Preliminary Report
Hurricane Andrew
16 - 28 August, 1992

STORM IMAGES

The last radar image taken from NHC before the WSR-57 radar was blown off the roof, 0835 UTC August 24, 1992. (120K JPEG)

Infrared image of Andrew over Dade county at 0900 UTC August 24, 1992. (566K GIF)
The landfall pressure has since been revised to 922 mb.

NWS Melbourne WSR-88D image of Andrew over South Florida, 1011 UTC August 24, 1992. (99K GIF)

Map of selected observations in Dade county. (24K GIF)

Visible image of Andrew approaching south Louisiana at 2000 UTC August 25, 1992. (349K GIF)

NWS Houston WSR-88D image of Andrew making landfall over Louisiana, 0746 UTC August 26, 1992. (99K GIF)

The roof sign says it all! (555K GIF)

Lakes by the Bay (751K GIF)

Naranja Lakes (524K GIF)

Naranja Lakes (563K GIF)

Naranja Lakes (520K GIF)

Homestead Fl. (530K GIF)

Homestead Fl. (434K GIF)

Homestead Fl. (324K GIF)

Additional damage images courtesy of Roger Edwards of the Storm Prediction Center.

Andrew was a small and ferocious Cape Verde hurricane that wrought unprecedented economic devastation along a path through the northwestern Bahamas, the southern Florida peninsula, and south-central Louisiana. Damage in the United States is estimated to be near 25 billion, making Andrew the most expensive natural disaster in U.S. history¹. The tropical cyclone struck southern Dade County, Florida, especially hard, with violent winds and storm surges characteristic of a category 4 hurricane (see addendum on upgrade to category 5) on the Saffir/Simpson Hurricane Scale, and with a central pressure (922 mb) that is the third lowest this century for a hurricane at landfall in the United States. In Dade County alone, the forces of Andrew resulted in 15 deaths and up to one-quarter million people left temporarily homeless. An additional 25 lives were lost in Dade County from the indirect effects of Andrew². The direct loss of life seems remarkably low considering the destruction caused by this hurricane.

a. Synoptic History

Satellite pictures and upper-air data indicate that Hurricane Andrew formed from a tropical wave that crossed from the west coast of Africa to the tropical North Atlantic Ocean on 14 August 1992. The wave moved westward at about 20 kt, steered by a swift and deep easterly current on the south side of an area of high pressure. The wave passed to the south of the Cape Verde Islands on the following day. At that point, meteorologists at the National Hurricane Center (NHC) Tropical Satellite Analysis and Forecast (TSAF) unit and the Synoptic Analysis Branch (SAB) of the National Environmental Satellite Data and Information Service (NESDIS) found the wave sufficiently well-organized to begin classifying the intensity of the system using the Dvorak (1984) analysis technique.

Convection subsequently became more focused in a region of cyclonic cloud rotation. Narrow spiral-shaped bands of clouds developed around the center of rotation on 16 August. At 1800 UTC on the 16th (UTC precedes EDT by four hours), both the TSAF unit and SAB calculated a Dvorak T-number of 2.0 and the "best track" (Table 1 and Fig. 1 [85K GIF]) shows that the transition from tropical wave to tropical depression took place at that time.

The depression was initially embedded in an environment of easterly vertical wind shear. By midday on the 17th, however, the shear diminished. The depression grew stronger and, at 1200 UTC 17 August, it became Andrew, the first Atlantic tropical storm of the 1992 hurricane season. The tropical cyclone continued moving rapidly on a heading which turned from west to west-northwest. This course was in the general direction of the Lesser Antilles.

