The Re-analysis of Hurricane Andrew (1992)
Submission to the NHC Best Track Change Committee
8 August 2002 - Revision

This document proposes a set of revisions to the intensity portion of the “best track” for 1992's Hurricane Andrew, which is currently classified as peaking at a Saffir-Simpson Category 4 hurricane on August 23 and 24, 1992.  The revisions would make Andrew a Category 5 hurricane twice, as it was moving through the northern Bahamas and at landfall in Florida.  The maximum sustained winds experienced in south Dade County, Florida are estimated to be about 140 kt (originally 125 kt).  Changes to the intensity portion of Andrew’s best track are provided for a five day period (22 to 26 August).  Documentation for these changes are provided in the writeup below.

This proposed revision is part of “The Atlantic Hurricane Database (HURDAT) Re-analysis Project” (NOAA/NASA Grant #GC02-093) that the author is Principal Investigator for (Landsea et al. 2002).  While Hurricane Andrew would have been examined sequentially in this project sometime in late 2004 or early 2005, a request was made by the Tropical Prediction Center Director Max Mayfield to expedite Andrew’s re-analysis because of the massive societal impacts that the hurricane caused in South Florida.  For example, insurance costs for residential homes have dramatically increased in the years since 1992 for residents of Southeast Florida as a direct result of this single hurricane and today it is extremely difficult to obtain coverage from a private insurer.  Additionally, we are at the point now where credible evidence from new meteorological studies has called into question the original estimation of Hurricane Andrew’s intensity and that these new conclusions are not likely to be significantly advanced and/or altered in the next two to three years of the Re-analysis Project.  It is crucial that an accurate as possible account be provided as to the characteristics of all tropical cyclones and Hurricane Andrew in particular, since society needs reliable information on the frequency and intensity of past catastrophic events to best plan for the future.

Best Track Changes:
 No changes are proposed for the track of Hurricane Andrew (Fig. 1), which formed in the tropical North Atlantic, moved west-northwestward then westward crossing the northern Bahamas and South Florida into the Gulf of Mexico before making final landfall over Louisiana and decaying over the Southeast United States during the period 17 to 28 August, 1992.  The proposed new wind history is given in Fig. 2 and Table 1 with alterations for the dates of 22 to 26 August; no changes are proposed to the best track pressure data (Fig. 3).  (Winds from Atlantic basin tropical cyclones are recorded in HURDAT in six-hourly intervals as the maximum sustained [1 min] surface [10 m] windspeed somewhere in the circulation of the tropical cyclone.)  The original wind best track is shown in Fig. 4.  The revisions make Andrew a Category 5 hurricane on the Saffir-Simpson Hurricane Scale (Saffir 1973, Simpson 1974) at landfall in both Eleuthera Island, Bahamas and in South Florida.  The maximum sustained surface wind for Hurricane Andrew at landfall in South Florida at 0905Z 24 August is analyzed to be 150 kt for open-ocean exposure and 140 kt at the coast (originally both were assessed at 125 kt).  The peak intensity of Andrew is assessed to be 150 kt at landfall in South Florida and at 1800 UTC 23 August (originally 135 kt).  The remainder of this report provides justification for these changes along with appropriate references.

Flight level wind extrapolation to the surface:
Studies by two independent groups - Franklin et al. (2002) and Dunion et al. (2002)/Dunion and Powell (2002) - provide strong evidence that the methodology used to estimate the maximum sustained surface wind in Hurricane Andrew (Rappaport 1994; Mayfield et al. 1994; Powell and Houston 1996) was too conservative for a good portion (~5 days) of its lifetime.  This new understanding of the wind structure in strong hurricanes is due to an advance in technology - the Global Positioning System (GPS) dropwindsonde (Hock and Franklin 1999) - that allowed for the first time a detailed look at the wind profile in a hurricane’s eyewall from flight level to the ocean’s surface.  While the GPS dropwindsondes are not perfect measures of the hurricane eyewall environment as they often do fail to provide winds near the ocean’s surface under extreme conditions, they have compared favorably with nearby moored buoys/CMAN stations (Houston et al. 2000) and co-collected Stepped Frequency Microwave Radiometer data (Uhlhorn and Black 2002) - though both of these studies have limited observations in the core of strong hurricanes.  The unique new GPS dropwindsonde observations in hurricane eyewalls, first collected in Hurricane Guillermo in 1997, suggest that all hurricanes in the aircraft reconnaissance era including Hurricane Andrew should be re-examined for their intensity when the primary method for estimating surface winds were from flight-level winds extrapolated to the surface.

