Atmospheric and Oceanic Sciences

Study region The downstream sub-basin of the Lower Yellow River in Shandong Province. Study focus Flood risk in suspended-river systems remains poorly understood due to their anomalous geomorphology and hydrodynamic behavior. This study integrates a two-dimensional hydrodynamic model with a multi-source Background-Dynamics-Loss-Resilience (BDLR) framework to assess flood risk in suspended river systems. Flood inundation processes for return periods of 5–100 years were simulated to quantify the spatiotemporal responses of inundation depth and flow velocity. A hybrid weighting approach, combining Analytic Hierarchy Process (AHP) and Entropy Weight Method (EWM), was applied to determine the relative importance of various factors in flood risk. Spatial risk distribution maps were generated, and dynamic threshold analysis identified risk-sensitive areas for targeted flood prevention. New hydrological insights for the region Floodwaters in suspended river systems overtop topographic ridges in stages, resulting in diminishing marginal increases in inundated area with increasing return period. Flood risk is controlled by geomorphic constraints and energy redistribution. Key risk factors include slope (14.868%-15.899%), inundation depth (16.371%-17.004%), flow velocity (14.380%-15.695%), population and GDP density (27.382%-28.570%). About 8.09% of the region exhibits higher- and high-risk values (> 0.279) under low return period (5-year) flood scenarios, indicating high sensitivity to such floods, particularly in the Jindi River basin, the Yellow River estuary. These findings offer new insights for flood management in the Shandong reach.

Abstract Solar wind is a complex network of distinct magnetic flux tubes, each contained and separated by a current sheet. In this study, more than 50,000 current sheets in the solar wind are identified and characterized for the first time, using multi‐point Cluster observations during solar minimum and maximum intervals. Flux tubes at Earth are found to have an average diameter of ∼1.5 Earth radii, nearly 45 times smaller than previously reported. Furthermore, 6 years of NASA's ACE solar wind observations at Sun‐Earth Lagrange point L1 is used to show that only 30% of the observed flux tubes would directly impact the Earth's magnetosphere. It is shown that flying closer to the Sun‐Earth line could help to improve the prediction accuracy of space weather monitors at L1 by up to 2.5 times.

Abstract This study examines the performance of ensemble simulations in representing precipitation using the Weather Research and Forecasting (WRF) model in three recent Nor’easter events: March 14 th –15 th 2017, March 2 nd –3 rd 2018, and January 29 th –30 th 2022. Compared to observations, the model tends to overestimate total precipitation along the eastern New England coast, with varying accuracy across the events. The WRF model reasonably captures heavy precipitation in the March 2018 event, while it largely overestimates moderate and light precipitation during the March 2017 and January 2022 simulations, respectively. A detailed sensitivity analysis of the March 2018 event using four different microphysics schemes reveals distinct strengths: Morrison most accurately simulates total precipitation amount; Thompson captures peak precipitation intensity best; NSSL yields overall reliable performance; and WDM6 achieves the strongest spatial correlation with gridded observations. The multi-physics ensemble approach offers modest improvements that are comparable to the gains achieved by a five-member stochastic ensemble. These findings provide insights for refining Nor’easter modeling, ultimately advancing preparedness and risk management strategies in vulnerable coastal regions.

