Abstract CyclObs‐derived wind and SWH field are extracted from over 600 dual‐polarized Sentinel‐1 (S‐1) images of around 300 tropical cyclones (TCs) over the past eight years to investigate asymmetry of wind and wave fields during TCs. Fetch analysis and machine learning technique, eXtreme Gradient Boosting (XGBoost), is used to establish a relationship between TC wind speed and significant wave height (SWH). It was found that TC wind and SWH radii become asymmetric as sea states intensify. Notably, wind radii correlations (CORs) increase on the left‐right and left‐back quadrants for wind speeds larger than 20 m/s, while SWH radii exhibit the opposite trend. XGBoost is employed to obtain the improved relationship between wind fetch and SWH (COR < 0.17). Validation against buoys and Haiyang‐2 (HY‐2) observations of 20 TCs indicates that the root mean squared error in SWH predictions is reduced by up to 1.1 m using XGBoost instead of empirical model. The new TC wave model by XGBoost is particularly robust under high‐wind conditions, therefore vital for warning and mitigation of extreme storms and improved parameterizations of air‐sea interaction.

Read original article

Read original article