Franklin et al. (2002) examined several hundred over-ocean GPS dropwindsonde profiles in the hurricane eyewall and have shown that the mean ratio of surface to 700 mb winds was about 92% in the eyewall region.  However, because there was an effort to preferentially seek out the surface wind maximum at the expense of the flight-level maximum, this ratio may not be quite appropriate for estimating the peak surface wind from the peak flight-level wind.  Therefore, a separate analysis was conducted at the flight-level radius of maximum wind (RMW).  At the flight-level RMW (FL-RMW), the mean surface to 700 mb wind speed ratio was in fact a bit lower, 88%.  Since sondes dropped at the flight-level RMW generally encounter the peak flight-level wind but not the peak surface wind, the 88% ratio should represent a lower bound on the mean ratio of peak surface winds to peak 700 mb winds, if azimuthal variation is ignored.  There does exist some variability of the mean adjustment ratio with azimuth, but this variability does not change the recommended adjustment factors.   The strongest flight-level winds are generally found on the right-hand side of the storm.  The eyewall sondes show that the adjustment factor in the right quadrant is a little lower than in the left quadrant (0.89 vs. 0.93, for all eyewall sondes). Assuming this difference is applicable at the FL-RMW, a more accurate lower bound mean adjustment factor for the right quadrant FL-RMW would be just above 86%.  The upper bound on the mean adjustment factor for the right quadrant would be just under 90%.

Recent work by Dunion et al. (2002) and Dunion and Powell (2002) also support a revised flight-level to surface wind extrapolation in the context of the H*WIND surface wind analyses of tropical cyclones (Powell et al. 1996, 1998) due to the new GPS dropwindsonde observations.  Dunion et al./Dunion and Powell utilized the drop data to alter the H*WIND analysis in a two-step process.  First, analyses of the dropwindsondes shows that the assumption previously in H*WIND that 700 mb flight-level data were equivalent to a mean boundary layer (0 to 500 m) was an underestimation of the true boundary layer winds.  Secondly, the GPS dropwindsondes showed that the over-ocean surface to mean boundary layer wind ratio reached a minimum near 100 to 110 kt mean boundary layer winds and had an increasing ratio with stronger winds, in contrast to a steadily decreasing ratio with stronger boundary layer winds previously utilized in H*WIND.  The combined effect of these two changes to H*WIND produces maximum sustained surface winds for hurricanes with aircraft reconnaissance data at 700 mb substantially higher (10-20%) than the earlier version.  This new H*WIND methodology provides marine exposure surface wind analyses (from extrapolated flight-level wind observations) that agree within 5% of the Franklin et al. (2002) new estimates for major hurricanes.

Consequently, unless there is other information to the contrary, current operational practice at the National Hurricane Center (Franklin et al. 2001) is to assess the intensity of a hurricane at about 90% of the peak wind observed at the 700 mb level.  This factor is at the high end of the range established by the sondes (86%-90%), in part due to a simple rounding of the midpoint of this range, but also to account for the likelihood that the highest 700 mb wind speed was not sampled.

About an hour prior to Andrew’s landfall in South Florida, at 0810Z Air Force reconnaissance aircraft measured a 10 s wind at 700 mb of 162 kt, 90% of which is 146 kt.  Ten-second flight-level winds are customarily assumed to represent 1 min sustained winds in the inner core of hurricanes.  One minute averaging of the flight-level winds would tend to underestimate the true maximum 1 min wind because the aircraft does not remain in the peak gradient region that long on a radial flight track, especially in relatively small hurricanes like Andrew.  Peak 10 s flight level winds supporting Category 5 at the surface (at least 136 kt) were found in three consecutive minute observations (0809, 0810, and 0811Z) using the Franklin et al. [2002] adjustment of 90%.  Neither were these the only aircraft obs that would support Category 5 winds; on the next pass one 10 s report (over land, at 0918Z) yields 138 kt using the same reduction.  (However, it is not known what the wind profile and surface winds are for conditions well-inland, since the vast majority of GPS dropwindsondes were dropped over the open ocean.  It would not be appropriate to claim that these winds at 0918Z provided good evidence for maximum sustained surface winds of Category 5 conditions at that time.  The point is that the Category 5 winds over water around 0810Z were not momentary fluke conditions.)  Lacking compelling contradictory data, the flight-level observations would be accepted as representative and a hurricane presenting these data today would very likely be assigned an operational intensity of about 145 kt, well above the threshold for Category 5 hurricane conditions.

Analyses from the revised H*WIND surface wind analysis system (Dunion and Powell 2002) provide a marine exposure surface wind estimate of 150 kt from the 0810Z reconnaissance flight-level data, which is a 93% surface to flight-level ratio (Fig. 5).  This result is in agreement with the Franklin et al. (2002) estimate, perhaps not surprisingly since they are both based upon new formulations from the nearly the same set of GPS dropwindsondes.