Between the 17th and 20th of August, the tropical storm passed south of the center of the high pressure area over the eastern Atlantic. Steering currents carried Andrew closer to a strong upper-level low pressure system centered about 500 n mi to the east-southeast of Bermuda and to a trough that extended southward from the low for a few hundred miles. These currents gradually changed and Andrew decelerated on a course which became northwesterly. This change in heading spared the Lesser Antilles from an encounter with Andrew. The change in track also brought the tropical storm into an environment of strong southwesterly vertical wind shear and quite high surface pressures to its north. Although the estimated maximum wind speed of Andrew varied little then, a rather remarkable evolution occurred.

Satellite images suggest that Andrew produced deep convection only sporadically for several days, mainly in several bursts of about 12 hours duration. Also, the deep convection did not persist. Instead, it was stripped away from the low-level circulation by the strong southwesterly flow at upper levels. Air Force Reserve unit reconnaissance aircraft investigated Andrew and, on the 20th, found that the cyclone had degenerated to the extent that only a diffuse low-level circulation center remained. Andrew's central pressure rose considerably (Fig. 2 [87K GIF]). Nevertheless, the flight-level data indicated that Andrew retained a vigorous circulation aloft. Wind speeds near 70 kt were measured at an altitude of 1500 ft near a convective band lying to the northeast of the low-level center. Hence, Andrew is estimated on 20 August to have been a tropical storm with 40 kt surface winds and an astonishingly high central pressure of 1015 mb (Figs. 2 and 3 [87K GIF]).

Significant changes in the large-scale environment near and downstream from Andrew began by 21 August. Satellite imagery in the water vapor channel indicated that the low aloft to the east-southeast of Bermuda weakened and split. The bulk of the low opened into a trough which retreated northward. That evolution decreased the vertical wind shear over Andrew. The remainder of the low dropped southward to a position just southwest of Andrew where its circulation enhanced the upper-level outflow over the tropical storm. At the same time, a strong and deep high pressure cell formed near the U.S. southeast coast. A ridge built eastward from the high into the southwestern Atlantic with its axis lying just north of Andrew. The associated steering flow over the tropical storm became easterly. Andrew turned toward the west, accelerated to near 16 kt, and quickly intensified.

Andrew reached hurricane strength on the morning of 22 August, thereby becoming the first Atlantic hurricane to form from a tropical wave in nearly two years. An eye formed that morning and the rate of strengthening increased. Just 36 hours later, Andrew reached the borderline between a category 4 and 5 hurricane (see addendum on upgrade to category 5) and was at its peak intensity (Table 1). From 0000 UTC on the 21st (when Andrew had a barely perceptible low-level center) to 1800 UTC on the 23rd the central pressure had fallen by 92 mb, down to 922 mb. A fall of 72 mb occurred during the last 36 hours of that period and qualifies as rapid deepening (Holliday and Thompson, 1979).

The region of high pressure held steady and drove Andrew nearly due west for two and a half days beginning on the 22nd. Andrew was a category 4 hurricane when its eye passed over northern Eleuthera Island in the Bahamas late on the 23rd and then over the southern Berry Islands in the Bahamas early on the 24th. After leaving the Bahamas, Andrew continued moving westward toward southeast Florida.

Andrew weakened when it passed over the western portion of the Great Bahama Bank and the pressure rose to 941 mb. However, the hurricane rapidly reintensified during the last few hours preceding landfall when it moved over the Straits of Florida. During that period, radar, aircraft and satellite data showed a decreasing eye diameter and strengthening "eyewall" convection. Aircraft and inland surface data Fig. 4 [121K GIF]) suggest that the deepening trend continued up to and slightly inland of the coast. For example, the eye temperature measured by the reconnaissance aircraft was at least 1-2C warmer at 1010 UTC (an hour after the eye made landfall) than it was in the last "fix" about 15 n mi offshore at 0804 UTC. These measurements suggest that the convection in the eyewall, and the associated vertical circulation in the eye and eyewall, became more vigorous as the storm moved onshore. The radar data indicated that the convection in the northern eyewall became enhanced with some strong convective elements rotating around the eyewall in a counter-clockwise fashion as the storm made landfall. Numerical models suggest that some enhancement of convection can occur at landfall due to increased boundary-layer convergence in the eyewall region. That situation appeared to have occurred in Andrew. The enhanced convection in the north eyewall probably resulted in strong subsidence in the eye on the inside edge of the north eyewall. This likely contributed to a displacement of the lowest surface pressure to the north of the geometric center of the "radar eye" (cf., Fig. 4 and 6 [107K JPEG]). It is estimated that the central pressure was 922 mb at landfall near Homestead AFB, Florida at 0905 UTC (5:05 A.M. EDT) 24 August (Fig. 4).

