Search Results (2,318)

Atmospheric wet deposition represents a major pathway for the transfer of organic micropollutants into terrestrial and aquatic ecosystems. This study investigates the occurrence and spatial distribution of polycyclic aromatic hydrocarbons (PAHs), phthalate esters (PAEs), and BTEX compounds in rainwater across Northern Serbia (Vojvodina region). Rainwater samples were collected during the 2025–2026 heating season at three locations: a petrochemical site in Kikinda, a traffic- and residentially influenced site in Sremska Mitrovica, and an urban background site in Sombor. Total concentrations showed pronounced spatial variability, with the highest ΣBTEX and ΣPAE levels recorded in Kikinda (∑BTEX = 2.818 μg L^∑1; ∑PAE = 0.930 μg L^∑1). Diagnostic ratios identified a dominant petrogenic signature in Kikinda (LMW/HMW > 1), while pyrogenic sources prevailed in Sremska Mitrovica and Sombor ((Fla/Fla + Pyr) > 0.5). BTEX profiles across all sites were characterised by the absence of benzene and elevated toluene and xylene levels (B/T ≈ 0; T/X > 1). Health risk assessment indicated an acceptable but non-negligible carcinogenic risk from PAHs, particularly for children in industrial areas. These findings highlight the role of precipitation as an efficient scavenger of organic pollutants and emphasise the need for integrated atmospheric–hydrological monitoring frameworks in industrialised regions. Full article

Tianjin, nestled on the North China Plain, possesses abundant geothermal resources with tremendous potential for development and utilization. This study employs hydrogeochemical and isotopic analysis techniques to thoroughly explore the geochemical characteristics and circulation patterns of geothermal fluids in Tianjin, shedding light on the mechanisms underlying the formation and evolution of deep geothermal fluids. The findings reveal that atmospheric precipitation serves as the primary recharge source for the region’s geothermal fluids, with the calculated recharge heights coinciding with those of the Jixian mountainous area. This precipitation infiltrates through permeable layers and the deep, large faults surrounding the southern plain, entering relatively enclosed or semi-enclosed geothermal reservoirs. As they circulate, the geothermal fluids undergo intricate interactions with the surrounding rocks, including processes such as leaching, adsorption, carbonate reprecipitation, cation exchange, and decarbonation. The fluids circulate at depths ranging from 1.6 to 3.5 km, with temperatures spanning from 67 to 133 °C. Along the flow path, the anionic composition of the geothermal fluids shifts from HCO₃⁻ dominance in the north to a coexistence of Cl⁻ and SO₄²⁻, ultimately dominated by Cl⁻ in the south, accompanied by an increase in total dissolved solids (TDS). The results indicate that Tianjin geothermal fluids are mainly recharged by meteoric water and evolve along their flow paths through dissolution of evaporitic and carbonate minerals, cation exchange, and carbonate precipitation. Hydrochemical and Sr-isotope differences suggest generally limited vertical connectivity among the studied reservoirs, although local hydraulic interaction may occur near conductive faults. These results provide constraints on the hydrogeochemical evolution and management of geothermal resources in the Tianjin sedimentary basin. Full article

Machine learning approaches to wildfire spread prediction are constrained by the lack of standardized, multi-source, spatiotemporal datasets that fuse terrain, weather, and fire-state information into a single ML-ready format. We present WildfireCube, a reproducible event-centric pipeline and methodology for constructing dense fourth-order spatiotemporal tensors of shape (T, C, H, W) at 30 m spatial and 3 h temporal resolution. Following the analysis-ready data convention established in the Earth Observation community, the pipeline fuses four open data sources: the Copernicus GLO-30 Digital Elevation Model for static terrain derivatives, ERA5-Land reanalysis for hourly weather forcing, Sentinel-2 Level-2A imagery for spectral vegetation and burn-severity indices, and NASA FIRMS active-fire hotspot detections for fire-state reconstruction via ordinary kriging. The resulting 13-channel normalized tensor separates causal drivers into three physically motivated groups: static landscape controls (elevation, slope, aspect, fuel load), dynamic atmospheric forcings (wind components, temperature, precipitation), and evolving fire state (fire-front mask, burn severity, fractional burn, observation confidence). A physics-informed normalization framework maps all channels to bounded ranges using fixed physical constants rather than sample statistics, ensuring cross-event comparability and exact invertibility. We demonstrate the pipeline on 13 wildfire events across the United States, Canada, and Greece (2017–2023), producing a processed catalog exceeding 300 GB compressed and spanning a 14-fold range in burned area, a 27 °C range in mean temperature, and different fire regimes. Event tensors are stored in chunked Zarr archives with Zstandard compression, achieving a 2.58× compression ratio. As future work, the pipeline will be applied to a 40-event target catalog projected to exceed 2 TB of raw data, providing the multi-regime diversity and scale required for training robust deep learning models for spatiotemporal wildfire prediction. Full article

