Atmospheric correction (AC) of ocean color imagery faces significant challenges in regions with absorbing aerosols, adjacency effects, and optically complex waters, common in coastal and inland areas. Traditional algorithms, which extrapolate the atmospheric signal from near and shortwave infrared to shorter wavelengths, often fail in these conditions. To overcome this, the top-of-atmosphere (TOA) signal undergoes principal component (PC) decomposition, retaining only PCs sensitive to water signals, which allows for precise non-linear mapping of TOA to water reflectance. This methodology is tailored for PACE OCI imagery with a 5 nm resolution across 340 to 895 nm. The algorithm excludes OCI observations affected by strong gaseous absorption, but maps TOA PCs to water PCs defined across all spectral bands, enabling water reflectance estimates over the entire 340-895 nm range. In situ measurements and Hydrolight simulations are used to generate TOA reflectance ensembles, with pixel-wise uncertainty estimation integrated. The method’s performance under various angular and geophysical conditions is theoretically assessed and applied to OCI imagery of diverse oceanic regions. Results show that the PCA-based AC scheme effectively handles complex atmospheric settings, yielding realistic water reflectance. Comparisons with the operational OCI product (version 2) demonstrate improvements, particularly in mitigating negative or excessively low values at shorter wavelengths. Preliminary experimental validation using HyperNav measurements shows acceptable agreement between estimated and measured water reflectance values, but additional matchups are needed to confirm accuracy.

Local area optimizations of ocean color atmospheric correction were investigated for Global Change Observation Mission- Climate (GCOM-C)/Second-generation GLobal Imager (SGLI) observations in the north of Ariake Bay by using in-situ measurements of remote sensing reflectance (Rrs) by AERONET-OC on Ariake-Tower (33.104°N, 130.272°E, operated from 2018 to 2023): (1) dedicated vicarious calibration coefficients derived by Ariake-Tower 's R_rs, were applied to reduce bias error of the R_rs estimate; (2) the atmospheric correction (aerosol reflectance) look-up tables (LUTs) were modified by matching with Ariake-Tower 's Rrs statistically to reduce error dependencies on scattering angles and aerosol types; (3) viewing angle dependent error on R_rs was normalized statistically through a cross calibration with other sensor images; (4) the empirical red and Near-InfraRed (NIR) Rrs estimation was optimized by using inter-wavelength relationship in Ariake- Tower Rrs; and (5) aerosol absorption residuals (Δt_a) affecting on aerosol reflectance (ρa) and water-leaving reflectance (ρw) were estimate by fitting to the nearest day’s Ariake-Tower Rrs scene-by-scene, and extrapolating Δt_a to the surrounding several tens of kilometers. By the corrections of (1) to (4), root-mean-square difference (RMSD) from Ariake-Tower’s R_rs 443 nm could be reduced by about 20%. The direct adjustment (5) improves agreement with most of individual ship observation data, however, we have to take note that some cases apart from the Ariake-Tower became worse when Δt_a cannot be assumed to be constant according to its finer spatial scale than the distance. The point-by-point adjustment can be applied operationally when we can deploy in-situ continuous (sited or moored) R_rs measurements in the target area.

Current satellite radiation products from polar-orbiting sensors like MODIS, SGLI, VIIRS, and OCI provide only daily averaged quantities, missing critical diurnal variations. These fluctuations in light energy, ranging from none to ample throughout the day, significantly influence photosynthetic communities, impacting carbon export, nutrient cycling, and ecosystem functions. Incorporating hourly changes in phytoplankton light absorption and irradiance, as shown with GOCI data, has enhanced estimates of net primary productivity. However, the global oceans cannot be fully observed from a single geostationary platform, especially at high latitudes. EPIC, positioned at the first Lagrange point (L1) about 1.5 million km from Earth, offers a unique advantage by simultaneously capturing the entire sunlit ocean with high temporal resolution, enabling detailed observation of evolving systems and diurnal phenomena. Unlike polar orbiters, EPIC provides better coverage at low and middle latitudes, effectively mitigates Sun glint, and offers adequate views of polar regions. Utilizing EPIC data, we estimate hourly photosynthetically available radiation (PAR) over the global ocean. The algorithm is developed and validated through radiative transfer simulations under realistic conditions. Comparisons with geostationary AHI and GOCI data show strong performance. The diurnal PAR estimates, when combined with other ocean color products, are expected to advance studies of aquatic photosynthesis and biogeochemistry.