The maximum sustained surface wind speed (1-min average at 10 meters [about 33 ft] elevation) during landfall over Florida is estimated at 125 kt (about 145 mph), with gusts at that elevation to at least 150 kt (about 175 mph). The sustained wind speed corresponds to a category 4 hurricane on the Saffir/Simpson Hurricane Scale (see addendum on upgrade to category 5). It should be noted that these wind speeds are what is estimated to have occurred within the (primarily northern) eyewall in an open environment such as at an airport, at the standard 10-meter height. The wind experienced at other inland sites was subject to complex interactions of the airflow with trees, buildings, and other obstacles in its path. These obstructions create a turbulent, frictional drag that generally reduces the wind speed. However, they can also produce brief, local accelerations of the wind immediately adjacent to the structures. Hence, the wind speed experienced at a given location, such as at a house in the core region of the hurricane, can vary significantly around the structure, and cannot be specified with certainty. The landfall intensity is discussed further in Section b.

Andrew moved nearly due westward when over land and crossed the extreme southern portion of the Florida peninsula in about four hours. Although the hurricane weakened about one category on the Saffir/Simpson Hurricane Scale during the transit over land, and the pressure rose to about 950 mb, Andrew was still a major hurricane when its eyewall passed over the extreme southwestern Florida coast.

The first of two cycles of modest intensification commenced when the eye reached the Gulf of Mexico. Also, the hurricane continued to move at a relatively fast pace while its track gradually turned toward the west-northwest.

When Andrew reached the north-central Gulf of Mexico, the high pressure system to its northeast weakened and a strong mid-latitude trough approached the area from the northwest. Steering currents began to change. Andrew turned toward the northwest and its forward speed decreased to about 8 kt. The hurricane struck a sparsely populated section of the south-central Louisiana coast with category 3 intensity at about 0830 UTC on the 26th. The landfall location is about 20 n mi west-southwest of Morgan City.

Andrew weakened rapidly after landfall, to tropical storm strength in about 10 hours and to depression status 12 hours later. During this weakening phase, the cyclone moved northward and then accelerated northeastward. Andrew and its remnants continued to produce heavy rain that locally exceeded 10 inches near its track (Table 2b). By midday on the 28th, Andrew had begun to merge with a frontal system over the mid-Atlantic states.

b. Meteorological Statistics

The best track intensities were obtained from the data presented in Figs. 2, 3, 4, and 5 (95K GIF). The first two of those figures show the curves of Andrew's central pressure and maximum sustained one-minute wind speed, respectively, versus time, along with the observations on which they were based. The figures contain relevant surface observations and intensity estimates derived from analyses of satellite images performed by the TSAF unit, SAB and the Air Force Global Weather Central (USAF in figures). The aircraft data came from reconnaissance flights by the U.S. Air Force Reserve 815th Weather Reconnaissance Squadron based at Keesler AFB, Mississippi. Additional data were collected aboard a NOAA aircraft.

Table 2 lists a selection of surface observations. The anemometer at Harbour Island, near the northern end of Eleuthera Island in the Bahamas, measured a wind speed of 120 kt for an unknown period shortly after 2100 UTC on the 23rd. That wind speed was the maximum that could be registered by the instrument.