We explored the applicability of microwave satellite observations to characterize the carbon dioxide (CO 2 ) exchange processes in boreal agricultural fields for three regions in Finland. The sites under investigation included two grassland fields on mineral soil and one on peat soil. We evaluated whether the space-borne microwave radiometer data-derived information on freeze–thaw, snow accumulation and snow melt obtained at a landscape scale is useful in explaining carbon and nitrogen exchange processes at daily temporal and field scale spatial resolutions. Specifically, we focused on the annual balance and seasonal behaviour of CO 2 exchange, and the feasibility of microwave observations in explaining the observed nitrous oxide (N 2 O) emission rates from a peatland field. Additionally, the feasibility of space-borne synthetic aperture radar (SAR) data to support observation time series of optical satellite data on a field block level was investigated for the growing season. The results indicate the potential of space-borne microwave observations to provide information useful in assessing CO 2 exchange and N 2 O bursts from boreal agricultural fields. The results show that seasonal changes due to climate and management, detectable on a landscape scale from space-borne microwave radiometry, significantly contribute to the annual balance of CO 2 exchange of boreal fields. The results also suggest that SAR data can potentially be used to fill cloud cover induced gaps in leaf area index (LAI) time series available from optical-range satellite imaging, improving the usefulness of satellite data in the monitoring of field management. Results show that landscape-scale freeze–thaw and snow cover information derived from space-borne SMOS and SSMI/S observations distinguishes seasonal periods with distinct CO 2 exchange characteristics, indicating the potential of coarse-resolution microwave radiometer data to constrain carbon exchange models of boreal agricultural fields outside the summertime growing season. Further, C-band Sentinel-1 SAR backscatter observations were found to correlate significantly with GPP dynamics and enable detection of harvesting events, suggesting that combined SAR and radiometer time series can support year-round agricultural monitoring.

Abstract Using a list of sudden‐commencement storms, the ring‐current index, and 1‐h near‐Earth solar‐wind measurements from solar cycles 20–25, we develop extreme‐value statistical models relating storm intensity to the storm main‐phase maximum duskward interplanetary electric field . The conditional relationship is demonstrably sublinear—linear models are confidently rejected—indicating saturation of magnetospheric response under extreme solar‐wind forcing. An event like that of July 2012 ( mV/m), if Earth‐directed, would be associated with a median storm intensity of nT. Storms comparable to March 1989 ( nT) correspond to electric fields of mV/m, while Carrington‐class storms ( nT) correspond to mV/m—substantially lower than several previous estimates. These results indicate that solar‐wind conditions capable of driving extremely intense magnetic storms are less exceptional, and potentially more frequent, than previously thought.

The dynamics of methane (CH4), nitrous oxide (N2O) and carbon dioxide (CO2) in groundwater have rarely been investigated. As dissolved gases they may be transported to distant sites and, hence, to the atmosphere. Crumps Cave (CC) is located on a perched aquifer in south-central Kentucky. Water was sampled at a waterfall within the cave located 15 m below the surface, at two adjacent surface wells 15 m and 50 m deep, providing samples from the epikarst and regional aquifer, respectively. Dissolved gases and geochemistry parameters were analyzed for seasonal changes across three years of weekly monitoring (2015–2017) using Kruskal–Wallis H tests and Bonferroni-corrected pairwise comparisons. Dissolved CO2 concentrations are mainly controlled by percolation through the epikarst, influenced by soil respiration, and vary with rainfall and seasonal temperature fluctuations. CH4 showed a site-dependent pattern: concentrations were significantly elevated in warm seasons at the shallow and deep wells, where anaerobic conditions and agriculturally derived organic matter promote methanogenesis; no seasonal variation was detected at the cave site, where oxic conditions limit CH4 year-round. N2O was significantly elevated in cold seasons at all three sites, driven by cold-season denitrification of agriculturally derived nitrates. N2O did not differ between sites, indicating seasonal temperature-driven denitrification as the primary control rather than site hydrology, with cold-season denitrification of agriculturally derived nitrates from fertilizer application. Indirect gas emissions are characteristic of karst systems and may be transported or stored in aquifers through complex interactions of groundwater recharge, microbial activity, and seasonal land-use variability.

