Unmanned sailing vessels commonly rely on real-time wind-field measurements as key inputs for multi-level decision-making, including path planning and control. During long-endurance offshore missions, however, wind-sensing systems may be affected by harsh environmental conditions, external mounting constraints, and maintenance difficulties. These issues motivate complementary control strategies that can support autonomous navigation while reducing dependence on real-time wind-measurement channels. In this study, we formulate sail–rudder control under partial observability, where the policy uses onboard observable vessel states and target-relative information without receiving measured wind speed or wind direction. A recurrent SAC-LSTM policy is employed as a wind-input-free control implementation, in which a Long Short-Term Memory (LSTM) module is incorporated into the Soft Actor-Critic (SAC) framework to encode recent observation–action history. This design is used to examine whether temporal context from vessel-motion evolution and previous sail–rudder actions can support continuous sail–rudder decision-making when direct wind variables are excluded from the policy input. In randomized-wind simulations evaluated over five independent training seeds, SAC-LSTM achieved a task success rate of 99.36%, compared with 39.64% for the memoryless wind-input-free SAC baseline and 95.24% for the wind-informed SAC reference. Under the same wind-input-free current-observation setting, SAC-LSTM also reduced the average successful-episode navigation time and rudder-angle variation rate relative to the memoryless SAC baseline. In the present randomized-wind benchmark, SAC-LSTM obtained a higher task success rate than the wind-informed SAC reference without using measured wind speed or wind direction, while showing similar successful-episode navigation time and trajectory length. These results indicate that recurrent observation–action history can provide useful temporal context for wind-input-free sail–rudder control under the tested 4-DOF still-water simulation conditions.

Remote sensing reflectance (R-rs) is the first-level product of atmospheric correction in ocean color remote sensing, and it directly affects the accuracy of subsequent water color inversion products. The bidirectional effects of R-rs have been primarily explored in phytoplankton-dominated seawater, while limited efforts have been made to explore turbid water rich in sediments. In this article, based on the R-rs simulated using Hydrolight with the solar zenith angle (theta(s)) in a range of 0 degrees-75 degrees, viewing zenith angle (theta(v)) from 0 degrees to 87.5 degrees, relative azimuth angle (phi) from 0 degrees to 180 degrees, and wavelength in 400-900 nm, the bidirectional effects of R-rs in varying optical water types from clear to sediment-dominated turbid water were quantified. Generally, the greater bidirectional effects tend to occur at larger sun-viewing angles, clearer water, and shorter wavelengths. Among various factors affecting bidirectional reflectance effect, the angles have the greatest influence, among which, theta(v) has the largest mean relative difference (RD) of 13% +/- 11%, followed by phi (12% +/- 4%) and theta(s) (11% +/- 3%). The wavelength and water types account for a mean RD of 11% +/- 1%. In turbid water, bidirectional effects at red and near-infrared bands are also necessary to be considered. Furthermore, two commonly used bidirectional correction methods of MAG2002 and Lee2011 were applied and examined in this study. The Lee2011 generally outperforms MAG2002 from clear to turbid water. However, the Lee2011 tends to deteriorate at large solar-viewing angles and long wavelengths in turbid water with high-load sediments. This study contributes to understand the comparability of remotely sensed data measured under varying solar-viewing geometries and optical water types and can help in further reducing the uncertainty of bidirectional effects in coastal water color remote sensing.

During 1982-2018, eight significant spring marine heatwave (MHW) events occurred in the Yellow and East China Seas (YECS), each lasting 15-63 days. Anomalous descending motions and anomalous surface winds were the main local drivers of these MHWs. Anomalous descending motions increased incident shortwave radiation, and anomalous surface winds reduced oceanic latent heat loss and shoaled the mixed layer (ML). The resultant positive net surface heat flux (Q) anomalies directly warmed the ML, driving MHWs. In addition, the shoaled ML amplified the warming effects of both positive Q anomalies and positive climatological mean Q. Positive Q anomalies could also indirectly contribute to MHWs by shoaling the ML. Further analyses identified four types of large-scale atmospheric drivers that were important to these spring MHWs. Westward extension of the northwestern Pacific subtropical high and enhanced convection in the Indian Ocean and the Maritime Continent triggered the MHWs in the southern East China Sea (ECS). Atmospheric wave trains from the North Atlantic to the YECS played important roles in causing MHWs in the Yellow Sea and the northern ECS. Furthermore, enhanced atmospheric latent heating in the northern South China Sea and the subtropical northwestern Pacific supported the MHW around the boundary between the Yellow Sea and the ECS. These findings highlight key physical predictors that can improve the forecasting of YECS spring MHWs.