It  has been suggested that applying the 90% rule to the flight-level data may be inconsistent with available surface data in Hurricane Andrew, in particular the Fowey Rocks and R. Fairbanks observations.  The Fowey Rocks C-MAN weather station reported a peak 2 min wind of 122 kt at an elevation of 40 m, which adjusts to about 111 kt for a maximum 1 min surface [10 m] wind.  R. Fairbanks, an amateur weather observer located in Perrine, noted a peak gust of 184 kt from his home-based anemometer, which adjusts to a maximum 1 min open terrain surface wind of about 119 kt after accounting for an overestimation bias of his instrument and a typical gust factor [Rappaport 1994; Mayfield et al.1994, Powell et al. 1996].  The Fairbanks observation becomes approximately 127 kt, if it were converted to a marine exposure (M. Powell, personal communication).  While the Fowey Rock station reported windspeed and direction until its failure around 0800Z, the Fairbanks observation has no wind direction associated with it and only an approximate time (0830 to 0900Z).  Both of these instruments likely failed before the strongest winds of Andrew arrived, as they were in the northwest portion of the eyewall outside of the radius of maximum winds while the peak winds were closer to the storm’s center in the northern portion of the eyewall.  (The Fairbank observation may instead have been in the north eyewall, but its storm-relative location at the time of the peak gust in very uncertain.)  Thus neither of these observations represent the maximum sustained winds of Hurricane Andrew at landfall in South Florida.  Inspection of these surface observations in comparison with reduced-to-the- surface flight level data do not suggest any large inconsistency, though it is difficult to directly compare them for two reasons.  First, the flight level data primarily were in radial legs running north-south and east-west, which did not coincide well spatially with the Fowey Rocks and Fairbanks observations.  Secondly, because of the turbulent and transient nature of hurricane wind field, it is not straightforward to make direct comparisons from a storm-relative perspective of the adjusted flight-level winds to the couple of observations available.  It would take a systematic discrepancy over many observation points to warrant a departure from the mean relationship.  This does not appear to be the case in Andrew’s landfall in South Florida.

An additional surface observation that provides some insight into Andrew’s intensity is the Mara Cu sailboat measurement in the south eyewall, while the ship was moored north of Key Largo.  While the anemometer was pegged at 99 kt (the maximum that could be read on the dial) for several minutes not allowing for an exact measurement of the peak winds, Powell et al. (1996) estimated that these observations were on the order of 110 kt 1 min sustained surface winds.  These measurements do allow for a more direct comparison of the new flight-level extrapolation due to their relative positions.  At flight level, peak winds measured by the aircraft for a 10 s observations were 134 kt at 0801Z and 116 kt at 0802Z.  The new H*WIND scheme suggest that these would correspond to sustained marine exposure surface winds of 126 kt and 121 kt, respectively, which is about 15% and 10% higher than the sailboat’s estimated observations.  Uncertainties here, however, include include whether the 110 kt sustained winds for the sailboat is the most reasonable estimate, the lack of calibration of the sailboat instrumentation at these windspeeds, and imprecise knowledge of which flight level observation corresponds best spatially to the sailboat.  Another piece of information that may be relevant is that for westward-moving tropical cyclones the surface wind speed in the north eyewall usually exceeds that of the south eyewall by twice the storm’s translational velocity (Dunn and Miller 1964).  The 110 kt observation from the sailboat along with the 16 kt translational velocity of Andrew suggests a maximum sustained surface wind of at least 142 kt.  However, this line of evidence is considered indirect at best.

Assuming that the 162 kt aircraft wind is representative of the peak winds present at 700 mb, a surface adjustment factor of 83% or less is required to keep Andrew a Category 4 hurricane.  Of  the hurricanes examined in Franklin et al. (2002), only Bonnie, a weakly convective storm with a large eyewall, has been observed to have a mean ratio that low.   If the surface winds are kept at 125 kt, this implies an adjustment factor of 77%.  No storm has yet been observed with GPS dropwindsondes to have a mean adjustment factor this low.  (However, hurricanes like Gloria of 1985 and Bob of 1991 while moving north of the Gulf Stream over cool waters were observed to have surface observations from moored buoys to flight-level wind ratios as low as 55% [P. Black, personal communication]).  Furthermore, although sample sizes tend to get small when the dropwindsonde sample gets stratified, relative surface winds appear to be enhanced when the winds are very high (Dunion et al. 2002, Franklin et al. 2002), and when vertical motions are particularly vigorous (Franklin et al. 2002).  Andrew at landfall in South Florida satisfied both of these conditions.  Thus, there is little evidence to conclude to that Andrew had a lower than normal adjustment factor.

While there does appear to be a consistent answer for the intensity of Andrew while the storm was approaching South Florida, the question then arises: Are these winds also felt along the coast line itself?  The definition of the Saffir-Simpson Hurricane Scale characterization of a hurricane at landfall in the United States is the following:
 The highest maximum sustained [1 min] wind present at 10 m at the U.S. coast or inland attributable to the cyclone circulation as the hurricane is making landfall (or a close approach). This highest wind is used to assign the appropriate Saffir-Simpson Hurricane Scale category.
Thus it is possible for a hurricane to have higher winds over the ocean classifying it one way for its overall intensity at the time (say, Category 5) and weaker winds actually impacting the coast (say, Category 4).  For Andrew, it is possible (M. Powell, personal communication) that the winds in the north eyewall’s transit over Biscayne Bay before impacting land were weakened by increased roughness presented by the bay.  Powell suggests that Biscayne Bay in this case could be considered not a typical marine exposure with small roughness length, but instead similar in conditions to over-land, open exposure conditions with roughness lengths on the order of 0.01 m.  This is a plausible hypothesis as Biscayne Bay averages on 2-4 m in depth, so that in extreme hurricane conditions one would expect an increase in breaking and shoaling waves in the bay itself.  Indeed, breaking waves were able to be observed (as enhanced reflectivity values) from the Miami WSR-57 weather radar on the reef right directly adjacent to Biscayne Bay, which indicates the large size of the wave impacts on the bay. With the value of 0.01 m roughness length (consistent with that utilized in Powell et al. 1996 and Powell and Houston 1996), the 178 kt mean boundary layer wind value (after boosting the 162 kt 700 mb flight level data) suggests     1 min 10 m maximum sustained winds over Biscayne Bay and at the coast of approximately 132 kt.  This is just below the Category 4/5 threshold using this “over-land” exposure adjustment to the Dunion and Powell (2002) H*WIND methodology.  Figure 6 presents some preliminary evidence of this hypothesis with a comparison of available GPS dropwindsondes near shore and offshore in the right eyewall of hurricanes making landfall in the United States.  While this is admittedly an extremely small sample, are taken from Category 1 and 2 hurricanes and may be impacted by tropospheric vertical shear, it is suggestive that the near shore winds are 90-95% of the open-ocean values for near surface winds.  Using this ratio with the Franklin et al. (2002) surface wind methodology gives maximum sustained winds at the coast of 131 to 139 kt, near the Category 4/5 threshold.

