Hurricane Andrew in Florida: Dynamics of a Disaster ^ H. E. Willoughby and P. G. Black Hurricane Research Division, AOML/NOAA, Miami, Florida ABSTRACT Four meteorological factors aggravated the devastation when Hurricane Andrew struck South Florida: completed replacement of the original eyewall by an outer, concentric eyewall while Andrew was still at sea; storm translation so fast that the eye crossed the populated coastline before the influence of land could weaken it appreciably; extreme wind speed, 82 m s_1 winds measured by aircraft flying at 2.5 km; and formation of an intense, but nontornadic, convective vortex in the eyewall at the time of landfall. Although Andrew weakened for 12 h during the eyewall replacement, it contained vigorous convection and was reintensifying rapidly as it passed onshore. The Gulf Stream just offshore was warm enough to support a sea level pressure 20-30 hPa lower than the 922 hPa attained, but Andrew hit land before it could reach this potential. The difficult-to-predict mesoscale and vortex-scale phenomena determined the course of events on that windy morning, not a long-term trend toward worse hurricanes. 1. Introduction might have been a harbinger of more devastating hur- ricanes on a warmer globe (e.g., Fisher 1994). Here When Hurricane Andrew smashed into South we interpret Andrew's progress to show that the ori- Florida on 24 August 1992, it was the third most in- gins of the disaster were too complicated to be ex- tense hurricane to cross the United States coastline in plained by thermodynamics alone. the 125-year quantitative climatology. It destroyed $25 billion in property and killed 15 people directly (Mayfield et al. 1994; Rappaport 1994). Andrew's 2. Hurricane intensification landfall coincided with a period of historically high global temperatures (Follard et al. 1992) and followed Hurricanes, by definition, sustain surface winds two other memorable hurricanes: Gilbert in 1988, greater than 33 m s_1. Their intensity is measured in which took 318 lives and established 888 hPa as a terms of maximum wind or minimum sea level pres- record for the lowest hurricane sea level pressure in sure at the storm center/^. The Saffir-Simpson scale the Atlantic basin (Lawrence and Gross 1989); and (Simpson 1974) ranks increasing intensity in catego- Hugo in 1989, which took 49 lives and destroyed ries 1-5. The threshold of category 5 is sustained (1- $7 billion in property, the largest total up to that time min average) surface wind stronger than 69 m s_1 or (Case and Mayfield 1990). These events, reinforced Pc below 920 hPa. At inland sites, the surface winds by thermodynamic arguments that relate the hurri- used to assign categories are typically 65% of the wind canes' maximum intensity to sea surface temperature measured at several kilometers altitude by aircraft. (Emanuel 1987), raised public concern that Andrew The weaker wind inland in the frictional boundary layer is caused by the greater aerodynamic roughness of the land surface. The reduction occurs over the first Corresponding author address: Dr. H. E. Willoughby, Hurricane Research Division, AOML/NOAA, 4301 Rickenbacker Cause- few kilometers inland. At sea or on the coast, the sur- way, Miami, FL 33149. face winds are stronger, 70%-80% of the flight-level E-mail: [email protected] wind, but the strongest gusts may approach the flight- In final form 15 September 1995. level wind (Powell 1982, 1987; Powell et al. 1991; Bulletin of the American Meteorological Society 543 Unauthenticated | Downloaded 10/08/21 06:01 AM UTC Black 1993). Thus, the sustained surface wind equiva- fall, the smaller effective heat capacity of the solid lent to the 82 m s_1 flight-level wind observed in An- surface leads to boundary layer cooling, reduced drew would be 53 m s_1 ashore and 57-66 m s-1 at sea. evaporation, and weakening of the storm as a whole Independent estimates of Andrew's strongest winds (Miller 1964; Tuleya 1994). This thermodynamically agree generally with these figures (Rappaport 1994), induced spin-down of the vortex is more gradual than but debris patterns at some inland sites are consis- the frictional reduction of the surface wind at the coast. tent with surface winds near the flight-level value A hurricane that has achieved the greatest intensity (Wakimoto and Black 1994). possible at sea may be modeled as a heat engine oper- Category 5 hurricanes are rare in the Atlantic. The ating between a warm reservoir at the sea surface tem- historical record shows U.S. landfall by only two, perature TS and a cold reservoir at the tropopause Camille in 1969 and the Labor Day Hurricane in 1935. temperature Tr Figure 1 shows that Ps, the lowest pos- Category 4 hurricanes cross the U.S. coastline more sible/^, becomes lower as the thermodynamic efficiency often, every 6-7 years on average. No