Digital holographic imaging has become an effective technique for plankton observation due to its capability of capturing three-dimensional information with a large depth of field. However, holographic images contain complex interference patterns and noise that differ significantly from conventional microscopic images, making it difficult to directly apply existing datasets and deep learning models trained on microscopy imagery. Moreover, the scarcity of annotated holographic datasets further limits the development of automated detection methods. To address these challenges, this paper constructs a synthetic plankton hologram dataset generated through diffraction propagation simulation and proposes a lightweight detection framework named YOLO-Plankton. The proposed model introduces an improved convolutional block attention module to enhance feature representation and suppress background interference in holographic images. In addition, a lightweight detection head called LiteSPHead is designed to reduce computational complexity while maintaining effective detection capability. Experimental results show that YOLO-Plankton significantly outperforms the baseline YOLOv8n and several mainstream detection models, with the mean average precision (mAP) on the synthetic dataset increasing from 0.688 to 0.939, and demonstrates robust generalization when tested on images fused with real backgrounds.

High-frequency variations in bottom layer dissolved oxygen (DO) critically influence marine ecosystem metabolism and benthic habitats, yet their controlling mechanisms and spatial heterogeneity in shallow coastal oceans remain poorly understood. Based on four years of in-situ observations and high-resolution numerical simulations, this study systematically investigates the spatial diversity of high-frequency dynamics of bottom layer DO concentrations in the coastal oceans of China. The results show that the high-frequency fluctuations of bottom layer DO concentrations are most pronounced in summer and have two dominant frequency bands at diurnal and semi-diurnal periods. Further analyses indicate that bottom-layer DO concentrations are primarily influenced by diurnal variations in shallow waters with low turbidity, whereas semi-diurnal variations exert the strongest influence in high-turbidity environments. Model experiments further confirm turbidity as the key factor shaping the spatial diversity of high-frequency bottom DO dynamics by regulating underwater light availability and the relative importance of biogeochemical and physical processes. Under low-turbidity conditions, sufficient light penetration enhances near-bottom photosynthesis, leading to diurnal dominance. In contrast, under high-turbidity conditions, strong light limitation suppresses biological regulation, allowing semi-diurnal tidal mixing to become the dominant driver of bottom-layer DO variability. Overall, this study underscores the pivotal role of turbidity in mediating the balance between tidal dynamics and biogeochemical processes, offering new insights into high-frequency oxygen variability in shallow coastal oceans.

Efficient in-situ detection of marine microplastic pollution remains a critical challenge in environmental monitoring. While most existing detection technologies are lab-based and highly accurate, they are also time-consuming, costly, and labor-intensive. In contrast, portable aquatic microplastic detection systems can significantly reduce these overheads. Holographic polarimetry is a promising method with the potential for in-situ detection. However, current research in this area has largely focused on the laboratory-based analysis of microplastic features or utilized networks too computationally intensive for deployment on edge devices. This study aims to enable high-throughput, in-situ detection for multiple classes of microplastics. To this end, we built a portable holographic polarimetric microplastic imager (HPM imager) and developed WMViT3, a lightweight feature extraction model for microplastic classification. Unlike traditional methods, our approach bypasses complex reconstruction by directly decoding the raw polarimetric interference fringes. The model integrates a wavelet transform specifically designed to capture the high-frequency oscillatory features of holographic fringes, effectively extracting unique "optical fingerprints" of different materials. It achieves a classification accuracy of 97.97% on our standardized HPM-500 dataset while reducing computational load by 53.2%. Crucially, this efficiency enables real-time inference on embedded devices, providing a viable technical solution for portable, real-time environmental monitoring systems.