However, after the time of the 162 kt aircraft report, Andrew’s minimum pressure continued to fall up to and just after landfall, perhaps by as much as 10 mb.  It is quite possible that flight-level winds at the time of landfall were even higher than what was observed earlier.  Additionally, both aircraft-to-surface extrapolation methods rely upon obtaining the peak flight-level winds to make reasonable surface wind estimates.  If the 162 kt aircraft reconnaissance measurement under-reports the strongest winds at flight-level (and it is far from assured that the peak winds were monitored successfully), then the surface wind analysis will also be biased slightly low.  Because of both of these reasons, the maximum sustained surface wind for Hurricane Andrew is estimated to be 150 kt for the open ocean conditions and 140 kt at the coast based upon extrapolations of flight-level data.  These values are slightly higher than the values obtained solely from the aircraft observations taken an hour before landfall, though the value from the new H*WIND methodology for marine exposure is the same.  Thus both over the open- ocean and at the South Florida coast, the peak intensity of Hurricane Andrew is estimated to be Category 5 from this line of evidence.

Winds in HURDAT for Hurricane Andrew are also altered, as is seen in Table 1 and Figures 2 and 4, for the dates of 22 to 26 August based upon the Franklin et al. (2002) methodology, which has been found to be consistent with the work of Dunion et al. (2002) and Dunion and Powell (2002).  These changes are appropriate given that aircraft reconnaissance observations were available throughout this period and there were limited surface observations indicative of the maximum sustained surface winds.  The decrease in intensity late on August 23rd to early on August 24th is due to the formation of a concentric eyewall and its inward contraction (Willoughby and Black 1996).  By 0600Z on the 24th, the original eyewall had dissipated and the second eyewall became the feature that intensified and made landfall in South Florida.

Radar reflectivity feature track data:
Low-altitude radar feature tracking suggests surface winds similar to those implied by the flight-level data.  Recently, some vectors based on feature tracking from the Miami WSR-57 radar in and inside of Andrew's eyewall just prior to landfall in south Florida have been generated (P. Dodge, personal communication). Low-altitude radar feature tracks have been demonstrated to provide winds in the circulation of a hurricane that are comparable to measured winds from aircraft to within 10% (Tuttle and Gall 1999), though these are relatively noisy signals and must be quality controlled before use.  The three highest feature motions found were 172 kt (88 m/s) at 700 m at 0739Z, 176 kt (91 m/s) at 400 m at 0839Z, and 180 kt (93 m/s) at 1100 m at 0730Z (Figs. 7 and 8).  The strongest of these does not appear reasonable in comparison with other radar feature tracks and with the available flight-level data.  If we assume the most conservative estimate that the remaining two observations represent the maximum winds in the eyewall near the boundary layer top (BLT) with an average of 174 kt, we can adjust this wind to the surface using mean dropsonde profiles.  If we use the overall eyewall mean profile from Franklin et al. (2002) to go from the BLT to the surface (75%), we get a marine exposure surface-adjusted wind of 130 kt, below the threshold of Category 5.  However, as mentioned above, there is an apparent relationship between boundary layer adjustment factors and wind speed, with adjustment factors increasing with wind speed once the BLT speed exceeds about 105 kt (Fig. 12 of Franklin et al. 2002).  Franklin et al suggest that the surface to BLT adjustment factor increases to 82% for BLT wind speeds of 135-155 kt (this is their highest category).  Using this adjustment, the BLT wind of 174 kt converts to 143 kt at the surface for marine exposure.  The Dunion and Powell (2002) methodology also suggest that these low-altitude radar feature tracks correspond to winds of 148 kt at the ocean’s surface.  Converting these values to over-land, open exposure conditions as suggested above due to the impact of increased roughness in Biscayne Bay gives a wind of about 132 kt.  Thus these new radar-derived data for Hurricane Andrew provide support increasing the intensity of the storm offshore to Category 5 conditions and borderline Category 4/5 conditions impacting the coast.

