METEOROLOGY

STORM DEVELOPMENT AND HISTORY

The weather disturbance that later became Hurricane Alicia began to form in the north-central Gulf of Mexico during the night of August 14 and morning of August 15, 1983. Figures 2.lA-2.lD show a sequence of four regional surface weather maps that depict the development and intensification of Alicia into tropical storm and hurricane status. These analyses cover the four-day period from August 14 to 18. The first map (Figure 2.1A) shows the early stages of Alicia. It developed as a mesoscale (mid-sized) low-pressure area on the extreme western end of a frontal trough that extended from off the New England coast southwestward into the middle Gulf of Mexico on August 14-15. Figures 2.1B and 2.1C show the low-pressure area moving westward off the Mississippi and Alabama coasts into the north-central Gulf of Mexico, along with the remnants of the frontal system extending westward across Florida and into the western Atlantic. Though this developmental pattern may seem unusual, other similar cases of typhoon/hurricane development have been documented in the western Pacific and Atlantic. There, tropical cyclones have been known to develop on the southwestern end of old cold fronts that have moved off the east Asian and U.S. coasts, lose their thermal contrasts, and become quasi-stationary in the tropics or subtropics (Riehl, 1959; Simpson and Riehl, 1981).

 In the case of Alicia, as with typhoon developments in the western Pacific, the old frontal zone became an identifiable source of cyclonic vorticity, which was the embryo of the storm. This type of tropical storm development more often occurs much earlier or later in the storm season. This is well illustrated by the analysis in Figure 2.1A, which shows the earliest detectable tropical depression stage of Alicia on the evening of August 14, when the system was centered over the northern Gulf of Mexico south of New Orleans.

Alicia was upgraded to tropical storm status on August 15 (Figure 2.1B), even though the surface pressures over the entire region remained rather high. Note that the minimum pressure of the tropical storm at this time was only about 1014 mb (29.95 in.). However, ships in the central Gulf of Mexico clearly indicated a closed circulation at that time. With these relatively high environmental pressures (approximately 1015 mb) around Alicia's circulation, the storm remained quite small and therefore generated winds stronger than those usually observed in tropical storms with similar minimum central pressures. Other cases of small tropical storms with relatively high central pressures have been observed in the southwest Atlantic (Simpson and Riehl, 1981). This condition persisted through August 16 (Figure 2.1C), when the system became a minimal hurricane.

Hurricane forecasters often use "steering currents" around hurricanes to estimate or extrapolate their likely future direction and speed of motion. Steering currents over the storm were quite weak throughout most of Alicia's lifetime over the Gulf of Mexico. As shown in Figures 2.lA-2.lC, a high-pressure ridge was well established to the north of the storm. In fact, slight pressure rises were observed to the north of the storm center, with pressure falls along the Midwestern and Gulf states from August 15 to 17. The result of these pressure changes was that Alicia drifted toward the west through midday of August 16, when the storm turned toward the west-northwest.

The steering currents around the storm during this period are best illustrated by Figures 2.2A-2.2D, the first three of which correspond in times to Figures 2.1A-2.1C, showing the 500-mb analyses over the largescale regions surrounding Alicia (these 500-mb geopotential height analyses depict the large-scale flow around the storm at approximately 18,000 ft--i.e., the midtroposphere). The initial development of Alicia's circulation aloft was one that is very often observed and has been described for the Caribbean by Riehl (1954). In this case (data not shown), an upper trough broke into two parts near the eastern coast of the United States, the northern portion continuing east and the other part retrograding westward from the southeastern United States. A cyclonic envelope remains at the surface from an old front in these cases (Figures 2.1A and 2.1B), but the temperature contrast must disappear across the frontal zone, of course, or there will be no tropical cyclone, which is the most frequent case.

The final best-fit track of Alicia from tropical depression stage to after its landfall as a hurricane on the Texas coast is shown in Figure 2.3A. This track was derived from reconnaissance aircraft fixes (up to landfall), land-based radar fixes (Galveston, Lake Charles, Texas A&M-College Station, and Corpus Christi before and after landfall), and positions estimated from satellite data. Figure 2.3B shows the radar centers from the Galveston and Texas A&M radars. There are more than one apparent circulation centers in many hurricanes after landfall, and the radar center is not necessarily the center of circulation.

