Earth and Environmental Sciences

Land Use and Land Cover (LULC) mapping is a fundamental task in Earth Observation (EO). However, current LULC models are typically developed for a specific modality and a fixed class taxonomy, limiting their generalizability and broader applicability. Recent advances in foundation models (FMs) offer promising opportunities for building universal models. Yet, task-agnostic FMs often require fine-tuning for downstream applications, whereas task-specific FMs rely on massive amounts of labeled data for training, which is costly and impractical in the remote sensing (RS) domain. To address these challenges, we propose LandSegmenter, an LULC FM framework that resolves three-stage challenges at the input, model, and output levels. From the input side, to alleviate the heavy demand on labeled data for FM training, we introduce LAnd Segment (LAS), a large-scale, multi-modal, multi-source dataset built primarily with globally sampled weak labels from existing LULC products. LAS provides a scalable, cost-effective alternative to manual annotation, enabling large-scale FM training across diverse LULC domains. For model architecture, LandSegmenter integrates an RS-specific adapter for cross-modal feature extraction and a text encoder for semantic awareness enhancement. At the output stage, we introduce a class-wise confidence-guided fusion strategy to mitigate semantic omissions and further improve LandSegmenter’s zero-shot performance. We evaluate LandSegmenter on six precisely annotated LULC datasets spanning diverse modalities and class taxonomies. Extensive transfer learning and zero-shot experiments demonstrate that LandSegmenter achieves competitive or superior performance, particularly in zero-shot settings when transferred to unseen datasets. These results highlight the efficacy of our proposed framework and the utility of weak supervision for building task-specific FMs. The code and dataset are publicly available at https://github.com/zhu-xlab/LandSegmenter.git .

The Red Sea crisis has severely impacted global shipping networks, yet its causal effects on port operations remain inadequately quantified. This study examines 16 major container ports along the Red Sea-Mediterranean route, employing the difference-in-differences method to empirically assess the net impact of the crisis on port container throughput. The analysis covers the period from Q1-2021 to Q4-2025, encompassing 672 port-quarter observations. The baseline DID regression reveals a core interaction coefficient of -0.295 after controlling for port fixed effects, time fixed effects, and a series of covariates, indicating an average 29.5% decline in throughput at treatment group ports — a result that is both statistically significant (p < 0.01) and economically meaningful. Dynamic effect analysis confirms the parallel trends hypothesis, with the negative impact intensifying post-crisis. PSM-DID tests, placebo tests, and tail-trimming procedures all validate the robustness of the findings. Heterogeneity analysis shows that hub ports experienced only approximately 43% of the impact compared to non-hub ports, while ports in low-political-risk regions demonstrated stronger resilience. This study employs rigorous causal inference methods to quantify the dynamic cumulative effects of the Red Sea crisis on port operations, providing empirical evidence and decision-making references for port resilience to geopolitical risks.

Sea surface height around Taiwan Island shows complex multi-scale variability, posing a major forecasting challenge. To address this, the Taiwan Island Adjacent Seas Sea Level AI Forecaster (TAS-SLAF), a deep learning framework integrating convolutional layers, ConvLSTM, and the Convolutional Block Attention Module (CBAM) to capture spatiotemporal features, is applied for 15-day sea level anomaly (SLA) prediction in this sea area. Fed with 15-day combined SLA, sea surface wind speed and ocean current data, the TAS-SLAF achieves root mean square errors (RMSEs) of 2.38 cm, 4.83 cm and 5.95 cm on the 5 th , 10 th and 15 th forecast days, respectively, with higher RMSEs in shelf/slope regions than in the open ocean. Over a 3–10 day horizon, it reduces RMSE by 11.46%–50.76% compared to leading numerical models (GOPAF and ESPC-D-V02), with greater improvement in the open ocean. The TAS-SLAF also performs robustly during large Kuroshio intrusion events northeast of Taiwan Island, showing better agreement with observations than reanalysis data. This study highlights the TAS-SLAF’s application value in improving sea level predictability around Taiwan Island with markedly error reduction.

