Earth and Environmental Sciences

Coral reefs play the main role in marine ecology, as home and shelter to many aquatic species, and quite importantly, supporting local economies. The danger to coral reefs from coral bleaching is joined by Acanthaster planci invasions that can extensively harm these sensitive colonies. The normal practice has been the manual identification of these starfish by divers and snorkelers. This study proposes an automatic deep learning-based approach to detecting and classifying such species within their natural underwater environments. In its architectural setup, the system fashions advanced types of Convolutional Neural Networks (CNN) together with Vision Transformer (ViT) models. This study evaluated several pre-trained CNN and transformer models on the marine species dataset, achieving test classification accuracy of 91.80% for MobileNetV3-Small, 94.54% for VGG19, 94.54% for ResNet50, 95.90% for DenseNet121, 97.54% for Swin-Tiny, 97.54% for CrossViT, and 97.81% for FastViT-T8. The proposed Spatial-Attention Swin + FastViT + CrossViT framework achieved an overall test classification accuracy of 99.18%. It will also make the detection process much safer and faster, besides being more dependable in coral reef protection.

Landfast ice in Terra Nova Bay represents a dynamic ecological interface where sea-ice structure, meltwater inputs, and platelet-ice processes interact to shape microalgal communities across the ice–ocean boundary. During the austral spring–summer 2015/2016, we examined how seasonal transitions—from early-season to progressive melting and stratification—modulate the development of sympagic assemblages and their connection with the underlying water column. By integrating sea-ice thermodynamics, pigment signatures, nutrient variability, and water-column profiles, we describe how physical forcing and biogeochemical gradients interact to shape community composition and functional structure. Surface salinity decreased from a maximum of 36.67 to a minimum of 30.67, while surface chlorophyll a exceeded 40 µg L - ¹ during the early-December biomass maximum and subsequently declined to below 1 µg L - ¹ in late December and January. The pigment record showed changes in the relative contribution of fucoxanthin, 19′-hexanoyloxyfucoxanthin (19′HF), and chlorophyll c2, while the multivariate analysis identified a significant vertical effect (PERMANOVA, F = 1.7188, R² = 0.1629, p = 0.0003), with sampling date accounted for as a temporal blocking factor. Together, these findings highlight the meltwater-influenced layer beneath landfast ice as a biologically relevant interface linking sea-ice processes with phytoplankton dynamics in Antarctic coastal waters.

The Mediterranean Sea is threatened by ocean acidification and warming, in combination with habitat loss, pollution, invasive species, and overexploitation. Assessing local high-frequency pH variability in the context of relevant basin-scale time series data is essential to understand ecosystem vulnerability. We report the results of the Italian national pH monitoring based on the Marine Strategy Framework Directive, and the pH dynamics observed at Capo Carbonara. An experimental station equipped with a pH sensor and CTD probe was deployed at 32 m depth on a sandy bottom bordered by Coralligenous assemblages. Physical and biogeochemical variables were continuously recorded over 345 days and analyzed through additive time series decomposition. The benthic assemblages were surveyed by photographic and destructive sampling in summer and autumn 2016 and spring 2017 on algal reef and sandy bottom, to describe the biological components undergoing and contributing to the pH fluctuations. Seasonal pH variability ranged up to 0.2 units annually and 0.1 units daily, with maxima in January and minima in August, while the range of pH on the basin scale had higher mean values and exceeded local variability. Assemblage changes were driven by algal turf development and the invasive Caulerpa cylindracea . Coralligenous taxa appear stable, despite the substantial natural pH variability, shaped by meteo-marine conditions, water mass circulation, Saharan dust inputs, and benthic biological processes, with primary producers and calcifiers exerting a key influence. The ongoing trend of increasing temperatures and decreasing pH may threaten the Coralligenous ability to cope with the shifting future boundaries of such variability.