Hurricane Alicia's slow track toward the west-northwest continued at speeds less than 10 mph through the late afternoon hours of August 17. An important and unusual aspect of the storm's motion developed during the late night and early morning hours of August 17, when the eye of the hurricane began slow, erratic looping motions best characterized as cycloidal (see Figure 2.3B). These cycloidal motions in the eye's track occurred between about 7:00 a.m. and 9:00 p.m. CDT on August 17 and again during the two hours after landfall. Cycloidal motions in hurricane tracks are difficult to forecast and have only rarely been observed in the past. Shortly after 7:00 p.m. on August 17, Alicia turned rather sharply toward the north-northwest and began to gain forward speed toward its final landfall on western Galveston Island. There is no obvious explanation for this change in the storm's track from the available environmental data shown in Figures 2.2A-2.2D.

A sequence of NOAA geostationary satellite photographs (Figures 2.4A-2.4D), which correspond to the series of surface analyses presented earlier, illustrate the evolution of Alicia from weak depression to tropical storm and finally to a hurricane of moderate size and intensity at landfall (see Figure 2.4D, an enhanced-infrared satellite photograph). It should be noted that Alicia took a more northerly track as the high-pressure ridge to the north apparently weakened and broke apart into two cells (see Figures 2.lA-2.lD). At the same time, an upperlevel anticyclone became well established over the storm. A portion of the upper trough did not cross eastward but remained north of the Gulf Coast oriented east-west, with west winds at high levels above the coastal stations. Thus an upper clockwise circulation was already over the low-level storm on August 15. This factor, well known by hurricane forecasters to be conducive to storm strengthening, combined with the slow movement and long period over the warm (more than 290C) Gulf waters, resulted in Alicia's deepening at an apparently steady rate of 1 mb/h over the approximately two days prior to landfall (due to available potential energy for release). According to Riehl (personal communication), this long quasi-steady deepening from weak tropical disturbance to full hurricane strength is very unusual. Finally, it should be emphasized that the cycloidal motions noted above in Alicia's track on August 18 have been documented in a few other hurricanes before, most notably Hurricane Carla in 1961, which also devastated the Texas coast (for causal mechanism, see Yeh, 1950, and Novlan and Gray, 1974).

STORM CHARACTERISTICS AT LANDFALL

By most standards for Atlantic and Gulf hurricanes, Alicia was a mediumsized hurricane of only slightly greater than average intensity. It reached minimal category 3 status on the Saffir/Simpson scale at landfall (Simpson and Riehl, 1981, App. A). The Saffir/Simpson scale is a relative scale ranging in value from 1 to 5: 1 is a minimal hurricane and 5 is the strongest hurricane that could be expected (the best example of a 5 in this century was Hurricane Camille in 1969, which moved ashore in the Biloxi, Mississippi, area). Structural damage typically begins when winds exceed 90 to 100 mph. Therefore, a major hurricane is arbitrarily defined as a 3, 4, or 5 or one in which the winds exceed 110 mph. For reference, the 1900 hurricane that claimed 6,000 lives on Galveston Island (Tannehill, 1938) was a strong 4. However, most of the fatalities in the 1900 Galveston hurricane were a result of the high storm surge, which led the survivors to construct the 15-ft-high seawall that served so well during Alicia.

The eye of Hurricane Alicia made landfall on the extreme western tip of Galveston Island (about 25 miles southwest of the NWS radar site at Galveston) at approximately 1:45 a.m. CDT on Thursday, August 18. It should be emphasized, however, that most of the damaging effects of Alicia occurred over a much larger area. The minimum central pressure as determined by a NOAA P-3 reconnaissance aircraft at about the same time as landfall was 962 mb.