Mangrove ecosystems face critical challenges from climate change, anthropogenic pressures, and fragmented conservation paradigms. This study constructs a large, model-driven, multi-agent framework for mangrove protection and sustainable development in the China-ASEAN region, using Guangxi mangrove protection as an empirical case to explore the application pathways of technological innovation in cross-border ecological cooperation. Through the integration of literature review, case studies, and policy simulation, this study systematically analyzes the limitations of traditional mangrove protection paradigms and proposes a three-level “Perception-Decision-Execution” collaborative multi-agent architecture. It is projected that the application of this framework in typical regions, such as the Beilun Estuary and Qinzhou Bay in Guangxi, will lead to significant improvements in the mangrove ecosystem recovery rate, increased community economic income, and enhanced annual carbon sequestration. This study innovatively constructs a “Technology-Theory-Practice” three-dimensional collaborative theoretical system, providing a novel theoretical analytical framework for mangrove protection and sustainable development, and explores how emerging technologies can enhance mutual trust among China-ASEAN partners, particularly through blockchain-enabled transparent record-keeping and federated learning that respects data sovereignty. The policy recommendations section combines recent international developments, including the 2025 UN Ocean Conference and the International Court of Justice’s advisory opinion, to propose forward-looking and actionable policy frameworks. These technological foundations are critical for advancing bilateral marine cooperation and governance in the region. This study provides innovative practical pathways for promoting China-ASEAN marine economic cooperation and achieving the United Nations Sustainable Development Goals (SDGs), particularly “Life Below Water” (SDG 14) and “Climate Action” (SDG 13).

With climate warming, storm surge increasingly threatens coastal cities, underscoring the necessity for accurate and timely forecasts of peak surge height and timing. Among widely used neural network-based approaches in storm surge forecasting, the Gated Recurrent Unit (GRU), while strong in sequential prediction, tends to underestimate peak magnitudes and exhibits timing errors at extended lead times. To address the issue, this study developed a single-station GRU prediction model using storm surge data from 42 tropical cyclones (2000-2023) at nine stations in the Pearl River Estuary. To reduce peak underestimation, an iterative forecasting scheme was designed, incorporating tropical cyclone (TC) parameters selected via Accumulated Local Effects (ALE) and correlation coefficient. To mitigate peak timing errors, a Support Vector Regression (SVR) model based on TC peak intensity elements was proposed to predict and correct peak timing errors. Additionally, multi-station joint forecasting was explored by embedding spatial information between stations and TC tracks. Results based on leave−one−out cross−validation (LOOCV) indicated that integrating TC location, distance, and 34−kt wind radii in the key quadrants yielded a 27% reduction in Peak Error (PE) at 12–18 h and a 13% reduction in RMSE over 1–18 h for the 35 test cases. After applying the SVR−based phase correction, the forecast for all cases achieved an 55% reduction in Timing Error (TE) and a 2.81% reduction in RMSE relative to the uncorrected forecast. Compared to numerical model outputs, the RMSE of water level sequence during Typhoon Hato was decreased by 0.22 m. Multi-station joint forecast results showed that embedding such spatial information not only enhances forecast skill across multiple gauges but also provides benefits for stations with sparse historical data. This study offers a promising method that targets both the magnitude and timing of storm surge peaks, thereby improving the accuracy and reliability of storm surge forecasting.

Small-scale livelihood systems are recognized for simultaneously providing subsistence, employment, and income to the same communities and households. Small-scale fisheries exemplify this multifunctionality, deriving a comparative advantage over industrial fleets that maximize single-output efficiency. Yet prevailing frameworks acknowledge these multiple outcomes without operationalizing their interdependencies, leaving undiagnosed whether livelihood functions operate as a coupled system or in isolation. We introduce Functional Integrity—adapted from ecological integrity—to diagnose whether a livelihood system’s multiple functions operate as a coherent whole, operationalized through Level (average performance) and Balance (cross-functional evenness). Applied to 136 national fisheries systems as a diagnostic window on small-scale-anchored livelihood configurations, using five dimensions from the Illuminating Hidden Harvests and Ocean Health Index datasets, we find no significant trade-offs after controlling for GDP: all partial correlations are positive, although zero-order coupling strength varies across dimensions, and countries cluster into four distinct functional profiles. Governance quality is the most consistent positive institutional predictor of functional integrity, reaching conventional significance in Models A–C and remaining marginal in Model D (p = 0.054), with its association clearest in Level. Export orientation is associated with lower Balance (p = 0.005) without lowering Level, tilting livelihood profiles toward greater unevenness while leaving overall performance intact. Small-scale employment share shows no independent effect (p = 0.603). These findings suggest that the multifunctional realization of small-scale livelihoods is more institutionally conditioned than structurally guaranteed in these cross-national models, with implications for other contexts where multiple livelihood functions converge.