Introduction Geopolitical disruptions increasingly affect maritime transportation networks, yet their spatially heterogeneous impacts on global shipping activity remain insufficiently understood. This study develops a two-scale analytical framework to investigate the spatiotemporal evolution of global shipping activity and examine regional responses associated with geopolitical disruptions in the Red Sea and the Arctic Ocean. Methods Historical Automatic Identification System (AIS) data were processed to reconstruct vessel trajectories, aggregated into spatial grids, and analyzed using a space-time cube model and emerging hot spot analysis. Results At the global scale, the framework identifies persistent, intensifying, diminishing, and emerging patterns of shipping activity, revealing substantial spatial heterogeneity across major shipping corridors and maritime chokepoints. At the regional scale, the comparison between the Red Sea and the Northern Sea Route cases demonstrates differentiated response patterns under distinct geopolitical contexts. In the Red Sea region, AIS-derived observations show a rapid redistribution of vessel activity between the Red Sea-Suez Canal corridor and the Cape of Good Hope route. In contrast, Arctic shipping activity exhibited stronger temporal fluctuations and spatial heterogeneity, reflecting the combined influence of geopolitical conditions, environmental variability, infrastructure development, and resource-related activities. Discussion The findings indicate that geopolitical disruptions are associated with heterogeneous adjustments in maritime networks, with different regions exhibiting distinct patterns of route redistribution, activity concentration, and temporal persistence. The proposed framework provides a data-driven approach for monitoring changes in maritime activity, identifying vulnerable corridors, and improving understanding of shipping-network resilience under geopolitical uncertainty.

Microplastic (MP) contamination of the open ocean is well documented, yet the vertical structuring of MPs and the role of water masses in shaping deep-water inventories remain poorly resolved along the eastern boundary of the North Atlantic Subtropical Gyre. Here we present a high-resolution vertical profile of MPs in the Canary Basin, based on discrete water samples collected during the MSM/127 cruise (R/V Maria S. Merian, March 2024) at three offshore stations south of Gran Canaria, comprising one full-depth profile (seven depths, 5 to 2720 m) and two stations sampled at shallower horizons; consequently, the deep-water structure below 1500 m reflects a single-station profile and should be interpreted as indicative rather than basin-wide representative. Particles were characterised by stereomicroscopy, ATR-FTIR, micro-Raman spectroscopy and SEM-EDX. MPs were detected in 100% of samples (n = 33), with a mean abundance of 19.5 ± 14.9 MP L-1, the large dispersion reflecting a markedly non-monotonic vertical distribution: a subsurface maximum of 39.3 ± 4.0 MP L-1 at 500 m, coinciding with the upper boundary of the North Atlantic Central Water and the top of the regional Oxygen Minimum Zone; a vertical minimum of 2.0 ± 1.0 MP L-1 at 2000 m; and a secondary deep-water signal between 2500 and 2720 m. MP abundance was strongly correlated with temperature (Pearson r = +0.77), salinity (r = +0.66) and potential density (r = -0.76; all p < 0.001), whereas its association with dissolved oxygen was weak and non-monotonic (Spearman ρ = -0.06, ns), consistent with water-mass structure, rather than dissolved‑oxygen levels, as the principal physical correlate of MP vertical distribution. With depth, the assemblage converged toward fine (<500 μm), dark, fibrous polyethylene terephthalate (PET) particles, reaching 100% PET between 1500 and 2720 m. SEM-EDX revealed weathered, biofouled surfaces bearing Saharan mineral and biogenic carbonate residues, consistent with a Saharan dust and biogenic origin, suggesting that the deep Canary Basin acts as a potential accumulation zone delivering an aged PET-fiber signature to abyssal layers.

