Hurricane Model Predictions

Introduction

The Hurricane Research Division (HRD) (which is part of the Atlantic Oceanographic and Meteorological Laboratory (AOML) of the National Oceanic and Atmospheric Administration (NOAA)) produces strike probablity maps twice daily to help track the movement of hurricanes. Hurricane Isabel, which traversed the Atlantic during September of 2003 and made landfall in North Carolina on September 18, was chosen to produce a composite map that is described below. It is the purpose of the composite to give the viewer an idea of how accurate the predictions are and to facilitate a discussion of the predictions and what they mean to those potentially in the path of a storm.

The Strike Probabililty Maps

Below is a portion of the 5 PM EDT, September 8, 2003 strike probablity map for Isabel. It is also the first map incorporated into the composite. (All maps were taken from the Tropical Prediction Center site.)

Each probablility map produced by the NWS indicates the last charted position of the eye of the hurricane (colored circle), followed by a connected chain of 4 black dots representing the predicted location at 12-hour intervals. Subsequent dots in the chain represent additional 24 hour predictions. Note that each dot is labeled with the time predicted.

The tear drop envelope, filled in white, shows the danger area of the 3-day prediction. Danger areas are created by forming circles about the 1, 2, and 3 day forecast positions, that represent the 10-year average error in the predictions. These circles establish the outline for the envelope. On average, the hurricanes stay withing the envelpe about 60% of the time. (The average error values for the 1, 2, and 3 day forecast positions are, very roughly, 100, 200, and 300 nautical miles. These values are updated each year. Here is an illustration of how the envelope is created.)
(Special thanks to James L. Franklin, Hurricane Specialist at the National Hurricane Center.)

The extended portion of the envelope, filled in with hatched marks, is the additional 4 to 5 day range. Visit the Tropical Storm Center's animated archives to see animated sequences of the 3 and 5-day warnings, strike probablities, and wind swaths for Isabel. For more information on forecasts, check out the Hurricane Research Division's FAQ site.

The Composite Map

The Composite Probability Map for Hurricane Isabel - 2003 map shown below is a composite of a series of strike probablity maps. The series consists of seven 5 PM EDT maps, each two days apart, from Sept 8 to Sept 18.

As with the individual maps, the actual locations of the hurricane are given by the solid colored circle at the apex of each probablility envelope. The position that was predicted by the model 2-days earlier, has been added and is shown by the same colored "X" on top of the black dot. It should be noted that each actual position is for 5 PM EDT on that day, whereas the model prediction is for 2 PM of that same day. Hence the predicted position is for 3 hours before the indicated actual position. No attempt was made to adjust for this small discrepancy. For the marked positions #1, #5 and #7, this might have slightly improved the concurrence of the actual and predicted positions, while it would have exacerbated the difference for positions #3, #4 and #6.

The envelopes were overlapped such that later predictions are on top of earlier predictions. In addition, the date labels for the predicted positions (seen in the example above), have been removed from the composite for clarity.

Composite Probability Map for Hurricane Isabel - 2003

Details of the map construction .

How Good are the Predictions?

The composite map we can draw a few conclusions.

• First, note that all of the 3-day prediction envelopes lie within the previous 5-day prediction envelopes. All of the hurricane locations were within the 3-day prediction envelopes.

• The predictions were the least accurate during the first four days shown, when the hurricane was far out to sea. The primary reason for this is that oceanographic and atmopsheric data for this region is not as precise as that nearer to the US coast. The Aircraft Operations Center (an organization of the NOAA Marine and Aviation Office) does not begin its reconnaisance flights until the storm is nearing the Lessor Antilles island chain. Once those flights begin, information about the storm and the surrounding atmopshere becomes more precise.

• The predictions starting on September 10 were considerably better. While this was due in part to better data, it might also have been due to more stable conditions in the surrounding atmosphere. When major features, such as the Bermuda-Azores high, remain relatively constant, the models are more accurate.