Groundwater-potential assessment in karst aquifers is complicated by pronounced spatial heterogeneity driven by structural permeability, lithological variability, recharge redistribution, and unresolved subsurface conduit connectivity. Although machine-learning approaches have improved regional groundwater mapping, most existing models provide only deterministic predictions and offer limited information on predictive uncertainty and hydrogeological reliability. To address this limitation, we propose CT-TreeFlow. This probabilistic groundwater assessment framework goes beyond conventional machine-learning models by explicitly learning the full conditional probability distribution of groundwater favourability rather than a single deterministic estimate. The framework integrates sparse probabilistic environmental routing, conditional density estimation, hydrogeologically constrained pseudo-absence generation, geographically structured spatial validation, and explainability-driven interpretation within a unified modelling architecture, enabling simultaneous groundwater prediction, uncertainty quantification, and hydrogeological interpretation. The framework was applied to the Zagros karst system in Khuzestan Province, Iran, using remote-sensing-derived environmental predictors, Copernicus DEM-based morphometric variables, geological–structural datasets, and hydroclimatic indicators. Performance was evaluated against LightGBM and XGBoost using GroupKFold spatial cross-validation. CT-TreeFlow achieved a mean RMSE of 2.737 and a mean R2 of 0.852, while also providing spatially explicit uncertainty estimates and probabilistic prediction intervals. Explainability analyses identified fracture density, lithology, drainage organisation, and terrain-controlled recharge conditions as the dominant controls on groundwater favourability. Predicted high-favourability zones showed strong spatial correspondence with major carbonate formations and independent spring–cave inventories, supporting the hydrogeological plausibility of the mapped patterns. These results demonstrate that probabilistic modelling can provide more reliable and physically interpretable groundwater assessments than deterministic approaches in structurally complex karst environments. CT-TreeFlow offers a transferable framework for uncertainty-aware groundwater exploration and regional hydrogeological decision support in heterogeneous aquifer systems.

A new study looks at daily swings in sea surface temperatures for new insights.

Abstract Fog prediction remains challenging because the physical processes governing its life cycle evolve across time scales and do not follow reversible or stationary dynamics. Using high‐frequency visibility observations from Sable Island, Canada, this study analyzes fog intensity and turbulent kinetic energy (TKE) for their time irreversibility and causal relations. Fog intensity exhibits temporal asymmetry in all stages, while TKE remains nearly reversible. The lead–lag structure between the two variables is stage dependent: TKE leads fog intensity during formation, the coupling becomes symmetric during the mature phase, and fog intensity leads TKE during dissipation. Notably, during fog formation, the strength of fog's intrinsic irreversibility increases linearly with the strength of its causal linkage to TKE, revealing that fog initiation is governed by a directional sequence of turbulence–moisture interactions. These findings demonstrate that fog is a non‐equilibrium, time‐asymmetric system, and that capturing its stage‐dependent directionality is required for enhanced fog prediction.

Two major factors influence the expected recovery of the ozone layer: a decline in ozone-depleting substances (ODSs) due to implementation of the Montreal Protocol and stratospheric cooling due to increasing greenhouse gas (GHG) concentration. We investigate the largest spring Arctic ozone anomalies revealed in three ensemble calculations of the chemistry–climate model (CCM) SOCOLv3 under moderate (SSP2-4.5) and severe (SSP5-8.5) scenarios of GHG growth, accounting for the expected decline in ODS concentrations over 2015–2099. During the first half of the 21st century, the coldest winters could still produce Arctic total ozone content (TOC) anomalies comparable with the record spring 2020 values, despite the overall recovery of the ozone layer by mid-century. In March and April, TOC may occasionally drop below 220 Dobson Units. According to estimates from the Moscow State University (MSU) radiation model, the lowest TOC values under cloudless midday conditions could increase surface UV radiation by a factor of 1.5–2, reaching a UV index of 5–6 (and up to ~8 in April)—levels requiring sun protection measures.

PFAS form a class of synthetic chemicals that has become an area of increasing concern because of its impact on the environment and the threat it poses to human health. The structure of PFAS makes them highly resistant to degradation. As a result, they are highly effective at bioaccumulation. Certain water treatment technologies have been proven to remove PFAS from contaminated water sources. This study reviews the most promising treatment technologies used for the treatment of PFAS-contaminated waters. Well-established treatment technologies, such as granular activated carbon, ion exchange resin, reverse osmosis, and nanofiltration, were quantitatively compared. The removal efficiency was assessed by collecting the data of individual PFAS species from the literature and grouping them into five groups: PFAS (all species), PFSA, PFCA, long chain, and short chain. The results identified that, for all PFAS groups, the most effective treatment technologies were in the following order: reverse osmosis, nanofiltration, ion exchange resin, and granular activated carbon. The performance of reverse osmosis and nanofiltration did not appear to significantly differ between the different PFAS groups, as opposed to ion exchange resin and granular activated carbon, where there was a greater degree of variation in performance between different PFAS groups. Overall, it was identified that membrane technologies outperformed adsorbent technologies. However, the cost associated with membrane technologies may limit its economic viability when compared with adsorbent technologies, which are typically a more viable option except under specific circumstances. For example, contaminated water with high concentrations of other contaminants that need to be treated simultaneously. Lack of standardised experimental and operational conditions limited the available data. While this work provides guidance on which treatment is more likely to be appropriate based on the concentration and composition of different species of PFAS, more data are needed to conduct a more accurate statistical analysis and enable accurate modelling of treatment performance.