Radar eyewall reflectivity core velocities:
Radar echo core velocities and reflectivities were analyzed  for the north eyewall of Andrew as it made landfall. Cores were tracked using the Tampa WSR-57 radar with a mean beam altitude over Dade County of 8,200 m.  Seven cores were observed to develop at the coast as the eyewall transited across the coast. The lifetime of these cores averaged approximately 12 minutes and developed approximately every 10 minutes. The initial growing cores are considered to be accompanied by intense updrafts carrying low level momentum. Hence core velocities are assumed to be proportional to the winds near the top of the boundary layer at about cloud base. This assumption is supported by the results of Shenk (P. Black, personal communication) who tracked growing convective towers in the Caribbean with geostationary satellite imagery and found their translation speed to be nearly equal to the winds observed by aircraft at cloud base. This conclusion is also supported by radar studies undertaken by Senn and Stevens (1965).

The maximum echo core velocities in Andrew were 180 kt for four cores that developed in the north eyewall as it crossed the coast, and 170 kt for two other cores. As reflectivities approached a
maximum near 48 dBZ in the southwest eyewall, the cores decelerated rapidly to 50-60 kt. As the cores rotated around to the south (left) sector of the eyewall, reflectivities decreased rapidly as the cores became elongated patches of stratiform precipitation. Applying the same extrapolations to the surface as in earlier sections and then adjusting for inland terrain with open exposure, gives estimates of about 153 kt for the marine exposure and 137 kt for the over land maximum wind.  This supports the conclusion of Category 5 status offshore and borderline Category 4/5 status at the coast.

Pressure-wind relationships:
Andrew’s South Florida landfall pressure of 922 mb corresponds to a wind of 139 kt (Category 5), using the operational (Dvorak) pressure-wind relationship (OFCM 2002).  The regionally-based pressure-wind relationships developed by Landsea et al. (2002) suggest 133 kt for the southerly (25o N and equatorward) and 122 kt for the subtropical (25 to 35o N) relationships.  Tropical cyclones with a small radius of maximum wind (RMW) will tend to have higher winds for the same central pressure than large storms (Callaghan and Smith 1998).  Andrew’s surface RMW in the north eyewall was only about 9 nmi based upon the peak storm surge location, substantially smaller than the 14 nmi climatologically for a central pressure of 922 mb and a latitudinal position of 25o N (Vickery et al. 2000).  Additionally, Andrew was imbedded in a stronger than usual surface ridge.  Typically at Andrew’s position in August, climatology is for sea level pressures of around 1017 mb, while the environmental pressure at Andrew’s landfall was about 1019 mb.  This would also somewhat enhance the pressure gradient and thus the peak winds in the storm.  Thus given Andrew’s small size and the higher than usual surrounding pressure, its 922 mb central pressure does support maximum sustained surface winds of about 150 kt at landfall.  These proposed revisions to Andrew’s best track fall within the range of existing scatter in the operational pressure-wind relationship (Fig 9, Brown and Franklin 2002).  The current best track pressure-wind pair at landfall (922 mb-125 kt) also falls within the observed scatter around the operational pressure wind curve; however, it is near the edge on the light-wind (left) side of the curve.  A storm of Andrew’s size and structure should not be this far to the left of the pressure-wind curve.  Thus the central pressure of Andrew at landfall, allowing for its smaller than typical size and higher than normal environmental pressures, is consistent with assessing the intensity of Andrew over the ocean near South Florida to Category 5 status, but cannot provide guidance for the impact at the coast.

Maximum winds from surface pressure reports:
Over 67 reports of minimum surface pressure made by the public were compiled by Mayfield et al. (1994). The barometers for the lowest several of these were borrowed from their owners and calibrated at the NOAA Aircraft Operations Center Calibration Lab to insure accuracy. All of these observations were composited relative to the flight level center as it moved from the coast to Krome Avenue (US 27). A cubic spline fit to the observations as a function of radial distance was computed. The derivative of this fit was computed every 2 nmi to estimate the gradient wind. The gradient wind was assumed to be the mean wind at the top of the boundary layer. The peak gradient wind from the north-south pressure profile was 139 kt 7-8 nmi from the center, or roughly along 152nd St, or Coral Reef Drive. This corresponded to the location of the maximum damage swath north of the center and to the location of maximum storm surge near the Burger King headquarters building. However, using the pressure profile from the pressure center to the west of the center, the gradient was stronger with a maximum gradient wind of 165 kt located only 5 nmi from the pressure center. Applying the above reduction factors, this suggests 1 min sustained winds of 114 kt north of the center and 135 kt west of the center at 0930Z, the time of the minimum surface pressure observation near Krome Avenue using an over-water marine exposure. However, after accounting for inland conditions with open exposure, an estimated 1 min sustained overland surface wind of 106 kt and 125 kt were obtained. This analysis supports a conclusion that winds of borderline Category 4/5 for over open water conditions and Category 4 over land.  However, uncertainties in this technique are quite large as relatively small errors in the pressure gradient assignment can dramatically impact the winds obtained and the assumption that the gradient winds represent values appropriate to the top of the boundary layer is scientifically unsure.