Neither of the two conventional measures of hurricane intensity, central barometric pressure and maximum sustained wind speed, were observed at official surface weather stations in close proximity to Andrew at landfall in Florida. Homestead Air Force Base and Tamiami Airport discontinued routine meteorological observations prior to receiving direct hits from the hurricane. Miami International Airport was the next closest station, but it was outside of the eyewall by about 5 nautical miles when Andrew's center passed to the south of that airport.

To supplement the official information, requests for data were made to the public through the local media. Remarkably, more than 100 quantitative observations were received (see Figs. 4 and 5). Many of the reports came from observers who vigilantly took readings through frightening conditions including, in several instances, the moment when their instruments and even their homes were destroyed.

Some of the unofficial observations were dismissed as unrealistic. Others were rendered suspect or eliminated during follow-up inquiries or analyses. The remainder, however, revealed a physically consistent and reasonable pattern.

1. Minimum pressure over Florida

The final offshore "fix" by the reconnaissance aircraft came at 0804 UTC and placed the center of the hurricane only about 15 nautical miles, or roughly one hour of travel time, from the mainland. A dropsonde indicated a pressure of 932 mb at that time. The pressure had been falling at the rate of about 2 mb per hour, but the increasing interaction with land was expected to at least partially offset, if not reverse, that trend. Hence, a landfall pressure within a few millibars of 932 mb seemed reasonable.

Shortly after Andrew's passage, however, reports of minimum pressures below 930 mb were received from the vicinity of Homestead, Florida (Fig. 4). Several of the barometers displaying the lowest pressures were subsequently tested in a pressure chamber and calibrated by the Aircraft Operations Center (AOC) of NOAA. Two key observations came from a Mrs. Hall and Mr. Martens, sister and brother. They rode out the storm in residences about one-quarter mile apart. Mrs. Hall's home was built by her father and grandfather in 1945 to be hurricane-proof. Although some of the windows broke, the 22-inch thick concrete and coral rock walls held steady, allowing her to observe her barometer in relative safety. The AOC tests indicate that the minimum pressure at her home was near 921 mb. The barometer at her brother's home was judged a little more reliable and the reading there was adjusted to 923 mb. Based on the observations and an eastward extrapolation of the pressure pattern to the coastline, Andrew's minimum pressure at landfall is estimated to be 922 mb. This suggests that the trajectory of the dropsonde deployed from the aircraft did not intersect the lowest pressure within the eye.

In the United States, this century, only the Labor Day (Keys') Storm in 1935 [103K GIF] (892 mb) and Hurricane Camille in 1969 [122K GIF] (909 mb) had lower landfall central pressures than Andrew (Hebert et al. 1992).

2. Maximum wind speed over Florida (see addendum on upgrade to category 5)

The strongest winds associated with Andrew on 24 August likely occurred in the hurricane's northern eyewall. The relatively limited number of observations in that area greatly complicates the task of establishing Andrew's maximum sustained wind speed and peak gust at landfall in Florida. While a universally accepted value for Andrew's wind speed at landfall may prove elusive, there is considerable evidence supporting an estimate of about 125 kt for the maximum sustained wind speed, with gusts to at least 150 kt (Fig. 5). (Please see addendum on upgrade to category 5.)

The strongest reported sustained wind near the surface occurred at the Fowey Rocks weather station at 0800 UTC (Fig. 5). The station sits about 11 n mi east of the shoreline and, at that time, was within the northwest part of Andrew's eyewall. The 0800 UTC data included a two-minute wind of 123 kt with a gust to 147 kt at a platform height of about 130 ft. The U.S. National Data Buoy Center used a boundary-layer model to convert the sustained wind to a two-minute wind of 108 kt at 33 ft elevation. The peak one-minute wind during that two-minute period at Fowey Rocks might have been slightly higher than 108 kt.