Study Region : The Upper Yangtze River basin, comprising the Jinsha, Min, Jialing, and Wu tributaries and the mainstream, contains 38 large reservoirs with a live storage capacity (LSC) of 80.5 billion m 3 . Study Focus : This study couples hydrological simulation with the Soil and Water Assessment Tool (SWAT) and reservoir-operation optimization to quantify flow regime alteration across three development stages: a natural baseline, a 21-reservoir configuration, and a 38-reservoir configuration. SWAT simulated daily natural flows for 1965–2024, and the Progressive Optimality Algorithm-Dynamic Programming Successive Approximation (POA-DPSA) optimized reservoir operations under a common objective and constraint framework. This counterfactual design isolates reservoir-operation effects from climatic variability and enables comparison among sub-basins with contrasting reservoir layouts and coordination structures. New Hydrological Insights for the Region : Reservoir operations reduced wet-to-dry flow ratios by up to 64.6% through seasonal redistribution. Under the fixed-parameter baseline, DOR (storage-to-annual-runoff ratio) above about 10% was associated with substantial flood attenuation, whereas low-flow augmentation emerged above about 5%. Across the three development stages examined in each basin, DOR and the coefficient of variation (CV) of daily flow were well described by linear fits (R ≥ 0.98), but fitted slopes differed among basins, indicating spatially variable sensitivity to DOR increases across development stages. A coordinated-versus-non-coordinated comparison showed basin-dependent coordination gains, strongest in the Jinsha cascade and weaker where mainstream multi-year reservoirs already provided substantial regulation.

Assimilation of all-sky radiance (ASR) observations informs atmospheric states and cloud distributions; however, it does not always lead to improved analyses or forecasts. In particular, directly updating hydrometeor fields introduces substantial uncertainty into ASR assimilation. This study examines the impact of dynamic observation errors on analyses and precipitation forecasts under different hydrometeor control variable (HCV) configurations. Observation errors are prescribed using a fifth-order polynomial model as a function of a cloud impact parameter, allowing spatiotemporally varying (i.e., scene-dependent) errors that adapt to cloud conditions. Results indicate that dynamic observation errors generally improve cloud analyses and associated thermodynamic fields. By contrast, constant errors tend to overweight ASR observations in heavily cloud-affected regions, thereby degrading analysis quality. The advantages of dynamic errors are more pronounced when solid-phase hydrometeors are included in the HCV, as these strongly influence brightness temperature (BT) analysis and the representation of convective cloud tops. Among all experiments, those combining dynamic errors with direct updates of solid-phase hydrometeors produce the most realistic BT and reflectivity analyses, as well as the greatest improvements in precipitation forecasts. These results underscore the importance of cloud-dependent observation error modeling in ASR assimilation, particularly when solid-phase HCVs are employed.

