Upper Ocean Response to Hurricane Gilbert

Transcription

1 JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 97, NO. C12, PAGES 20,227-20,248, DECEMBER 15, 1992 Upper Ocean Response to Hurricane Gilbert LYNN K. SHAY, 1 PETER G. BLACK, 2 ARTHUR J. MARIANO, 1 JEFFERY D. HAWKINS, 3 AND RUSSELL L. ELSBERRY 4 The evolving upper ocean response excited by the passage of hurricane Gilbert (September 14-19, 1988) was investigated using current and temperature observations acquired from the deployment of 79 airborne expendable current profilers (AXCPs) and 51 airborne expendable bathythermographs from the National Oceanic and Atmospheric Administration WP-3D aircraft in the western Gulf of Mexico. The sea surface temperatures (SSTs), mixed layer depths, and bulk Richardson numbers were objectively analyzed to examine the spatial variability of the upper ocean response to Gilbert. Net decreases of the SSTs of 3ø-4øC were observed by the profilers as well as by the airborne infrared thermometer (AIRT) along the flight tracks and advanced very high resolution radiometer (AVHRR) imagery. The AXCPs indicated a marked cooling from 29øC to about 25.5øC on September 17, 1988, which was about 1.2 inertial periods (IP) following storm passage. This pool of cooler water (3.5 ø) was located further downstream in the hurricane wake by September 19 (2.7 IP following the storm) as a result of the near-inertial currents in the mixed layer. While there was a bias of about 0.6øC and 1.7øC between the in situ and AVHRR-derived SSTs, respectively, both the AVHRR images and the objectively analyzed fields indicated a rightward bias in the upper ocean cooling that extended from the storm track to about 4Rmax (where Rmax, the radius of maximum winds, is equal to 50 km). The larger SST offset of 1.7øC was due to the difference between the time of the AVHRR image and the time of the aircraft experiment on September 19. The SSTs derived from the AVHRR images and the AIRT also indicated large gradients between the cold wake and the warm eddy in the central Gulf of Mexico. The mixed layer deepened by about m on the right side of the track during the storm and 1.2 IP later, with little evidence of continued deepening afterward. The mixed layer current vectors demonstrate that a strong, near-inertially rotating current was excited by the passage of Gilbert, with maxima of about m s -1. The currents, observeduring and subsequent to (1.2 IP) the storm, diverged from the storm track, whereas the mixed layer current vectors 2.7 IP after storm passage converged toward the track, with relative maxima of m s -1. This alternating pattern of convergence and divergence of the mixed layer current was associated with the upwelling and downwelling cycles of the baroclinic response. Considerable current shear existed between the mixed layer and the thermocline currents in the cool wake between the storm track and the 4Rmax. Estimates of the bulk Richardson numbers ranged between 0.2 and 1.0 during Gilbert and at 1.2 IP, which suggests that enhanced current shears were responsible for some of the mixed layer deepening. 1. INTRODUCTION The passage of a tropical cyclone over a warm ocean represents one of the most extreme cases of air-sea interaction. Warm, preexisting sea surface temperatures (SSTs) are a necessary but not sufficient condition for the development and maintenance of a tropical cyclone [Palmen, 1948]. Although the time scale is relatively short, O(f-1), several important physical processes are enhanced in the oceanic and atmospheric planetary boundary layers (PBLs) during such events, including the sensible heat flux that is proportional to the air-sea temperature difference. The most apparent response of the tropical cyclone-ocean interaction is the marked cooling of the upper ocean. Price [1981] summarized the extensive hydrographic studies of the oceanic response to hurricanes and included measurements acquired prior and subsequent to the passage of tropical cyclones between 1975 and 1978 by Soviet 1Division of Meteorology and Physical Oceanography, Rosenstiel School of Marine and Atmospheric Science, University of Miami, Florida. 2Atlantic Oceanographic and Meteorological Laboratories, Hur- ricane Research Division, NOAA, Miami, Florida. 3Remote Sensing Branch, Naval Research Laboratory, Stennis Space Center, Mississippi. 4Department of Meteorology, U.S. Naval Postgraduate School, Monterey, California. Copyright 1992 by the American Geophysical Union. Paper number 92JC /92/92JC ,227 scientists. The "Typhoon 75" experiment studied the ocean response to typhoon Tess [Pudov et al., 1978], and the "Typhoon 78" experiment investigated typhoon Virginia [Pudov, 1980]. As part of POLYMODE in the Atlantic Ocean, the wake of hurricane Ella was examined [Federov et al., 1979]. In addition to finding maximum SST decreases of 4ø-5øC subsequent to typhoon passage and large SST gradients on the periphery of the wake, these studies have suggested that tropical cyclones induced a well-defined modulation of the SST decreases parallel to the track in the wake of the hurricane. Another important result observed during these Soviet experiments was the modification of the hurricane PBL (HPBL) processes. These measurements suggested that air temperatures decreased by 3ø-5øC and caused a deceleration of the surface wind speed by as much as 65% across the wake. That is, the cool wake may influence the latent and sensible heat fluxes in the atmo- spheric PBL (APBL) that are necessary to sustain or intensify the tropical cyclone. Black [1983] combined 10 years of ocean temperature observations from a combination of airborne expendable bathythermographs (AXBTs), airborne infrared thermometer (AIRT), and buoy measurements to document the ocean thermal response to the passage of a hurricane. The observations were stratified in relation to the maximum wind stress (tmax), radius of maximum winds (Rmax), translation speeds (Uh), oceanic mixed layer depth (MLD), and temperature structure (T(z)). A key result was that the SST decreases in the storm have a crescent-shaped pattern, with

2 20,228 SHAY ET AL.' UPPER OCEAN RESPONSE TO HURRICANE GILBERT the largest SST decreases located in the right rear quadrant between R max and 2R max' In some instances the SSTs on the right of the track continued to decrease by 1ø-2øC between 0.6 and 1.4 days after hurricane passage. The spatial distribution of the SST decreases in the wake of tropical cyclones has also been observed [Stramma et al., 1986] with highresolution satellite instruments such as the advanced very high resolution radiometer (AVHRR). Although this rightward bias in the maximum sst response was found in previous studies [Chang and Anthes, 1978; Price, 1981], the physical mechanisms associated with this pattern remain an important issue in understanding the upper ocean response to the passage of tropical cyclones. Another important contribution to the upper ocean cooling is the vertical mixing at the base of the mixed layer. Observational and numerical studies suggest that vertical mixing at the base of the mixed layer may account for over 80% of the observed SST decreases in the wake of a hurricane [Black, 1983]. As cooler thermocline water is mixed with the warmer water from the mixed layer by entrainment mixing (w' T ), the SST decreases. This entrainment mixing is generated either by the vertical shear of the horizontal currents [Pollard et al., 1973; Price, 1981] or by surface-generated turbulence within the mixed layer [Elsberry et al., 1976]. In association with the entrainment mixing, the depth of the mixed layer increases at radial distances of 2Rmax-4Rmax from the storm track. Oceanic current measurements during the passage of tropical cyclones have been obtained through fortuitous encounters with moorings deployed in support of other field experiments [Mayer et al., 1981; Brooks, 1983; Shay and Elsberry, 1987; Brink, 1989]. Although these measurements have described the evolution of the current response at fixed vertical levels, the current meter moorings have not provided the spatial sampling necessary to examine the mesosynoptic-scale oceanic response over the scales of the atmospheric forcing. During the summers of 1984 and 1985 a series of airborne expendable current profilers (AXCPs) were successfully deployed from the National Oceanic and Atmospheric Administration (NOAA) WP-3D aircraft during PBL experiments in hurricanes Norbert and Josephine (1984) [Sanford et al., 1987] and Gloria (1985) (J. F. Price, T. B. Sanford, and G. Z. Forristall, Observations and simulations of the forced response to moving hurricanes, submitted to Journal of Physical Oceanography, 1992; hereafter Price et al., submitted, 1992). For the first time these AXCPs provided a three-dimensional description of the vertical structure of the ocean currents and temperatures underneath these hurricanes. Sanford et al. [1987] fit the velocity observations to a three-layer model to describe the orbital velocities of surface waves and the steady and shear components in each layer. Shay et al. [1989] removed these orbital velocities and examined the forced, baroclinic velocity structure excited by Norbert within the context of linear, near-inertial wave dynamics. About 70% of the observed velocity structure was explained by a model based on the first four baroclinic modes. However, the time dependent three-dimensional near-inertial response [Gill, 1984] in the wake could not be resolved with only one flight into Norbert. To examine the evolving three-dimensional response, an upper ocean response experiment was conducted in the Gulf of Mexico from NOAA WP-3D research aircraft by successfully deploying 79 AXCPs and 51 AXBTs prior to, during, and subsequent to the passage of hurricane Gilbert from September 14 to 19, 1988 (Plate 1). This experimental effort improved upon previous aircraft experiments in measuring both the prestorm and in-storm currents and temperatures and then revisiting the same area 1 and 3 days following storm passage (Figure 1). The objective of this descriptive study is to document the magnitude of the evolving upper ocean current and temperature patterns from the point measurements (AXCPs and AXBTs) and the remotely sensed SSTs derived by the AIRT and he AVHRR imagery. The paper is organized as follows: a brief description of the data, including a chronology of hurricane Gilbert and the relevant air-sea parameters, is given in section 2; the experimental sampling strategy is described in section 3; the upper ocean current and temperature response is discussed in section 4; a comparison of the remotely sensed and in situ SSTs in the wake of Gilbert using regression techniques is given in section 5; and the results of the study are summarized in section Hurricane Gilbert 2. DATA DESCRIPTION As Gilbert moved through the western Caribbean Sea at about 7 m s - the hurrican explosively deepened to a record 885 mbar on September 13, 1988, and formed a strong outer eye wall structure [Black and Willoughby, 1992]. After this deepening process the flight-level winds exceeded 80 m s - prior to crossing land near the Mexican islands of Cancun and Cozumel (Table 1). These maximum winds were located at a Rmax of km. Near-hurricane-strength winds (32 m s -1) extended beyond 160 km to the north of the eye, whereas winds ranged from 20 to 30 m s- at distances of 120 km south of the eye. After Gilbert moved over the Yucatan Peninsula in a northwestward direction at about 5.6 m s - from 1600 to 2300 UTC, September 14 (Figure 1), the storm was significantly weakened as it entered the Gulf of Mexico. The central pressure increased to about 940 mbar; maximum sustained winds of m s - were found near a new primary Rma x at about km, and a secondary wind maximum was located at about 90 km or about 2Rma x from the eye because of the contracting outer eye wall structure [Black and Willoughby, 1992]. The /3 spline analysis [0oyama, 1987; Lord and Franklin, 1987] of the flight-level wind fields from September 15 (Figure 2) indicates a cyclonically rotating flow and broad wind structure. Thus the hurricane was characterized as an intense, fast-moving system with a broad and complex wind regime, because gale force winds extended beyond 100 km. Gilbert made landfall at 0100 UTC, September 17, approximately km south of Brownsville, Texas Air-Sea Parameters A series of scaling parameters (Table 2) derived from Price [1983] is useful for comparing oceanic response studies with different characteristics and ocean conditions. Flight-level meteorological observations and the distribution of the surface wind stress, as inferred from the stepped frequency microwave radiometer (SFMR) [McLaughlin et al., 1991], were obtained during Gilbert. The SFMR is a remote-sensing tool in the microwave spectrum that utilizes the extension of

