Application of a random forest algorithm that considers barometric pressure in GNSS-R sea surface wind speed retrieval

• Published in: Journal of physics. Conference series Vol. 2935; no. 1; pp. 12014 - 12025
• Main Authors: Ma, Yi, Wang, Yifan, Wang, Guofang, Geng, Hao, Chen, Yuan
• Format: Journal Article
• Language: English
• Published: Bristol IOP Publishing 01.01.2025
• Subjects: Algorithms; Atmospheric pressure; Energy sources; Error analysis; Multistatic radar; Ocean currents; Radar cross sections; Radar scattering; Regional development; Retrieval; Scattering cross sections; Surface wind; Wind power; Wind speed; Winter
• ISSN: 1742-6588; 1742-6596
• DOI: 10.1088/1742-6596/2935/1/012014

This paper aims to rely on CYGNSS data to propose a random forest-based wind speed retrieval method that considers barometric pressure as a pivotal factor. Taking the Hawaiian Islands and the peripheral waters as the subject, the research conducts a systematic analysis of the spatial-temporal variation characteristics of wind speeds. In this case, an innovative random forest (RF) model that considers barometric pressure is built and trained, with the longitude, latitude, time, normalized bistatic radar scattering cross section (NBRCS), leading-edge slope (LES), and barometric pressure data as the input features, and the measured buoy wind speeds as the target variable. The results show that the introduction of barometric pressure can significantly improve the accuracy of wind speed retrieval, raising the correlation to above 0.8 and reducing the root mean square error (RMS) by more than 40%. The RF (pressure) method performs best when the pressure system changes dynamically, such as in winter. In the Hawaiian Islands, moreover, the wind speeds exhibit notable spatial variations across seasons. The wind speeds are generally stable and moderate in spring and autumn. In winter, the wind speeds in the northern and northwestern regions can reach the highest level. In summer, the wind speeds in the southeastern region come to a significant decrease. These findings reveal the complex influences of barometric gradients like the subtropical high and the Aleutian Low, as well as topography and ocean currents, on wind speeds, providing a scientific basis for understanding regional climatic dynamics and wind energy resource development.