Releases: GOFUVI/hf_eolus_wind_inversion
Release list
v0.1.1
The HF-EOLUS toolkit consolidates all command-line pipelines and artefact definitions required to retrieve near-surface wind speed and direction from HF-radar observations via artificial neural networks. It couples radar-derived backscatter statistics with two supervisory references: Sentinel-1 Level-2 OCN products aggregated on regular spatial grids and collocated in-situ winds measured by an oceanographic buoy embedded in the same footprint. The workflow respects the frequency-dependent radar operating band by combining gated regression heads with a complementary range-classification branch, thereby preserving the physical constraints of the inversion problem. For comprehensive documentation and reproducibility artefacts, consult the project repository on GitHub (https://github.com/GOFUVI/hf-wind-inversion).
Data preparation stages pivot the radar spectra around Bragg peaks, validate the geometrical consistency of each station, enrich the features with bearings, distances and maintenance metadata, and standardise schema variations across radar sites. Stratified partitions, cross-domain concatenation, and maintenance-aware normalisation ensure that model training, validation and inference rely on reproducible datasets that capture both spatial variability and temporal stability.
Training routines span hyper-parameter optimisation, final non-cross-validated retraining, and multiple fine-tuning strategies—plain adaptation, L2-SP anchoring, and knowledge distillation—to explore cross-domain transfer between the spatially rich SAR corpus and the buoy’s high-fidelity point measurements. Inference jobs cover grid-wide predictions, domain-specific test sets, and maintenance-interval diagnostics, delivering quantitative metrics and provenance-rich artefacts that can be catalogued alongside the input datasets. By packaging the entire pipeline, the repository enables systematic experimentation with HF-radar wind inversion models while providing the documentation necessary for reproducibility and scientific auditing.
What's New
A buoy wind height correction utility (scripts/aggregation/apply_buoy_wind_height_correction.sh) now accompanies the pipeline. The helper encapsulates the neutral logarithmic wind-profile adjustment inside a containerised execution environment, accepting configurable source and target heights so deployments can harmonise buoy winds to any reference level without manual scripting. Documentation and the illustrative data-preparation pipeline reflect this addition, ensuring that downstream partitions ingest height-standardised buoy tables while preserving the raw measurements for provenance and auditing workflows.
Acknowledgements
This work has been funded by the HF-EOLUS project (TED2021-129551B-I00), financed by MICIU/AEI /10.13039/501100011033 and by the European Union NextGenerationEU/PRTR - BDNS 598843 - Component 17 - Investment I3. Members of the Marine Research Centre (CIM) of the University of Vigo have participated in the development of this repository.
Disclaimer
This software is provided "as is", without warranty of any kind, express or implied, including but not limited to the warranties of merchantability, fitness for a particular purpose, and noninfringement. In no event shall the authors or copyright holders be liable for any claim, damages, or other liability, whether in an action of contract, tort, or otherwise, arising from, out of, or in connection with the software or the use or other dealings in the software.
v0.1.0
HF-EOLUS HF-Radar Wind Inversion Toolkit for Artificial Neural Networks Training and Inference
The HF-EOLUS toolkit consolidates all command-line pipelines and artefact definitions required to retrieve near-surface wind speed and direction from HF-radar observations via artificial neural networks. It couples radar-derived backscatter statistics with two supervisory references: Sentinel-1 Level-2 OCN products aggregated on regular spatial grids and collocated in-situ winds measured by an oceanographic buoy embedded in the same footprint. The workflow respects the frequency-dependent radar operating band by combining gated regression heads with a complementary range-classification branch, thereby preserving the physical constraints of the inversion problem. For comprehensive documentation and reproducibility artefacts, consult the project repository on GitHub (https://github.com/GOFUVI/hf-wind-inversion).
Data preparation stages pivot the radar spectra around Bragg peaks, validate the geometrical consistency of each station, enrich the features with bearings, distances and maintenance metadata, and standardise schema variations across radar sites. Stratified partitions, cross-domain concatenation, and maintenance-aware normalisation ensure that model training, validation and inference rely on reproducible datasets that capture both spatial variability and temporal stability.
Training routines span hyper-parameter optimisation, final non-cross-validated retraining, and multiple fine-tuning strategies—plain adaptation, L2-SP anchoring, and knowledge distillation—to explore cross-domain transfer between the spatially rich SAR corpus and the buoy’s high-fidelity point measurements. Inference jobs cover grid-wide predictions, domain-specific test sets, and maintenance-interval diagnostics, delivering quantitative metrics and provenance-rich artefacts that can be catalogued alongside the input datasets. By packaging the entire pipeline, the repository enables systematic experimentation with HF-radar wind inversion models while providing the documentation necessary for reproducibility and scientific auditing.
Acknowledgements
This work has been funded by the HF-EOLUS project (TED2021-129551B-I00), financed by MICIU/AEI /10.13039/501100011033 and by the European Union NextGenerationEU/PRTR - BDNS 598843 - Component 17 - Investment I3. Members of the Marine Research Centre (CIM) of the University of Vigo have participated in the development of this repository.
Disclaimer
This software is provided "as is", without warranty of any kind, express or implied, including but not limited to the warranties of merchantability, fitness for a particular purpose, and noninfringement. In no event shall the authors or copyright holders be liable for any claim, damages, or other liability, whether in an action of contract, tort, or otherwise, arising from, out of, or in connection with the software or the use or other dealings in the software.