Introduction: By using the latest patents and technology progress in maritime hydrodynamics, this research aims to improve the maneuvering ability of an Unmanned Surface Vehicle (USV) through the method of carrying out systematic optimization on rudder configuration parameters. Methods: Numerical simulation works were carried out via Unsteady Reynolds-Averaged Navier– Stokes (URANS) equations coupled with shear transport k-ω turbulence model, which was implemented through Computational Fluid Dynamics (CFD) software STAR-CCM+ 2310. Results: The main parameters: horizontal position, vertical position, aspect ratio, and area were optimized, which led to a decrease in the turning radius by 21.2% with an insignificant effect on propulsive performance. Discussion: Such results prove that the specified parameter optimization of the rudders may enhance the level of steering greatly without resorting to deteriorating propulsion efficiency. Conclusion: The rudder design optimization proposed methodology is a transferable and robust proposal applicable to the USV and could be extrapolated and applied to other marine systems.

Mesoscale eddies profoundly modulate air–sea turbulent heat fluxes (THF). Their long-term trends and drivers under a changing climate remain uncertain. Using six high-resolution flux products, satellite SST datasets and a global eddy trajectory atlas, we isolate mesoscale thermodynamic footprints from the large-scale background. We reveal that eddy-induced THF is disproportionately concentrated in the mid-latitude SST fronts, especially in the western boundary currents and Antarctic Circumpolar Current, where eddies contribute 10–30% of climatological THF. These hotspots exhibit a robust, dataset-independent intensification, with decadal growth rates surging to 28–43% in the Gulf Stream, significantly outpacing the global mean (~20%). Linearized attribution indicates that this intensification is dominated by ocean-surface SST and humidity anomalies associated with sharpening of background SST gradients, rather than by enhanced eddy kinetic energy. Our results identify frontal sharpening as a key control on mesoscale air–sea heat exchange in a warming climate.

Although the frequency and intensity of marine cold spell (MCS) events in the South China Sea have declined under global warming, their impacts on regional ocean environments and ecosystem processes remain substantial. This study examined how cyclonic mesoscale eddies (CEs) modulate the vertical structure of MCSs, using a combination of satellite observations, reanalysis datasets, and in situ temperature profiles based on data acquired from 1993 to 2022. Using the self-organizing map classification method, two typical vertical structures of MCS within CEs were identified: shallow MCSs, characterized by surface cooling that weakens with depth, and subsurface-intensified MCSs, featuring a pronounced cold anomaly peaking near the thermocline. The radial position of MCSs relative to an eddy center was found to be the dominant factor controlling the vertical cooling intensity—subsurface anomalies become more pronounced when an MCS occurs closer to the eddy core. Further analysis suggested that this subsurface cooling is primarily driven by an uplift of the thermocline associated with strong upwelling near the center of a CE. These findings reveal the important role of mesoscale eddy dynamics in shaping extreme cold events and will enhance our understanding of the subsurface extreme low temperatures associated with MCS events.

Intensified acidification has emerged as an increasing threat to the health of global coastal ecosystems,and decreasing pH trends have been observed in Chinese coastal waters.However,the drivers in complex regions like Hangzhou Bay remain poorly understood due to the lack of high-resolution time-series data.Thus,utilizing two years of hourly data(2019-2020)from two buoy stations,this study analyzes the key variables driving pH variability and develops a high-performance support vector regression(SVR)model for pH estimation,which achieves a coefficient of determination(R2)of 0.72 on an independent testing dataset.This model was then used to reconstruct a continuous daily sea surface pH dataset,which reveals a clear seasonal pH cycle with minimums in summer and maximums in winter.This pattern arises from a complex interplay among multiple factors.River runoff is the dominant influence in summer,whereas temperature and biological activity primarily shape the pH patterns in other seasons.The pronounced pH variability and the strong,runoff-driven summer minimum highlight the region's heightened vulnerability to episodic acidification.This work provides key insights into the drivers of sea surface pH in Hangzhou Bay,improving the understanding of the nearshore carbonate system and its ecological responses to global change in a complex estuarine environment.