Satellite estimates of intensity:
The proposed peak wind for the South Florida landfall is 20 kt above the operational Dvorak (1984) estimates.  However, the proposed points fall within the envelope of previously observed scatter, as 10% of all Dvorak intensity estimates are in error by 20 kt or more (Brown and Franklin 2002). More importantly, Dvorak estimates of Andrew’s intensity were consistently too weak, as can be seen in Fig 3.  Beginning at 0000Z on the 23rd, measured reconnaissance pressures were running 10-15 mb lower than the Dvorak estimates.  Dvorak pressure estimates remained consistently too weak until after the Florida landfall.  Given this, as well as the likelihood of Andrew’s winds exceeding the Dvorak pressure-wind expectation because of Andrew’s size as discussed above, the Dvorak wind estimates for Andrew cannot reasonably be expected to yield an accurate intensity estimate.

Recently a retrospective objective Dvorak technique (ODT - Velden et al. 1998) analysis for Andrew was run (T. Olander and C. Velden, personal communication).  The ODT shows an intensity spike (T6.7) at 0900Z corresponding to winds increasing by over 30 kt to 132 kt in a short period of time (T5.4 at 0800Z), in agreement with the pressure tendency after the last reconnaissance  pressure.  Additionally, because the oblique look-angle associated with single GOES operations in 1992 which would give an eye temperature cooler than what would have been measured directly overhead, the ODT value provided  is likely an underestimate of the true intensity.  However, operational usage of the ODT usually involves averaging values over at least a three hour interval, as these measurements tend to be somewhat noisy.  Thus if the ODT value at 0900Z were the only evidence available, little to no change in the intensity would be noted.  Despite these limitations, given the systematic bias of the operational Dvorak technique to underestimate the intensity of Andrew and the new analysis from the objective Dvorak technique, satellite-based intensity estimates of Hurricane Andrew at South Florida landfall are at least consistent with an assessment of Category 5 intensity over the open ocean, but are at best indirect measures of the peak intensity and provide little guidance for the over-land conditions.

Consistency with SLOSH model runs:
The first run of the Sea, Lake, and Overland Surge from Hurricanes (SLOSH - Jelesnianski et al. 1992) for a simulation of Andrew’s impact shown in Powell and Houston (1996) utilized a pressure difference of 87 mb between the eye and the periphery, an RMW of 8 nmi, and an 8 m/s motion of the storm.  SLOSH-derived maximum sustained surface winds of about 125 kt helped produce a storm tide of 15.2 feet.  While this original SLOSH run was unable to obtain the peak storm tide observed (16.9 feet), it also significantly overestimated the storm tide outside of the peak area.  These biases are likely due to a pressure gradient/wind field that is not sufficiently concentrated in the eyewall (B. Jarvinen, personal communication) and a use of lake conditions for Biscayne Bay rather than an oceanic exposure (S. Houston, personal communcation).  Houston et al. (1999) compared the H*WIND surface winds from the Powell and Houston (1996) paper with the windfield that is an output from the SLOSH model.  They found that the SLOSH maximum wind was about 4% stronger than the H*WIND analysis.  A newer version of SLOSH can run with forcing from an observationally-based wind field (J. Chen and W. Shaffer, personal communication).  Using the H*WIND output with maximum sustained surface winds of 150 kt based upon the GPS dropwindsonde parameterizations, this newer version of SLOSH does give a more realistic 15.8 foot storm tide, but at the expense of having an even larger overestimation bias outside of the peak wind and storm tide region.  It is possible that the new wind analysis from H*WIND is utilizing a RMW (11 nmi) that is somewhat too large and a dropoff in the winds outside of the peak wind region that is too weak (B. Jarvinen, personal communication and Dunion et al. 2002).  Thus while the current configuration of SLOSH has difficulties replicating the details of a storm tide of a very small hurricane like Andrew, these various runs of SLOSH suggests that a Category 5 impact at landfall for Hurricane Andrew in South Florida is reasonable.

Estimates of intensity from damage surveys:
Analyses of structural damage that occurred due to winds in Hurricane Andrew has been utilized to provide intensity estimates (Fujita 1992; Wakimoto and Black 1994).  These identify two main regions of peak structural damage:  a larger area in the Naranja Lakes neighborhood and a smaller area farther north in Cutler Ridge neighborhood near the peak storm tide region.  While the Fujita analysis did not provide a numerical guide for his revised f-scale analysis, the Wakimoto and Black analysis suggests F3 conditions corresponding to peak gusts of 135 to 175 kt.  Powell and Houston (1996) converted these to a range of 105 to 135 kt for a maximum 1 min sustained surface wind, utilizing a standard overland gust factor of 1.3.  These values are in the Category 3 or 4 range for Andrew’s landfall in South Florida.

Another interpretation (P. Black, personal communication) is that these F3 damage regions, especially in the Naranja Lakes area, were caused not by 2-5 s gusts, but a longer timescale feature because of the extended lateral extent - a few to several hundred meters - on the order of about   20 s.  With a longer timescale, a factor of 1.12 is utilized to convert this to an approximate 1 min sustained surface wind, giving a range of 120 to 155 kt. These values would instead suggest Hurricane Andrew’s impact in South Florida would be consistent with either a Category 4 or 5 hurricane.