3 SHAY ET AL.' UPPER OCEAN RESPONSE TO HURRICANE GILBERT 20,229,, URR C NE GILB -T ' 1 EP 1' 88 NO -10 ^VHR- CH 1355 SENSING Plate 1. Visible image of hurricane Gilbert before it moved over the Yucatan Peninsula on September 14, 1988, and the actual storm track (red) in the western Gulf of Mexico. Notice the well-developed eye of the storm with a radius of about 15 km. relationships between surface emissivity and surface stress ature difference and a specific humidity gradient of about 5 g (as well as backscattering cross section from lower to higher kg - associated with a relative humidity of about 85-90%, wind speed range) to make surface stress estimates along the common in tropical cyclones [Frank, 1977]. flight legs. Since Gilbert was moving faster than the first mode Using a wind speed dependent formulation of the drag internal wave phase speed c, the Froude number (2) excoefficient [Large and Pond, 1981], the maximum surface ceeds unity, which indicates that the oceanic response to wind stress was 3.3 N m -2 at 50 km, with a secondary Gilbert is expected to be predominantly baroclinic. Shay et maximum in the wind stress (2.4 N m -2) at 90 km from the storm track. This "double eye" structure is indicative of collapsing eye walls, presumably due to the interaction of the hurricane with a land mass [Willoughby et al., 1984]. Over the period of the storm forcing O(12 hours) the maximum heat flux from the ocean to the atmosphere was about 1000 cal cm -2 at Rma x [Large and Pond, 1982]. This heat flux was based on observations of a 2øC air-sea temper- al. [1990] demonstrated that a barotropic current response may exist in water depths of less than 1000 m but that for the oceanic depths where the AXCPs were deployed (3000 m), the barotropic current response induced by Gilbert was relatively small ( 2-4 cm s-i). Since the barotropicurrent response is small, the upper ocean baroclinic response is the dominant physical process that will be determined from the AXCPs and AXBTs.

4 20,230 SHAY ET AL.' UPPER OCEAN RESPONSE TO HURRICANE GILBERT 98 ø 28ø! 25 ø 20 ø 95 ø 90ow,', ' / 9/28, ' ß i,,,,,: ' 9/10 'J'"x m '..' i ' ß ' "It / 9/28 -- m, it I I l I I... " m m, : 8/28 = = =.. : ".. :. 98 ø 95 ø 90ow _ J!.."/ --\ A. Prestorm [..' / \ B. Storm \ ß 9/28..' :, I, ß 9/10' r..l. F ß :' ß.. "' 8/28 ' ß =..... :". '. i i.''',, : 28 ø,'. " UTM : 20 ø N 25 ø 28 ø N 25 ø 20 ø 98 ø 95 ø 90ow 98 ø 95 ø 90ow Fig. 1. Locations of AXCPs (solid circles) and AXBTs (solid squares), tracks of SAIC drifting buoys (dashed curve), and the National Data Buoy Center buoys (triangles) relative to the track of hurricane Gilbert (solid curve) in the western Gulf of Mexico from (a) September 14 (Prestorm),(b) September 16 (Storm), (c) September 17 (Wake I), and September 19, 1988 (Wake II). The 200-m bottom contour is depicted as a dotted line AXCPs The AXCPs descended at a rate of 4.5 m s- ] and measured the motionally induced electric fields between two horizontally separated sensors spaced about 5 cm apart [Sanford et al., 1982]. This electromagnetic current measuring technique acquires observations of the baroclinic currents at intervals of 0.3 m relative to an unknown reference velocity. The root-meansquare errors in the velocity measurements are associated with electronic noise and typically range from 1 to 2 cm s -] over 3-m intervals in nonstorm deployments. The depth independent offset of the horizontal velocities was a maximum of about 5 cm s- ] estimated from the deep AXCP signals, which agreed with deep ocean current measurements of 2-4 cm s -] at 1650 and 2500 m sponsored by the Minerals Management Service TABLE 1. Times, Geographical Coordinates, Flight Level Wind Speeds (Ws), and Sea Level Pressures (SLP) September for Hurricane Gilbert Time, Ws, SLP, Date UTC Latitude Longitude m s - l mbar Sept ø24.0'N 82ø30.0'W Sept ø54.0'N 86ø48.6'W Sept ø51.6'N 91ø40.2'W Sept ø59.4'N 94ø41.4'W Sept ø56.4'N 97ø54.6'W during the summer of 1988 [Science Applications International Corporation (SAIC), 1989]. This offset should not be confused with the barotropic current response induced by the gradients in the sea surface depression in the wake of a tropical cyclone [Shay et al., 1990]. The 79 profiles of AXCP data were recorded by the NOAA data acquisition system, and the velocity profiles have been processed following the procedures described by Sanford et al. [1982]. The times and positions of the AXCP deployments are given in Tables 3-5. The failures were attributed to radio frequency (RF) (13 cases) and audio frequency (AF) (nine cases) problems (Table 6). Most of the AF failures were due to the squib misfiring, preventing the release of the probe from the airborne casing. Two cases were observed where the wire broke between the XCP (shipboard version of the AXCP) and the sea-keeping spool, which is located in the surface unit. D'Asaro et al. [1990] noted that about 10% and 7% of the AXCPs in the Ocean Storms Experiment were lost because of RF and AF failures, respectively. Similar RF and AF failure rates of the AXCPs have been observed in a series of aircraft experiments in the Subarctic Front of the northeast Pacific Ocean [Shay, 1992] and more recently in the Surface Wave Dynamics experiment (L. K. Shay, E. J. Walsh, and P. C. Zhang, Surface wave induced orbital velocities, submitted to Journal of Atmospheric and Oceanic Technology, 1992; hereafter Shay et al., submitted, 1992).

5 SHAY ET AL.' UPPER OCEAN RESPONSE TO HURRICANE GILBERT 20,231 2'! ' 93 Fig. 2. (a) Wind fields and (b) streamline (solid curve)/isotach (dased curve) fields resulting from/3 spline analysis for hurricane Gilbert on September 15, 1988, in the vicinity of the AXCP/AXBT measurements. Sanford et al. [ 1987] developed a three-layer, least squares model to isolate the effects of the surface wave induced orbital velocities on the current profiles, plus the shear and steady components: V m --e khz C cos o.}, q-s sin o 2 q- Sn(Z- (Zn-1 q- Zn) ) q- Vn q- where C and S are the amplitude coefficients of the surface wave with frequency o and horizontal wavenumber kh, W is the fall rate of the profiler, S n represents the shear component, Vn is the vertically averaged velocity in the nth layer, Zn is the depth of each layer, z represents the depth, and Vr is the residual currents that cannot be explained by the ( ) model. (Note that o 2 = k hg, which is the linear deepwater and 17 research flights. Vertical shear of the horizontal dispersion relation.) For a 10-s wave (Ah 160 m), one velocity components in the upper two layers (Z, Z2) was realization of the surface wave's orbital velocities was quite large, O(10-2) s -1, whereas the shear in the lower observed from the velocity profiles for typical mixed layer depths of m. A similar expression holds for the u (east-west) component of velocity. To ensure continuity layer (Z3) decreased by an order of magnitude. All of the observed current profiles were rotated by 4.5 ø to place the profiles into a true north and east coordinate system. between layers, additional constraints are prescribed in the three-layer model. The model was fit to the current profiles using least squares beginning at depth Z0 where the electronic noise 2.4. AIRT SSTs TABLE 2. Air-Sea Parameters for Hurricane Gilbert Parameter Variable Value Radius of maximum winds, km R ma x 50 Maximum wind stress, N m -2 n'ma x 3.3 Maximum heat flux, cal cm -2 Q Speed of the hurricane, m/s Uh 5.6 Wavelength, km A 600 First mode phase speed, m/s C l 2.8 Froude number (Uh/ C 1 ) 2. Inertial period, days IP 1.25 Coriolis parameter, s -1 x 10-4 f 0.6 Deformation radius, km A1 48 Mixed layer depth, m h 30 was less than 15 cm s-1. The coefficients based on (1) for the Gilbert current profiles are listed in Tables 7-9 for the AXCPs deployed during the September 16 (Storm), September 17 (Wake I), and September 19 (Wake II) research flights, respectively. Since o can vary from 5 to 15 s -1 the model has been solved using a nonlinear least squares technique [Marquardt, 1963]. The observed u component of AXCP 209, which was located within Gilbert's eye, illustrates the minimization of the residual variance (Figure 3, right panel) between the observed current profiles and the three-layer model. The amplitudes of the orbital velocities associated with surface waves ranged between 1 and 1.5 m s -1, which agree with estimates from the previous experiments [Sanford et al., 1987]. Since the surface wave signals in the AXCP data were not present during the September 19 research flight, only the orbital velocities have been removed from the current profiles acquired during the September 16 The AIRT-derived SSTs were acquired along the flight tracks using a modified PRT 5 radiometer [Black and Schricker, 1978]. This remote-sensing device has the advantage of providing continuous profiles along the aircraft legs and complements point measurements from the AXBT and AXCP soundings. However, the disadvantage is that a correction must be applied to account for the water vapor attenuation in the intervening atmospheric layer, which can be significant in the tropical storm environment. On the basis of several case studies, one of the more effective methods of determining the empirical atmospheric corrections to the AIRT-derived SSTs as a function of flight altitude is to compare the values to the observed SSTs from AXBTs using multivariate regression techniques. Typical values for atmospheric corrections in the regions of hurricanes are 0.8øC for flights near the level of the cloud base. This correction factor

