Wind speed estimation for tropical cyclone from combined active and passive measurements

Sea wind speed (SWS) plays a key role for understanding air-sea interactions and marine physical processes. Both microwave radiometer and scatterometer can provide SWS products. However, under high wind speeds (HWS), especially during tropical cyclones (TCs), these products have larger errors. This paper proposes a method of combining active and passive observations to invert the SWS based on random forest (RF) regression during the global TCs. The estimation results indicated the good accuracy of the RF-AP model with the root mean square error (RMSE) about 2.09m/s and coefficient of determination was 0.89 for validation dataset. In contrast, the estimation results of the RF-active regression model (RMSE-3.63m/s) and the RF-passive regression model (RMSE-2.38m/s) alone are not as good as the RF-AP model. Overall, the proposed method in this study shows the potential for application in SWS inversion, and the results will be helpful for the inversion of SWS under TCs.

I. Introduction

Sea wind speed (SWS) is one of the most useful indicators of ocean dynamic parameters, especially in the case of high wind speed (HWS) of tropical cyclones (TCs). HWS and storm surge during TCs will cause serious disasters and heavy losses. For example, Super Typhoon Haiyan (2013) was the strongest typhoon ever to land on record in the Northwest Pacific. It had a maximum wind speed of 170 knots before landing, causing 6,300 deaths, 1061 missing, and 28,689 injuries [1]. In addition to the impact on human activities, HWS is also one of the important parameters of complex ocean-atmosphere interaction [2]. Therefore, it is of great significance to obtain accurate HWS.