GUST: A General Utility Scatterometry Tool for Multi-Platform Ocean Wind Retrieval

Wind vector retrieval from radar measurements plays a fundamental role in ocean remote sensing, providing essential information for weather forecasting, climate studies, and maritime operations. Traditional scatterometer missions, including the ASCAT series aboard METOP satellites, the Chinese HY-2 constellation, and CFOSAT-SCAT on the CFOSAT mission, have established robust operational wind processing chains that have been successfully deployed for decades. These systems rely on well-validated inversion algorithms and ambiguity removal techniques to derive ocean surface wind vectors from radar backscatter measurements. However, the expanding diversity of spaceborne radar systems capable of providing wind information—such as Synthetic Aperture Radars (Sentinel-1), altimetry missions (Sentinel-3, Sentinel-6, SWOT), and emerging constellation concepts—creates a growing need for flexible processing frameworks that can accommodate different instrument geometries, viewing configurations, and measurement characteristics.

In this work, we present GUST (General Utility Scatterometry Tool), a modern Python-based framework designed to provide a unified approach to wind vector retrieval from diverse radar observations. GUST implements established wind-retrieval algorithms, including maximum-likelihood estimation (MLE) inversion with configurable geophysical model functions, and advanced ambiguity-removal techniques such as the two-dimensional variational method (2DVAR), originally developed for operational scatterometer processing. The implementation leverages PyTorch for GPU-accelerated computations, enabling efficient processing of large satellite datasets while maintaining algorithmic transparency and modularity.

In its baseline configuration, GUST fully reproduces the functionality and accuracy of the KNMI Advanced Wind Data Processor (AWDP), the operational standard for ASCAT data processing. Validation against AWDP reference products demonstrates excellent agreement: wind speed RMSE of 0.02 m/s, correlation exceeding 0.99, and approximately 98% agreement in ambiguity selection. These results confirm that the Python implementation maintains the scientific rigor of established Fortran-based processors while providing a more accessible and modifiable codebase.

The key advantage of GUST lies in its flexible architecture. The processing workflow—comprising data reading, quality control, inversion, ambiguity removal, and output generation—is organized as independent, configurable modules. This modular design enables rapid implementation of new retrieval algorithms tailored for emerging observation techniques: cross-polarisation measurements, Doppler-based retrievals, multi-band and multi-angle configurations, or bi-static observation geometries. Furthermore, GUST supports the development of multi-parametric geophysical model functions that incorporate additional ocean state variables—such as sea surface currents and sea state conditions—into the wind retrieval process, advancing beyond traditional wind-only inversions toward more comprehensive air-sea interaction characterisation.

Looking forward, GUST provides a foundation for processing data from next-generation missions with enhanced sensing capabilities, such as Metop-SG and the Harmony mission concept. The Python-based implementation enables straightforward incorporation of new algorithms, including machine learning approaches, and supports collaborative development within the scientific community. By bridging traditional operational processing with modern software practices, GUST aims to accelerate the development of wind-retrieval capabilities in the evolving landscape of ocean observing systems.

Acknowledgement This work was performed as a part of CFOSAT IFREMER Wind and Wave Operation Center (IWWOC) development. The IWWOC is co-funded by CNES and IFREMER.