6 20,232 SHAY ET AL.' UPPER OCEAN RESPONSE TO HURRICANE GILBERT TABLE 3. Times, Geographical Positions, Flight Legs, Storm Coordinate Positions in Terms of Rma x and A, and Depth-Integrated Measurement Errors of AXCPs Deployed on Storm II Research Flight, September 16 Time, Flight Rmax, Error, Drop UTC Latitude Longitude Leg km A, km cm s -1 AF* ø23.50'N 89ø18.40'W ø51.20'N 90ø11.40'W ø32.70'N 90ø37.60' W ø11.00'N 91ø14.00'W ø44.70'N 91ø54.30'W ø22.30'N 92ø27.70'W ø57.10'N 94ø41.20'W ø24.20'N 93ø53.40'W ø09.50'N 92ø29.90'W ø52.50'N 91ø09.00'W ø03.60'N 90ø47.80'W ø18.30'N 90ø45.00'W ø01.50'N 90ø00.90'W ø41.40'N 91ø36.50'W ø09.20'N 92ø37.80'W ø50.70'N 93ø11.60'W ø24.50'N 94ø01.40'W ø24.60'N 91ø54.10'W ,2 *AF failures represent (1) wire break and (2) squib misfire.?negative values for A indicate AXCPs deployed to the left of the storm track or in front of the storm center. TABLE 4. Times, Geographical Positions, Flight Legs, Storm Coordinate Positions in Terms of Rma x and A, and Depth-Integrated Measurement Errors of AXCPs Deployed on Wake I Research Flight, September 17 Time, Flight Rmax, Error, Drop UTC Latitude Longitude Leg km A, km cm s -1 AF* ø18.00'N 90ø11.90'W ø29.90'N 90ø46.10'W ø47.70'N 91ø15.00'W ø05.70'N 91ø43.40'W ø35.50'N 92ø03.50'W ø05.80'N 92ø22.80'W ø19.50'N 92ø56.60'W ø44.20'N 93ø51.60'W ø24.10'N 93ø28.80'W ø00.10'N 93ø05.70'W ø30.50'N 92ø46.90'W ø00.00' N 92ø28.70' W ø48.40'N 91ø51.30'W ø31.10'N 91ø32.40'W ø52.90'N 92ø30.50' W ø06.00'N 92ø59.80'W ø24.70'N 93ø24.20'W ø53.20'N 93ø42.70'W ø23.50'N 93ø59.80'W ø38.10'N 94ø26.50'W ø01.30'N 94ø50.60'W ø32.00'N 95ø52.40'W ø12.60'N 95ø30.70'W ø42.40'N 95ø12.00'W ø12.40'N 94ø53.40'W ø42.20'N 94ø35.08'W ø18.00'N 94ø13.90'W ø01.50'N 93ø52.50'W ø10.20'N 94ø48.10'W ø36.90' N 95ø10.80'W ø06.40' N 95ø30.80' W ø33.70'N 96ø30.30'W *AF failures represent (1) wire break and (2) low signal/noise ratio.?negative values for Rma x indicate AXCPs deployed to the left of the storm track or in front of the storm center.

7 SHAY ET AL.' UPPER OCEAN RESPONSE TO HURRICANE GILBERT 20,233 TABLE 5. Times, Geographical Positions, Flight Legs, Storm Coordinate Positions in Terms of Rma x and A, and Depth-Integrated Measurement Errors of AXCPs Deployed on Wake II Research Flight, September 19 Time, Flight R max, Error, Drop UTC Latitude Longitude Leg km A, km cm s-1 AF* ø59.90'N 93ø48.40'W ø17.90'N 94ø14.50'W 'N 94ø36.40'W ø11.50'N 94ø54.40'W ø41.70'N 95ø 12.40'W ø26.80'N 95ø21.20'W ø11.40'N 95ø30.40'W ø29.90'N 95ø52.60'W 1-1 t ø00.90'N 94ø48.70'W 2 - It ø40.20'N 94ø24.70'W ø04.70'N 94ø10.50'W ø24.80'N 93ø58.90'W ø54.20'N 93ø41.50'W ø24.10'N 93ø24.80'W ø06.10'N 92ø59.40'W ø47.30'N 92ø14.00'W ø47.90'N 91ø57.50'W ø59.50'N 92ø28.40'W ø25.40'N 92ø51.00'W ø00.10'N 93ø06.40'W ø24.70'N 93ø28.00'W ø41.90'N 93ø52.20'W 3 - It ø19.30'N 92ø54.10'W 4 -It ø06.20'N 92ø23.00'W ø44.50'N 91ø56.80'W ø06.70'N 91ø41.90'W ø50.40'N 91ø12.20'W ø33.00'N 90ø43.10'W *AF failure (1) represents a T(z) with noise. tnegative values for Rma x indicate AXCPs deployed to the left of the storm track or in front of the storm center. was derived from the Gilbert flights and from other research flights where AXBTs and the AIRT have been used [Black, 1983] A VHRR Imagery Infrared imagery from the NOAA 10 and NOAA 11 AVHRR can be used to map the temporal evolution of the SST changes induced by the passage of hurricances [e.g., Stramma et al., 1986]. These 1-km high-resolution picture transmission (HRPT) images were collected by the Naval Oceanographic and Atmospheric Research Laboratory Satellite Digital Receiving and Processing System [Hawkins et al., 1985]. The AVHRR sensor is well suited to monitor cold wakes generated by hurricanes because of the thermal sensitivity of 0.12øC via 10-bit digitization and highresolution (1 km at nadir and 4 km at the edge of the swath) images; however, the AVHRR sensor suffers like all infrared TABLE 6. Summary of Ocean Research Flights Into Hurricane Gilbert, Including Success Rates and Number of RF and AF Failures Date Flight AXCPs AXBTs RF AF Sept. 14 Prestorm 1/6 (17%) 31/38 (82%) 2 3 Sept. 15 Storm I 17/20 (78%) Sept. 16 Storm II 18/24 (75%) 3/4 (75%) 4 2 Sept. 17 Wake I 32/36 (88%) 3 1 Sept. 19 Wake II 28/35 (78%) 4 3 imagers from cloud obscuration. Fortunately, hurricane Gilbert had a clear zone just beyond the main region of intense convection. The radial SST profiles were smoothed using a 15-km triangular window to facilitate direct comparisons to the SSTs derived from the AXCPs and AIRT. The SST patterns derived from the AVHRR imagery during the research flights on September 17 and 19 illustrate the pronounced temperature contrast between the regions directly affected by the hurricane (cold wake), by a warm eddy in the central Gulf of Mexico, and by the far-field temperatures (Plate 2). The lateral extent of the large pool of cooler water of 3ø-4øC was confined to the area between the central eddy and the storm track. On the basis of the conceptual model of Black [1983], the wavy nature of these SST decreases along the storm track may have been associated with a standing wave pattern of forced, near-inertial currents in the mixed layer. The image from September 20 (not shown) also indicated cooler temperatures with a similar pattern in the wake of Gilbert. Thus the AVHRR detected both the large gradients at the edges of the cool wake and this along-track modulation in the SSTs. 3. EXPERIMENTAL SAMPLING STRATEGY The goal of the aircraft-based experiment was to measure the evolution of the three-dimensional current and temperature structure excited by the passage of a hurricane over the scales of the atmospheric forcing. The sampling strategy was designed to acquire upper ocean current and temperature