The hurricane's rainfall structure during the 12 hours prior to landfall is strikingly illustrated by the radar reflectivity maps in Figures 2.5A-2.5H. This unusual composite, obtained only once before from an NWS coastal radar for Hurricane Frederic in 1979 (Parrish et al., 1982), shows a shaded digitized sequence of radar photographs developed by NOAA hurricane research scientists from the NWS Galveston 10-cm radar and from a 10-cm radar at Texas A&M University. (Digital radar data were also acquired by University of Miami researchers on a weakened Hurricane David in 1979.) The different shades refer to reflectivity values, which are strongest in the eyewall and intense rainbands of the hurricane (proportional to rain rates of 2.0 to 4.5 in./h or more). The highest wind speeds tend to occur under or inside the highest radar reflectivities in the eyewall and major rainbands (Parrish et al., 1982).

The hurricane exhibited a very unusual "double eye" structure from about 0300Z (10:00 p.m. CDT) to 0600Z (1:00 a.m. CDT), just prior to its landfall, and surprisingly again during the two hours after landfall. This double-concentric eyewall structure was also documented by NOAA research aircraft in several earlier Atlantic hurricanes, including Hurricane Allen in 1980 when it was rapidly deepening as it approached the Yucatan Peninsula in the northwest Caribbean Sea (Willoughby et al., 1981). Some other recent storms exhibiting this structure include Anita in 1977, David in 1979, and a number of western Pacific typhoons investigated by instrumented aircraft.

The double-concentric eyewall structure in Hurricane Alicia is most apparent in Figures 2.5C and 2.5D, where the highest reflectivities occur in the northeast and north quadrants of the storm. During the three to six hours prior to landfall (Figures 2.5C and 2.5D), the outer eyewall collapses in the southwest quadrant of the storm (i.e., reflectivities diminish rapidly there). Moreover, the outer eyewall appears to begin to dominate the convective structure of the hurricane after landfall (Figures 2.5E and 2.5F). Finally, it should be noted that the distribution of maximum reflectivity in the forward portion of the eyewall was very similar to the reflectivity patterns documented by NOAA research aircraft during the most intense phases of Hurricane Allen and other intensifying hurricanes.

The interested reader should carefully consider the distribution of rainfall in Hurricane Alicia and its temporal evolution as the hurricane approached and made landfall, as shown in Figures 2.5A-2.5H. The evolution of the hurricane's precipitation distribution should be compared with the storm track shown in Figure 2.3B to infer the general surface rainfall maxima relative to Alicia's track. In Figures 2.5C and 2.5D the inner eye is approximately 25 km in diameter while the outer eye is approximately 80 km in diameter. Note that the double eyewall surprisingly redeveloped shortly after landfall (Figure 2.5F).

The NOAA P-3 aircraft was flying through the storm nearly continuously during the last 6 to 10 hours prior to landfall. Composites of the NOAA P-3 flight-level winds along its track at 5,000 ft are shown in Figures 2.6A and 2.6B. The first composite covers the period from 2200Z (5:00 p.m. CDT) to 0300Z (10:00 p.m.) on August 17, ending approximately four hours prior to the hurricane's landfall. Superimposed on the analysis of winds (the stream lines are solid and the isotachs, in meters per second, are dashed) is the hurricane's track as it approached landfall. A noteworthy and unusual feature is the strong wind maximum in the northern semicircle of the eyewall, where sustained wind speeds measured by the aircraft reached slightly over 100 knots (Figure 2.6A). Downstream from this wind maximum--i.e., in the northwest quadrant of the storm--the flow diverges markedly and resembles the "downburst" phenomenon documented by Fujita (1978, 1980) beneath some severe thunderstorms over land (small, very intense downdrafts that impinge on the surface and spread out rapidly). Also noteworthy is the fact that the analysis clearly indicates a double wind maximum in the northern semicircle of the storm, corresponding to the double eyewall noted earlier in the composite radar sequence (Figure 2.5C).

