Determining surface winds from doppler radar data during hurricane passages over Florida

The destructive nature of hurricane force winds requires their diagnosis on as small a scale as possible. This small scale resolution is needed because recent studies have revealed that various types of mesoscale features are embedded within the storm's overall wind field. Through a joint effort between Florida Power and Light Corp. (FPL) and Florida State University, a technique has been developed to calculate fields of high resolution surface wind speeds in hurricanes crossing Florida and thereby identify areas of extreme winds.

The surface winds are derived from Level II Doppler radar data and analyzed onto a high resolution (1 degree radial, 0.5 km gate) gridded domain. This is done for each radar volume scan while a storm is affecting Florida. We utilize the Warning Decision Support System – Integrated Information (WDSS-II) software to analyze the data and quality control the reflectivity fields. The algorithm removes bad data associated with anomalous propagation, ground clutter, and clear-air impulse returns. The procedure has demonstrated skill in removing bad data in various types of precipitation regimes.

The first step in our methodology was to develop an algorithm that estimates the total wind speed from a combination of the radial velocity and quality controlled reflectivity values. Once the total wind fields were computed at each scan, reduction factors were developed to transpose the winds at varying altitudes down to the surface. Multiple factors were considered in calculating the reduction factors, including vertical shear, translation speed, gridpoint quadrant location, elevation, surface variability, and instability of the boundary layer.

Case studies of Hurricanes Jeanne (2004) and Wilma (2005) will be presented at the conference, with results in the forms of 1 min winds, 3 s gusts, and maximum values of these parameters during the period that each storm is over Florida. Computed wind speeds will be compared with the sparse amount of observed data