Regardless of which timescale and gust factors are chosen to convert the F-scale values to a sustained wind, less weight is placed here upon these estimates of windspeed because of other factors besides the peak wind in contributing toward structural damage including variation in construction type and quality, “duration of strong winds, structural sensitivity to wind direction, local topographic effects, and subgrid-scale convective downdrafts or microbursts” (from Powell and Houston 1996).  Indeed, the larger of the two F3 regions (Naranja Lakes) occurred well away from the location of the estimated maximum sustained surface winds via other methodologies.  In this region, winds were first experienced from the northwest as the west eyewall went through and then the winds shifted to be from the southeast as the east eyewall passed overhead.  It may very well be for this situation that the duration of winds and 180 degree shift in the wind direction has as much to do with the structural damage that occurred there as the peak wind itself.

Therefore, due to major uncertainties in the application of this methodology and the wide range of possible maximum sustained winds obtained, no firm conclusion can be drawn from this evidence except to say that either Category 4 or 5 is consistent with the structural damage caused by Hurricane Andrew’s winds at landfall in South Florida.

Discussion of uncertainties:
The purpose of the National Hurricane Center Best Track Change Committee is to ensure the most accurate historical hurricane record possible, consistent with contemporary science.  It has been suggested that the record in the case of Andrew should not be changed, in part because of the uncertainty surrounding the maximum wind, and an analogy has been made to the rules of instant replay in the National Football League, where conclusive evidence is required to overturn an official’s call.  Such a standard seems inappropriate for this Committee’s deliberations, though, and is inconsistent with the Committee’s mission statement.  No storm’s intensity can be determined with complete accuracy; the surface observations are almost never sufficiently comprehensive, and indirect measures must always be used.  Uncertainty in a wind speed estimate should not be an obstacle to revising an earlier estimate that is inconsistent with the observations, interpreted using current understanding.

It is acknowledged that the re-analysis presented here of Andrew’s intensity at landfall in South Florida (and elsewhere in its lifetime) is not known with exact certainty, nor will it ever be.  However, it is concluded here that Hurricane Andrew’s intensity is VERY LIKELY to be in the range of 133 to 148 kt (borderline Category 4/5 to solid Category 5) for the maximum sustained surface winds to impact the coast at landfall in South Florida with a best single estimate of 140 kt (Category 5) and that it is UNLIKELY that it was a 125 kt hurricane at landfall (Category 4) as originally thought.  Peak gusts over land were likely to be on the order of 165 to 170 kt.

It should be noted that these Category 5 conditions likely only occurred on land in a small region in south Dade County close to the coast in Cutler Ridge.  The vast majority of the region in the county south of Kendall Blvd. (88th Street) received Category 4 or Category 3 hurricane conditions.

This proposed revision is based upon substantial research that the author has only had a peripheral role in.  In particular, thanks are primarily due to the recent research on the structure of the hurricane’s inner core conducted by Mike Black, Jason Dunion, Little Jimmy Franklin, and Mark Powell.  Their work has provided the key advance in our knowledge of hurricane that has allowing this re-analysis to take place.  Jimmy Franklin also provided a writeup that served as much of the basis of the document found here.  Peter Dodge is gratefully acknowledged for providing the “new” radar reflectivity-based cell track winds utilized in this study.  Chris Velden also provided a retrospective objective Dvorak technique analysis for Hurricane Andrew that provided insight from the satellite intensity perspective.  Brian Jarvinen provided a special run of the SLOSH model to assist in the analysis of Hurricane Andrew’s intensity from a storm surge perspective.  Kind suggestions and recommendations in this re-analysis were given by many people in addition to those mentioned above:  Pete Black, William Bredemeyer, Steve Feuer, Paul Hebert, Paul Leighton, Charlie Neumann and Hugh Willoughby.

Atkinson, G. D., and C. R. Holliday, 1977: Tropical cyclone minimum sea level pressure/maximum sustained wind relationship for the western North Pacific. Mon. Wea. Rev., 105, 421-427.

Brown, D. P. And J. L. Franklin, 2002: Accuracy of pressure-wind relationships and Dvorak satellite intensity estimates for tropical cyclones determined from recent reconniassance-based “best track” data.  Preprints of the 25th Conference on Hurricanes and Tropical Meteorology, American Meteorological Society, San Diego, 458-459.

Callaghan, J. And R. Smith, 1998: The relationship between maximum surface wind speeds and central pressure in tropical cyclones.  Aust. Met. Mag., 47, 191-202.

Dunion, J. P., C. W. Landsea, and S. H. Houston, 2002:  A re-analysis of the surface winds for Hurricane Donna of 1960.  Accepted to Mon. Wea. Rev.

Dunion, J. P. and M. D. Powell, 2002: Improvements to the NOAA Hurricane Research Division’s surface reduction algorithm for inner core aircraft flight-level winds.  Preprints of the 25th Conference on Hurricanes and Tropical Meteorology, American Meteorological Society, San Diego, 581-582.