The present study investigates the statistical distribution of radar reflectivity slopes [S-Ze] in the lower troposphere along the west coast of India using a C-band radar during the pre-monsoon and monsoon seasons in 2024. The study period spans a range of meteorological conditions, from a drier atmosphere during pre-monsoon months to a moist atmosphere during the monsoon months, with varying updraughts and downdraughts. To investigate the S-Ze, we calculated the difference in Ze between 4 km and 2 km altitudes in the lower troposphere. The S-Ze could be either positive or negative, where, in a positive [negative] S-Ze, the Ze decreases [increases] towards the surface. The monthly variations in S-Ze from the pre-monsoon to monsoon months are observed in the lower troposphere and are higher in monsoon months compared to pre-monsoon months, which are too near the coast. The land–ocean contrasts of the vertical profiles contributing to +ve and −ve S-Ze are lower compared to north–south gradients and higher in monsoon months. The average S-Ze shows the highest +ve and −ve S-Ze magnitude near the coast among all the months. The highest magnitude in S-Ze is observed in March and April and is associated with the lower and higher numbers of vertical Ze profiles. The increase or decrease in hydrometeor size is less during the monsoon months (June, July, August, and September) compared to pre-monsoon months, where the March–April months have the highest increase or decrease in the hydrometeor’s size in the lower troposphere. The variations in the S-Ze are the combined effect of the atmospheric, thermodynamic (relative humidity (RH) and moisture flux), and dynamic conditions (zonal, meridional, and vertical velocity). Strong updraughts that carry RH to higher altitudes make the lower atmosphere drier and contribute to a +ve S-Ze; Ze tends to decrease in the lower troposphere. However, a weaker updraught or a moderate downdraught with sufficient RH provides sufficient time for hydrometeors to grow and contributes to −ve S-Ze, and Ze tends to increase in the lower troposphere. For example, in March and April, the atmosphere is dry, and we observe the largest decrease in hydrometeors near the coastal boundary. However, we also see significantly higher negative radar reflectivity slopes, and weak downdraughts provide enough time for hydrometeors to grow. In June and July, there are strong updraughts (downdraughts) with high (low) RH, making the atmosphere more conducive to a decreasing tendency in Ze and contributing to a higher fraction of +ve S-Ze. The results presented here would be an extension of the study from the satellite-based observations, revealing the extension of climatology for the inclusion of stratiform precipitation. Full article

Precipitation is among the most uncertain and poorly predicted weather products in earth system science. Local convective precipitation is particularly sensitive to strong land–atmosphere coupling. Two indices derived from atmospheric thermodynamic vertical profiles, convective triggering potential (CTP), a measure of the temperature lapse rate between approximately 1 and 3 km above the ground surface, and low-level humidity (HI_low), have become preferred measures of land–atmospheric coupling strength. To complement previous studies that primarily relied on limited station observations or regional analyses, this study provides a 20-year assessment of the CTP-HI_low framework for a wide area of the Continental United States (CONUS) using integrated satellite observations, reanalysis products, and surface datasets. The study further identifies important regional limitations in the framework’s predictive skill and demonstrates the influence of mid-level vertical wind shear on precipitation occurrence during both wet and dry soil advantage conditions. These findings provide new insight into why the framework performs inconsistently across different climate regions and suggest pathways for improving land–atmosphere coupling-based precipitation prediction. The objective is to determine the atmospheric and land-surface factors that control the regional performance of the CTP-HI_low framework and to identify how additional datasets that include more atmospheric variables can improve precipitation prediction skill. Full article

