Simulating the trajectory of surface drifters in the ocean matters for search and rescue, pollution tracking, oil and chemical spill response, and marine risk analysis. Accurate prediction remains difficult, as widely acknowledged in the literature, and also illustrated by the ``Forecasting Floats in Turbulence''challenge issued by the US Defense Advanced Research Projects Agency (DARPA) in 2021, and which ultimately led to this paper. The main source of error usually comes from uncertain ocean currents, while errors in wind forcing and object drift properties are often smaller [Dagestad and R\"ohrs, 2019]. Here, we use an open one-year dataset of Sofar Spotter trajectories together with several ocean and atmospheric hindcast products to test data-driven drift models at scale. We compare three approaches: i) a standard (baseline) drifter trajectory simulation based on one ocean model and one atmospheric model, ii) linear regression (LR) models that fuse all available predictors, and iii) neural networks (NN) using similar inputs. A simple LR model that combines all predictors performs equally well as the NN. Because LR is simpler, cheaper, and more robust, we retain it as the preferred approach. In 2-day trajectory prediction, this improves the Liu-Weisberg skill score by around 40\% relative to the baseline. These findings apply to hindcast mode; applying this methodology for forecast mode remains for future work.
Accurate multi-day forecasting of floating-object trajectories on the ocean surface is critical for applications ranging from search-and-rescue to environmental tracking. This task remains however challenging due to the complex interplay of influencing factors such as ocean currents and winds. In this work, we frame trajectory prediction as a denoising task and present Conditional Diffusion models for Trajectories (CoDiT), which adapts the denoising diffusion framework, originally developed for image synthesis, to the problem of trajectory forecasting. CoDiT generates realistic trajectory forecasts, conditioned on heterogeneous context data: ocean currents and winds from reanalysis products, bathymetry, and the initial position. We train and evaluate CoDiT on two global, specialized datasets focusing on the open ocean and coastal regions, using GPS trajectories from the Global Drifter Program as ground truth. We compare CoDiT rigorously against various baselines, including a convolutional neural network that predicts velocity fields, and physical forecasts generated directly from the current and wind fields. Quantitative evaluations show that CoDiT achieves the lowest position error across both datasets and all forecast horizons, and the best probabilistic forecast quality among all methods, as measured by the energy score. Notably, in the coastal setting, CoDiT is the only method to surpass the naive persistence baseline in position error.
Christian Donner, Shirin Goshtasbpour, Emanuele Dalsasso et al.· Machine Learning: Earth· 0 citations
OceanBench is a benchmark designed to evaluate and accelerate global short-range data-driven ocean forecasting, constructed from a curated dataset comprising first-guess trajectories, nowcasts, and atmospheric forcings from operational physical ocean models, typically unavailable in public datasets due to assimilation cycles.
Anass El Aouni, Quentin Gaudel, J. Emmanuel Johnson et al.· Neural Information Processin...· 7 citations· ⚡2
Abstract Accurate short-term tide forecasting is essential for flood early warning, port operations, coastal infrastructure protection, and adaptation planning under accelerating sea-level rise and intensifying storms. Traditional methods remain limited: harmonic analysis neglects surge contributions, while hydrodynamic models are computationally prohibitive for real-time applications. To address these gaps, we propose TideFusionNet, a lightweight, physics-anchored multi-source network that integrates tide-gauge records, hourly meteorology from ECMWF Reanalysis v5 (ERA5), harmonic constituents, satellite sea level anomaly (SLA) data, surface water and ocean topography (SWOT) snapshots, and static digital elevation model (DEM) data. A cross-source attention mechanism, anchored by gauge observations, ensures consistent information fusion, while spectral–phase regularization stabilizes tidal-band energy. Evaluations using one year of observations from Sippican Harbor (Buzzards Bay) demonstrate that TideFusionNet achieves near-zero bias, tight identity-line scatter, and smoothly increasing errors from 12 to 48 h horizons. Ablation experiments reveal a clear information hierarchy: harmonics and meteorology provide the largest contributions, satellites and DEM yield complementary gains, and the freshness–proximity gate effectively regulates SWOT inputs. Under typical missing-source conditions, the forecast skill degrades gracefully, confirming the robustness of the architecture. This study demonstrates how a compact, physics-anchored framework can deliver real-time, physics-constrained tide forecasts. The framework can support operational flood warning, port scheduling, infrastructure risk assessment, and local coastal digital twins by providing fast, interpretable, and physics-constrained sea-level forecasts.