Figure 2.6B is a later composite windfield from the NOAA P-3 flights at the 5,000-ft level between 0500Z (midnight August 17) and 1200Z (7:00 a.m., August 18), the times closest to and just following landfall. The wind patterns around the hurricane at this time have changed drastically (that is, they have become "normal" again after some hours of unsteady readjustment). These patterns indicate that the maximum wind speeds occurred to the northeast, or to the right of the storm center, as it made landfall. Again, there was a double wind maximum corresponding to the double-concentric eyewall structure noted in the sequence of radar maps, with an inner wind maximum of at least 85 knots a short distance southeast of the center and a higher maximum of at least 100 knots, from the south-southeast, in the outer eyewall of the hurricane. Also noteworthy in Figure 2.6B are the strong westerly wind components to the south of the recurving storm center.

Finally, Figure 2.6C gives a composite of surface winds, converted from time to space, relative to the hurricane after it made landfall. The composite extends from 0300Z (10:00 p.m. CDT, August 17) to 1500Z (10:00 a.m. CDT, August 18). Again, the highest wind speeds reported were to the east of the storm center over the eastern portion of Galveston Island, extending north-northwestward along the western portions of Galveston Bay and inland. In addition, while direct measurements were lacking, another region of damaging winds and surge levels occurred just to the right of Alicia's landfall over western Galveston Island. A secondary wind maximum of more than 60 knots is noted to the west and southwest of the storm center in the Freeport area.

Figures 2.7A and 2.7B shows radial profiles of horizontal winds through Alicia's eye outward to the northeast (2.7A) and southwest (2.7B). Note especially the distinct double maximum wind speed in these figures, at radii of about 18 and 35 nautical miles from the eye, with the outer wind maximum the strongest at landfall. Powell (1982) used 10-m-level wind data over water (VO) and at coastal stations (VL) to formulate approximate relationships of the low-level (500 to 1,500 m) aircraft wind (Va) to the mean coastal wind and peak gust (VLG) at the same place relative to the storm center. For Hurricane Frederic in 1979, Powell found VLG = 0.8Va and VL = 0.56 Va. These relationships may vary from storm to storm and with the altitude of the aircraft, but they are useful to forecasters in their assessments of low-level aircraft reconnaissance data for issuing warnings.

The hurricane as a whole produced only average amounts of hurricaneassociated rainfall after it made landfall. Figure 2.8 gives a preliminary analysis of the rainfall pattern over the two-day period following landfall. Maximum rainfall totals occurred over extreme eastern Harris County northeast of downtown Houston and ranged upward to 10 to 11 in. Somewhat lesser rainfall amounts, about 8 in., were reported in the Galveston area, and secondary maximum of 9 in. on the Sabine River north of Orange, Texas, and 8 in. in Leon County northeast of College Station are noteworthy.

Rainfall totals in the areas near landfall are suspect because of the well-known tendency for rain gages to underestimate hurricane rainfalls due to eddy currents around gages during high winds. Experimental evidence gathered by Larson and Peck (1974) shows that gages underestimate the true rainfall by approximately 20 percent at wind speeds of 9 m/s. Moreover, Dunn and Miller (1960) speculated that rain gages probably catch less than 50 percent of the actual rain when wind speeds are greater than 25 m/s. The hurricane's pressure distribution near and following landfall may be discerned by studying the four barograph traces shown in Figures 2.9A-2.9D. The eye of the storm passed over the NWS office at Alvin, Texas, which is well inland, with a minimum of 967 mb. The steepness of the pressure fall and subsequent rapid pressure rise after the eye's passage are clearly shown in the Alvin trace (Figure 2.9A) as contrasted with the microbarograph traces from Baytown, NWS Galveston, and Ellington Air Force Base (Figures 2.9B, 2.9C, and 2.9D), all of which were located to the east and northeast of the inner eye depicted in the radar composites of Figure 2.5. The Alvin pressure trace is the only one of the four shown that was clearly affected by the eye of the hurricane, and in fact Alvin was probably influenced by the western portion of the eye as it moved northnorthwest. Ellington may have been briefly affected by the eye during the slow looping period in the track during the few hours after landfall.

Another interesting feature of the Alvin pressure trace is the indication of a weak high-pressure ring surrounding the core of the hurricane, which is evident in the minor pressure rises in the morning of August 17 and again after the eye's passage in the night hours of August 18. The hurricane near landfall appears to have been in what has been called in the literature the "immature stage" (Dunn and Miller, 1960). The Alvin barograph trace is very impressive in that respect, showing a sudden drop to a central pressure of 967 mb but a period of only 10 hours with pressures below 1000 mb. Forecasters seldom have to contend with this kind of central pressure tendency in storms around landfall.