It is unlikely that this point observation was so fortuitously situated that it represents a sampling of the absolute strongest wind. The Fowey Rocks log (not shown) indicates that the wind speed increased through 0800 UTC. Unfortunately, Fowey Rocks then ceased transmitting data, presumably because even stronger winds disabled the instrumentation. (A subsequent visual inspection indicated that the mast supporting the anemometer had become bent 90 degrees from vertical.) Radar reflectivity data suggests that the most intense portion of Andrew's eyewall had not reached Fowey Rocks by 0800 UTC and that the wind speed could have continued to increase there for another 15 to 30 minutes. A similar conclusion can be reached from the pressure analysis in Fig. 4 which indicates that the pressure at Fowey Rocks probably fell by about another 20 mb from the 0800 UTC mark of 968 mb.

Reconnaissance aircraft provided wind data at a flight level of about 10,000 ft. The maximum wind speed along 10 seconds of flight track (often used by the NHC to represent a one-minute wind speed at flight level) on the last pass prior to landfall was 162 kt, with a spot wind speed of 170 kt observed. The 162 kt wind occurred at 0810 UTC in the eyewall region about 10 n mi to the north of the center of the eye. Like the observation from Fowey Rocks, the aircraft provided a series of "point" observations (i.e., no lateral extent). Somewhat higher wind speeds probably occurred elsewhere in the northern eyewall, a little to the left and/or to the right of the flight track. A wind speed at 10,000 ft is usually reduced to obtain a surface wind estimate. Based on past operational procedures, the 162 kt flight-level wind is compatible with maximum sustained surface winds of 125 kt.

One of the most important wind speed reports came from Tamiami Airport, located about 9 n mi west of the shoreline. As mentioned earlier, routine weather observations ended at the airport before the full force of Andrew's (northern) eyewall winds arrived. However, the official weather observer there, Mr. Scott Morrison, remained on-station and continued to watch the wind speed dial. Mr. Morrison notes that around 0845 UTC (0445 EDT) the wind speed indicator "pegged" at a position a little beyond the dial's highest marking of 100 kt, at a point that he estimates corresponds to about 110 kt. (Subsequent tests of the wind speed dials observed at the airport indicate that the needles peg at about 105 kt and 108 kt, respectively). He recounts that the needle was essentially fixed at that spot for three to five minutes, and then fell back to 0 when the anemometer failed. Mr. Morrison's observations have been closely corroborated by two other people. He has also noted that the weather conditions deteriorated even further after that time and were at their worst about 30 minutes later. This information suggests that, in all likelihood, the maximum sustained wind speed at Tamiami Airport significantly exceeded 105 kt.

A number of the wind speeds reported by the public could not be substantiated and are therefore excluded from Fig. 5. The reliability of some of the others suffer from problems that include non-standard averaging periods and instrument exposures, and equipment failures prior to the arrival of the strongest winds.

The only measurement of a sustained wind in the southern eyewall came from an anemometer on the mast of an anchored sailboat (see Fig. 5). For at least 13 minutes the anemometer there showed 99 kt, which was the maximum that the readout could display. A small downward adjustment of the speed should probably be applied because the instrument was sitting 17 m above the surface rather than at the standard height of 10 m. On the other hand, the highest one-minute wind speed during that 13-minute period could have been quite a bit stronger than 99 kt. Again, there may have been stronger winds elsewhere in the southern eyewall. For a westward-moving hurricane the wind speed in the northern eyewall usually exceeds the wind speed in the southern eyewall by about twice the forward speed of the hurricane (Dunn and Miller 1964). In the case of Andrew, that difference is about 32 kt, and suggests a maximum sustained wind stronger than 130 kt.

Several indirect measures of the sustained wind speed are of interest. First, a standard empirical relationship between central pressure and wind speed (Kraft 1961) applied to 922 mb yields around 135 kt. Second, the Dvorak technique classification performed by the NHC Tropical Satellite Analysis and Forecast unit using a 0900 UTC satellite image gives 127 kt. Also, an analysis of the pressure pattern in Fig. 4 gives a maximum gradient wind of around 140 kt.