Hypoxia is a major threat to marine aquaculture, yet operational early warning remains constrained by a fundamental gap: the physiological critical dissolved oxygen threshold ([O-2](crit)) is a laboratory quantity obtained from controlled respirometry and cannot be transferred directly to open marine ranching, where DO, temperature, currents, and fish behavior co-vary continuously. We address this gap by deriving a complementary in-situ indicator from macroscopic fish kinematics. A self-supervised Fish Behavior Classification Model (FBCM) with memory prototypes learns recurrent trajectory patterns without behavior-class labels; expert annotations are used only for held-out evaluation and post-hoc prototype-to-semantics mapping. A Behavioral Critical Dissolved Oxygen Threshold (CDT) algorithm couples these behavioral features with continuous DO records to estimate a Behavioral Critical Dissolved Oxygen Threshold ([O2]behcrit), and a rule base translates DO, abnormal behavior frequency, and event duration into operational warning levels. On in-situ data from the Yutai Marine Ranching, FBCM reached 92.55% multi-class accuracy for Lateolabrax maculatus, outperforming hand-crafted, recurrent, and self-supervised time-series baselines under the same data split. CDT estimated [O-2](crit)(beh) at 3.23 +/- 0.12 mg/L for L. maculatus and 2.86 +/- 0.22 mg/L for Sebastodes fuscescens (leave-one-day-out cross-validation; day-level block permutation p = 0.012 and 0.018). On an August-September temporal hold-out, the behavior-augmented rule issued earlier alerts than a sensor-only severe-DO rule (mean lead 28.0 h vs. 6.5 h) at a lower operational false-alarm rate, though the limited held-out event count requires cautious interpretation. The framework should be regarded as a site-specific behavior-derived early-warning indicator complementary to laboratory [O-2](crit), not a substitute for it.

Body coloration is a highly plastic trait in Plectropomus leopardus, playing essential roles in ecological adaptation and economic value. Spectrum is a pervasive ecological signal influencing coloration in fishes, yet how spectral cues are systemically coordinated across tissues remains unclear. In this study, we exposed P. leopardus to four spectral conditions (white, blue, green and red) and integrated short-term (48 h) and long-term (60 d) transcriptomic analyses of the skin, retina, and brain to elucidate spectrum-dependent cross-tissue regulatory mechanisms. Short-term spectral stimulation was primarily detected by skin-localized opn5, which rapidly activated local circadian components and pigment-migration-related genes, driving immediate color changes. Under prolonged exposure, spectral information was integrated through a cross-tissue vision-neural-skin network, with opn3 acting as a central regulatory node coordinating circadian rhythms, melanin synthesis, carotenoid metabolism and pigment cell differentiation. Although all spectral conditions recruited largely conserved molecular pathways, their activation strengths differed markedly: blue light induced the strongest transcriptional responses and most intense pigmentation, whereas green light elicited minimal pigment-related gene expression and the dullest coloration. Together, our findings reveal an opsin-mediated, spectrum-dependent cross-tissue signaling framework, providing new insights into the molecular basis of color adaptation and guiding spectral optimization in aquaculture.

High spatiotemporal-resolution observations of the oceanic partial pressure of carbon dioxide (pCO2) are essential for advancing our understanding of the global carbon cycle. Nonetheless, oceanic in situ pCO2 sensors mounted on mobile platforms frequently encounter challenges such as a physical response lag when crossing concentration gradients. To address this issue and consider the inherent noise of the sensors as well as data heterogeneity, including inconsistent data lengths, we propose a physics-informed deep learning framework for sensor time-lag correction. A single-layer bidirectional long short-term memory (BiLSTM) network coupled with an attention mechanism was designed to capture transient concentration gradients without oversmoothing the signal. A Kalman filter was then applied as a post-processing module to suppress the prediction noise without inducing a secondary physical lag. The experimental results demonstrate that this framework corrects the phase delay while minimizing the residual errors. Quantitative results show that the proposed method outperforms the physical linear time-invariant (LTI) model, linear autoregressive with exogenous input (ARX) model, and traditional statistical baselines (e.g., support vector regression, SVR). It dynamically corrects phase delay and reduces system response time by more than 80

