Hurricane Watch: Operational analysis of hurricane characteristics using synthetic aperture radar

The University of Miami and the Canadian Space Agency jointly implemented a Hurricane Watch program in 2004 at the Center for Southeastern Tropical Advanced Remote Sensing (CSTARS) by capturing and processing in near real time synthetic aperture radar (SAR) data from RadarSat-1. SAR provides a unique opportunity to study the impact of tropical cyclones at the ocean surface while electro-optical sensors are limited to viewing hurricanes at the top of the clouds. Many features at the ocean surface in the visible range are masked by the cloud bands. In contrast SAR penetrates cloud layers and rain bands and thus is capable of measuring properties at the sea surface. The goal of this project is to deliver in near real time new information about hurricane characteristics to the National Hurricane Center. Specifically we are extracting size, shape and location of the eye, radius to maximum winds as well as identify changes in the eye wall replacement cycle. Furthermore, we are examining the presence of thunderstorms and possibly tornadoes embedded in convection bands.

A modified algorithm based on CMOD5 has been applied to retrieve high resolution ocean surface wind fields from SAR images on an operational basis. The algorithm is applicable for both RADARSAT ScanSAR and Envisat ASAR data which operate at C-band in either vertical (VV) or horizontal (HH) polarization. Derivation of the sea surface wind field is a two step process: In the first step wind directions are retrieved based on first guess estimates to minimize directional ambiguity problems. SAR-based wind directions are retrieved from wind-induced phenomena or “streaks”, which are aligned in wind direction at a typical spacing of 200 m to 3 km. The orientation of these features is derived by determining local gradients from smoothed SAR amplitude images.

In the second step the wind speeds are retrieved utilizing the CMOD5 model function that describes the dependency of the NRCS of the ocean surface on the local near-surface wind speed, wind direction versus antenna look direction and incidence angle for VV polarized images. For wind speed retrieval from C-band HH polarized SAR images a hybrid model function is applied consisting of the CMOD5 and a C-band polarization ratio.

Simple gradient techniques are applied to the NRCS values for estimating the diameter and shape of the eye. This technique is also used to locate local pockets of high wind features or thunderstorms and tornadoes.

This study will present results from the 2005 hurricane season and compare SAR derived parameters with other estimates from airborne observations, scatterometry and in-situ sensors when available.