The Yangtze River Estuary is a highly dynamic river–sea transition zone in which freshwater discharge, tidal exchange, saltwater intrusion, plume dispersion, turbidity, and shallow-water habitat mosaics jointly shape fish assemblages. To provide a synoptic autumn baseline for this estuarine–coastal system, we analysed bottom-trawl survey data collected at 89 stations during October–November 2024 across three spatially defined sectors: the Yangtze River Estuary moratorium zone (EMZ), offshore waters outside the estuary mouth (OFF), and the northern waters of Hangzhou Bay (HB). Species composition, dominant taxa, catch density, catch per unit effort (CPUE), station-level alpha diversity, abundance–biomass comparison (ABC) curves, and Bray–Curtis-based multivariate structure were compared among sectors. A total of 71 fish species belonging to 12 orders and 29 families were recorded. Perciformes was the most species-rich order, and Collichthys lucidus was the only dominant species shared by all three sectors. The EMZ showed the highest abundance density (24.49 × 10³ ind./km²), biomass density (216.03 kg/km²), and mean CPUE (7.9 kg/h), whereas OFF had lower density values and the most negative ABC pattern. Station-level alpha diversity differed significantly among sectors (Kruskal–Wallis, all P < 0.001), with the EMZ showing the lowest diversity and evenness. Cluster analysis, NMDS, and PERMANOVA revealed statistically significant multivariate separation both within and among sectors. These results indicate that autumn fish assemblages in the Yangtze River Estuary and adjacent waters are strongly structured by spatial heterogeneity along the estuarine–offshore gradient. The higher fish density observed within the EMZ should be interpreted as a spatial ecological pattern of the inner estuary rather than as direct evidence of management effectiveness. The consistently negative ABC curves across all sectors further suggest that the assemblages remain dominated by small-bodied fishes, highlighting the need to evaluate density, size structure, and community composition jointly. This study provides an updated autumn reference for future monitoring and ecosystem-based management in the Yangtze estuarine–coastal system.

The global decline of marine biodiversity has intensified the need for international efforts to expand and strengthen marine protected areas (MPAs). In Macaronesia, tourism represents a key driver of economic growth and social prosperity. However, a rapid increase in visitor numbers, combined with ecological sensitivity inherent to island systems, has heightened pressure on marine environments and highlighted the need for sustainable tourism models. Ecotourism represents a potential way to balance conservation, economic development, and community well-being, especially within MPAs. Meanwhile, the success of this balance depends heavily on effective governance; yet limited attention has been paid as to how governance structures in Macaronesia influence the integration of ecotourism into MPAs. This study analyses the role of institutional and legal governance in supporting ecotourism development in the MPAs of the four Macaronesian archipelagos (Azores, Madeira, the Canary Islands and Cabo Verde). A comparative qualitative analysis of institutional and legal governance frameworks was conducted for each archipelago, examining relevant legislation, institutional structures and maritime tourism regulations to identify strengths, limitations and administrative conditions affecting the integration of ecotourism in MPAs. The analysis reveals varying governance capacities across the archipelagos, all of which are characterised by complex legal systems and fragmented management of MPAs and maritime tourism. Despite these barriers, ecotourism remains a viable pathway within Macaronesia’s MPAs, provided that the governing institutional and legal structures are refined to prioritise conservation outcomes.

Kelp forests depend on the successful settlement of microscopic zoospores, the motile dispersal stage of the kelp life cycle. Settlement begins with arrival at a surface and culminates in stable attachment through adhesive contact and curing. This is a critical transition between zoospore dispersal and gametophyte development, during which individual swimming behavior and zoospore traits may shape settlement success. Despite its importance, settlement metrics are rarely considered in kelp ecology, restoration, and ecotoxicology frameworks. Here, we explore and quantify how zoospore swimming behavior relates to settlement in three habitat-forming kelp species, Laminaria hyperborea , L. digitata , and Saccharina latissima (Laminariales, Phaeophyceae). Zoospores were cultured under controlled conditions and analyzed using high-resolution video microscopy at both individual zoospore and population levels. Across species, motile fraction declined with culture age, and cultures with higher proportions of motile zoospores exhibited substantially greater settlement. Motile fraction exerted a strong multiplicative effect on settlement, independent of species, season, and site effects. Species also differed significantly in zoospore size, aspect ratio, and surface ornamentation, revealing a previously under-quantified axis of early-life functional differentiation. Integrating swimming speed and directional persistence of individual zoospores into an encounter metric revealed a clear behavioral hierarchy, with highly progressive trajectories exhibiting the greatest encounter potential and immobile zoospores the least. These results provide a mechanistic link between zoospore swimming behavior, encounter rates, and settlement success, demonstrating that a simple motile/non-motile classification captures the likelihood of settlement, while kinematic metrics explain how sublethal impairment of motility translates into reduced recruitment. Therefore, zoospore motility and morphology are sensitive, mechanistically-grounded functional metrics for kelp ecology, restoration, and ecotoxicology.

