
Accurate prediction of surface ocean transport is critical for applications such as search and rescue, oil spill response, and marine hazard mitigation. We present a new tool developed to evaluate and compare ocean surface trajectory estimates from multiple numerical models using high-frequency radar (HFR) derived surface currents and in situ drifter data as observational references. The tool supports model intercomparison by aligning model output with virtual drifter release points and computing trajectory skill metrics including separation distance after 12 and 25 hours as well as a Lagrangian skill score. This framework has been applied in the Mid-Atlantic Bight from October 2024 to present using data from recent deployments of surface drifters and regional HF radar measurements. Forecasts from an ensemble of four ocean modeling systems-Short Term Prediction System (STPS), Regional Ocean Modelling System (ROMS) Doppio, Copernicus Marine Environment Monitoring Service (CMEMS), and the Earth System Prediction Capability (ESPC)-were evaluated over a range of oceanographic conditions. Results reveal variability in model performance, with some models excelling in surface drift assessment after 25 hours. HF radar data provided highresolution surface current fields served as an independent, synoptic benchmark across all models. This modular, open-source tool enables real-time and retrospective model evaluation, supports operational decision-making, and can be integrated into other ocean observing systems. Ongoing development aims to extend functionality to include uncertainty quantification, multimodel ensemble weighting, and support for additional observational datasets.
This paper presents a low-cost and scalable localization framework for autonomous underwater vehicles using one-way acoustic communication in shallow water environments. The system relies solely on received signal strength (RSS) from four fixed acoustic beacons and a heading measurement to estimate the vehicle's position. The signal field of each beacon is modeled independently using Gaussian Process regression, enabling a spatially-smooth data-driven representation of the signal landscape, even in environments prone to severe multipath effects. Motion is sampled probabilistically based on commanded velocity and heading measurements, then fused with RSS observations using a particle filter. Unlike conventional acoustic positioning systems that require synchronized two-way communication, our approach leverages one-way broadcasts and passive localization of the agent, reducing system complexity and enabling scalable multi-agent deployments without added infrastructure cost. Experimental results demonstrate sub-20 meter localization accuracy, outperforming a bearing only localization approach and highlighting the viability of RSS-based acoustic localization in shallow, cluttered, GPS-denied marine environments.
This study proposes a Grey-box system identification framework for Unmanned Surface Vehicles (USVs) operating under complex environmental disturbances. Planar motion dynamics (surge, sway, and yaw) of a low-speed flat-bottom USV are explicitly derived from physical principles and augmented with propulsion and wind effects. These dynamics are embedded in a Nonlinear AutoRegressive with eXogenous inputs (NARX) model, where the system accelerations are decoupled and formulated in terms of measurable variables and identifiable parameters. A custom USV testbed for experiment was developed and instrumented with an inertial navigation system, a data acquisition system, and a weather station. Sensor data were synchronized at 100 Hz and preprocessed to remove noise and ensure consistency. Field experiments under various maneuvers and environmental conditions demonstrated high prediction accuracy of the proposed model. However, its performance exhibited moderate sensitivity to the tuning of parameters within the nonlinear output function. Future research will focus on enhancing robustness and adaptability under broader operational scenarios.
Sea surface temperature (SST) and chlorophyll-a (Chl-$a$) are key indicators of ecosystem variability in temperate coastal-open ocean environments, reflecting both physical and biological responses. This study investigates the spatiotemporal variability of SST and Chl-$a$ in Sagami Bay using level 2 ocean color satellite data from Second generation Global Imager (SGLI)/Global Change Observation Mission-Climate (GCOMC) from 2018 to 2024. To address substantial data gaps caused by frequent cloud cover and sensor limitations such as sun glint, atmospheric interference, and land adjacency effects, we reconstructed missing values using the Data INterpolating Empirical Orthogonal Functions (DINEOF) method. Validation against randomly masked original satellite pixels and in-situ measurements confirmed that DINEOF shows significant correlation in filling gaps in both SST band Ch-$a$. Empirical Orthogonal Function (EOF) analysis was conducted on the DINEOF-reconstructed monthly datasets to extract dominant spatiotemporal modes of variability. SST showed strong seasonal signals in the leading EOF mode, with subsequent modes reflecting mesoscale variability, including Kuroshio intrusions. EOF modes of Chl-$a$ revealed seasonal bloom patterns and spatially complex responses to physical forcing, particularly in nearshore and river-influenced areas. The results highlight the importance and effectiveness of satellite data gap-filling for capturing short-term events and mesoscale dynamics in coastal waters. Our findings provide an insight into the spatiotemporal variability of physical and biological parameters in Sagami Bay and demonstrate a scalable approach for monitoring coastal systems under environmental change.
The Dense Oceanfloor Network System for Earthquakes and Tsunamis (DONET) and the Long-Term Borehole Monitoring System (LTBMS) have been deployed along the Nankai Trough to monitor seismic activity, crustal deformation, and tsunamis in real-time. Accurate temperature measurement is critical for calibrating hydraulic pressure sensors within these systems, as sensor readings can be affected by heat generated from electrical power consumption and environmental variations. In this study, we evaluated the reliability of DONET and LTBMS temperature sensors by comparing their measurements with reference temperature data obtained from CTD sensors deployed by the AUV-NEXT near the D-node. Results indicate that both oscillator temperature (OYT) and external thermometers (LKO) are influenced by electrical power heat, with DONET and LTBMS sensors recording higher temperatures than the reference CTD measurements despite being at greater depths. Additionally, we estimated the spatial distribution of seafloor electrical conductivity and derived seafloor temperatures independent of electrical heating effects, providing a reference for further calibration. The findings highlight the importance of accounting for thermal influences in long-term oceanographic and geophysical monitoring using subsea networks. Future work will focus on classifying long-term temperature variations into components attributable to global ocean warming and oceanic processes, such as sea surface height changes and baroclinic responses affecting seafloor temperature and pressure measurements toward precise monitoring of seafloor crustal deformation.