Implications for the Caribbean Islands vs the Continental US

The US Virgin Islands were within the 5-day envelope from September 6 until Sept 10. That spans the first three envelopes shown in the composite map. The Virgin Islands were just outside the 3-day envelope of the September 10 prediction. It wasn't until around Sept 10 that area residents began to relax. At that time, Isabel was nearly 200 miles north of our latitude. Hurricanes rarely travel south in this region, but rarely does not mean never. It is important to keep abreast of the latest HRC updates.

Comparing the coastal US mainland to the Caribbean islands, there are significant differences in the preparation for approaching hurricanes and subsequent recovery. Consider the following:

• Due to the implications of heavy road traffic, residents of the more populated areas along the US mainland coast must start evacuations soon. Major cities, such as Miami, must begin evacuation more than 24 hours in advance. Residents of small cities or rural areas may be able to wait until a few hours before hurricane force winds begin.

  For the islands, driving away is not an option. Most residents will ride out the storm due to the high cost of flying to a safe(r) destination. For those few who chose to leave, they often must make their decisions more than 24 hours in advance, since airports will be closed long before hurricane force winds are imminent.

• Mainland residents preparing for a hurricane know that once the hurricane passes, help typically arrives within 24 hours. For islanders, it is understood that it will be many days before stores will be reopened and restocked. It will be perhaps weeks before major equipment can be moblilized to begin reconstruction. The recovery period is considerably longer for islands far from the mainland.

• The southern and eastern coasts of the US mainland are relatively flat. They are also the boundary between the ocean and the large land mass of the continental US. This has several implications, some that are similiar to and some that are distinct from the Caribbean islands.
  1. The winds encountering the relatively flat land along the US coasts are less likely to be accelerated due to topological features. In contrast, many Caribbean islands are mountainous. Steep hills and valleys form channels through which winds are accelerated. Evidence of the occurrence of tornadoes during hurricanes is well documented on Caribbean islands. In addition, sustained wind speeds in hurricanes increase with elevation. Hence, residents who live at higher elevations are generally subjected to winds considerably greater than the sustained surface wind speeds assigned to hurricanes.

     It should be noted that the US mainland is not entirely immune to higher elevation and accelerated wind effects. For large hurricanes, storm force winds can occur many miles inland. The mountainous inland areas of the Carolinas are ocassionally subject to acclerated and high elevation winds.

  2. When a hurricanes moves over a large land mass, it quickly begins to dissipate. Moisture, which was provided by the warm ocean waters, is not available from the land. This greatly reduces the process of condensation that is the major source of energy for hurricanes. The net result is that when a storm moves inland, the winds and rain begin to diminish at a much faster rate than would occur solely due to the passing of the storm.

     For the majority of the Caribbean islands, the land mass area is too small to have any significant effect on hurricane strength. Local research indicates that islands of roughly the size of Puerto Rico and larger can decrease wind speeds by 5 to 10 mph on hurricanes that move directly over them. But islands the size of St. Croix and St. Thomas have no statistical effect on the strength of hurricanes.

  3. The most devastating damage from a hurricane can be caused by storm surge. The surface winds create a "piling up" of ocean water. (To a lesser extent, low pressure also contributes to the surge.) Storm surge can be several meters deep and can completely inundate low lying areas. The force of a moving wall of water can decimate even the strongest structures. The US coast is vulnerable in many regions. The Carolina coast and the greater New Orleans area are such regions.

     In the Caribbean, there is a wide range in island topologies and in the distribution of populations. Relative flat islands such as Anegada, BVI are particularly vunerable. But mountainous islands are still at risk if a significant portion of the residences and businesses are located near sea level. In 1995, downtown Charlotte Amalie on St. Thomas suffered a storm surge from hurricane Marilyn that left a waist-high water mark on most buildings. Many sea side hotels were also severely damaged.

Prepared by Dr. David J. Smith. Email to dsmith@uvi.edu