8 20,234 SHAY ET AL.: UPPER OCEAN RESPONSE TO HURRICANE GILBERT TABLE 7. Coefficients From Fits With the Sanford et al. [1987] Model and the Storm AXCP Profiles in the Upper 200 m Z0, C, S, V1, _,, ZI' V2' -S22' -1 ;-1 Z2, V3, _S23 Z3, Drop Variable m T, s cm/s cm/s cm/s l0 s -1 m cm/s l0 s m cm/s l0 m 201 u v u v u v u v u v u v u v u '" v "' 213 u v u v u v u v u v u v u v u v u v Z 0 is the start depth of the good data used in the fit, T is the period of the surface wave with coefficients of C and S, and Z1,2,3, V1,2,3, and S 1,2,3 representhe layer depth, the layered-averaged currents, and the current gradients, respectively, in each layer. observations to evaluate the effects of mixing and horizontal advection over near-inertial time scales relative to storm passage. A storm track prediction, which depends on the large-scale atmospheric circulation [Harr and Elsberry, 1991] and the departures from this steering flow [Carr and Elsberry, 1990], had to be made to establish the experimental grid coordinates for the prestorm measurements to achieve the experimental goal. These observational data points were to be revisited as the hurricane passed through the domain and again 1 and 3 days following storm passage Prestorm Experiment (1845 UTC, September 14, to 0005 UTC, September 15) To define the preexisting currents and temperatures in the projected path of hurricane Gilbert, a prestorm research flight was conducted while Gilbert was moving northwestward over the Yucatan Peninsula (Figure l a). At that time the storm was forecast to take a slightly more northerly track into southern Texas. The left edge of the prestorm soundings was along the eventual storm path, and no prestorm soundings were acquired to the left of the path. Prior to the flight the trajectory of a satellite-tracked drifing buoy indicated the presence of a warm, anticyclonically rotating eddy northeast of the projected path of Gilbert (E. Waddell, personal communication, 1988). The sampling strategy was to deploy 38 AXBTs and 10 AXCPs to profile the ocean current and temperature fields surrounding the eddy and in the projected path of hurricane Gilbert. The profilers were deployed in advance of the outer rainband of the approaching Gilbert where the flight-level wind speeds ranged from 20 to 30 m s -. On the prestorm flight, 31 of the 38 AXBTs deployed in the vicinity of the ocean eddy were successful; however, only one of the first six AXCP deployments acquired usable observations. This problem was determined to be associated with the speed of the aircraft during the deployments. On the basis of experiences in the Ocean Storms Experiment [D'Asaro et al., 1990], the true air speed (TAS) of the aircraft should be maintained at 200 knots (367 km/h) or less Storm I Experiment ( UTC, September 16) The first storm flight involved the deployment of 20 AXBTs and mapped the distribution of the hurricane winds using the SFMR (Figure lb). Since the updated hurricane track was more westward than was previously forecasted, these AXBTs were deployed in the vicinity of the prestorm AXBTs but were skewed further westward, toward the actual storm track Storm H Experiment ( UTC, September 16) During this flight, 24 AXCPs and four AXBTs were deployed on three legs oriented normal to the track (Figure lb). These

9 SHAY ET AL.: UPPER OCEAN RESPONSE TO HURRICANE GILBERT 20,235 TABLE 8. Coefficients From Fits With the Sanford et al. [1987] Model and the Wake I AXCP Profiles Z0, C, S V1, -S21', -1 Zl, V2, _S22, S -1 Z2, V3, _S23;_ 1 Z3, Drop Variable m T, s cm/s cm/s cm/s l0 s m cm/s l0 m cm/s l0 m 301 u v u v t u v u t v u v u v u v u -t v u v u v u v t u v u v u v u v u v u v u -! v too -lt u v u v u v u v u v u v u v u v u v u v t u v u v t u v too Zo is the start depth of the good data used in the fit, T is the period of the surface wave with coefficients of C and S, and Z 1,2,3, V1,2,3, and S 1,2,3 represent the layer depth, the layered-averaged currents, and the current gradients, respectively, in each layer. three legs were located about 200 km behind the storm center, through the eye, and 100 km in front of the eye. Since the TAS was adjusted to be about 200 knots (367 km/h) for the AXCP deployments, success rates improved to about 75%, even where the surface wind speeds were of hurricane strength. Most of the failures occurred in areas of atmospheric turbulence rather than high wind speeds. The data from this flight will be referred to as the Storm throughout the remainder of this paper.

10 20,236 SHAY ET AL.: UPPER OCEAN RESPONSE TO HURRICANE GILBERT TABLE 9. Coefficients From Fits With the Sanford et al. [1987] Model and the Wake II AXCP Profiles Drop Variable Zo, m cm/s Vl, 10 _S21, s -1 cm/s Zl, cm/s V2, 10 _S22, S -1 Z2, m cm/s V3, 10 _ 23-1 Z3, m 401 u ! v u ! v u v u v u v u v u v u v u v u v ! u v u v u v u v u v u ! v u v u v u !9-200 v u v u v u v u v u v u v u v u v Z0 is the start depth of the good data used in the fit, T is the period of the surface wave with coefficients of C and S, and Z1,2,3, V1,2,3, and S 1,2,3 represent the layer depth, the layered-averaged currents, and the current gradients, respectively, in each layer Wake I Experiment ( UTC, September 17) The first wake experiment (Wake I) started about 36 hours after the storm experiment, which equates to about 1.25 inertial periods (IP = 29.5 hours at 24øN). Surface wind -1 speeds during the research flight were only about 3-5 m s as the storm had moved inland and decreased in intensity by this time. The grid pattern for the AXCP deployments was oriented normal to the storm track (Figure 1 c), with a total of 36 AXCPs deployed on five flight legs. For a storm moving at 5.6 m s - the oceanic wavelength, which is proportional to the product of the storm translation speed (Uh) and IP [Geisler, 1970], was about 600 km. The cross-track spacing of the drop sites was adjusted to ensure that the velocity and temperature data were acquired at distances proportional to

11 SHAY ET AL.' UPPER OCEAN RESPONSE TO HURRICANE GILBERT 20,237 o oo 150 u component (cm s -1) Uobs ß l Uobs ' Umodel where X d - X e represents the along-track distances be- (cm s -1) tween the geographical positions of the AXCP (and AXBT) drop site and the eye of Gilbert at 0614 UTC, September 16 (22ø56.9'N, 94ø44.0'W), and At represents the time difference between deployment of the profiler and the eye (Price et al., submitted, 1992). The AXCP drop sites in this coordinate system are given in terms of A (the alongtrack coordinate) and Rma x (cross-track coordinate) in Ta bles 3-5. Spatial patterns of the SSTs, MLDs, and R/bulk were objectively analyzed using the algorithm of Mariano and Brown [1992]. The objective analysis consists of decomposing a scalar observation (subscript o) into three components, To(x, y, t)= Tin(x, y, t) + Te(x, y, t) + es(x, y, t), 200 Fig. 3. ([,eft) Observed (solid curve) and modeled (dotted curve) u components from AXCP 209 (located within the eye along the storm track) and (right) the absolute difference between the modeled and the observed currents in centimeters per second. the primary radius of maximum winds (R max = 50 km). The horizontal resolution was also increased where significant ocean features were observed and at the secondary radius of maximum winds (=2Rmax). On the fifth leg of the experiment, three consecutive AXCPs failed to telemeter any data in the vicinity of an oceanic eddy impinging on the continental shelf off Mexico Wake H Experiment ( UTC, September 19) The second wake experiment (Figure 1 d) started about 1.5 IP after the Wake I experiment and deployed a second grid of 35 AXCPs. Since the experimental flight time was limited because the aircraft was required to return to Miami, the fifth leg from the Wake I grid pattern where three consecutive AXCP failures occurred was eliminated, and the Wake II pattern was flown in the reverse order. 4. UPPER OCEAN RESPONSE The temperatures from the AXBTs and AXCPs are measured with a thermistor that has an accuracy of 0.2øC. The SSTs from the AXCPs and AXBTs are defined to be the temperature of the mixed layer, which is defined as the depth at which the temperature decreases by more than 0.2øC. While microlayer temperature effects at the surface are not included here, during periods of strong wind forcing the upper ocean will be well mixed in temperature. Thus the SSTs were quite similar to the mixed layer temperatures. The analyses are even more complete, because the hurricane moved steadily at a nearly constant speed and direction, allowing along-track distance to be converted into time [Geisler, 1970; Price, 1981]. In this coordinate system the AXCPs and AXBTs deployed at different times have an along-track position defined by X s = X d - X e + At Un, (2) where T m is the contribution of the large-scale or trend field, T e is the natural field variability on the mesoscale or synoptic time scale, and es is the combined effects of unresolved scales, i.e., subgrid-scale noise, and error from the particular sensor. The trend was calculated using a least squares plane fit to each variable from each flight. The resulting deviations from this trend were interpolated using the objective analysis technique [Mariano and Brown, 1992]. Final field estimates were the sum of the trend field and the objectively mapped deviation field. The anisotropic and time dependent correlation model for estimating T e from the detrended data (T O - T m) is R(dx, dy, dt)= C1 (3) ß exp - + +, (4) where DX = dx - C2dt; D Y = dy - C3dt; dx, dy, and dt are the east-west lag, north-south lag, and time lag; C1 is the correlation at zero lag and equals 1 minus the normalized (by the field variance) subgrid-scale variance (0.8); C2 and C3 are the mean phase speeds in the east-west and the northsouth direction, respectively (0.0, 0.0); C4 and C5 are the zero-crossing scales in the east-west and the north-south direction, respectively (1.6, 1.6); C6 and C7 are the spatial decay scales (or the e-folding scales) in the east-west and the north-south direction, respectively (1.1, 1.1); and C8 is the temporal decay scale (20). The units of the parameter values are days and degrees longitude and latitude with their numerical values given above in parentheses. Using the procedure described by Robinson et al. [1987], a set of maps were generated for a range of correlation parameter values. The median set of correlation parameter values from a subset of similar maps was selected. The maps were insensitive to the choice of phase speeds and the temporal e-folding scale, since the data from each flight were practically synoptic with the length scales and error levels governed by the data spacing. The analysis was performed for each variable one flight at a time. Given the strong nonstationarity in these fields and the goal of examining the temporal evolution, it was not advisable to combine data from different flights.