The strongest gust reported from near the surface occurred in the northern eyewall a little more than a mile from the shoreline at the home of Mr. Randy Fairbank. He observed a gust of 184 kt moments before portions of a windward wall failed, preventing further observation. The hurricane also destroyed the anemometer. To evaluate the accuracy of the instrument, three anemometers of the type used by Mr. Fairbank were tested in a wind tunnel at Virginia Polytechnic Institute and State University. Although the turbulent nature of the hurricane winds could not be replicated, the results of the wind tunnel tests suggest that the gust Mr. Fairbank observed was less than 184 kt and probably near 154 kt. Of course, stronger gusts may have occurred there at a later time, or at another site. Damage at that location was significantly less than the damage to similar structures located about 2 miles south of this neighborhood, implying even stronger winds than observed at this location.

Strong winds also occurred outside of the eyewall, especially in association with convective bands (Fig. 6). A peak gust to 139 kt was observed at a home near the northern end of Dade County (Fig. 5) on an anemometer of the brand used by Mr. Fairbank. Applying the reduction suggested by the wind tunnel tests to 139 kt yields an estimate close to the 115 kt peak gust (a five-second average) registered on a National Ocean Survey anemometer located not far to the east, at the coast.

3. Storm surge

During the afternoon of 23 August, Andrew crossed over the north end of the island of Eleuthera in the Bahamas and generated significant storm surge flooding. Two high water marks were recorded and referenced to mean sea level. The first mark of 16 ft was recorded in a house in the town of Little Bogue. The second mark of 23 ft was recorded in a damaged house in the town of The Current several miles west of Lower Bogue. Since this structure was located near the shoreline it suggests that battering waves riding on top of the storm surge helped to create this very high water mark.

During the morning hour of 24 August, Andrew generated storm surge along shorelines of southern Florida (Fig. 7) (103K GIF). On the southeast Florida coast, peak storm surge arrived near the time of high astronomical tide. The height of the storm tide (the sum of the storm surge and astronomical tide, referenced to mean sea level) ranged from 4 to 6 ft in northern Biscayne bay increasing to a maximum value of 16.9 ft at the Burger King International Headquarters, located on the western shoreline in the center of the bay, and decreasing to 4 to 5 ft in southern Biscayne Bay. The observed storm tide values on the Florida southwest coast ranged from 4 to 5 ft near Flamingo to 6 to 7 ft near Goodland.

Storm tides in Louisiana were at least 8 ft (Table 2a) and caused flooding from Lake Borgne westward through Vermillion Bay.

4. Tornadoes

There have been no confirmed reports of tornadoes associated with Andrew over the Bahamas or Florida. Funnel sightings, some unconfirmed, were reported in the Florida counties of Glades, Collier and Highlands, where Andrew crossed in daylight. In Louisiana, one tornado occurred in the city of Laplace several hours prior to Andrew's landfall. That tornado killed 2 people and injured 32 others. Tornadoes in the Ascension, Iberville, Baton Rouge, Pointe Coupee, and Avoyelles parishes of Louisiana reportedly did not result in casualties. Numerous reports of funnel clouds were received by officials in Mississippi and tornadoes were suspected to have caused damage in several Mississippi counties. In Alabama, the occurrence of two damaging tornadoes has been confirmed over the mainland while another tornado may have hit Dauphin Island. As Andrew and its remnants moved northeastward over the eastern states, it continued to produce severe weather. For example, several damaging tornadoes in Georgia late on 27 August were attributed to Andrew.

5. Rainfall

Andrew dropped sufficient rain to cause local floods even though the hurricane was relatively small and generally moved rather fast. Rainfall totals in excess of seven inches were recorded in southeast Florida, Louisiana, and Mississippi (Table 2b). Rainfall amounts near five inches occurred in several neighboring states. Hammond, Louisiana reported the highest total, 11.92 inches.

c. Casualty and Damage Statistics

Table 3 lists a count of casualties and damages associated with Andrew. The number of deaths directly attributed to Andrew is 26. The additional indirect loss of life brought the death toll to 65 (see footnote 2). A combination of good hurricane preparedness and evacuation programs likely helped minimize the loss of life. Nevertheless, the fact that no lives were lost in the United States due to storm surge is viewed as a fortunate aberration.