We investigated the isotopic variation of particulate and dissolved organic matter (OM; δ13CPOC, δ15NPN, δ13CDOC) associated with the frequent occurrence of hypoxia in a semi-enclosed bay (Jinhae Bay, South Korea). During the hypoxic period (August-October 2024), surface-water physical properties (temperature: 25.8 ± 3.0 °C; salinity: 30.5 ± 1.9 psu; dissolved oxygen: 7.0 ± 2.0 mg/L) coincided with significantly elevated concentrations of particulate and dissolved organic components (POC: 0.5 ± 0.2 mg/L; PN: <0.1 mg/L; DOC: 1.3 ± 0.3 mg/L; DTN: 0.3 ± 0.1 mg/L) throughout the water column. Under these conditions, the isotopic signatures of particulate and dissolved OM displayed distinct ranges (δ13CPOC: -20.7 ± 1.8 ‰; δ15NPN: 4.8 ± 1.7 ‰; δ13CDOC: -22.1 ± 0.9 ‰), reflecting in situ remineralization of autochthonous sources such as phytoplankton. By integrating apparent oxygen utilization with a Bayesian end-member mixing model (C₃ terrestrial plants, marine phytoplankton, and tidal-derived coastal water), we suggest potential seasonal shifts in dominant mode of OM production from new production to regeneration, under progressively declining oxygen conditions. This transition may be linked to the utilization of reduced nitrogen compounds. In particular, autumn isotopic variations (Δδ13CPOC: 2.4 ± 1.8 ‰; Δδ15NPN: -2.4 ± 2.4 ‰; Δδ13CDOC: 1.0 ± 1.4 ‰) may be consistent with a potential shift in nitrogen assimilation strategies, possibly involving the incorporation of regenerated dissolved nitrogen (i.e., DIN and DON) by phytoplankton under persistent hypoxia. Consequently, the isotopic signatures characterizing OM dynamics in this semi-enclosed bay provide a framework for interpreting OM interactions within ecological models.

Mesoscale eddies play a crucial role in the ocean, and accurate detection of their three-dimensional (3D) structures is essential for understanding marine dynamics. However, 3D eddy reconstruction remains challenging: traditional methods provide limited pixel-level boundary delineation, while existing deep learning approaches suffer from insufficient physical constraints and unrealistic eddy geometries. To address this issue, we propose a surface-to-volume physics-informed model (StV-PI-EddyNet) for extracting 3D mesoscale eddies. We construct a high-quality 3D eddy dataset covering the North Pacific (100°E–100°W, 5°N–65°N) down to 1000 m for model training. The model takes sea surface height (SSH), sea surface temperature (SST), and surface current components (U/V) as inputs and consists of three core modules: 2D feature extraction, 3D feature inversion, and 3D semantic segmentation. It embeds a vorticity prediction branch for physical constraints and uses U-Net-like architectures with spatial and channel attention mechanisms. The loss function combines weighted Dice loss and a vorticity-driven physical loss (mean square error + gradient consistency loss) to ensure high segmentation accuracy and physical consistency. Experimental results show StV-PI-EddyNet model achieves 96.97% global accuracy, 84.80% weighted Dice coefficient, and 0.8955 macro-averaged F1-score. It effectively mitigates boundary distortion issues of existing methods, generating 3D eddy masks with high spatial precision. This model enjoys good scalability and can be further optimized by integrating earth observation advances in architecture design, data sources and application generalization, promoting its in-depth applications in marine and especially underwater dynamic research.

Differing interpretations of the strength of interseismic plate coupling along the Main Himalayan Thrust (MHT), particularly the possible existence of weak coupling zones, have led to widely varying assessments of the associated seismic hazards along the megathrust with potential impacts for more than half a billion people. Here, we focus on the kinematic status of the MHT in the Uttarakhand Himalaya, the location of one of the proposed low-coupling regions. We use an updated compilation of Global Navigation Satellite System (GNSS) data, along with Interferometric Synthetic Aperture Radar (InSAR) satellite imagery, to estimate the slip deficit and characterize the width of the MHT that is accumulating elastic strain. GNSS-derived horizontal displacements indicate a slip deficit of ∼18 mm/year, with an MHT that is locked up to a width of ∼115 km. We use ALOS-2 InSAR imagery to quantify the interseismic vertical deformation in the same region and identify a peak uplift of 4–6 mm/year. We apply the Elastic Subducting Plate Model (ESPM) to characterize the fault slip responsible for this vertical deformation while avoiding the vertical motion artifacts introduced by backslip or deep dislocation models. Both the GNSS and InSAR measurements are consistent with a shallow MHT that is fully locked from the surface to a depth of ∼20 km. Our results also indicate that the megathrust in the Uttarakhand Himalaya is highly coupled (>0.8) and the accumulated strain energy is equivalent to one Mw 8.1 megathrust earthquake every 100 years along this ∼300 km section of the megathrust.

