Empirical Models for Estimating Significant Wave Height Using RADARSAT-2 Data

Abstract

The images captured by Synthetic Aperture Radar (SAR) are useful for retrieving ocean wave parameters. The objective of this study is to establish empirical algorithms to estimate significant wave height (Hs) from RADARSAT-2 (RS2) fine-quad (FQ) beam mode images. This is the first time to retrieve Hs by utilizing around 1,400 collocated samples over the eastern and western North America. The fitting procedures are based on linear regression and neural network. The models are validated against to buoy observations. Unlike the most related work, this study explores the effects of incidence angle and polarization on the estimation of Hs. I report five accuracy metrics and introduce a novel cost function to assess and compare the models' performance. Finally, the proposed two types of models both show a good agreement with buoy observations. The RMSE and R could reach to 0.26m and 0.97, respectively.