hurricane with (TS - TT)/TS increases (Emanuel 1988). An independent Pc < 900 hPa has reached the U.S. mainland since empirical estimate of Ps can be made by tabulating the 1 1900, although three have occurred in the Atlantic greatest intensity ever observed as a function of Ts basin (Hebert et al. 1992). Northwestern-Pacific ty- (DeMaria and Kaplan 1994). An example of this es- phoons reach category 4 and 5 intensities more fre- timate also appears in Fig. 1. The thermodynamic es- quently. Rapid deepening, defined as a rate of pressure timate of Ps falls 10 hPa for a 1°C increase in Ts near fall > 42 hPa in 24 h, is the mode of intensifica- 30°C, but the empirical estimate falls 17 hPa per de- tion in three quarters of the typhoons that reach Pc gree. The more rapid fall of empirical Ps stems from < 920 hPa and in all that reach Pc < 900 hPa (Holliday a climatological correlation between TS and TT\ the and Thompson 1979). The most intense recent Atlan- tropopause is higher and colder in low latitudes where tic hurricanes (Camille, Allen, Gloria, Gilbert, and the sea is warmer. Figure 1 conveys an illusory impres- Hugo) reached their lowest surface pressures after a sion of precision. Sea surface temperature is typically day or two of rapid deepening. Rapid deepening of- known to ±1°C, and its range in the Tropics is 4-5°C. ten follows interactions between tropical cyclones and The high-intensity ends of the curves are supported midlatitude troughs or subtropical lows. The mecha- by only a handful of cases. Indeed, one interpretation nism involves secondary circulations forced by upper- of the observations is that Ts is not the controlling fac- tropospheric eddy convergence of angular momentum tor in intensity of most hurricanes (Evans 1993). On (Pfeffer 1958; Pfeffer and Challa 1981,1992; Holland any given day during hurricane season, much of the and Merrill 1984; Molinari and Vollaro 1989; DeMaria et al. 1993). The dynamics of the eddies are not well understood, but it is clear that the momen- tum convergence forces axisymmetric outflow just below the tropopause with deep compensating inflow in the lower troposphere joined to the outflow by broad ascent around the hurricane's center. The as- cent causes intensification by destabilizing the column and enhancing convective heat release. Hurricanes' low pressures are driven by thermody- namic disequilibrium between the ocean and the at- mosphere. Evaporation from the sea supplies latent heat to the surface boundary layer that, in turn, feeds cumulonimbus convection around the eye. Entrain- ment into the convection forces a midlevel flow that concentrates angular momentum and spins up the FIG. 1. Minimum sea level pressure attainable in hurricanes vortex (Ooyama 1969, 1982). After hurricane land- as a function of sea surface temperature for tropopause temperatures -65°, -70°, and -75°C based upon Emanuel's (1988) argument (dotted curves) and derived form DeMaria and Kaplan's (1994) observational study (solid curve). Actual 1 The Labor Day Storm, 892 hPa; Allen of 1980, 899 hPa; and minimum sea level pressures at greatest intensity for Hurricanes Gilbert. Gilbert, Hugo, and Andrew are designated G, H, and A. 544 Vol. 77, No. 3, March 7 996 Unauthenticated | Downloaded 10/08/21 06:01 AM UTC tropical Atlantic is warm enough to produce a cat- Ps for the prevailing Ts, 30°C. Over the Bahamas at egory 4 or 5 hurricane, but few occur. 2045 UTC on 23 August, a reconnaissance aircraft The most important factor that limits both inten- reported a second eyewall at 20-km radius concen- sity of individual hurricanes and total activity on a sea- tric with the original eye, a double flight-level wind sonal basis is vertical shear of the environmental wind maximum (Fig. 3c), and a rise of Pc to 927 hPa. The (Gray 1968, 1984). Shear raises Pc by "ventilating" Miami radar, operating at extreme range, confirmed the warm vortex core that supports lowered hydro- the concentric eyewalls. By 0410 UTC on 24 August, static surface pressure (Simpson and Riehl 1958). when only the outer eyewall remained (Fig. 3d), Pc Another factor that limits intensity in low-shear situa- had risen to 941 hPa. Andrew then tracked due west tions is weakening due to replacement of a preexis- toward landfall over Homestead, Florida; the new eye tent eyewall by a newly formed outer concentric contracted to 13 km; and Pc fell to 936 hPa. Sea level eyewall. Both inner and outer eyewalls are contract- pressure at the storm center continued to fall, reach- ing rings of deep cumulonimbus that coincide with ing 922 hPa again after landfall. The aircraft encoun- local maxima of the swirling wind. Convergence in tered graupel, severe turbulence, and spectacular the frictional boundary layer localizes convection near electrical displays—both meteorological and from the the wind maximum.