Iberian reservoirs are highly vulnerable to droughts, warming temperatures, and agricultural runoff, which accelerate eutrophication. Monitoring these dynamics is crucial for sustainable management. This study investigated long-term trends in chlorophyll-a (Chl-a) and water transparency Secchi depth and developed empirical models for the Alto Rabagão (Rb) and Aguieira (Ag) reservoirs in Portugal. We used Sentinel-2 Level-2A reflectance data coupled with 153 in situ observations (2014–2024) for model calibration (n = 95) and validation (n = 58). Temporal trends were assessed using linear regression and Mann–Kendall analyses. Empirical models based on spectral indices (TBDO1, TBDO, MCI, NDWI) were evaluated using walk-forward time-series cross-validation. Results revealed a significant Chl-a increase (0.38 µg L−1 year−1; p = 0.016) and a simultaneous decline in transparency (p < 0.001) in Rb, indicating progressive eutrophication. In contrast, no significant trends were detected in Ag. Reservoir-specific models achieved moderate-to-high predictive performance, particularly for Chl-a (R2 up to 0.75; cross-validated R2 = 0.67–0.68, RMSE = 1.1 µg L−1, MAE = 0.82 µg L−1). Models using combined datasets showed lower accuracy, highlighting the importance of site-specific calibration. Wilcoxon signed-rank tests confirmed the absence of systematic bias between observed and predicted values. Ultimately, Sentinel-2 imagery combined with time-series cross-validation provides a reliable and cost-effective framework for the long-term monitoring of inland water quality.

The urban heat island (UHI) effect is one of the most studied phenomena in urban climatology. Numerous studies have revealed the heterogeneous nature of air temperature within cities, manifesting as an urban heat “archipelago” with small-scale air-temperature differences and multiple hot and cold spots rather than a single, uniform hot spot in the city core. With the introduction of the local climate zones (LCZs) scheme, close attention has been paid to the definition and description of “urban” and “rural” sites. However, what remains understudied and inconsistent across studies is the “sea level” of the “archipelago,” i.e., defining the air-temperature conditions of the surroundings, unaffected by the city. Here, we compare definitions and requirements of that “sea level,” and investigate multiple possible data sets for UHI calculation. Most typically, single weather stations, often at airports, are used as a “rural” reference. However, these stations are not ubiquitously available and typically suffer from effects such as a high fraction of impervious surfaces or urban heat advection when located downwind of the city. Crowd weather stations (CWS), which have gained attention in recent years in urban climate studies due to their abundance, are often affected by nearby buildings and are mostly located in urban areas. Besides station-based data, reanalysis products such as ERA5-Land could provide an independent reference, as they are available globally and often do not consider urban areas. In this study, all three data sources were compared against an ideal case of multiple professional weather stations placed around the city. We investigate the two temperate European cities of Paris (France) and Berlin (Germany) during six years (2019–2024), focusing on crowdsourced data from CWS. We find that ERA5-Land air-temperature data is a consistent and ubiquitously available reference for the definition of “rural,” providing UHI-calculations most comparable to the ideal case of having multiple professional weather stations. Using it as a universal rural reference allows for comparison between cities and further enables exploiting the potential of CWS, even in regions with few stations and a lack of professionally-operated rural weather stations. Establishing a consistent “sea level” for urban air-temperature and UHI-studies enables comparison between various cities globally and allows for the integration of different data sources, such as local networks of weather stations or CWS, on a larger scale.

Study region This study focuses on 118 tile-drained agricultural watersheds in Iowa, USA, a region characterized by intensive agricultural management and widespread subsurface drainage systems. Study focus Hourly precipitation and discharge data from 2002 to 2023 were used to identify 7621 flood events. Based on hydrograph characteristics, these events were classified into three flood types: small-fast floods, sustained low-crest floods, and high-crest, high-volume floods. Interpretable machine-learning models were employed to identify the climatic, soil, hydrological, and topographic controls on flood-type occurrence, peak discharge, and runoff volume. Correlation and multiple regression analyses were further used to examine how flood-type frequency and runoff contributions vary across tile-drainage gradients. New hydrological insight Flood-type occurrence reflected the combined influences of climatic, soil, hydrological, and topographic factors, with precipitation duration emerging as one of the strongest controlling factors. Short-duration events were primarily associated with small-fast floods, whereas longer-duration events increasingly favored sustained low-crest floods. Most importantly, flood-type composition varied systematically across tile-drainage gradients. Watersheds with higher tile-drainage percentages exhibited greater volume (R = 0.50) and frequency (R = 0.55) contributions from sustained low-crest floods, but lower contributions from high-crest, high-volume floods (R = −0.34 and −0.43, respectively). These relationships remained significant after controlling for climatic, topographic, and soil-related factors. Together, these findings demonstrate the importance of considering flood-type-specific characteristics when developing watershed-management strategies in tile-drained agricultural landscapes.