The wind field in Hurricane Alicia, as obtained by NOAA's P-3 aircraft at 5,000 ft in the hours up to landfall, has been described earlier. Figures 2.10A-2.10D present a sample of anemometer records depicting some typical and some unusual wind regimes in Hurricane Alicia (locations of the anemometers are indicated in Figure 1.1). Two traces from the Alvin NWS and Galveston NWS offices should be taken as representative samples of the winds at landfall in the inner eyewall (Figure 2.10A) and in the space between the inner and outer eyewalls (Figure 2.10B). Likewise, Figures 2.10C and 2.10D are the anemometer traces from Dow Chemical Plants A and B at Freeport, Texas, which were located in the southwestern semicircle of Alicia's inner eyewall at landfall. Note especially the strong evidence of high sustained winds and peak gusts, with westerly components at the Freeport site, which are rather unusual in that normally weaker side of hurricanes. Another recent example of anomalously high west winds in hurricanes making landfall was documented by Fujita (1980) for Hurricane Celia in 1971. Many intense damage swaths were produced at Celia's landfall over Corpus Christi, Texas, with west-to-southwesterly winds up to 120 knots.

A synthesis of all available wind data obtained thus far for Alicia near landfall is given in Table 2.1, which includes both sustained winds and peak gusts for various locations in the Houston-Galveston area (for locations of the sites listed in Table 2.1, see Figure 1.1). Some of these wind records were obtained at nonstandard mast heights, as noted in the table. Therefore a map plot of Alicia's wind speeds is not being attempted at this time (Richard Marshall of the National Bureau of Standards is doing more detailed analysis and research on the wind records obtained from many sites during Alicia in southern Texas). No wind records are available in the region of Galveston Island between Sea Isle and Jamaica Beach, which includes the boundary of the inner eyewall of Alicia at landfall (see Figure 2.5). It is therefore suspected that the sustained winds and peak gusts given for those locations nearest the coast at landfall are low estimates of Alicia's actual maximum winds. Moreover, sufficient data in the Baytown region indicate gusts ranging upward to 110 to 120 mph, marking this area as one of anomalously strong winds that seem to be associated with the outer eyewall on the radar composites at landfall (Figure 2.5). The radar film from Texas A&M University (Figures 2.5F and 2.5G) gives additional evidence that the outer eyewall became the dominant convective band after landfall (with the highest associated wind speeds).

One of the major problems encountered during the team's survey was the large number of anemometers in the Houston-Galveston area, both private and state or federally owned, that had no recording capability or backup power for emergencies. In particular, the small network of anemometers that comprise the Federal Aviation Administration-sponsored LLWAS (low-level wind shear alert system) at Hobby Airport provided no recorded wind data from Alicia.

Twenty-three tornadoes were reported to the National Severe Storms Forecast Center in Kansas City in association with Hurricane Alicia. Fourteen of these were reported to have occurred between 8:00 a.m. CDT on August 17 and 8:00 a.m. on August 18. This first group of tornado reports were concentrated in the area just southeast of Alvin, near the Hitchcock-Arcadia areas, and in the small coastal community of Baycliff on the western side of Galveston Bay. Less than half of these reported tornadoes could be corroborated by the study team's subsequent aerial damage surveys. All of the supportable tornadoes were apparently associated with a pronounced outer convective rainband and wind speed maximum, and were north and east of the storm's center during landfall on August 17-18. The other nine tornadoes reported occurred during the following 24 hours and were scattered over an area north of Houston to Tyler, Texas. Figures 2.11A-2.11C give aerial damage photographs in the area between Hitchcock and Baycliff from a NOAA helicopter at 1,000 ft over some of the suspected tornado tracks and one microburst (Fujita, 1980). All of these were embedded in the more general, spotty hurricane damage.

OFFICIAL FORECAST PERFORMANCE FOR ALICIA

In general, the forecasts issued by the National Weather Service's Miami National Hurricane Center for Gulf coastal areas threatened by Alicia were state of the art.