Extreme weather events exacerbated by global warming pose severe threats to urban safety, underscoring the urgent need for highly accurate precipitation nowcasting. Short-term local heavy precipitation remains a particular challenge for traditional forecasting due to its suddenness and high disaster potential. To address this, we propose a multi-modal fusion framework that integrates ground-based GNSS-derived Precipitable Water Vapor (GNSS-PWV) and ground-based Radar Composite Reflectivity (CR). While GNSS-PWV keenly captures pre-convective atmospheric water vapor accumulation, radar CR details the morphological distribution of hydrometeors. Specifically, we developed the Spatio-Temporal Enhanced Attention Swin U-Net (STEA-Swin) model to synergize these heterogeneous datasets over the Beijing–Tianjin–Hebei region. High-precision PWV was retrieved from 250 Continuously Operating Reference Stations (CORS) using the dual-frequency ionosphere-free Precise Point Positioning (PPP) method, achieving a strong correlation (>0.97) with ERA5 reanalysis data. Validated against measured data from the 2025 flood season, the STEA-Swin model achieved a Probability of Detection (POD) of 0.68 for torrential rain events at a +1 h forecast lead time. Notably, compared to single-source models, the Critical Success Index (CSI) and POD for torrential rain improved by 18.5% and 21.5%, respectively. These findings demonstrate that coupling deep learning with ground-based GNSS-derived atmospheric thermodynamic information can significantly enhance early warning capabilities, providing a promising technical approach for regional disaster prevention and climate resilience. Full article

Assessing the contribution of tropical cyclones to regional precipitation variability is essential for understanding the associated hydrometeorological benefits and risks. This study quantifies the contribution of tropical cyclones to annual precipitation in the northernmost part of South America from 2016 to 2025, utilizing data from surface rain gauges. Simulations using the Weather Research and Forecasting (WRF) model, configured with 2 km grid spacing and 38 vertical levels, estimate the influence of relative humidity at 850 hPa and ambient temperature at 500 hPa on precipitation over the continental region when each convective system is nearest to the coastline. During Hurricanes Matthew (2016) and Melissa (2025), contributions to the annual average precipitation reached 51% and 47%, respectively, with the highest values observed near the northern South American coastline. The contributions of Harvey (2017), Iota (2020), Julia (2022), and Beryl (2024) to annual precipitation were 0–26%, 0–18%, 0–12%, and 0–19%, respectively. Precipitation distribution was heterogeneous during the passage of tropical storms. The extent of accumulated precipitation was influenced by the cyclone’s trajectory and proximity to mountainous regions. Patterns of relative humidity at 850 hPa did not correspond to a uniform precipitation distribution. Between 6% and 30% of rain gauges did not record precipitation during the analyzed tropical cyclone events. These findings highlight that tropical cyclone-induced precipitation is strongly influenced by complex interactions between atmospheric dynamics and topography. Future research should assess the contributions of these systems to groundwater and surface reservoirs that support indigenous communities in rural areas. Full article

Drought is a slowly evolving, multi-driver hydro-meteorological hazard whose accurate early prediction is a cornerstone of climate-smart agriculture and water-resource planning. Existing data-driven drought forecasting frameworks suffer from three persistent limitations: (i) most models concatenate heterogeneous climate variables into a single flat feature vector, implicitly assuming a single dominant driver such as precipitation, even though atmospheric moisture demand, radiation and wind-mediated evapotranspiration co-determine drought onset; (ii) wavelet preprocessing is typically applied to the full series, introducing future-information leakage that violates the operational causality requirement of forecasting; and (iii) most architectures predict a single horizon and provide no causal attribution explaining when, where and which climatic variables initiated the event. This study proposes WaveDroughtNet, a multi-modal, multi-horizon deep-learning framework that addresses these limitations through five integrated components: (a) a strictly causal Daubechies-4 wavelet decomposition computed in a rolling fashion; (b) six modality-specific encoders with stochastic modality dropout (p = 0.15); (c) cross-modal multi-head attention with four heads; (d) a four-layer temporal convolutional network (TCN) backbone with dilation factors yielding a 240-step receptive field; and (e) a post hoc DroughtOriginTracer that combines temporal attention, modal-attribution and inter-district propagation scans. The Standardised Precipitation Evapotranspiration Index (SPEI), used as the supervisory target, is computed following the canonical Vicente-Serrano formulation. water balance $D=P-\mathrm{P}\mathrm{E}\mathrm{T}$ (Hargreaves PET) at a 4-week (≈1-month) timescale, fitted with a three-parameter log-logistic distribution via L-moments, validated by Kolmogorov–Smirnov goodness-of-fit testing ($\alpha =0.05$) per district, and standardised through the inverse-normal cumulative distribution function. Trained on 18,304 weekly district records from NASA POWER reanalysis (2014–2025) covering all 32 districts of Tamil Nadu, India, WaveDroughtNet uses only 256,869 parameters and produces, in a single forward pass, four forecasts (1 week, 1 month, 3 months, 1 year). On the held-out 2024 test partition ($N=1728$), the model attains weighted $F1=0.9221$ and $R^2=0.8512$ at the 1-week horizon, and weighted $F1=0.8498$ and $R^2=0.6812$ at the 1-year horizon. Diebold–Mariano tests confirm that WaveDroughtNet significantly outperforms naive persistence, seasonal naive, LSTM, ConvLSTM and a vanilla Transformer at the 3-month and 1-year horizons (p < 0.001). The DroughtOriginTracer successfully back-projects 15 Coimbatore events to causal origins 29–41 weeks prior to onset. We explicitly acknowledge three limitations that constrain operational deployment in its current form—zero severe events in the 2024 test partition ($F{1}_{severe}$ = 0.000), static inter-district modelling, and absence of vegetation-index supervision—and propose concrete mitigation pathways in the Discussion. Full article

