Louis Bowers, Michael Schaffer, and Brian Sullivan
 
 

Several recent experiments have been conducted to show the impact of an increased number of observations on hurricane model forecast track error.  The first studied the impact of collecting data about the environment ahead of a hurricane by the use of dropwindsondes on the Geophysical Fluid Dynamics Lab's GFDL hurricane model (Aberson and Franklin 1999).  The second studied the impact of assimilating satellite-derived wind data from GOES imagery into the GFDL model (Soden et al. 2000).  Both studies showed that the inclusion of the data in the hurricane model improved the accuracy of the track forecast.   With this in mind, this paper discusses some of the other remotely sensed data that can be possibly assimilated into forecast models to improve both their track and intensity forecast accuracy.  This data includes sea surface temperature (SST), hurricane heat potential, rainfall intensity, upper level divergence, and wind shear.

This paper will also briefly discuss the instruments and techniques that are used to retrieve this data, as well as discuss the GFDL model and how the data is assimilated.  Also some of the implications of this data to nowcasting tropical cyclones will be discussed.  The more information that is available for both the hurricane models to ingest and hurricane forecasters to go on will result in better tropical cyclone forecasts of both track and intensity.  Recent improvements in remote sensing technology will provide more data then ever before about tropical cyclones and their environments, and therefore with this information people will have more warning time to prepare and to evacuate before the storm makes landfall.

• Introduction

• Background
• Remote Sensing of the Oceans
• Remote Sensing of the Atmosphere
• Satellite-Derived Winds

• Methodology

• Results

• Discussion

• Bibliography and Acknowledgments

Introduction

Hurricanes are one of the most powerful forces on earth.  They may persist for weeks and travel thousands of miles.  The potential for death and destruction is immense.  There is no stopping tropical cyclones from striking the coast, there is only the hope to get out of its way before it strikes.  In 1900 a hurricane killed an estimated 6,000 people in Galveston, Texas.  The population had no idea that a hurricane was about to strike.  A similar situation occurred in 1938 when a hurricane killed more than 600 in New England.  Hurricane forecasters at the time made a forecast that followed climatology when they said that the storm would move out to sea.  Before weather satellites were launched into space in the 1960's meteorologist had only sparse reports from ships to tell what a storm was doing.  As satellite technology has increase so has hurricane forecasting.  Until recently, hurricane forecasters have only been able to use satellite images to "nowcast" hurricanes and the only indications about the future track and intensity of the storm would come from "hurricane hunter" aircraft reconnaissance.  This information is then put into special hurricane numerical weather prediction models, such as the Geophysical Fluid Dynamics Lab's (GFDL) model.  But now at the turn of the century remote sensing technology has advanced enough so that the information retrieved from satellite passes can be assimilated directly into hurricane models.  With more information about the hurricane's environment hurricane model track and intensity forecasts will improve as more remotely sensed data is input into hurricane models.

These satellites remotely sense many variables crucial in forecasting a hurricane's track and intensity.  These variables include sea surface temperature (SST), hurricane heat potential, the El Nino/La Nina phenomenon, cloud top temperature, rainfall intensity, the environmental steering flow, upper level divergence, and wind shear to name a few.  The potential applications of remotely sensed data to hurricane forecasting are shown in the figure below.

 Source

Satellite data will have the greatest impact on hurricane modeling.  The importance of hurricane modeling in tropical cyclone forecasting is shown in the figure below.

Source

Hurricane models have increased in accuracy greatly since aircraft reconnaissance began, however 72 hour forecasts still have a forecast error of about 250 nautical miles as shown in the figure below.

 Source
 

Even though track forecast errors have decreased with time, large errors are still common.  The cost for warning US coastal residents for hurricane landfall is roughly $600,000 per mile of coastline (McAdie and Lawrence 2000).  The current overwarning ratio of US coastline is roughly 3:1.  The goal of the National Hurricane Center (NHC) is to decrease the overwarning ratio to 2:1 (Aberson and Franklin 1999).  We believe that this goal can be more easily met by the further assimilation of remotely sensed data into hurricane model forecasts.
 
 
 

Hypothesis:More accurate tropical cyclone forecasts of track and intensity can be achieved through the improvement of remote sensing observations and by further assimilation of this data into hurricane models.

NOTE: Despite the evidence to suggest that remote sensing has greatly enhanced hurricane forecasting, it is important to note that there is no significant scientific proof.  All of this research is a hypothetical work in progress.
 

 Remote Sensing of the Oceans