There are currently seven operational hurricane prediction models available to hurricane forecasters at the NHC. Only two of these models are dynamical--i.e., are derived from fundamental physical principles and the equations of motion and thermodynamics. The other models depend heavily on statistical approaches (e.g., regression equations) to predict the future track of hurricanes in the Atlantic, the Gulf, and the Caribbean. Neumann and Pelissier (1981a) have thoroughly described each of the seven operational models used by the NHC to derive the 'official forecasts" of hurricane motion and changes in intensity. They have also provided an operational evaluation of the seven prediction models. They remark that "none of the models can be singled out as clearly superior or inferior, each having at least one temporal, spatial, economic or utilitarian advantage. In practice, it is difficult to combine these advantages into one all-purpose model." Neumann and Pelissier therefore conclude that for some time into the future official forecasts and operational guidance for hurricanes will likely have to be subjectively synthesized from a number of different models, both statistical and dynamical.

A series of sample, yet typical, runs of the seven operational hurricane prediction models used at the NHC during the 24 hours before Alicia's landfall are shown in Figures 2.12A-2.12C. The starting point on each of the model plots is the position of the hurricane's eye, as determined by the hurricane forecaster in a best-fit fashion from aircraft reconnaissance fixes, satellite images, ship reports, and coastal radar. In a companion paper that analyzed forecast errors in Atlantic tropical cyclones, Neumann and Pelissier (1981b) point out that the most important forecast for the issuance of hurricane warnings along a coastal segment is the 24-hour projection. Of the various models whose results are shown in Figure 2.12A, which were run using the initial position of Alicia and other data at 0600Z (1:00 a.m. CDT, August 17), most forecast landfall in the Corpus Christi area about 24 hours later. The best forecast was made by the NHC-67 statistical synoptic model, which put landfall on western Galveston island, although its forecast of Alicia's 24-hour displacement was too great, putting the storm just to the northwest of Houston.

Similarly, Figure 2.12B indicates that the forecast models run from data available a little more than 18 hours prior to landfall (1200Z or 7:00 a.m. CDT, August 17) also :tended to move the hurricane too far to the left of its actual track. This kind of model bias is typical for hurricanes over the northern Gulf of Mexico and is reflected in the divergence of hurricane tracks from past climatology (Neumann and Pryslak, 1981). Even the more sophisticated dynamical models, such as the medium-fine mesh numerical forecast model and the Navy's nested-grid model, had Alicia moving much too fast and well to the left of its actual path. Again, for this time period, the NHC-67 model had the closest projection to the actual track, although it too predicted a track faster and to the west of Alicia's actual path.

Finally, the NHC model runs made at midday, just a little more than 12 hours prior to landfall (1800Z or 1:00 p.m. CDT, August 17) had biases and errors similar to the earlier runs. Again, the statistical models NHC-67 and CLIPER gave the best results. These results are consistent with the findings of Neumann and Pelissier (1981b) that there is statistically significant bias for translation speed in the 12-hour projection, and that large errors are principally related to the recurvature situation (when a hurricane's track acquires a northerly component of motion and "recurves" into extratropical latitudes), which was the case for Alicia after 7:00 a.m., August 16. However, these model runs for Alicia are not in concert with the finding by Neumann and Pelissier (1981b) that, for short-range projections, forecast errors for storms initially located in the Gulf of Mexico tend to be lower than average for all periods. Neumann and Pelissier also found that the mean forecast errors for short-term predictions vary considerably over the northern Gulf of Mexico, and that the 24-hour forecast error in the area around Galveston during the period considered (1970-79) is greater than average (109 nautical miles).

Another major factor contributing to the forecast errors for Alicia (and other storms in the Gulf of Mexico) is the paucity of surface and upper-air data needed to more adequately define the environmental structure of the hurricane, and especially the midtropospheric "steering flow" around it. Most vexing for this particular hurricane were its rather sharp turn to the right late in the afternoon of August 17, less than 12 hours prior to landfall, and its simultaneous though temporary forward acceleration.