PREVENIR—Forecast and Warning of Flash Flood Events is an Argentina-Japan cooperation project for five years from 2022 that aims to develop an early warning system for heavy rainfall and urban floods. The current operational numerical weather prediction system of the Argentine National Meteorological Service consists of deterministic and probabilistic forecasts at 4 km resolution. Since PREVENIR aims at more accurate and timely precipitation forecast, we developed a 2 km resolution 5-min update data assimilation system with the Local Ensemble Transform Kalman Filter (LETKF) that assimilates observation data from an automated weather station network and C-band Doppler weather radars. Independent LETKF systems have been coupled with the regional models’ Weather Research and Forecasting (WRF) and Scalable Computing for Advanced Library and Environment (SCALE), for two target basins: a mountainous region of the Suquía Villa Paez in Córdoba Province and a flat region of the Sarandí Santo Domingo in Buenos Aires. This study investigates the performance of the two systems in two extreme rain cases. We show that the forecasted precipitation benefits from the high-resolution rapid-update process, which improves its location and intensity with respect to coarser forecasting resources available in the region and also non-data assimilation systems. The results suggest that the prototypes could offer several advantages for hydrological applications, representing a potentially valuable tool for the region.
M. E. Dillon, A. Amemiya, P. Maldonado et al.· Journal of the Meteorologica...· 0 citations
Ocean forecasting is crucial for both scientific research and societal benefits. Large artificial intelligence (AI)-based models have recently boosted forecasting efficiency and accuracy. However, it remains challenging to develop a comprehensive AI-driven ocean forecasting system capable of integrating cross-spatiotemporal and atmospheric forcing. This study introduces LangYa, a cross-spatiotemporal and atmospheric forcing ocean forecasting system featuring: (1) a large-language-model-based (LLM-based) time embedding to explicitly represent forecast lead times, (2) an asynchronous cross-iterative random sampling strategy to represent the impacts of atmospheric forcing on ocean processes, (3) an ocean self-attention module to enhance network stability and accelerate training convergence, and (4) an adaptive loss function to capture ocean dynamics in the thermocline, at depths ranging from tens of meters to about 300 m. LangYa is trained on 27 years of global ocean data from the Global Ocean Reanalysis and Simulation, version 12 (GLORYS12). Using reanalysis and observational data, compared to existing open-source AI-based forecasting systems and numerical models, LangYa enables a single model to produce forecasts with lead times of 1 to 7 d (1/12°, daily) and achieves 7 d RMSEs below 0.0736 m/s, 0.0701 m/s, 0.4376 ℃, and 0.1302 psu for global currents, temperature, and salinity respectively. These quantitative results indicate that LangYa provides clear advantages in forecast accuracy, lead-time robustness, and stability for global OSV forecasting, demonstrating its potential for real-time operational deployment.
Nan Yang, Chong Wang, Zi-Meng Zhao et al.· Science Bulletin· 0 citations
Accurate sea surface wind speed fields are essential for marine navigation, offshore operations, and air–sea interaction studies. However, limited communication bandwidth makes it difficult to receive forecasts from land-based centers, motivating wind speed reconstruction using sparse observations. To address this challenge, we propose SwiftWind, a coordinate-based deep learning framework for sea surface wind speed reconstruction at arbitrary locations by fusing multi-source observations. SwiftWind embeds non-gridded, variable-length observations through adaptive latent representations and latitude–longitude coordinate encoding. We conduct Observing System Simulation Experiments (OSSEs), real-world observational experiments, and arbitrary-location inference experiments. Under ERA5-based evaluation, SwiftWind consistently outperforms existing data-driven baselines, including Fourier Neural Operator (FNO) and Vision Transformer (ViT) models, demonstrating robustness to observation number, noise level, and spatial distribution. Compared to the GFS 6 h forecast fields, SwiftWind achieves approximately 20–23% reductions in RMSE and 19–22% reductions in MAE under real-world observational settings. In independent buoy validation, SwiftWind performs comparably to ViT and slightly worse than FNO, likely due to differences in scattered-point processing and buoy distribution. These findings indicate that SwiftWind is suitable for near-real-time onboard wind speed reconstruction under sparse-observation conditions.
Ruisheng Hu, Jiaqi Ding, Jinhui Yang et al.· Remote Sensing· 0 citations
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.