12 20,238 SHAY ET AL.' UPPER OCEAN RESPONSE TO HURRICANE GILBERT Plate 2a. Satellite AVHRR imagery for the Wake I experiment at 2213 UTC, September 17. The AXCPs (blue dots), drifting buoy trajectories (yellow lines), and National Data Buoy Center buoys (purple boxes) are shown relative to the storm track (red) in the western Gulf of Mexico. The 200-m isobath is depicted by the green line Sea Surface Temperatures Figure 4 shows the Prestorm, Storm, Wake I, and Wake II objectively analyzed SST fields derived from the data distributions with the corresponding 0.6 error contour. On September 14, prestorm SSTs of 28.5ø-29øC were fairly uniform over the experimental domain without any evidence of a warm core eddy in the central Gulf of Mexico (Figure 4a). During the storm (September 16) the SST cooling pattern began in advance of the eye of Gilbert, with significant SST decreases in the wake (Figure 4b). These measurements in the storm represent the response during the first quarter of the inertial cycle after converting space into time [Geisler, 1970]. This initial SST decrease agreed with previous observational [Black, 1983] and numerical [Price, 1981] studies. Farther back in the wake of hurricane Gilbert, the observed SSTs acquired from the Wake I (September 17) and Wake II (September 19) experiments indicated a broad cooling pattern with a cool pool of water (25.5øC) located at about Rmax-2Rma x (Figures 4c and 4d). This pool of cooler water was elongated along the storm track during Wake I. From the Wake II observations this pool of cooler water was displaced downstream from its previous location observed in Wake I. This effect may have been induced by the vertical motion associated with upwelling and downwelling cycles, which result from the divergence and convergence of the near-inertial currents in the mixed layer [Shay and Elsberry, 1987; Shay et al., 1989]. The pattern of these objectively analyzed SSTs was consistent with the AVHRR images in

13 SHAY ET AL.' UPPER OCEAN RESPONSE TO HURRICANE GILBERT 20,239 Plate 2b. Same as Plate 2a except for the Wake II experiment of 2151 UTC, September 19. Plate 2 and the conceptual model of Black [1983]. The SST gradients on the poleward side of the cool pool (4Rma x- 5R max) increased near the warm, anticyclonic eddy centered at approximately 25øN, 91øW. The temperature gradients between the warm pool and the eddy were about 3øC over a 50-km distance. These large horizontal temperature gradients indicate that the potential exists for fairly strong advection. The SST difference fields between Storm, Wake I, and Wake II and the Prestorm objectively analyzed fields illustrate the net cooling induced by Gilbert (Figure 4b). Within a quarter of an inertial cycle in back of the eye (September 16), maximum cooling was 3ø-4øC, centered at R ma x on the right side of the track. Similarly, maximum cooling of about 3.5øC was located at R max-2r max from the storm track in the Wake I observations (Figure 4c). The downstream displace- ment of this pool of cooler water that occurred over 1.5 IP, as shown in the Wake II observations, was consistent with the phase of the inertial cycle (Figure 4d). Since heat was extracted from a much deeper layer, the maximum SST decrease within the central eddy was about 0.5ø-1øC Mixed Layer Depths On the basis of Prestorm MLDs of m (Figure 5a), the cool wake was associated with deeper mixed layers of about 60 m, which were induced by enhanced mixing processes due to a combination of surface wind stress and cooling [Elsberry et al., 1976] and vertical current shear at the base of the mixed layer [Price, 1981]. The passage of the hurricane core increased the MLD (see Figure 5b) along the storm path. However, at about,v4 behind the eye of the

14 20,240 SHAY ET AL..' UPPER OCEAN RESPONSE TO HURRICANE GILBERT x 4 - SST DATA POINTS ' ' 28.5,-l J[. 0;;6, ß *** ß ß _ / 29 ASST x ;6'6 6/ STORM o -1 -o.15), o I 0.4), -0.15), 0 0.4), -0.15), 0.4), -1 E2 I 26.5\ k.'/..5/ 1.25 IP ), 1.75), _-1 X 1.75), ), 1.75), Z,, --/"" '1-5/ J//'u).1'3 5, L4. """ I Inertial Wavelength ( ) Fig. 4. (Left) Objective analysis of SSTs (degrees Celsius), (center) data points and normalized error, and (right) ASST (degrees Celsius) derived from the (a) Prestorm, (b) Storm, (c) Wake I, and (d) Wake II observations of hurricane Gilbert in the storm coordinate system of A (600 km) and R ma x (50 km) for the along-track and cross-track distances. The location and direction of movement of Gilbert are depicted in the Storm frames, with the shaded area in the Storm, Wake I, and Wake II panels depicting regions of maximum cooling. Note that the 0.6 contour in the center panels represents the normalized error, which corresponds to a standard deviation (rr) for SST of 0.8øC, for MLD of 2.5 m, and for R/bulk of hurricane the anticyclonically rotating current vectors in the mixed layer diverged from the track with relative maxima of about 1 m s -. This divergence of the mixed layer currents induced the upwelling of the cooler thermocline water underneath the storm track that decreased the MLD. Farther to the right (4Rmax-5Rmax), the mixing effects were evidently enhanced by downwelling processes as the MLD increased to approximately 60 m. While the region of maximum MLDs was not well defined because of the presence of the warm eddy, the MLDs were deeper (70 m) and farther to the fight during the Wake I experiment with large gradients in the MLDs (Figure 5c). The mixed layer currents diverged from the storm track as part of the next cycle of the near-inertial rotation 1.2 IP after storm passage, with relative maxima of about m s -1. The current field associated with the warm core eddy was

15 SHAY ET AL.' UPPER OCEAN RESPONSE TO HURRICANE GILBERT 20,241 MLD 6. 4 I o -1 -, AMLD 6 k, S.,. TOR M Ri bulk t S[I'ORM 4 E 2 0 n- 40._ / B i -0.15% 0 0.4, -0.15, 0.4% -0.15% 0.4% IP 1.25 IP % 1.75% 1.75% IP 2.75 IP 2.75 IP 4 -, /' -lo -1 D 2. ),, 3.25% 2.5% 3.25% 2.5% 3.25% Inertial Wavelength ( %) Fig. 5. (Left) Objective analysis of MLDs (meters), (center) AMLDs (meters), and (right) the Ribu]k numbers derived from the AXCPs and AXBTs in the (a) Prestorm, (b) Storm, (c) Wake I, and (d) Wake II experiments in the storm coordinate system, as in Figure 4. The depth-integrated mixed layer currents are superposed on the MLDs (left panels) with maxima of 1.1, 1.4, and 1.0 m s - in Storm, Wake I, and Wake II, respectively. The differences between the mixed layer and thermocline currents are superposed on the Ribu]k panels with maxima of 1., 1.5, and 0.6 m s also anticyclonic, but with a smaller scale, 0(200 km), as compared to the forced, near-inertial wavelength of about 600 km. During Wake II (1.5 IP later) the MLD pattern indicated a region of shallower water along the track of less than 30 m as a result of the convergent cycle of the mixed layer currents and the end of the upwelling or the beginning of the downwelling cycle (Figure 5d). The first two legs (2.5 and 2.8 IP) coincided with the convergent cycle (west of 93øW), while the remaining legs (3. and 3.2 IP) were in the area of transition from a convergent to a divergent flow pattern in the storm coordinate system. The differences in the MLDs between the Prestorm and Storm and the Wake I and Wake II experiments indicate that the mixed layer deepened by m. As shown in Figure 5b, the relative maximum occurred at about 3Rmax-4Rmax to the right of the track. Notice that the orientation of the mixed layer difference field is quite different from that of the