Satellite-based land surface temperature (LST) products are frequently affected by cloud cover and atmospheric conditions, resulting in missing data that significantly limits the continuous monitoring of the thermal environment in complex terrains, such as the Tibetan Plateau. Existing spatiotemporal interpolation methods face clear accuracy limitations when addressing extensive data gaps, while physical models often struggle due to insufficient meteorological inputs in complex landscapes. Moreover, conventional data-driven approaches usually overlook local spatial variations, resulting in smoothed thermal patterns and systematic errors. To overcome these issues, we propose a Physically Constrained Spatial Residual Learning framework. In this framework, we use the Enhanced Annual Temperature Cycle (EATC) model to capture the temporal baseline of LST first. Then, we integrate multi-source auxiliary data into the Geographical-XGBoost (G-XGBoost) algorithm to model spatial nonlinear residuals. Using simulated cloud masks on the 2017 MODIS LST dataset from the Qinghai–Tibet Engineering Corridor, we show that the hybrid model outperforms both individual physical models and global machine learning models in accuracy and spatial detail recovery. Validation results yield an R2 of 0.88, an RMSE of 1.92 K, and a mean bias of 0.07 K. Seasonal evaluations indicate best performance in winter (RMSE = 1.19 K) with robust performance in summer. Furthermore, the framework reduces boundary artifacts and accurately reproduces thermal spatial patterns in complex terrain through adaptive local bandwidth and weight adjustments. This approach provides a reliable method for high-precision LST reconstruction over heterogeneous alpine surfaces.

Evapotranspiration (ET) estimation in desert-oasis ecotones remains challenging due to sparse meteorological observations and the coarse spatial resolution of satellite remote sensing, which limit the ability to resolve highly heterogeneous surface conditions. To address this issue, this study develops a high-resolution ET estimation framework by integrating unmanned aerial vehicle (UAV)-based thermal infrared remote sensing with a three-temperature (3T) model in the Hexi Corridor. UAV-derived land surface temperature (LST) at meter-scale resolution, together with meteorological and vegetation data, was used to drive the model and generate high-resolution ET maps. The model’s performance was validated spatially against the Surface Energy Balance Algorithm for Land (SEBAL) model and at the point-scale against a two-source model. The results show that: (1) The 3T model effectively captured the spatial gradient of decreasing ET from cropland (3–10.69 mm d−1), through shelterbelts (3–6 mm d−1), to desert areas (<3 mm d−1). (2) Spatial validation against the SEBAL model was conducted using stratified pixel-wise comparisons across four land-cover types over 14 UAV transects, showing strong agreement (R2 = 0.90–0.95; RMSE = 0.22–0.43 mm d−1). The model achieved highest accuracy in cropland (R2 = 0.92; RMSE = 0.24 mm d−1), with slight overestimation in shelterbelts. (3) Point-scale validation against the two-source model yielded an MAE of 0.38 mm d−1. This study demonstrates the effectiveness of combining UAV thermal infrared data with the 3T model for high-resolution ET simulation in complex ecological transition zones, offering a promising technical approach for ecohydrological monitoring and water resource assessment in arid regions.

A new study suggests warming temperatures and increased solar radiation have boosted carbon fixation in tidal wetlands across the country.