Introduction Estimating the genetic diversity of cetaceans at sea, particularly abundant social delphinids, can be difficult with traditional biopsy sampling of individuals. Environmental DNA (eDNA) metabarcoding has been shown to be a powerful tool for the identification of species assemblages and estimation of genetic diversity, especially in aquatic environments. Methods We collected 126 samples of seawater from within the immediate vicinity of schools of dolphins during 15 encounters with the four most common delphinid taxa in the waters around Santa Catalina Island, California, USA: long-beaked common dolphins ( Delphinus delphis bairdii , n = 8), short-beaked common dolphins ( D. d. delphis , n = 3), common bottlenose dolphins ( Tursiops truncatus , n = 2), and Risso's dolphins ( Grampus griseus , n = 2). Next-generation sequencing was used to assign Amplicon Sequence Variants (ASVs) of mitochondrial DNA to species using GenBank and a region-specific reference database. Results A total of 240 ASVs were resolved for the four species. ASV richness and the effective number of ASVs, or true diversity, measured as Hill numbers of order 1, were consistent with known characteristics of the four species. Despite collecting up to 12 samples from a single group, a rarefaction analysis indicated that the population diversity was not fully represented for the more abundant species (genus Delphinus), but were closely approximated for G. griseus . Discussion This study demonstrates the application of eDNA for estimating population genetic diversity of abundant species and makes recommendations for improving future studies to better capture this diversity in wild delphinid populations. This provides a more solid foundation for studies using eDNA to monitor these species, which often include those in close proximity to anthropogenic threats.

Danish coastal waters were severely affected by eutrophication in the 1980s, prompting the implementation of national nutrient reduction plans. The aim of this study was to explore the recovery of Danish coastal waters and report trends for selected indicators of eutrophication. Dissolved inorganic nitrogen (DIN), dissolved inorganic phosphorus (DIP), chlorophyll a, light penetration depth, eelgrass main depth limit, a benthic fauna index (DKI), and bottom water oxygen concentration. We report on the temporal and spatial trends of eutrophication in 109 coastal waterbodies from 1980 to 2023. The analysis revealed that although improvements have been observed in several waterbodies, eutrophication continues to affect large areas. The national eutrophication status improved until the early 2000s, followed by stagnation. This trend was primarily driven by changes in winter DIN and DIP concentrations, whereas oxygen concentrations, eelgrass main depth limits, and DKI remained relatively stable throughout the study period. The means of chlorophyll a and light penetration depth showed signs of recovery until 2012 and 2021, respectively, followed by trend reversals. These findings suggest that reductions in nutrient levels have been pivotal in driving Danish coastal waters towards recovery; however, current mitigation efforts are insufficient to ensure that Danish coastal waters can achieve status as ‘not affected by eutrophication’.

Solenogastres (Mollusca, Aplacophora) are shell-less, vermiform marine molluscs found across a wide range of marine habitats and depths. Despite their broad distribution, their diversity remains poorly documented. Because Solenogastres tend to be externally cryptic, morphological identification requires trained specialists and time-consuming processes (e.g., histology). DNA barcoding has emerged as a powerful tool to complement traditional taxonomy, yet Solenogastres remain underrepresented in public genetic datasets. Here, we present the first global DNA barcode reference library for Solenogastres, comprising 655 newly generated sequences combined with previously published sequences, resulting in 531 COI, 483 16S, and 129 CytB barcodes representing an estimated 325 aplacophoran species (287 solenogaster species) from 642 specimens. The included taxa broadly encompass the diversity and geographic distribution of the group. Specimen identification was achieved through an integrative approach that combined molecular data with morphological information from histology, light microscopy, and scanning electron microscopy. Phylogenetic analysis of the concatenated dataset recovers major clades consistent with transcriptome-based studies and reveals several lineages that may require taxonomic revision. This curated dataset, available on GenBank, provides an essential resource for future efforts in systematics, ecology, and biodiversity research of Solenogastres.