Table 3a reveals that more than one-half of the fatalities were indirect. Many of the indirect deaths occurred during the "recovery phase" following Andrew's passage.

Damage is estimated at $25 billion. Andrew's impact on southern Dade County, Florida was extreme from the Kendall district southward through Homestead and Florida City, to near Key Largo (Table 3b). Andrew reportedly destroyed 25,524 homes and damaged 101,241 others. The Dade County Grand Jury reported that ninety percent of all mobile homes in south Dade County were totally destroyed. In Homestead, more than 99% (1167 of 1176) of all mobile homes were completely destroyed. The Miami Herald reported $0.5 billion in losses to boats in southeast Florida.

The most devasted areas correspond closely in location to the regions overspread by Andrew's eyewall and its accompanying core of destructive winds and, near the coastline, decimating storm surges. Flight-level data about an hour prior to landfall places the radius of maximum wind at 11 n mi (in the northern eyewall at 10,000 ft altitude). The radius of maximum wind at the surface was likely a little less than 11 n mi. (Figure 6) displays a radar reflectivity pattern (similar to rainfall intensity) about 30 minutes prior to landfall, superimposed on a map of southern Florida, from which it can be seen that the average diameter of the "radar" eye was about 11 n mi at landfall.)

The damage to Louisiana is estimated at $1 billion.

Damage in the Bahamas has been estimated at $0.25 billion.

Andrew whipped up powerful seas which extensively damaged many offshore structures, including the artificial reef system of southeast Florida. For example, the Belzona Barge is a 215 ft, 350-ton barge that, prior to Andrew, was sitting in 68 ft of water on the ocean floor. One thousand tons of concrete from the old Card Sound bridge lay on the deck. The hurricane moved the barge 700 ft to the west (50-100 tons of concrete remain on deck) and removed several large sections of steel plate sidings.

Damage in the Gulf of Mexico is preliminarily estimated at $0.5 billion. Ocean Oil reported the following in the Gulf of Mexico: 13 toppled platforms, five leaning platforms, 21 toppled satellites, 23 leaning satellites, 104 incidents of structural damage, seven incidents of pollution, two fires, and five drilling wells blown off location.

Hurricanes are notoriously capricious. Andrew was a compact system. A little larger system, or one making landfall just a few nautical miles further to the north, would have been catastrophic for heavily populated, highly commercialized and no less vulnerable areas to the north. That area includes downtown Miami, Miami Beach, Key Biscayne and Fort Lauderdale. Andrew also left the highly vulnerable New Orleans region relatively unscathed.

d. Forecast and Warning Critique

Track forecast errors by the NHC and by the suite of track prediction models are given in Table 4. On average, the NHC errors were about 30% smaller than the current 10-year average. The most significant changes in Andrew's track and intensity (see Fig. 1, Table 1) were generally well anticipated (noted in NHC's Tropical Cyclone Discussions) and the forecast tracks generally lie close to the best track. However, the rate of Andrew's westward acceleration over the southwestern Atlantic was greater than initially forecast. In addition, the NHC forecast a rate of strengthening that was less than what occurred during Andrew's period of rapid deepening.

Several of the dynamic track models had stellar performances during this hurricane. The Aviation Model and a tracking routine that follows a simulated hurricane vortex (AVNO) performed especially well. However, this was the first storm for which AVNO output was available to NHC forecasters. Hence, its operational reliability was not established. The GFDL and QLM models also had small errors. It should be pointed out, however, that the NHC works on a six-hourly forecast cycle and that the models mentioned above are run just once per 12 hours. Moreover, the output from these models becomes available to forecasters no earlier than the following six-hour forecast cycle.