Traditional zooplankton net sampling provides well-established approaches for species identification and density estimation; however, it is often subject to physical filtration, sample compression, and behavioral avoidance, which can lead to underestimation of small-sized or fragile taxa. Moreover, such methods have limited capacity to resolve fine-scale spatial distributions and aggregation patterns of zooplankton under in situ conditions. In recent years, underwater in situ imaging has emerged as a promising tool for zooplankton investigations owing to its ability to provide non-invasive, real-time observations in the natural environment. In this study, underwater in situ zooplankton imaging data collected during winter 2023 and spring 2024 in the Ningyuan River estuary, Hainan Island, were used to examine seasonal variations in community structure, niche differentiation, and environmental responses of zooplankton in this tropical estuary. The results revealed pronounced seasonal differences in zooplankton communities. Winter assemblages were characterized by higher overall abundance and were dominated by Chaetognatha and Copepods, whereas spring assemblages exhibited lower abundance but a more diverse taxonomic composition, with Copepods and tunicates as the dominant groups. Chaetognatha and Copepods displayed relatively broad niche breadths, while most other taxa exhibited intermediate to narrow niches. Niche overlap and variance-ratio results suggested that overall interspecific association was weak and non-significant in both seasons, with a slight segregation tendency in winter (VR = 0.76) and a slight co-occurrence tendency in spring (VR = 1.79). Multivariate analyses further indicated that water temperature, salinity, dissolved oxygen (DO), and pH were the primary environmental factors associated with the spatial patterns of zooplankton communities. These findings improve our understanding of zooplankton responses to environmental gradients in tropical estuarine systems and highlight the value of integrating in situ imaging into multi-source monitoring frameworks.

The hourly L2-level chlorophyll-a (CHL-a) concentration spatial energy spectra of GOCI-II from 2021 to 2023 are employed to investigate the characteristics of the CHL-a spatial energy spectrum slopes in three regions of the East China Sea, namely nearshore, offshore, and open ocean. The seasonal trends of the spatial energy spectrum slopes are also examined for the nearshore and offshore regions. It is observed that the slopes of the CHL-a spatial energy spectrum are −2 at scales larger than 5 km, whereas at smaller scales, they are −5/3, −1, and −0.3 from the nearshore region to the open sea, respectively. On the larger scales, the spatial energy spectrum slopes are consistent with surface quasi-geostrophic (sQG) theory, but this is not the case on smaller scales. An insufficient regional CHL-a concentration leads to a flattening of the slope at the smaller scales. On the submesoscale, the slope of the nearshore CHL-a concentration spatial energy spectrum is steeper in summer and flatter in winter, a pattern that contrasts with changes observed offshore. This seasonal variation is attributed to the southward flow of ZheMin Coastal Current (ZMCC) during winter, which carries freshwater and enhances the horizontal buoyancy gradient in the nearshore region.

In the past decades, the California Current Ecosystem has experienced intense marine heatwaves, which have induced significant disruptions to local phytoplankton communities. Here, using 30-year cruise observations, we identify a previously undocumented vertical structure in chlorophyll-a concentration response to marine heatwaves, characterized by reductions in the surface layer coupled with increases in the subsurface. By integrating observations and coupled physical-biogeochemical model products, we demonstrate that declines of surface chlorophyll-a are primarily attributed to suppressed nutrient upwelled to the upper ocean. Although surface irradiance increased modestly (+3.5%), light availability in the subsurface layer improved substantially (+21.7%) due to reduced phytoplankton shading at the surface. Concurrent with enhanced lateral nutrient transport, phytoplankton growth at depth was promoted during heatwave events. This study highlights the pivotal role of subsurface phytoplankton dynamics in shaping the vertical chlorophyll-a concentration structure and its variability under extreme events.