Remote sensing satellites generate a large amount of data that is conventionally down-linked to ground stations for further processing, but the time from acquisition until the satellite makes contact to the nearest ground station incurs additional latency, ultimately increasing the delay before the desired processing results are available to stakeholders. For certain applications though, latency is critical as the contained information decays rapidly. For example, the location of a moving ship on satellite imagery becomes outdated after a short time. Moving data processing from the ground to on-board of the satellite has the potential to significantly speed up product delivery. This contribution introduces developments regarding satellite on-board processing of synthetic aperture radar (SAR) data, addressing products concerning maritime safety and security in particular. Focusing of raw SAR data into interpretable images is implemented on specialized hardware to keep within the constraints on-board of satellites. The focused SAR images are processed further using conventional algorithms and artificial intelligence to extract desirable information, such as ship location, ocean wave height, and sea ice classification. Compared to the raw SAR echoes or focused image, the data size of derived products is so small that low data-rate transfers via satellite data relays become a possible path of delivering the information directly to the users, further facilitating rapid delivery. The on-ground prototype developed in this work intends to serve as a first step towards high-resolution satellite on-board data processing in near-realtime by using hardware representative for use in low Earth orbit.

This study aims to improve the online Quality Control (QC) process for Taiwan's Very High-Frequency (VHF) coastal radar system, specifically for ocean current mapping. Implementing a robust multi-level data system ensures the accuracy and reliability of real-time data collected step-by-step for ocean monitoring. The methodology combines automated monitoring, statistical analysis, and manual inspections to classify data quality using several QC measures. These measures are organized into levels, including raw data verification, initial data calibration, and advanced spatial-temporal data integration. This approach enhances the precision of ocean current measurements for maritime navigation, resource management, and scientific research.

The marine habitats of the Philippines are diverse and ecologically important and while they support a wide range of species and provide important ecosystem services, they remain inadequately mapped on a nationwide scale. This study maps sand, corals, and seagrass in Boracay Island (Malay, Province of Aklan) and Sta. Cruz Islands (Zamboanga City) using deglinted and land-masked Sentinel-2 imagery and Random Forest classification in Google Earth Engine. Ground truth data for Boracay Island were gathered using geotagged validation points, while for Sta. Cruz Islands, a continuous video from a towed camera and GPS tracks were collected and subsequent geotagged images were produced. These data were used for training and testing the benthic classification model. Results indicate high coral coverage in both locations, with more than 90% overall accuracy. This methodology demonstrates an automated benthic habitat classification framework that can be adapted to the entirety of the Philippines.

Variations in seagrass percent cover is one of the key indicators of the health of a seagrass ecosystem. Given that seagrass is regarded as one of the most effective and efficient nature-based solutions for the mitigation and adaptation to climate change, there is a clear need for spatially explicit and extensive information on seagrass percent cover in order to inform management, restoration and conservation efforts. This paper presents a comparison of the accuracy of seagrass percent cover mapping derived from the sunglint and water column corrected surface reflectance (SR), principal component analysis (PCA), and kernel principal component analysis (KPCA) bands of Sentinel-2 L2A. The comparison is based on the application of various regression models, including stepwise regression (SWR), automated partial least squares regression (AutoPLSR), random forest regression (RFR), and support vector machine regression (SVR). The metrics for evaluating the accuracy of different input-model combinations are R², RMSE, and plot 1:1 between reference and estimated percent cover. The findings indicate that incorporating PCA or KPCA into deglint bands and DII within the SWR, SVR, RFR, and AutoPLSR models has varying effects on accuracy. Notably, accuracy consistently improved when deglint bands or DII were combined with both PCA and KPCA, suggesting that this combination is preferable for enhancing model performance and mapping accuracy.

Thermal discharges from coastal power plants can significantly impact marine ecosystems, particularly in environmentally sensitive regions. This study focuses on the Barakah Nuclear Power Plant, situated in the hypersaline and thermally extreme waters of the Arabian Gulf. By leveraging satellite observations and advanced machine learning techniques, this research aims to enhance the monitoring and understanding of thermal plume dynamics, providing valuable insights into their spatial and temporal characteristics. In particular, the study seeks to downscale sea surface temperature (SST) data from a coarse resolution of ~2 km (GHRSST) to a high resolution of 100 m, enabling improved spatial and temporal analysis of thermal anomalies. Ambient seawater temperatures in the region, reaching ~36°C during late summer, underscore the importance of accurately assessing thermal plume dispersion. Landsat 8/9 imagery was utilized to derive SST from Band 10 and Band 11, with validation against in-situ measurements confirming its reliability. However, the limited 8-day revisit interval of Landsat satellites restricts continuous monitoring. To address this, the study employed the Extreme Gradient Boosting (XGBoost) algorithm, incorporating GHRSST SST, wind data, and SST biases as input features. The downscaling from ~2 km to 100 m resolution provided enhanced spatial detail of SST patterns. The model was trained on data spanning 2017–2021 and validated with 2022 data, achieving an R² of 0.94 and an RMSE of 1.23°C. The downscaled SST accurately resolved thermal plumes, demonstrating strong agreement with reference data and enabling finer characterization of plume dispersion patterns. This integrated approach highlights the potential of combining satellite remote sensing and advanced machine learning techniques to monitor thermal discharges at high spatial and temporal resolutions, offering a robust framework for assessing environmental impacts in ecologically sensitive coastal regions.