Abstract Chemical exposure causes adverse neurological outcomes, highlighting the need for high-throughput screening of chemical neurotoxicity. However, existing computational methods rely on single-modality inputs and lack complementary molecular information to sufficiently enhance the accuracy and efficiency of neurotoxicity risk assessment. Here, we present NeuroToxPredictor, a scalable and interpretable multimodal framework built on a curated chemical neurotoxicity dataset for accurate and efficient chemical neurotoxicity prediction. It integrates graph attention networks and molecular fingerprints with ChemBERTa, a pretrained chemical language model that provides contextual encoding of SMILES. A gated fusion module adaptively weights these complementary representations according to their predictive relevance to enhance modality-specific weight allocation through a gating fusion mechanism. NeuroToxPredictor achieved competitive performance across five independent training seeds (AUC = 0.919 ± 0.004; accuracy = 0.845 ± 0.014; F1-score = 0.861 ± 0.011; recall = 0.884 ± 0.013; precision = 0.839 ± 0.024; MCC = 0.688 ± 0.028), outperforming both single-modality and reduced-modality baselines. It demonstrates strong generalizability with an AUC of 0.861, as externally validated on an independent set of 279 compounds. NeuroToxPredictor is deployed on a publicly platform (https://www.ai4environ.cn/NeuroToxPredictor), enabling an end-to-end intelligent workflow from chemical retrieval, neurotoxicity prediction, applicability-domain assessment, and large language model-assisted result interpretation.

ABSTRACT Extreme precipitation and atmospheric moisture transport are closely linked components of the Tibetan Plateau (TP) hydroclimate, but statistical covariability should be distinguished from causal control. We analyse extreme‐precipitation and integrated vapour transport (IVT) indices over 1979–2015 using APHRODITE precipitation and ERA5 reanalysis. The precipitation and moisture‐transport products used in the original workflow were evaluated against independent station and radiosonde observations: APHRODITE showed the best overall performance among the evaluated precipitation products, while ERA5‐derived IVT agreed substantially better with radiosonde observations than NCEP–NCAR. Regional‐mean R95p, R10mm and Rx1day increased by 7.02 mm decade −1 , 0.40 days decade −1 and 0.77 mm decade −1 , respectively. IVT90p increased by 1.34 kg m −1 s −1 decade −1 ( Z = 1.79), but the regional‐mean trend did not reach the two‐sided 0.05 significance level; R150IVT and IVTmax likewise showed non‐significant increases of 0.65 days decade −1 and 1.61 kg m −1 s −1 decade −1 . The leading EOF explains 66.0%, 52.6% and 26.3% of the variance in R95p, R10mm and Rx1day, respectively, compared with 23.6% for IVT90p and 21.9% for R150IVT. These results quantify historical spatial and temporal covariability over 1979–2015 without implying hydrological risk, statistical non‐stationarity, or deterministic climate control.

ABSTRACT This study evaluates whether an encoder‐decoder deep neural network with multi‐head attention can bias‐correct and downscale ERA5 daily rainfall to Taiwan's 5‐km TCCIP gridded observations. The Encoder–Decoder with Attention (EDA) model ingests ERA5 rainfall, 10‐m winds, coarsened TCCIP rainfall, and 5‐km topography, is trained with consecutive‐year splits and a weighted MSE loss, and is benchmarked against bias correction spatial disaggregation (BCSD) and a rainfall‐only variant. Across Taiwan's five seasonal rainfall regimes, EDA improves the placement of orographic rainfall, reduces the low‐intensity wet bias in ERA5, and better reproduces RX1day (annual maximum 1‐day precipitation), RX10mm (number of days with daily precipitation ≥ 10 mm), CDD (maximum number of consecutive dry days), and interannual variability than BCSD. The largest gains occur in seasons dominated by monsoon and typhoon forcing, indicating that auxiliary wind information helps the network learn the statistical relationships that better reproduce rainfall patterns associated with synoptic flow interacting with complex topography. The results show that attention‐based statistical downscaling can improve the reproduction of high‐resolution gridded rainfall fields for regions with steep terrain.