Hydroacoustic remote sensing represents a non-invasive, repeatable approach to monitoring benthic communities, supporting sustainable management, conservation efforts, and the detection of environmental change. Mussel clusters on a sandy seafloor in the Oder Bank area (10 to 15 m water depth, southwestern Baltic Sea) were detected in high-frequency backscatter data recorded with a Norbit STX multibeam echosounder in 2019. The blue mussel ( Mytilus edulis ) complexes appear as narrow bands in backscatter mosaics, showing average backscatter intensity increases between 0.2 and 1.0 dB, with localized peaks of up to 2 dB compared to the surrounding sand. Mussel coverage, verified by underwater video sledge observations, reaches up to 50% in isolated patches but typically remains below 15%. The acoustic response of the mussel clusters shows a weak to moderate yet significant correlation with mussel cover, independent of frequency (tested at 200, 400, and 700 kHz). This response is observed at incidence angles greater than 40°. A persistent shell hash layer found at 8 cm depth in sediment cores was not detected acoustically at any frequency. Comparison with data from 2024 suggests that these mussel clusters are ephemeral. Due to the shallowness of the Oder Bank they are influenced by natural processes such as wind and current-driven circulation. This causes the Mytilus bands to roll back and forth on the sand, but can also lead to the dissolution and dispersal of the clusters. The mussel complexes cannot be reliably captured using point-based sampling and short video transects. • Mytilus edulis clusters on sand are detectable in multibeam backscatter data. • Change in mussel abundance explains changes in backscatter intensity. • The clusters are short-lived features and not detected in repeated surveys.

Under future climate change, seagrasses may benefit from the effect of ocean acidification, although their response may be species-specific and may also depend on several ecological factors. Using a field manipulative approach, this study investigated the morphology and physiology of the seagrass Zostera noltii under the effect of increased CO 2 concentrations, taking advantage of the presence of natural shallow submarine CO 2 . Comparisons between transplanted plants from control pH sites to low pH site with translocated control pH plants allowed investigating the short-term acidification effect. Moreover, an observative approach was used to assess the long-term acidification by comparing untouched plants from control pH with the low pH site. The photosynthetic pigments were increased under short-term and long-term acidification. Furthermore, the above-ground morphology of the seagrass may be improved by the short-term acidification, although these differences were no more evident on the long-term acidification. Therefore, Z. noltii could respond positively to the low pH conditions that will occur in the future, as the observed rapid plasticity response may lead to an initial acclimation. • Zostera noltii response under natural field CO 2 increase has been assessed • Short-term acidified plants exhibited higher morphology and photopigments • Long-term acidified and control plants showed a similar morphology • The higher plant performance to short-term acidification was attenuated through time

Loss of sea ice is both a response to climate change and a driver of further warming, due to the lower albedo of open water. Covering strategic areas of Arctic sea ice with a thin layer of Hollow Glass Microspheres (HGMs) has been proposed as an intervention to slow, or even reverse, loss of ice. Toxicological studies have identified effect limits for HGMs in both pelagic and benthic species. The current study provides context for those toxicological studies, by estimating environmental concentrations if HGMs are deployed as suggested. Two different models were used, a vertical-dimension water-column model and a three-dimensional transport model. With the water-column model, we assume an initial surface concentration of 70 g/m 2 of HGMs, and horizontally homogeneous conditions over large areas. HGMs were found to distribute throughout the surface mixed layer with concentrations from around 5 mg/L at the surface, to below 1 mg/L at 25 m depth. In the three-dimensional model, we considered HGMs applied to an ice floe of area 20 km 2 , to correspond to a field trial in the Arctic. Near the initial location, average concentrations in the top 20 m of the water column were around 0.5 mg/L after 5 days. In both cases considered, concentrations of HGMs in the surface mixed layer are such that adverse effects on copepods are likely. Given that the melting season in the Arctic also coincides with a period of high biological activity, it is possible that application of HGMs could have detrimental effects on the Arctic ecosystem. • Use of HGMs to reduce melt by increasing the albedo of sea ice, has been proposed. • We model the fate HGMs in the environment, and predict likely concentrations. • Tox studies found effects on benthic and pelagic species at predicted concentrations.

-derived OM from rainforest or mangroves and upstream anthropogenic inputs. The Bayesian mixing model (MixSIAR) indicated clear contrasts in OM sources between basins. In the CR basin, sedimentary OM was predominantly derived from mangroves (median = 0.76), with a secondary contribution from sewage (0.18). In contrast, in the PSR basin, sewage was the dominant OM source (0.65), followed by mangrove contributions (0.28). Organic pollutants correlated with OC, showing that contaminant distribution is partly governed by OM dynamics. PAH diagnostic ratios revealed the sewage, biomass/coal burning as the main source in PSR, with additional fossil-fuel inputs, while CR reflected mixed combustion sources linked to agricultural fires and boat traffic. Organochlorine patterns in CR suggest historical pesticide use and diffuse atmospheric input, while PSR shows lindane and local/industrial influences. These results highlight contrasting OM dynamics under different land-use intensities and the dual role of estuaries in storing carbon and pollutants.