Marine oil pollution causes severe ecological and environmental damage, which underlines the importance and necessity of an oil spill monitoring system. With the advantage of its wide coverage and the capability to operate at night and during cloudy weather, spaceborne synthetic aperture radar (SAR) has been extensively used. This study demonstrates a near real-time automated oil slick detection system using SAR imagery from the European Copernicus Sentinel-1 mission. The system uses a two-step approach consisting of detection and segmentation. A deep learning-based object detection algorithm was custom-trained with images collected in the Southeastern Mediterranean Sea. To extend the ability of the detector to target oil slicks in other oil pollution hotspots where ocean properties and types of oil spills are different, an additional dataset from another oil pollution hotspot in the North Sea was collected. After the detector has returned bounding boxes of possible oil slicks, a segmentation algorithm is applied to extract the exact areas covered by oil. Based on our test, it took around 6 minutes from providing two Sentinel-1 scenes to delivering possible oil binary masks. This study provides a near real-time oil spill monitoring system and discusses its feasibility for application in oil-polluted waters worldwide.

Tropical islands ‘s oceanic biogeochemical environments may be influenced by continental forcing. The impact of the large Rewa River plume South of the main island of Viti Levu (Fiji) on the microbial processes i.e. carbon and nitrogen fluxes, fisheries richness or coral resources of the Fijian archipelago is unknown. SGLI imagery is helpful in describing ocean color South of Fiji. In order to examine the applicability of ocean colour algorithms for GCOM-C, in situ bio-optical and radiometric data were collected during the March-April 2022 SOKOWASA cruise. Typical mesotrophic waters with ranges of chlorophyll a concentration (Chla, 0.10-0.41 mg·^m−3), MES (0.2 to 1.62 g.^m-3), b_bp550 (5.7 10^-4-0.0037 ^m-1) and a_CDOM440 (0.013- 0.047 ^m-1) are found on a South-North gradient between open ocean and the coast. The remote sensing reflectance obtained using HOCR were inverted to retrieve Inherent Optical Properties (IOPs) for the empirical and semianalytical (QAA, GIOP) algorithms and for OC3 for Chla. All IOPs could be derived from Chla (Case 1 waters). Results showed the capacity and usefulness of the derived products of SGLI (GCOM-C) to monitor the water quality of the Southern ocean South of Fiji. As concluded by bio-optical analysis, during the dry period experienced during SOKOWASA, mesotrophic waters extend far south of the Kadavu Island and small phytoplankton plumes issued from land are mostly formed by organic matter and may feed coral reefs.

High-frequency radar (HFR) observations of meteotsunamis offer real-time measurements of surface ocean currents across extensive spatial areas. On the northern shore of Taiwan, HFR remote sensing systems operating in the high-frequency band have successfully detected meteotsunami-induced surface current triggered by convective storms. During identified meteotsunami occurrences, the amplified radial velocity of 0.1-0.2 m/s is observed and closely followed by the maximum wave height. Changes of the radial velocity between 3 and 18 km is crucial for early meteotsunami detections.

Global climate change poses severe challenges to marine ecosystems, particularly in western Taiwan, where coastal and offshore areas are susceptible to natural and anthropogenic influences. These ecosystems are crucial for biodiversity support, climate regulation, and fisheries. Our study examines the complex coastal waters of northwestern Taiwan, spanning a 95 km coastline extending 40 km offshore and including ecologically sensitive zones. We assessed Sea Surface Height (SSH), Sea Surface Temperature (SST), Sea Surface Salinity (SSS), and Sea Surface Current (SSC) using both satellite and in situ observations. Additionally, we employed the Diffuse Attenuation Coefficient at 490 nm (Kd490) and chlorophyll-a (Chla) as indicators of water turbidity and biological productivity. Analysis from 1993 to 2023 revealed increasing trends in SSH and SST, with SST rising by 0.02°C and SSH by 0.003 m annually—trends consistent with regional and global warming patterns. SSH was highest in autumn (0.63 m) and lowest in spring (0.49 m), while SST peaked in summer (29.89°C) and dropped in winter (19.39°C). SSS declined by 0.006 psu per year, likely due to increased freshwater input and changes in circulation. Seasonal fluctuations in Kd490 and Chla concentrations, with lower values in summer (0.06 m^-1, 0.47 mg m^-3) and higher in autumn and winter (0.19 m^-1, 3.83 mg m^-3), suggest that turbidity and productivity are closely linked to seasonal oceanographic processes and external inputs.