16 20,242 SHAY ET AL.' UPPER OCEAN RESPONSE TO HURRICANE GILBERT SST differences in Figure 4b. This may have been due to the presence of the eddy field. However, the location of enhanced mixing area relative to the storm center agreed well with the previous studies [Black et al., 1988]. A somewhat closer similarity in the wavy nature of MLD differences and the SST response is observed during Wake I and II (Figures 5c, 5d, 4c, and 4d), respectively, but the agreement is not exactly one-to-one. Thus both advective and mixing effects contributed to the SST differences Bulk Richardson Number The bulk Richardson number was estimated from the expression ghaat R/bulk = V 2, (5) where a(2 x 10-4 øc -1) is the coefficient of thermal expansion, h is the mixed layer depth, AT is the temperature difference between the mixed layer and the thermocline, V is the difference between the depth-integrated mixed layer and the upper thermocline current velocity over a vertical scale of about m, and # is the acceleration of gravity. Note that vertical current shear estimates from the XCP (shipboard version of the AXCP) are considered to be accurate over vertical scales of 7-10 m [Gregg et al., 1986]. Pollard et al. [1973] used a value of unity for the bulk Richardson number as a condition for the onset of mixing processes at the base of the mixed layer, whereas Ellison and Turner [1959] found that more appropriate values ranged from 0.4 to 0.8 for the initiation of vertical mixing. To assess the effect of vertical current shear on the deepening of the mixed layer, the bulk Richardson numbers estimated from (5) were objectively analyzed. During the storm period (Figure 5b), the Richardson numbers ranged from 0.2 to 1 between the storm track and the ocean eddy (typically 0-4Rmax). These values were within the critical values described by Pollard et al. [1973] and Ellison and Turner [1959] and were likely a result of the enhanced current shears in the cool wake region. The current shears remained large in the Wake I observations (Figure 5c) with pockets of low bulk Richardson numbers of extending to about 3Rmax-4Rma x. While in the Storm experiment it is difficult to separate the effects of shear instabilities from stress-induced turbulent mixing, the continued SST cooling and mixed layer depth changes 1.2 IP following Gilbert were associated with a combination of the shear-induced mixing and vertical advection resulting from the downwelling/upwelling cycle. Although the bulk Richardson numbers suggest continued deepening of the mixed layer, these upwelling and downwelling cycles may have been the dominant factor in the MLD changes after removal of the surface stress. The maximum current differ- ences between the mixed layer and the thermocline (over vertical scales of m) were quite large, ranging in value from 0.6 to 1.0 m s -1 and were located in the region of low bulk Richardson numbers, as expected from (5). In the Wake II experiment the bulk Richardson numbers exceeded unity throughout most of the domain (Figure 5d), including the region of the warm core eddy, and far from the storm track because of deeper mixed layers. In other words, the vertical current shears would have to be substantially larger to induce any further layer deepening Velocity Profiles The cross-track sections of the velocity profiles from the wake I and II experiments indicated considerable vertical structure in the upper ocean (Figure 6). The upper ocean cross-track current components were larger than the alongtrack components from -Rma x to 3Rma x in Wake I, which indicates the divergence phase of the upper ocean current field. With the exception of the current profiles at about 5Rma x of this cross-track section (Figure 6a) the phase of the velocity components indicated energetic near-inertial motions in the vertical structure. Between R ma x and 4Rma x there was also considerable current shear that corresponded to the cøol pool of water in Figure 5 at the depths of m. The velocity Profiles from the: Wake II experiment at nearly the same geographical position were acquired about 2.8 IP following the storm, or 1.4 IP later than the Wake I current profiles. While this was about 0.1 IP following the phase for maximum convergence (2.75 IP), the negative cross-track currents (with a weaker along-track flow) indicated that a large part of the upper ocean currents were directed toward the track, except for the two profiles influenced by the eddy at 5R max-6r max. This result suggested that a large fraction of the upper ocean currents in the cold wake were near-inertial, in agreement with earlier findings [Shay and Elsberry, 1987; Shay et al., 1989]. The velocity structure remained complex between Rma x and 3Rmax, with large changes of m s over m Temperature Profiles Considerable differences were found in the thermal struc- ture changes from Wake I and II between the left side and the right side of the trac k (Figure 7). Between -Rma x and the storm track the mixed layer decreased to about 20 m, whereas on the fight side of the track (Rmax-3Rmax) the divergence and convergence cycle induced an upwelling and downwelling of the isotherms, respectively. The MLDs of m in the Wake I experiment were increased to m in Wake II. Thus higher temperatures were observed at greater depths during Wake II as a result of the convergent flow and downwelling processes modulating the vertical mixing events. These oscillations in the thermal structure with approximately 20-m deflections in the mixed layer depths were consistent with the expected amplitudes (9-10 m) of the isopycnal displacements based on a scaling with the storm intensity and translation speed ( 'max/po Uh) [Price, 1983]. From 4Rma x to 5Rma x the upper ocean thermal structure in the eddy indicated a mixing signal between the two snapshots of the wake of Gilbert. In light of the bulk Richardson numbers exceeding unity in the warm eddy field with similar current shears observed from the Wake I and II experiments the changes in the upper ocean temperatures and mixed layer depths were due to the interactions of the mixing effects and the dynamically induced vertical and horizontal advection fields Radial Profiles 5. SST COMPARISONS Remotely sensed SSTs from the AIRT and AVHRR are compared to the in situ SSTs from the AXCPs and AXBTs for the purpose of determining any systematic bias in the -1

17 SHAY ET AL.' UPPER OCEAN RESPONSE TO HURRICANE GILBERT 20, loo Rrnax loo OO -50 o 50 loo Velocity (cm s -x) Fig. 6. Vector stick plot of the observed velocity profiles (centimeters per second) from (a) Wake I (1.4A) and (b) Wake II (2.8A) with the cross-track distance relative to the storm track scaled by Rma x. Each current profile is displaced by 85 cm s -1, going from -Rma x to about 5Rmax, with a velocity scale in the lower left corner. remotely sensed data. Since the AIRT and AVHRR data provide the larger-scale context of the SST response, the SST differences are examined using regression techniques to determine the slope and variance between the in situ and remotely sensed observations. The purpose here is to correct for the bias in the remotely sensed data to match the observations. AIRT. The radial profiles of the AIRT-derived SSTs also indicated a relative minimum of about 25.5øC located at about 2R max to the right of the storm track in both the Wake I and Wake II experiments (Figure 8). This SST minimum and its location relative to the storm track agreed well with the results from the objective analyses of the AXCP and AXBT data. The large SST gradient regions were located between 5Rma x and 6Rma x in the wake of hurricane Gilbert and were also well documented by the AIRT observations. The SSTs changed by 3øC over 75 km in the Wake I hurricane passage that was not corrected for in the retrieval of the AIRT-derived SSTs. A common feature in the AIRT observations was the lower values of the SSTs in Wake II relative to the Wake I experiment. The most pronounced cooling of 1ø-1.5øC occurred from 2Rma x to 4Rma x on the right side of the storm track. On the basis of previous observational [Black, 1983] and numerical investigations [Price, 1981], maximum cooling occurs at 2Rma x to the right of the storm track. In fact, the SSTs from the AIRT corroborate these findings with the AXCP and AXBT measurements. Along leg 2 the SST gradients were located at 3R max, which was primarily due to the juxtaposition of the anticyclonically rotating eddy and the cold wake. A VHRR. Since the Wake I and II experiments were at the same time of the day on two separate days ( 1.5 IP), any diurnal influence can be removed by taking the difference experiment and by as much as 4øC over a similar horizontal between the fields. By contrast, the SSTs derived from scale in Wake II. Along leg 1 the differences between the AVHRR imagery were from a different time of the day AIRT- and AXCP-derived SSTs were Within _+0.5øC, between the two experiments. The AVHRR instrument whereas these differences were about IøC along the same leg in Wake II (not shown). However, this larger offset may have been due to more atmospheric moisture 3 days after responds only to the temperature of the microlayer, which may differ considerably under light winds from those derived from the AXCPs. Price et al. [1986] showed that there can be

18 20,244 SHAY ET AL.' UPPER OCEAN RESPONSE TO HURRICANE GILBERT -1 0 I ma= / Temperature (øc) Fig. 7. Temperature profiles (degrees Celsius) from Wake I (solid curve) and Wake II (dashed curve) at the same positions as in Figure 6. Each temperature profile is displaced by 4øC going from -Rma x to about 5Rmax, with a temperature scale in the lower lef t corner. detectable changes in the mixed layer thermal structure as a result of the diurnal warming and cooling cycle, especially during periods of light winds. Other physical processes contributing to the SST offsets are cloud cover and precipitation. Satellite-derived SST profiles along flight legs 1 (1.7A) and 4 (1.2A) in the Wake I experiment, which were located at the east and west ends of the analyzed domain, are compared in Figure 9 with the SSTs observed from the AXCPs. Both sources indicated temperatures in the pool of cooler water of about 25.5øC at 3Rma x from the storm track along leg 4. AVHRR detected slightly higher temperatures along flight leg 1 than along leg 4. The maximum gradient regions indicated SST changes of 3.5øC over a scale of km. These large gradients suggesthere may be significant crosstrack advective effects, because the mixed layer currents were O(1 m s-1) during the first few inertial periods following Gilbert Regression Analyses AXCP versus AIRT. Using a linear regression model, the SSTs derived from the AXCP, AIRT, and AVHRR data were analyzed to provide a quantitative measure of the statistical quantities such as the slope of the regression line, bias, and variance (Table 10). The bias derived from the Wake I observations was 0.23øC, which represents the value that has to be subtracted from the AIRT-derived SSTs to coincide with those observed from the AXCPs. The physical significance is that the atmosphere can hold increasing amounts of moisture with higher SSTs and air temperatures, so that the atmospheric correction error should account for moisture variations [Black and Schricker, 1978]. The slopes of the regression line were 0.16 and 0.33 for the Wake I and II experiments, respectively, over the 25ø-30øC range of the SSTs. These slopes were not statistically different from those estimated from the AVHRR and AXCP fits (described below) within the error bounds of the regression line fit, 0(0.05). However, the variances of regression lines about the means from both experiments (0.18øC, 0.28øC) were below the variance of the SST differences from the AXCPs and AXBTs. TABLE 10. Regression Analysis of AXCP-, AVHRR-, and AIRT-Derived SSTs in Wake I and II AXCP and AXCP AVHRR AVHRR and AIRT and AIRT Parameter I II I II I II Bo, øc B SSTs - SST o, øc , (øc) ,, (øc) m, (øc) B o is the y intercept, B1 is the slope of the regression curve, 0-0 is the observed SST variance, 0-a = 0-SST,-SSZo is the difference between the remotely sensed and directly observed SSTs, and 0-m represents the variance about the mean of the regression curve.