This study presents a dataset of meteorological and soil-hydrological instrumental observations collected at three agrometeorological stations in the East Kazakhstan Region during the growing seasons of 2022–2025. The dataset includes time series from automatic weather stations: WS “OCES-1” (Solnechnoe village) provides hourly data over four years (2022–2025; 14,614 records; 65 variables), while WS “OCES-2” (Lugovoe village; 203,279 records) and WS “Altyn Kazan” (Sulusary village; 207,115 records) provide minute-resolution data for 2025 (49 variables each). Measured parameters at 200 cm height include air temperature and humidity, atmospheric pressure, precipitation, wind speed and direction; soil measurements down to 100 cm depth include temperature and moisture. Also, field-based express measurements of volumetric soil moisture within a 1 m profile (every 10 cm) were collected during three campaigns (May–August 2025), resulting in a total of 253 measurements. The stations are located across steppe and forest-steppe landscapes of the transboundary Altai–Sayan mountain region on active agricultural lands under diverse soil–climatic conditions. Climate types correspond to Dfb and Dfa per the Köppen–Geiger classification. Soils are classified under WRB as Chernozems and Calcic Chernozems. The dataset is published in CSV format on Zenodo under a CC-BY 4.0 license. Full article

Soil surface moisture (SSM) is a critical indicator of agricultural drought, yet high-resolution projections under climate change remain scarce. This study develops a machine learning framework to predict and project SSM at 1 km resolution across five European Living Labs (LLs), encompassing vineyards, olive groves, and fruit tree systems. Historical Sentinel-1 SSM observations (2014–2024) were used to train ensemble models (Random Forest, XGBoost, ExtraTrees, LightGBM) incorporating climate variables, soil texture, topography, and land use. Tree-based models achieved R² values of 0.63–0.87. Vineyards showed the highest predictability (R² ≈ 0.87), reflecting their sensitivity to short-term atmospheric demand and surface water availability, whereas olive groves were the least predictable (R² ≈ 0.63–0.68), consistent with deeper rooting systems and greater drought buffering capacity. When forced with bias-corrected CMIP6 projections under SSP1-2.6 and SSP5-8.5 for 2041–2070, models indicate minimal changes under SSP1-2.6 but pronounced SSM declines of 8–24% under SSP5-8.5, with historically wetter regions experiencing the largest absolute losses. SHAP analysis confirmed precipitation and potential evapotranspiration as dominant predictors across all crops. This framework provides spatially explicit, crop-relevant SSM projections to support climate adaptation in European agricultural landscapes. Full article