ABSTRACT Warm‐season extreme precipitation over North and Northeast China is a major regional hazard and is strongly influenced by large‐scale atmospheric circulation. However, how large‐amplitude wave activity varies among different spatial patterns of extreme precipitation remains insufficiently understood. Using precipitation observations and atmospheric reanalysis data for May–August 1961–2024, this study investigated the spatial patterns of extreme precipitation using a structural self‐organising map and examined their large‐scale circulation characteristics within the local finite‐amplitude wave activity (LWA) framework. Four dominant precipitation patterns were identified, with their primary precipitation centres distributed progressively farther south from Northeast China to the Jiang–Huai region. These meridional differences were accompanied by corresponding shifts in both the mid‐tropospheric troughs and enhanced LWA, with the precipitation centres consistently located on the equatorward side of the intensified LWA. The northern precipitation pattern was associated with a baroclinic trough–ridge configuration, pronounced mid‐tropospheric LWA and moisture convergence involving southwesterly transport via eastern China and southeasterly transport from the East China Sea–Yellow Sea region. In contrast, the southern precipitation pattern was related to a more equatorward and meridionally extended wave structure, enhanced lower‐tropospheric LWA and continuous southwesterly moisture transport from the subtropical oceans towards a zonally elongated precipitation band. The vertical LWA analysis revealed a bimodal structure with upper‐ and lower‐tropospheric maxima in both patterns, and the southern pattern featured more pronounced lower‐tropospheric LWA but weaker middle‐tropospheric LWA than the northern pattern.

ABSTRACT In August 2022, the South Asian High (SAH) deviated from its typical northwest–southeast displacement pattern and exhibited a record‐breaking northeastward shift over the period 1980–2024. Accompanying this displacement, a deep and persistent high‐pressure centre developed over mid‐latitude East Asia, directly leading to extreme heat over the Yangtze River Valley (YRV) in August 2022. Our results indicate that the northeastward shift of the SAH is significantly associated with the negative phase of the Silk Road Pattern (SRP) wave train that is a major atmospheric dynamical process over summer Eurasia. Compared with the weak cold sea surface temperature anomalies (SSTA) over the tropical Indian Ocean and Pacific, the unprecedented negative‐phase SRP played a more critical role in driving the shift of the SAH in August 2022. Furthermore, the strongest warming over the mid‐latitude North Atlantic contributed to triggering and maintaining the unprecedented negative‐phase SRP since 1980, ultimately facilitating the occurrence of the record‐breaking northeastward shift of the SAH in August 2022. The extreme warming in the mid‐latitude North Atlantic is attributed to the northward shift and strengthening of the Gulf Stream. This study provides a novel explanation of the linkage between the Gulf Stream and the northeastward shift of the SAH in August 2022 via the warming North Atlantic, which carries a potential source of predictability of the SAH location and climate change in East Asia.

ABSTRACT Topographic conditions are key factors governing surface energy distribution in high‐alpine mountainous areas, determining the dynamic variations of local climate and the thermal state of permafrost, which holds profound implications for understanding global climate system responses in extreme terrestrial environments. However, the in situ thermal impact mechanisms on air temperature and permafrost under complex topographic configurations require further investigation. Based on a decadal (2015–2024) in situ observational dataset of near‐surface air temperature and deep ground temperature on the southern and northern slopes of the Kunlun Mountains, this study evaluates the asymmetric variations and decoupling effects between near‐surface thermal conditions and the underlying permafrost. The results demonstrate that: (1) Comprehensive topographic configurations modulate the expected distribution patterns of air temperature. Influenced by cold‐air pooling, the cold‐season air temperature on the southern slope remains consistently lower than that on the northern slope. However, such near‐surface atmospheric thermal characteristics do not strictly determine the subsurface thermal evolution. (2) The near‐surface air temperature and ground thermal processes are decoupled. Although the Thawing Degree Days (TDD) are comparable between the two slopes, the Ground Thawing Degree Days (GTDD) on the southern slope is approximately four times that of the northern slope. Under conditions of intense heat input, the downward geothermal flux on the southern slope noticeably deepens the permafrost Active Layer Thickness (ALT). (3) Against the backdrop of generalized climate warming, the high‐intensity effective heat input during the warm season drives permafrost degradation. By 2024, the ground temperatures at depths of 8 and 15 m on the southern slope increased by 0.21°C–0.29°C and 0.19°C–0.20°C, respectively, compared to 2015, which noticeably exceeds predicted levels. This reflects the high sensitivity of local permafrost to heat input. This study provides critical in situ evidence for correcting topographic thermal offsets in high‐alpine land surface models.