19 SHAY ET AL..' UPPER OCEAN RESPONSE TO HURRICANE GILBERT 20, , I I I I Normalized Distance Fig. 8. Comparison of AIRT SST profiles (degrees Celsius) from Wake I (solid curve) and Wake II (dashed curve) scaled in relation to the storm track Rma x at 1.7A and 3.2A, respectively. The circles and triangles represent the SSTs from AXCPs acquired during Wake I and II, respectively. AVHRR versus AXCP. The results of the regression fits for both the Wake I and II experiments indicated that the slopes of the regression lines for the experiments were similar, with values from 0.2 to 0.3 over a range of SSTs from 25 ø to 30øC. More importantly, the biases between the observed SSTs from the AVHRR images and the AXCPs were approximately 0.6øC and 1.7øC for Wake I and II, respectively. Considering the amount of corrections for various atmospheric processes that have to be included in the satellite processing algorithms, the difference of 0.6øC represents a fairly realistic result. Since the thermistors on the AXCPs have an accuracy of 0.2øC, this correction may further reduce the SST bias between the AVHRR images and the AXCPs. The requirement to acquire simultaneous satellite and in situ observations is underscored by the large bias between the observing systems of 1.7øC during the Wake II experiment. These results indicate a better fit between the AIRT and the AXCPs than for the AVHRR. The variance of the regression line about the mean (trm), estimated from the sum of the squares and the number of degrees of freedom, was about 0.28øC. This value was again below the total variances of the SST differences (tr o) between the AXCPs and the AVHRR images during Wake I and indicates statistical confidence. Despite the large bias in Wake II, the variance of the regression line about the mean (0.35øC) also suggested statistical confidence in the results from the regression fits. The SST patterns derived from the AVHRR were very consistent with the objectively analyzed SST fields in Figure 4, even though the absolute magnitudes were not correct. A VHRR versus AIRT. Radial SST profiles from the AVHRR imagery were regressed with the AIRT observations that were subsampled at 10-kin intervals. The slopes of the regression curves were 0.14 and 0.33 for the Wake I and II experiments, respectively. Given the error limits in estimating the slope of the regression line, these slopes were not statistically different from those discussed above, but there were more independent values ( 120), which strengthens the results. The bias between the AVHRR and the AIRT SSTs was 0.29øC for the Wake I experiment, whereas the SST bias in the Wake II experiment exceeded 2.0øC. This large bias during the Wake II experiment suggested that the AIRT detected more cooling compared to the AVHRR imagery, which was acquired in the late afternoon. That is, the AVHRR may have sensed the skin temperatures at the end of the diurnal heating cycle [Price et al., 1986]. 6. SUMMARY The goal of the aircraft-based experiment was to measure the evolution of the three-dimensional current and temperature structure excited by the passage of a hurricane over the scales of the atmospheric forcing. For the first time the aircraft-based sampling strategy resulted in snapshots of upper ocean current and temperature structure that are required to examine the three-dimensional, time dependent near-inertial velocity response to a hurricane. Given the inherent uncertainties of storm track prediction for the preevent flight [e.g., Harr and Elsberry, 1991], the experimental objective was achieved with a high degree of success. The upper ocean response to the passage of hurricane Gilbert was investigated using in situ observations from

20 20,246 SHAY ET AL.: UPPER OCEAN RESPONSE TO HURRICANE GILBERT 30. ' I I [ I [ I t I [ ] I [! t I I I I Normalized Distance (R. =) Fig. 9. Comparison of SST profiles (degrees Celsius) scaled in relation to the storm track by R max from AVHRR imagery during Wake I at 1.7A (solid curve) and 1.2A (dashed curve). The circles and triangles represent the SSTs from corresponding AXCPs along each leg in Wake I. AXCPs and AXBTs and remotely sensed SSTs from the AVHRR and the AIRT. Hurricane Gilbert induced significant decreases in the SST by as much as 3.5ø-4øC. The region of enhanced cooling extended to about 2Rmax-3Rma x (150 km) to the right of the storm, which is consistent with the conceptual model of Black [1983] and the numerical simulations of Chang and Anthes [ 1978] and Price [ 1981]. On the northern periphery of the cold wake between 4Rma x and 5Rma x, a significant SST gradient of 3ø-4øC occurred over horizontal scales proportional to R ma x. In addition to the cold wake an anticyclonically rotating eddy was located just beyond the maximum SST gradient regime. The SST within a warm eddy decreased by approximately 0.5øC as a result of Gilbert. However, its juxtaposition with the cold wake appeared to limit the horizontal extent of the cooling, because warm eddies have a deeper mixed layer depth so that the SST decrease was considerably smaller. Mixed layer deepening on the right side of the storm track by m during the Storm and Wake experiments was a result of the strong atmospheric forcing by the wind stress and the current shear between the mixed layer and the top of the thermocline. Although the MLD patterns varied between the three observing periods, the MLD did not significantly change between the Wake I and II experiments. Mixed layer current vectors indicated a strong divergent and convergent cycle with a predominant anticyclonic rotation associated with forced near-inertial motions [Shay and Elsberry, 1987; Shay et al., 1989]. In the mixed layer the maximum currents ranged between 1 and 1.4 m s -1 for the Storm and Wake I experiments, whereas the maximum currents during Wake II ranged between 0.8 and 1 m s -1. The shear between the currents in the mixed layer and top of the thermocline yielded bulk Richardson numbers below unity between the storm track and 4Rma x (on the right side) during the Storm and Wake I experiments. During Wake II there was very little evidence of continued mixed layer deepening as the bulk Richardson numbers exceeded unity over most of the experimental domain. The across-wake velocity structure also illustrated the convergent and divergent cycle of the upper oceanic response to hurricane Gilbert with large changes of m s -1 over a 10- to 20-m layer between the mixed layer and the thermocline. Regression analyses between the AXCP-, AIRT-, and AVHRR-derived SSTs indicated consistent results with re- gard to bias, slope, and variance of the SST differences. The slopes of the regression curves ranged from 0.14 to 0.20, and the variances of the SST differences about the mean were less than the total variances. During Wake I the bias between the satellite-derived SSTs and the in situ observations was 0.6øC, which indicates that the AVHRR-derived SSTs have to be corrected by this amount to coincide with the AXCP observations. Although the bias for the Wake II observations was considerably larger (1.7ø-2øC), the slopes of the regression curves ranged from 0.27 to Even though quantitative differences existed, many of the major features of the ocean influenced by this hurricane were well revealed by the AVHRR imagery [Stramma et al., 1986]. The large offsets between the AVHRR-derived SSTs and the AXCP- and AIRT-derived SSTs underscore the need for simulta- neous satellite- and aircraft-based measurements. Because