Historically, the NHC90 statistical-dynamical model has been the most accurate of NHC's track guidance models. The NHC90 errors were rather large during Andrew. Because the NHC90 uses output from the Aviation Model it is possible that the recent changes in the latter model may be responsible for the NHC90's degraded performance.

Table 5 lists a chronology of watches and warnings issued by the National Hurricane Center and the Government of the Bahamas. The associated lead times (based on landfall of the eye) are given in Table 6.

Massive evacuations were ordered in Florida and Louisiana as the likelihood of Andrew making landfall in those regions increased (Table 7). About 55,000 people left the Florida Keys. Evacuations were ordered for 517,000 people in Dade County, 300,000 in Broward County, 315,000 in Palm Beach County and 15,000 in St. Lucie County. For counties further west in Florida, evacuation totals exceeding one thousand people are Collier (25,000), Glades (4,000) and Lee (2,500).

It is estimated that 1,250,000 people evacuated from parishes in southeastern and south-central Louisiana.

About 250,000 people evacuated from Orange and Jefferson Counties in Texas.

The winds in Hurricane Andrew wreaked tremendous structural damage, particularly in southern Dade County. Notwithstanding, the loss of life in Hurricane Andrew, while very unfortunate, was far less than has previously occurred in hurricanes of comparable strength. Historical data suggests that storm surge is the greatest threat to life. Some lives were likely saved by the evacuation along the coastline of southeast Florida. The relatively small loss of life there serves as testimony to the success and importance of coordinated programs of hurricane preparedness.

References

Dunn, G. E. and B. I. Miller, 1964: Atlantic Hurricanes.
     Louisiana State University Press, Baton Rouge, LA. 326 pp.

Dvorak,  V.  F., 1984: Tropical  cyclone  intensity  analysis using 
     satellite  data.   NOAA  Technical  Report NESDIS 11, National 
     Oceanic  and  Atmospheric  Administration, U. S. Department of 
     Commerce, Washington, DC, 47 pp.

Hebert, P. J., J. D. Jarrell, and M. Mayfield, 1992: The deadliest, 
     costliest, and most intense  hurricane  of this century (and 
     other frequently requested facts).  NOAA Technical Memorandum 
     NWS NHC-31, National Oceanic and Atmospheric Administration, 
     U.S. Department of Commerce, Washington, DC, 40 pp.

Holliday,  C.  R.,  and   A.  H.  Thompson,  1979:   Climatological 
     characteristics  of  rapidly  intensifying typhoons. Mon. Wea. 
     Rev., 107, 1022-1034.

Kraft, R. H., 1961:  The hurricane's  central  pressure and highest 
     wind. Mar. Wea. Log., 5, 157.

Acknowledgments

Much of the data in this summary was provided by NWS WSFO/WSO reports from MIA,EYW, MLB, PBI, TBW, SIL, BTR, LCH, JAN, BHM, MOB, MEM, BPT and ATL. Sam Houston of the AOML Hurricane Research Division collected additional observations. Jerry Kranz of the NOAA Aircraft Operations Center performed the barometer calibrations. Martin Nelson provided a summary on the damages to artificial reefs adjacent to the southeast Florida coast. Joan David, Stan Goldenberg and Mike Black developed several of the figures. Sandra Potter helped prepare the manuscript.

---

 

          [1] When indirect and continuing costs are considered,
the total could ultimately rise to $40 billion, according to a
personal  communication  from William E. Bailey, Co-Director,
Hurricane Insurance Information Center. Mr. Bailey indicates that
Floridians filed more than 725,000 insurance claims related to
Andrew.

---

 

          [2] Based on data from the Dade County Medical Examiner. 
The Miami Herald reported on 31 January 1993 that it could relate
at least 43 additional (indirect) deaths in Dade County to Hurricane
Andrew.