Rice cultivation commonly employs the continuous flooding (CF) method, which depends heavily on water availability creating anaerobic conditions for methane (CH₄) emissions. Rainfed rice areas rely on precipitation for irrigation, making the system sensitive to climatic variability. This study examines associations between ENSO phases and satellite-observed atmospheric XCH₄ variability over Thailand using GOSAT as the primary long-term dataset from 2012 to 2022, with Sentinel-5P/TROPOMI used as a supporting dataset for recent spatial patterns. The analysis conducted covers three cropping seasons: (1) January–April, (2) May–August, and (3) September–December. The results indicate comparable average atmospheric methane concentrations of 1787.94 ± 11.50 XCH₄ (ppb) during El Niño, 1788.8 ± 11.22 XCH₄ (ppb) in neutral conditions, and 1793.45 ± 10.93 XCH₄ (ppb) during La Niña. The obtained data indicate a seasonal variability, with the highest satellite-observed XCH₄ values found during September–December, corresponding to the main growing period of wet-season rice. The results suggest that climate change amplifies these anomalies through altered precipitation patterns and water availability. Current rice cultivation practices warrant reconsideration, in particular the alternate wetting and drying (AWD) method, offering reduced CH₄ emissions while conserving water resources. This underscores the importance of water management strategies for sustainable rice production and resilience to climate variability. Full article

The accurate characterization of global precipitation regimes, encompassing not only the mean quantities but also the seasonal structure, concentration, and spatial heterogeneity, is essential for understanding the hydroclimatological dynamics and supporting climate-sensitive applications. This study presents a multi-scale precipitation climatology based on the IMERG Final Run V06B dataset (2001–2021) integrating satellite-derived monthly precipitation fields, unsupervised K-means clustering, Walsh–Lawler concentration metrics, and pixel-scale regime-dynamics indicators. The analysis identifies eight physically interpretable global precipitation regimes and six Italian sub-regional regimes characterized by distinct seasonal structures and precipitation persistence patterns. The resulting classifications exhibit a strong consistency with major atmospheric circulation domains, including monsoonal, mediterranean, continental, and equatorial precipitation regimes. A Hovmöller diagram highlights the seasonal northward migration of the Intertropical Convergence Zone (ITCZ) from approximately 5° S in January to 10° N in August. The K-means classification identifies eight physically interpretable global regimes, including a perhumid equatorial regime, a South-Asian monsoonal regime, a Southern-Hemisphere Mediterranean type, and a transitional autumn-peaked Mediterranean–Atlantic regime covering most of Italy and the broader Mediterranean basin. At the Italian scale, a dedicated K = 6 clustering reveals six distinct precipitation regimes, characterized by contrasting seasonal structures: the Alpine Convective regime, unique to the Alps and pre-Alpine foothills; the Po Valley Padano regime, the least seasonal regime in Italy; the Apennine Hybrid; the Tyrrhenian Mediterranean; the Adriatic Transition; and the Semi-arid Mediterranean regime, dominant across Sicily, Sardinia, and coastal southern Italy. The Walsh–Lawler Concentration Index increases markedly from north to south (~0.58), indicating a pronounced intensification of the temporal concentration of precipitation toward the Mediterranean climatic extreme. Overall, the study demonstrates the capability of high-resolution satellite climatologies to identify dynamically coherent precipitation-regime structures across multiple spatial scales and provides a quantitative baseline for future applications in hydrology, climate-risk assessment, and climate-change impact analysis. Full article

To elucidate the solute sources, migration and enrichment mechanisms of water bodies in the endorheic lake region of the Qiangtang Plateau on the Tibetan Plateau and clarify the hydrogeochemical cycling patterns in alpine arid environments, this study focuses on two core scientific objectives: quantitative identification of the multi-source contributions of aquatic solutes, and revelation of the key processes governing the enrichment of strategic elements including lithium (Li) and boron (B). To achieve these goals, we conducted systematic hydrogeological field investigations and collected 28 multi-type water samples, covering springs, rivers, thermal springs, freshwater lakes, salt lake brines, atmospheric precipitation, and glacial meltwater. The physicochemical properties, major ions, and trace elements of all samples were comprehensively analyzed. On this basis, the hydrogeochemical characteristics, evolutionary processes, and solute origins of regional waters were systematically explored. Combined with PHREEQC numerical simulation, principal component analysis (PCA), and Pearson correlation analysis, the dominant controlling factors of water geochemistry were quantified, and a conceptual hydrogeochemical evolution model was established. The results reveal a clear hydrogeochemical evolutionary gradient across the study area: water bodies evolve from low-salinity HCO₃-Ca recharge end-members and transitional HCO₃·SO₄-Ca(Mg) type water to highly mineralized Cl-Na (SO₄·Cl-Na) salt lake brines, accompanied by synchronous enrichment of Li, B, arsenic (As), and other characteristic elements. Solute accumulation in regional waters is governed by the ternary coupling effects of evaporative concentration, rock weathering and leaching, and deep geothermal fluid input, while cation exchange and mineral dissolution–precipitation reactions further modulate ionic composition and ratios. Elements including As, Li, B, and chloride (Cl⁻) exhibit conservative migration behaviors in non-hydrothermal waters, whereas thermal springs possess unique geochemical signatures driven by deep fluid recharge. PCA results indicate that evaporative concentration serves as the primary controlling factor with a contribution rate of 55.39%; rock weathering provides the basic solute load (17.09%); and the coupled processes of deep fluid mixing and carbonate precipitation regulate elemental fractionation (14.21%). These findings systematically clarify the hydrogeochemical evolution laws and multi-source coupling mechanisms of inland lake waters in the Qiangtang Plateau. Furthermore, this study establishes a conceptual framework of “multi-source recharge–water–rock interaction–evaporative concentration”, advances the understanding of alpine hydrological cycling under climate change, and provides a solid scientific foundation for hydrological cycle research and green exploration of strategic mineral resources in endorheic salt lake regions. Full article

