Earth and Environmental Sciences

Accurate fractional snow cover (FSC) mapping requires combining the complementary strengths of optical imagery (spectral discrimination) and Synthetic Aperture Radar (all-weather capability). Yet existing fusion approaches employ simplistic concatenation strategies that fail to capture complex inter-modal relationships, while taking into account the resolution mismatch between high-resolution inputs and coarse reference labels. This paper introduces XFuse , a hybrid CNN-Transformer architecture that addresses these challenges through three key contributions: (1) bidirectional cross-attention fusion that explicitly models SAR-optical interdependencies via query-key–value transformations, enabling each modality to selectively attend to complementary information from the other; (2) a multi-resolution training strategy ( 10 m → 60 m → 10 m ) that fuses Sentinel-1/2 features at native resolution, aggregates to 60 m for supervision against Gap-Filled Fractional Snow Cover (GFSC) labels, and disaggregates through guided upsampling to restore fine spatial structure at 10 m ; and (3) quality-aware curriculum learning that incorporates multi-source metadata (cloud masks, atmospheric parameters, quality control flags) through adaptive loss weighting, enabling robust training from heterogeneous satellite data while providing calibrated uncertainty estimates. Evaluation on 29 Sentinel-1/2 tiles from Northern Finland (18,117 test patches from 4 geographically unseen tiles) demonstrates state-of-the-art performance: 13.77% MAE and 94.66% classification accuracy, representing 38.7% MAE reduction over the best baseline. Ablation studies confirm cross-attention as the largest individual contributor (5.16 percentage point accuracy gain), while uncertainty quantification achieves excellent calibration (ECE=0.025). The methodology generalizes to earth observation applications requiring fusion of heterogeneous multi-resolution satellite data

• We analysed plastics in stomachs of 507 fulmars from bycatch. • 81% of birds contained plastics. • 20% of stomachs contained >0.1 g plastics. • Number but not total mass of plastic particles differed among regions. • Sex- and age differences in plastic loads were not consistent across regions.

Abstract This study investigates spatiotemporal changes in socioeconomic exposure to tropical cyclone-induced precipitation (TCP) extremes across China using high-resolution rainfall and gridded population datasets. Results show that extreme TCP exhibits a significant increasing trend over central–southern China, while decreasing trends are observed in parts of the Yangtze River Delta. These contrasting patterns are linked to changes in TC characteristics, including intensified precipitation rates and reduced translation speeds over coastal regions, which enhance rainfall persistence and accumulation and thereby contribute to increasing extreme TCP events inland. Regions with high socioeconomic exposure are mainly concentrated in southeastern and central–eastern China. Based on a modified exposure index that integrates physical hazard and socioeconomic factors, we find that although exposure is jointly determined by these components, its recent increase is primarily driven by intensified TCP intensity and duration, with population growth and economic development further amplifying the impacts. Population and Gross Domestic Product exposure to extreme TCP events (>200 mm) increase by approximately 27% and 17%, respectively, with pronounced spatial heterogeneity. Rapid urbanization and economic expansion in recent decades have further elevated vulnerability to TC-induced precipitation extremes. This multiscale framework provides actionable insights for prioritizing mitigation and adaptation strategies, enhancing climate resilience in high-risk zones.

Abstract Climate conditions worldwide are influenced by the mean and variability of the tropical Pacific zonal sea surface temperature (SST) gradient. How this gradient responds to greenhouse gas forcing is therefore critical for accurate future climate projections. The nature of the response, however, remains debated: historical model simulations favor a weakening trend, whereas observational records from the same period are characterized by a strengthening trend. To explain this model–observation discrepancy, some attribute the observed trend to internal variability or observational uncertainties, while others suggest that models may inaccurately simulate the radiatively forced response. Past studies have analyzed different trend intervals and observational datasets, potentially contributing to conflicting conclusions about whether observations reflect the forced response. We present a comprehensive analysis of observed zonal SST gradient trends and their statistical significance. We estimate observed trends over all 20-year or longer intervals within the 1870–2024 period and subsequently evaluate these trends against a series of null hypotheses using bootstrapped ensembles of various statistical, conceptual, and geophysical models. Our analysis reveals that both strengthening and weakening trends are observed, depending on the analyzed intervals; however, intervals extending into the 21st century, particularly those since 1950 or those over a century or longer, exhibit statistically significant strengthening trends, suggesting that such trends are unlikely to have emerged from internal variability alone. This finding has implications for the historical and probable near-term transient responses, indicating they are likely radiatively forced. We confirm these findings with multiple observational datasets, demonstrating that data uncertainties minimally influence our conclusions.

Abstract Oceanic currents redistribute nutrients, phytoplankton, and other biogenic materials, fundamentally shaping marine biodiversity and ecosystem functioning. Yet, the topology of fine‐scale material transport remains poorly resolved due to limitations in high‐resolution flow observations. Here, by constructing Lagrangian flow networks from the Surface Water and Ocean Topography (SWOT) observations, we analyze surface fine‐scale transport features in the South China Sea in 2023. Compared with networks derived from conventional altimetry products, SWOT‐derived networks identify more sinks, sources, and transport gateways at 2–10‐day timescales and spatial scales below ∼60 km (90 km) in summer (winter). As such, SWOT resolves hydrodynamic provinces that remain invisible to conventional altimetry, revealing previously undetected corridors and barriers of surface exchange. This advantage also provides better dynamic explanations for complex phytoplankton community structures and evolution. Our results highlight SWOT's transformative capacity to improve the diagnosis and prediction of ocean material transport, opening new avenues for interdisciplinary oceanographic and ecological applications.