We have developed a portable fluorescence lidar system for chlorophyll monitoring in surface coastal waters. The portable LD-based fluorescence lidar system uses an excitation wavelength of 405 nm with a pulsed width of 11 ns and a repetition frequency of 500 kHz. A campaign was conducted to take lidar measurements and collect seawater samples at Inage Beach, Chiba, Japan during the summer season. In this study, we have compared the chlorophyll distribution measured by the lidar data with the regional Chl-a derivation algorithms (MODIS OC3M and the Three-band reflectance model). A higher positive correlation is observed between the lidar data and the three-band reflectance model (R2 = 0.75) than the MODIS OC3M (R2 =0.17). Other water and weather parameters were also measured to determine the environmental impacts on the phytoplankton community. This study provides information on the importance of real-time in-situ monitoring, such as fluorescence lidar systems, for monitoring water quality in coastal waters.

The global seaweed aquaculture, particularly Eucheuma spp. cultivation has experienced rapid growth, particularly in Southeast Asia, driven by the increasing demand for products such as carrageenan, agar, and alginate [1, 2]. Indonesia, a leading producer, faces challenges in optimizing site selection for seaweed farming, especially in regions like Lombok Island where environmental conditions vary significantly [3].

This study aims to assess the site suitability for Eucheuma spp. cultivation around Lombok Island using ocean color remote sensing and GIS-based spatial analysis, addressing the need for a systematic and scalable approach to site selection in seaweed aquaculture [1]. The objective is to identify optimal locations for seaweed farming that maximize productivity while minimizing environmental impacts, thereby supporting the livelihoods of local communities [4].

We integrate multiple environmental parameters, including sea surface temperature, sea surface salinity, bathymetry, total suspended matter, dissolved oxygen, pH, and chlorophyll-a concentration. Data from various satellite sensors including GOCI-II, were preprocessed, classified, and scored based on their suitability for Eucheuma spp. cultivation. A weighted scoring system was employed for all the parameters to evaluate both annual and seasonal variations, which produced a comprehensive suitability map for Eucheuma spp. cultivation around Lombok Island. The result indicates that highly suitable areas are predominantly located in shallow coastal waters near the southeastern and southwestern coastlines of Lombok Island.

The findings are able to provide useful information for policymakers, farmers, and stakeholders involved in the long-term growth of seaweed aquaculture in Indonesia and potentially other parts of the world.

To examine the spatial distribution of mackerel catches within South Korea's Exclusive Economic Zone (EEZ) from 2011 to 2019, we evaluated the habitat suitability index (HSI) for mackerel using satellite-derived environmental data combined with government-provided catch data. The HSI is widely recognized for its capability in identifying and predicting potential fishing areas. By integrating mackerel catch records with satellite-based environmental variables such as Chlorophyll-a concentration (CHL), Sea Surface Temperature (SST), Sea Surface Height (SSH), and Primary Productivity (PP), we established optimal environmental ranges: CHL (0.32 to 1.6 mg m⁻³), SST (14.45 to 26.72°C), SSH (0.61 to 0.84 m), and PP (654.94 to 1,731.3 mg C m⁻² d⁻¹). Using these thresholds, we generated seasonal habitat suitability maps for mackerel, which were validated against catch data. The findings suggest that our HSI model is a reliable tool for predicting mackerel fishing grounds within the South Korean EEZ. This research offers valuable insights into the spatial distribution and environmental preferences of mackerel in these waters, contributing to improved fisheries management strategies.

Coral reefs play a significant role in marine ecosystems, but are increasingly facing degradation, primarily due to ocean warming. Effective monitoring and management methods are crucial for the conservation and rehabilitation of these ecosystems, and accurate mapping is essential for monitoring and conserving coral reef ecosystems. Utilizing remote sensing data is an effective approach to achieve high-quality maps. However, water column interference can disturb the reflected signal in remote sensing data. In this study, depth-invariant index (DII) transformation was applied to perform water column correction on high-resolution multispectral data from the QuickBird sensor, focusing on the coral reefs around Weno Island, Micronesia. Habitat classification was performed using an object-based image analysis method with additional in-situ measurements. The classification performance was evaluated using error matrices and kappa statistics.