Accurate spatial measurement of aboveground biomass (AGB) is essential for assessing carbon stocks in the forest ecosystem. To enhance this estimation, integrating active and passive Earth Observation data with advanced machine learning techniques offers a promising approach. This study presents an integrated HybridEnsemble model with golden jackal optimization for AGB estimation and evaluates its predictive performance against individual base-learners, including categorical boosting (CatBoost), extreme gradient boosting (XGBoost), light gradient boosting machine (LightGBM), and adaptive boosting (AdaBoost) via the synergistic application of Explainable Artificial Intelligence (XAI) and active and passive datasets. The findings revealed a clear performance ranking among these models, with the HybridEnsemble Golden Jackal Optimization (HGJO) model identified as the most effective, which yielded a correlation coefficient ( R 2 ) of 0.821 and a Relative Root Mean Square Error (rRMSE) of 16.30%. This performance was followed by CatBoost ( R 2 = 0.816, rRMSE = 16.51%), LightGBM ( R 2 = 0.804, rRMSE = 17.05%), XGBoost ( R 2 = 0.802, rRMSE = 17.13%), and AdaBoost ( R 2 = 0.731, rRMSE = 19.97%), with all comparisons reported at 95% confidence intervals. XAI revealed that predictors from optical sensors (passive) were strongly correlated with AGB and played a significant role in predicting AGB, while features derived from SAR (synthetic aperture radar, an active sensor), less influential, provided unique backscatter and context-specific insights that enhanced the model’s performance. Forecast results indicate an increasing trend. Additionally, the analysis revealed that future AGB accumulation in the subtropical forest of Hong Kong will be highly variable and strongly dependent on initial biomass levels, with high-biomass plots likely to see the greatest gains. However, spatial uncertainty in AGB predictions varied across the study area, with higher uncertainties observed in forested areas and lower uncertainties in urban areas. Overall, this study not only enhances understanding of optimized hybrid ensemble models for biomass prediction but also offers valuable insights for forecasting forest dynamics, supporting sustainable forest management and carbon stock monitoring globally.

Abstract Anthropogenic pressures from irrigation and human usage (i.e., domestic and industrial use) pose growing challenges to river sustainability under future global change. However, how these pressures evolve across river networks at fine spatial scales remains poorly understood. Here we project the anthropogenic pressures on 129,429 river reaches in China from 2020 to 2050 under the four CMIP6 scenarios (SSP1-2.6, SSP2-4.5, SSP3-7.0, and SSP5-8.5), using a proximity-based index that integrates built-up and irrigated areas. The results show that the river reaches in the Hai and Huai River basins are projected to face high and intensifying anthropogenic pressures, which are further exacerbated by meteorological drought threats under the SSP1-2.6 and SSP2-4.5 scenarios. In contrast, the Southwest and Inland River basins exhibit approximately 5% of the anthropogenic pressures observed in the Huai River basin, but meteorological drought remains an important stressor threatening the potential future river sustainability. Furthermore, anthropogenic pressures also vary by river size. Large rivers (stream orders >6) are projected to experience the highest anthropogenic pressures, followed by medium rivers (stream orders 4-5), whereas small rivers (stream orders 1-3) remain comparatively less impacted across all scenarios. Our reach-scale analysis provides a refined understanding of human-river interactions and offers critical insights for targeted water management under climate change.

Accurate crop mapping in fragmented agricultural landscapes is challenged by overlapping crop calendars and redundancy among multi-source time-series variables. Using Sentinel-1/2 imagery from December 2022 to December 2023, we constructed 275 season-specific spectral–phenological feature–month variables (125 for summer crops and 150 for winter crops) for rice, maize, soybean, winter wheat, and winter rapeseed in Jiangsu Province, China. An auxiliary binary Random Forest (RF) was used to estimate out-of-bag (OOB) permutation-based predictive contributions and construct search priors. A prior-guided genetic algorithm (GA) then identified compact subsets, with crop-specific five-class RF models used both to evaluate candidate subsets and to produce the final classifications. A fixed stratified 80/20 development–validation split was maintained throughout the analysis, with the validation subset reserved for final assessment. August and April were the principal discriminative periods for summer and winter crops, respectively, while VH backscatter and SWIR-related indices, particularly STI and NDTI, showed recurrent predictive contributions across crops. On the independent validation subset, the optical/vegetation-index scheme, SAR-only scheme, and the complete feature library achieved mean target-crop F1-scores of 78.42%, 83.74%, and 86.96%, respectively. The GA-selected subsets retained 9–39 variables and achieved a mean five-class overall accuracy of 91.77% and a mean target-crop F1-score of 93.95%. After non-target classes were merged into a single background class, the integrated seasonal maps achieved overall accuracies of 81.20–95.03% on the same validation subset. Supplementary classifier comparisons indicated that subset effects depended on the crop and learning algorithm. The findings support crop-specific, interpretable dimensionality reduction within the RF workflow, while broader transferability requires multi-year and multi-region evaluation.