21 SHAY ET AL.: UPPER OCEAN RESPONSE TO HURRICANE GILBERT 20,247 of the diurnal heating and cooling cycle [Price et al., 1986], there can also be significant significant temperature differences between the satellite and in situ observing systems. Other possible sources of discrepancies are due to the amount of moisture in the intervening atmosphere, which will be affected by clouds and precipitation. Substantive scientific issues remain to be resolved using the hurricane Gilbert data set. First, the time evolution (or along-track distance) of the anticyclonically rotating current vector indicates not only that the hurricane excited strong near-inertial current in the mixed layer but also that the vertical current shear of the horizontal velocities over a 10- to 20-m layer was quite large between Rma x and 4Rma x. A key question that has emerged from previous studies is how much of the observed current variability can be ascribed to linear, near-inertial wave dynamics given the presence of the warm core eddy in the Gulf of Mexico and the complicated flow fields. Second, the relative importance of vertical mixing and horizontal advection can be formally addressed from a combined observational and modeling approach. In previous simulations of the upper ocean response, models generally treated only the temperature structure, thereby allowing the numerical solution to be tuned to maximize the agreement with the observed temperature profile. However, the addition of the current profile places more rigorous constraints on the model physics to include the correct thermal and momentum balance to agree with the observed temperature and current structure. Finally, as in the hurricane Norbert and Josephine profiles [Sanford et al., 1987] in Ocean Storms [D'Asaro et al., 1990] and the Surface Wave Dynamics experiment (Shay et al., submitted, 1992) the deployment of AXCPs provides a unique and powerful data set for understanding the upper ocean response during and subsequent to impulsive atmospheric forcing events. The simultaneous deployment of aircraft expendables with measurements of the turbulent fluxes in the APBL over the scales of the atmospheric forcing events [Bane and Osgood, 1989] and remote-sensing observations of ocean surface waves and winds by the NASA surface contour radar [Walsh et al., 1985] and the C band scatterometer [McLaughlin et al., 1991] provides valuable measurements to use in combination with coupled tropical cyclone-ocean models [Chang, 1991]. This combination of evolving snapshots of the three-dimensional structure with the high-resolution numerical models can exploit important physical processes that can only be inferred from point measurements. This technology does not alleviate the need for high-quality time series measurements from the conventional measurement techniques from research vessels and moored buoys. However, this suite of aircraft instrumentation will improve our understanding of the spatial variations of the upper ocean response and the coupling between both geophysical fluids subjected to a broad spectrum of atmospheric conditions. Acknowledgments. The Office of Naval Research supported the AXCP part of the experiment with additional support from Navy Direct Research Funds at the Naval Postgraduate School. Flight hours were provided by the Hurricane Research Division (HRD) of NOAA as part of the Hurricane Planetary Boundary Layer research program, and the Naval Oceanographic and Atmospheric Research Laboratory also provided AXBT's for the experiment. We gratefully acknowledge the drifting buoy data provided by Evans Waddell of SAIC to locate the position of the warm eddy in the Gulf of Mexico during planning stages of the research flights into hurricane Gilbert. Jim Trout of HRD prepared the AXCPs for shipment to New Orleans for the experiments. Duane Hicks of the Delta Freight Office at the New Orleans Airport contributed significantly to the success of the experiment by expediting the AXCP shipments, which were inadvertently rerouted through the Atlanta Airport. Bob Burpee of HRD and Jim McFadden of the Office of Aircraft Operations (OAO) orchestrated the aircraft experiments. We appreciate the extraordinary efforts of the OAO pilots, engineers, and technicians during the experimental effort. The authors would also like to thank Jim Hannon of Sippican Ocean Systems for his ideas on the deployment of the AXCPs and Eric D'Asaro of the Applied Physics Laboratory/University of Washington for providing guidance on the air speed problems encountered during the Ocean Storms Experiment. Fred Abell of Sverdrup Technology Inc. processed the numerous satellite AVHRR images. Jean Carpenter carefully redrafted the objective analysis figures, and Penchen Zhang assisted in processing the AXCP data. This research has been supported by the Office of Naval Research and the National Science Foundation. REFERENCES Bane, J. M., Jr., and K. E. Osgood, Wintertime air-sea interaction processes across the Gulf Stream, J. Geophys. Res., 94, 10,755-10,772, Black, M. L., and H. L. Willoughby, The concentric eyewall of hurricane Gilbert, Mon. Weather Rev., 120(6), , Black, P. G., Ocean temperature changes induced by tropical cyclones, Ph.D. dissertation, 278 pp., Pa. State Univ., State College, Black, P. G., and T. Schricker, Direct and remote sensing of ocean temperature from an aircraft, in Proceedings of the Fourth Symposium on Meteorological Observation and Instrumentation, pp , American Meteorology Society, Boston, Mass., Black, P. G., R. L. Elsberry, L. K. Shay, R. P. Partridge, and J. D. Hawkins, Atmospheric boundary layer and oceanic mixed layer observations in hurricane Josephine obtained from air-deployed drifting buoys and research aircraft, J. Atmos. Oceanic Technol., 5, , Brink, K. H., Observations of the response of the thermocline currents to a hurricane, J. Phys. Oceanogr., 19, , Brooks, D., The wake of hurricane Allen, J. Phys. Oceanogr., 13, , Carr, L. E., III, and R. L. Elsberry, Observational evidence for predictions of tropical cyclone propagation relative to environmental steering, J. Atmos. Sci., 47(4), , Chang, S. W., Development of a 3D coupled tropical cyclone-ocean model (extended abstract) in Nineteenth Conference on Hurricanes and Tropical Meteorology, pp , American Meteorological Society, Boston, Mass., Chang, S. W., and R. A. Anthes, Numerical simulations of the ocean's nonlinear baroclinic response to translating hurricanes, J. Phys. Oceanogr., 8, , D'Asaro, E. A., T. B. Sanford, R. X. Drever, and M. O. Morehead, Air-expendable current profiling during the Ocean Storms experiment, Tech. Rep. APL-UW 8916, App. Phys. Lab., Univ. of Wash., Seattle, Ellison, T. H., and J. S. Turner, Turbulent entrainment in stratified flows, J. Fluid Mech., 6, , Elsberry, R. L., T. Fraim, and R. Trapnell, A mixed layer model of the ocean thermal response to hurricanes, J. Geophys. Res., 81, , Fedorov, K. N., A. A. Varfolomeyev, A. J. Ginzburg, A. G. Zatsepin, A. Yu. Krasnopevtsev, A. G. Ostrovskiy, and V. Ye. Sklyarov, Thermal response of the ocean to the passage of hurricane Ella, Oceanology, Engl. Transl., 19, , Frank, W. M., The structure and energetics of the tropical cyclone, II, Dynamics and energetics, Mon. Weather Rev., 105, , Geisler, J. E., Linear theory on the response of a two layer ocean to a moving hurricane, Geophys. Fluid Dyn., 1, , Gill, A. E., On the behavior of internal waves in the wakes of storms, J. Phys. Oceanogr., 14, , 1984.

22 20,248 SHAY ET AL.: UPPER OCEAN RESPONSE TO HURRICANE GILBERT Gregg, M., E. A. D'Asaro, T. J. Shay, and N. Larson, Observations of persistent mixing and near-inertial internal waves, J. Phys. Oceanogr., 16, , Harr, P. A., and R. L. Elsberry, Tropical cyclone track characteristics as a function of large scale anomalies, Mon. Weather Rev., 119(6), , Hawkins, J., et al., Remote sensing at NORDA, Eos Trans. AGU, 66(23), , Large, W. G., and S. Pond, Open ocean momentum flux measurements in moderate to strong wind, J. Phys. Oceanogr., 11, , Large, W. G., and S. Pond, Sensible and latent heat flux measurements over the ocean, J. Phys. Oceanogr., 12, , Lord, S. J., and J. L. Franklin, The environment of hurricane Debby (1982), I, Winds, Mon. Weather Rev., 115, , Mariano, A. J., and O. B. Brown, Efficient objective analysis of heterogeneous and nonstationary fields via parameter matrix, Deep Sea Res., Part A, 39(7), , Marquardt, D., An algorithm for least squares estimation of nonlinear parameters, J. $oc. Ind. Appl. Math., 11, , Mayer, D. A., M. O. Mo0eld, and K. D. Leaman, Near-inertial internal waves on the outer shelf in the Middle Atlantic Bight in the wake of hurricane Belle, J. Phys. Oceanogr., 11, , McLaughlin, D. J., R. E. Mcintosh, A. Pazmany, L. Herizi, and E. Bottniew, A C-band scatterometer for remote sensing the air-sea interface, IEEE Trans. Geosci. Remote $ens., 29, , Ooyama, K. V., Scale controlled objective analysis, Mon. Weather Rev., 115(10), , Palmen, E., On the formation and structure of tropical cyclones, Geophysics, 3, 26-38,1948. Pollard, R. T., P. B. Rhines, and R. Thompson, The deepening of the mixed layer, Geophys. Fluid Dyn., 3, , Price, J. F., Upper ocean response to a hurricane, J. Phys. Oceanogr., 11, , Price, J. F., Internal wave wake of a moving storm, I, Scales, energy budget and observations, J. Phys. Oceanogr., 13, , Price, J. F., R. A. Weller, and R. Pinkel, Diurnal cycling: Observations and models of the upper ocean response to diurnal heating, cooling, and wind mixing, J. Geophys. Res., 91, , Pudov, V. D., Mesostructure of the temperature and current velocity fields of a baroclinic ocean layer in the wake of typhoon Virginia, Oceanology, Engl. Transl., 20, , Pudov, V. D., A. A. Varfolomeyev, and K. N. Fedorov, Vertical structure of the wake of a typhoon in the upper ocean, Oceanology, Engl. Transl., 18, , Robinson, A. R., A. Hecht, N. Pinardi, Y. Bishop, W. G. Leslie, Z. Rosentroub, A. J. Mariano, and S. Brenner, Small synoptic/ mesoscale eddies: The variability of the Eastern Levantine Basin, Nature, 327, , Sanford, T. B., R. G. Drever, J. H. Dunlap, and E. A. D'Asaro, Design, operation and performance of an expendable temperature and velocity profiler (XTVP), Rep. APL-UW 8110, 83 pp., Appl. Phys. Lab., Univ. of Wash., Seattle, Sanford, T. B., P. G. Black, J. Haustein, J. W. Fenney, G. Z. Forristall, and J. F. Price, Ocean response to hurricanes, I, Observations, J. Phys. Oceanogr., 17, , Science Applications International Corporation, Gulf of Mexico physical oceanography program, Final report: Year 5, vol. II, 0C$ Rep./MM$ , 325 pp., Raleigh, N. C., Shay, L. K., Airborne expendable velocity profiling in the Subarctic Front during the Northeast Pacific Ocean Experiment, Tech. Rep , 99 pp., Rosenstiel Sch. of Mar. and Atmos. Sci., Univ. of Miami, Miami, Fla., Shay, L. K., and R. L. Elsberry, Near-inertial ocean current response to hurricane Frederic, J. Phys. Oceanogr., 17, , Shay, L. K., R. L. Elsberry, and P. G. Black, Vertical structure of the ocean current response to hurricanes, J. Phys. Oceanogr., 19, , Shay, L. K., S. W. Chang, and R. L. Elsberry, Effects of the free surface on the near-inertial response to a hurricane, J. Phys. Oceanogr., 20, , Shay, L. K., P. G. Black, J. D. Hawkins, R. L. Elsberry, and A. J. Mariano, Sea surface temperature response to hurricane Gilbert (extended abstract), in Nineteenth Conference on Hurricanes and Tropical Meteorology, pp , American Meteorological Society, Boston, Mass., Stramma, L., P. Cornilion, and J. F. Price, Satellite observations of sea surface cooling by hurricanes, J. Geophys. Res., 91, , Walsh, E. J., D. W. Hancock III, D. E. Hines, R. N. Swift, and J. F. Scott, Directional wave spectra measured with the surface contour radar, J. Phys. Oceanogr., 15, , Willoughby, H. E., F. D. Marks, Jr., and R. W. Feinberg, Stationary and moving convective bands in hurricanes, J. Atmos. $ci., 41(6), , P. G. Black, Atlantic Oceanographic and Meteorological Laboratories, Hurricane Research Division, NOAA, Miami, FL R. L. Elsberry, Department of Meteorology, U.S. Naval Postgraduate School, Monterey, CA J. D. Hawkins, Remote Sensing Branch, Naval Research Laboratory, Stennis Space Center, MS A. J. Mariano and L. K. Shay, Division of Meteorology and Physical Oceanography, Rosenstiel School of Marine and Atmospheric Science, University of Miami, 4600 Rickenbacker Causeway, Miami, FL (Received January 21, 1992; accepted April 9, 1992.)