Accurate forecasting of fine particulate matter (PM2.5) remains a global challenge due to spatial gaps, data imbalance, and limited representation of extreme events. This study presents an enhanced Deep Imbalanced Regression (DIR) framework that integrates NASA’s GEOS-FP forecasts with global ground-based PM2.5 observations using a Temporal Convolutional Network (TCN) and a Residual Mixture-of-Experts (ResMoE) architecture. The model was trained on 378,000 samples (2021–2025) from U.S. Embassy AirNow sites and OpenAQ sensors, increasing geographic diversity. To address the imbalance, Label Distribution Smoothing (LDS) and weighted loss were applied, while ResMoE adaptively routed samples to specialized experts across meteorological-aerosol regimes. This configuration achieved strong performance (R² = 0.88, MSE = 23.4, MAE = 2.86 µg/m³) and generalized well across polluted and clean regions, including unseen sites. During the May 2025 Minnesota wildfire, the model captured both temporal evolution and peak magnitude missed by the TCN baseline, demonstrating improved responsiveness to extreme events. Uncertainty quantification and sensitivity analysis confirm model consistency. Beyond forecasting, the framework enables spatiotemporally consistent PM2.5 reconstruction for exposure assessment and policy analysis in data-scarce regions. This study provides a scalable and interpretable pathway for next-generation global air-quality forecasting.

Mesoscale Convective Systems (MCSs) are the primary drivers of extreme rainfall and flood hazards. Understanding their response to global warming is therefore important for improving climate projections and disaster risk reduction. In East China, however, evidence remains largely limited to case studies, and long-term projections of MCSs are scarce due to their poor representation in coarse-resolution climate models. Using convection-permitting model simulations with a pseudo–global warming approach, this study presents 22-summer projections of future MCSs over East China under the SSP5-8.5 scenario. Results show that future MCSs exhibit higher precipitation intensity and wider convective areas. Convective precipitation intensifies more than stratiform precipitation, indicating a transition toward more convection-dominated systems, accompanied by enhanced mid-level updrafts and super–Clausius–Clapeyron scaling of extreme rainfall. Increases in convective available potential energy (CAPE) and total column water vapor (TCWV) largely explain the intensification of maximum convective precipitation, with vertical zonal wind shear acting as a key dynamic modulator. The product of TCWV and vertical velocity is identified as an effective predictor of MCS peak precipitation. These findings imply heightened risks of intense rainfall from MCSs in East China and highlight the combined roles of thermodynamic and dynamic processes in shaping future MCSs.

Abstract Metasomatized lithospheric mantle plays a critical role in the petrogenesis of CO 2 -rich magmas, which are important hosts of rare-earth element deposits. However, the relationship between the structure of the lithosphere and the global distribution of CO 2 -rich magmas remains poorly quantified. Here we analyse the locations of young (<200 million years ago) continental intraplate CO 2 -rich silicate magmas and magmatic carbonatites in conjunction with upper-mantle shear-wave velocity anomalies and lithospheric thickness estimates. Our results document systematic increases in lithospheric thickness with estimated magma CO 2 content from basanites (<5 wt% CO 2 ), which erupt through seismically slow and thin non-cratonic lithosphere, to nephelinites, melilitites and ultramafic lamprophyres, which occur within progressively faster, thicker lithosphere and, finally, to lamproites and kimberlites (<20 wt% CO 2 ), which are emplaced on thick cratonic lithosphere. Carbonatites are associated with lithospheric thicknesses similar to those of nephelinites, melilitites and ultramafic lamprophyres, implying the derivation of carbonatites from these mafic CO 2 -rich silicate magmas via liquid immiscibility and/or fractional crystallization. We illustrate our lithospheric thickness–magma type relationship using Cretaceous–Pleistocene alkaline magmatism across western North America, ultimately demonstrating how lithospheric thickness controls the global occurrence of CO 2 -rich magmas and, consequently, their associated rare-earth element deposits.

Abstract Satellite and reanalysis rainfall datasets are crucial for meteorological research, hydrological applications, and validating numerical weather prediction models. However, their reliability must be carefully assessed before operational use. The Mumbai MESONET, a high-density network of automatic rain gauges providing minute-level temporal resolution, offers a unique and robust ground-based reference for such validation efforts. This study uses MESONET observations to evaluate the accuracy and consistency of various satellite-based precipitation estimates (MPEs) and reanalysis products across daily and sub-daily timescales. Our findings indicate that the IMERG-F and ERA5 reanalysis datasets have the best performance on daily scales, successfully capturing the overall variability of rainfall, albeit with some discrepancies in magnitude. While satellite and reanalysis products can capture diurnal rainfall cycles, they often differ in their representation of rainfall peaks, particularly during heavy precipitation events. Bias-corrected products, such as GSMaP-MVK and IMERG-F, demonstrate improved accuracy but continue to underestimate extreme rainfall events. Although satellite-based products show limitations in detecting light rainfall at shorter timescales, they perform reasonably well for moderate to heavy rainfall events. Statistical indices show that the performance of the target datasets considered for evaluation improves significantly after a 12-hour accumulation period. Based on the normalized composite score derived from all evaluation metrics, IMERG-E emerges as the most suitable product for real-time applications, such as flood forecasting and hydrological modeling. IMDAA outperforms INSAT-based products (i.e., INSAT-HE and IMSRA) in terms of overall accuracy over the study region.