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

Dvorak, V.F., 1984: Tropical cyclone intensity analysis using satellite data. NOAA Technical Report. NESDIS 11, 47 pp.

Franklin, J. L., L. A. Avila, J. L. Beven, M. B. Lawrence, R. J. Pasch, and S. R. Stewart, 2001: Atlantic hurricane season of 2000.  Mon. Wea. Rev., 129, 3037-3056.

Franklin, J. L., M. L. Black, and K. Valde, 2002:  GPS dropwindsonde wind profiles in hurricanes and their operational implications.  Submitted to Wea. Forecasting.

Fujita, T. T., 1992: Damage survey of Hurricane Andrew in south Florida.  Storm Data, 34, 25-30.

Hock, T. R, and J. L. Franklin, 1999: The NCAR GPS dropwindsonde.  Bull. Amer. Meteor. Soc., 80, 407-420.

Houston, S. H., W. A. Shaffer, M. D. Powell, and J. Chen, 1999: Comparisons of HRD and SLOSH surface wind fields in hurricanes: Implications for storm surge modeling.  Wea. Forecasting, 14, 671-686.

Houston, S. H., P. P. Dodge, M. D. Powell, M. L. Black, G. M. Barnes, and P. S. Chu, 2000: Surface winds in hurricanes from GPS-sondes: Comparisons with observations.  Preprints of the 24th Conference on Hurricanes and Tropical Meteorology, American Meteorological Society, Ft. Lauderdale, 339.

Jelesnianski, C. P., J. Chen and W. A. Shaffer, 1992: SLOSH: Sea, lake and overland surges from hurricanes.  NOAA Tech. Rep. NWS 48, Silver Spring, MD, 71 pp.

Kraft, R. H., 1961: The hurricane’s central pressure and highest wind.  Mar. Wea. Log, 5, 155.

Landsea, C. W., C. Anderson, N. Charles, G. Clark, J. Dunion, J. Partagas, P. Hungerford, C. Neumann and M. Zimmer, 2002: The Atlantic hurricane database re-analysis project: Documentation for the 1851-1910 alterations and additions to the HURDAT database.  Hurricanes and Typhoons:  Past, Present and Future, R. J. Murnane and K.-B. Liu, Eds., Columbia University Press, in press.

Mayfield, M., L. Avila, and E. N. Rappaport, 1994: Atlantic hurricane season of 1992. Mon. Wea. Rev., 122, 517-538.

OFCM, 2002:  National Hurricane Operations Plan. Office of the Federal Coordinator for Meteorological Services and Supporting Research (OFCM), FCM-P12-2002, Washington, D.C.

Powell, M. D., and S. H. Houston, 1996: Hurricane Andrew's landfall in South Florida.  Part II: Surface wind fields and potential real-time applications.  Wea. Forecasting, 11, 329-349.

Powell, M. D., S. H. Houston, L. R. Amat, and N. Morisseau-Leroy, 1998: The HRD real-time surface wind analysis system. J. of Wind Engng. and Indust. Aero., 77 & 78, 53-64.

Powell, M. D., S. H. Houston, and T. A. Reinhold, 1996: Hurricane Andrew's landfall in South Florida. Part I: Standardizing measurements for documentation of surface wind fields.  Wea. Forecasting, 11, 304 - 328.

Rappaport, E. N., 1994: Hurricane Andrew.  Weather, 49, 51-61,

Saffir, H. S., 1973: Hurricane wind and storm surge.  The Military Engineer,  423, 4-5.

Senn, H. V., and J. A. Stevens, 1965: A summary of empirical studies of the horizontal motions of small radar precipitation echoes in Hurricane Donna and other tropical storms.  National Hurricane Research Laboratory Tech. Note 17-NHRL-74, 55 pp.
Simpson, R. H, 1974: The hurricane disaster potential scale. Weatherwise, 27, 169 & 186.

Tuttle, J. And R. Gall, 1999: A single-radar technique for estimating the winds in tropical cyclones.  Bull. Amer. Meteor. Soc., 80, 653-668.

Uhlhorn, E. W., and P. G. Black, 2002: Verification of remotely sensed sea surface winds in hurricanes.  Accepted to J. Atmos. Ocean. Tech.

Velden C. S., T. L. Olander, and R. M. Zehr, 1998: Development of an objective scheme to estimate tropical cyclone intensity from digital geostationary infrared imagery. Wea. Forecasting, 13, 172-186.

Vickery, P. J., P. F. Skerlj, and L. A. Twisdale, 2000:  Simulation of hurricane risk in the U. S. using empirical  track model.  J. of Structural Engineering, 1222-1237.

Wakimoto, R. M. and P. G. Black, 1994: Damage survey of Hurricane Andrew and its relationship to the eyewall.  Bull. Amer. Meteor. Soc., 75, 189-200.

Willoughby, H. E. and P. G. Black, 1996: Hurricane Andrew in Florida: Dynamics of a disaster.  Bull. Amer. Meteror. Soc., 77, 543-549.

Chris Landsea