Climate change remains among the most pressing environmental challenges confronting the world, exerting profound pressure on both ecological systems and socio-economic development. To advance understanding of the evolution patterns and driving mechanisms governing hydroclimatic systems in arid and semi-arid regions, this study employed an integrated framework encompassing trend testing, change-point detection, periodicity and persistence analysis, and machine learning-based attribution. Focusing on the Mu Us Sandy Land from 1982 to 2023, we systematically investigated the spatiotemporal evolution, periodic characteristics, and driving mechanisms of hydroclimatic factors. Furthermore, future climate risks were assessed using CMIP6 multi-model data. The results showed that: (1) All four variables exhibited positive slopes, but only soil moisture showed a statistically significant long-term wetting trend (β = 0.025 × 10⁻³, p = 0.0008) and a clear global abrupt change in 2011; the upward tendencies of precipitation (p = 0.3946), potential evapotranspiration (p = 0.4970), and surface runoff (p = 0.1097) did not reach the 0.05 significance level. (2) Meteorological elements showed weak periodicity and strong anti-persistence (mean Hurst index = 0.379 for precipitation and 0.222 for PET), whereas hydrological elements exhibited clear seasonal–interannual periods and more random future variability with greater spatial heterogeneity (mean Hurst index = 0.436 for runoff and 0.414 for soil moisture). (3) Monthly changes were mainly associated with local surface processes. Vegetation dynamics were key predictors of precipitation, runoff, and soil moisture, while potential evapotranspiration was dominated by atmospheric demand, with limited influence from large-scale climate indices. (4) Under high-emission scenarios, imbalanced water–heat increases may lead to a higher likelihood of drought conditions. Full article

Accurate precipitation forecasting on the Qinghai-Tibet Plateau (QTP) remains a significant challenge due to complex terrain and nonlinear atmospheric dynamics. This study evaluates an XGBoost-SHAP framework for 24 h precipitation forecasting at Maqu Station, leveraging multi-source observations from 2020 to 2022. Vertical profile analyses via microwave radiometer (MWR) indicate that moisture is predominantly confined to altitudes below 4 km (AGL), with Integrated Water Vapor (IWV) and Liquid Water Path (LWP) typically varying between 0–15 mm and 0–2.5 mm, respectively. The optimized XGBoost model achieves an annual R² of 0.872 and a Root Mean Square Error (RMSE) of 1.609 mm, showing improved statistical consistency compared with standard Random Forest baselines. While the framework maintains robust performance for winter stratiform precipitation (RMSE = 0.32 mm), predictive variance increases during summer convective periods (RMSE = 3.26 mm). SHAP diagnostic analysis identifies Dew Point Temperature (DPT) as a consistent year-round predictor. Feature sensitivity analysis further reveals shifting seasonal driving mechanisms: spring precipitation appears sensitive to mid-tropospheric geopotential height, whereas summer forecasts are more strongly modulated by 500 hPa specific humidity and lower-level water vapor density. Overall, the XGBoost-SHAP framework serves as a transparent and physically plausible diagnostic tool for examining seasonal moisture–dynamic coupling. While these site-specific results are encouraging, they represent a localized empirical baseline; further cross-site validation is required to assess regional generalizability. Full article