mountainous areas and holds great significance for mine ecological restoration and regional disaster prevention and mitigation. Full article

Combining remote sensing data with field inventory data is a common practice in estimation of forest Aboveground Biomass (AGB). However, in areas with well-established ground monitoring systems, the actual added value of this combination has not been fully quantified. This paper aims to critically assess the marginal contribution of optical remote sensing data to AGB estimation when the forest inventory data are already available, and further investigates error sources and residual distributions. Using Longyan City, Fujian Province as a case study, we systematically tested the effects of data types (inventory only, Landsat 8 only, inventory + Landsat 8, and inventory + Landsat 8 + meteorological data), sampling methods, and statistical models on plot-scale AGB estimation accuracy, along with uncertainty distribution across biomass levels. Inventory data alone (stand age and canopy cover) achieved acceptable accuracy, with ${\mathit{R}}^2$ = 0.60 and RMSE = 35.78 t/ha under the optimal configuration (RandomForest + ShuffleSplit_5). Adding Landsat 8 data yielded only modest improvements: RMSE decreased by 5.3% and ${\mathit{R}}^2$ increased by 6.7% relative to the inventory-only baseline. When the optimal combination for each data type was evaluated on the independent test set, the highest ${\mathit{R}}^2$ reached only 0.56, leaving approximately 44% of the observed variation in plot-level AGB unexplained. ANOVA showed that data type was the dominant factor, accounting for $73.0\%$ of the variation in RMSE, followed by model choice ($15.1\%$) and their interaction ($8.5\%$), whereas the contribution of the validation scheme was less than $3.0\%$. We further examined residual distributions across biomass levels and found that residual skewness shifted systematically: negative skewness (overestimation) prevailed in low-biomass plots, near-zero skewness in medium-biomass plots, and positive skewness (underestimation) in high-biomass plots. Among all data types, the Landsat-only configuration exhibited the most severe bias at both low and high biomass extremes; adding remote sensing and climate data to inventory data did not substantially correct this bias structure. Overall, these findings demonstrate that in regions with high-quality ground inventory, the marginal gains from integrating optical remote sensing are limited—both in terms of overall accuracy and the structure of prediction errors. The unexplained variance and systematic residual biases across biomass gradients highlight the need for better representation of high-biomass stands and suggest that future efforts should prioritize sample augmentation in under-represented biomass classes, rather than relying solely on multisource data fusion for accuracy improvement. Full article