Modification of the loop current warm core eddy by Hurricane Gilbert (1988)

Transcription

1 DOI /s ORIGINAL PAPER Modification of the loop current warm core eddy by Hurricane Gilbert (1988) Xiaodong Hong Æ Simon W. Chang Æ Sethu Raman Received: 9 April 2005 / Accepted: 3 August 2006 Ó Springer Science+Business Media B.V Abstract Numerical investigation of Hurricane Gilbert (1988) effect on the Loop Current warm core eddy (WCE) in the Gulf of Mexico is performed using the Modular Ocean Model version 2 (MOM2). Results show that the storm-induced maximum sea surface temperature (SST) decrease in Gilbert s wake is over 2.5 C, as compared with the 3.5 C cooling in the absence of the WCE. The near-inertial oscillation in the wake reduces significantly in an along-track direction with the presence of the WCE. This effect is also reflected between the mixed layer and the thermocline, where the current directions are reversed with the upper layer. After two inertial periods (IP), the current reversal is much less obvious. In addition, it is demonstrated that Hurricane Gilbert wind stress increases the current speed of the WCE by approximate 133%. With the forcing of Gilbert, the simulated translation direction and speed of the WCE towards the Mexican coast are closer to the observed (42% more accurate in distance and 78% more accurate in direction) compared with the simulation without the Gilbert forcing. The simulated ocean response to Gilbert generally agrees with the recent observations in Hurricane Fabian. Keywords Hurricane Gilbert Æ Warm core eddy Æ Near-inertial oscillation Æ Ocean response 1 Introduction Hurricane Gilbert attained a record minimum sea level pressure (MSLP) of 888 hpa at 2152 UTC 13 Sept 1988 when traveling toward west-northwest (WNW) in the western X. Hong (&) Æ S. W. Chang Naval Research Laboratory, 7 Grace Hopper Ave., Monterey, CA 93943, USA xd.hong@nrlmry.navy.mil S. Raman North Carolina State University, Raleigh, NC, USA

2 Yearday /00 16/06/ 15/12 13/18 Fig. 1 Sea surface temperature (shaded), surface height (contour) and surface current (vector) at yearday 257 as initial condition in the simulations. The track of Hurricane Gilbert (dark solid line) is marked with date and time in UTC Caribbean. Gilbert entered the Gulf of Mexico after landfall on the Yucatan Peninsula, moving in a northwestward direction at about 5.6 m s 1 (Fig. 1). Although the storm was much weaker when it entered the Gulf of Mexico, Gilbert maintained nearly constant intensity (MSLP of 940 hpa) before the final landfall in Mexico approximately 200 km south of Brownsville, Texas on 17 Sept Analyses of the AXBT and drifting buoy data and satellite images indicated that there was a Loop Current warm core eddy (WCE) F located within 200 km to the right of Gilbert s track. Objectively analyzed temperature and estimated geostrophic flow fields from AXCPs deployed in the western Gulf of Mexico one day after the passage of Gilbert (Wake I Experiment in Shay et al. 1998) indicated that there was interaction between hurricane-induced flows and the background eddy field. Maximum geostrophic currents in the upper m were approximately m s 1 rotating anticyclonically around the WCE F. In the core of the WCE F, observed temperature was C, about 1 C cooler than the observed undisturbed Gulf water temperatures of C. Upper ocean cooling reached 3 4 C with a right-side bias was observed along the storm track to about 200 km (Shay et al. 1992). However, cooling at 50 m depth in the WCE F was not as evident. There is a large thermal gradient between the cold wake and the warm eddy in the central Gulf of Mexico as indicated by the AVHRR images and the Airborne InfraRed Thermometer (AIRT). The purpose of this study is to discuss the simulated ocean response to Hurricane Gilbert using the GFDL Modular Ocean Model version 2 (MOM2). The simulated model results will be compared with observation. The model results, as it will be shown, provide further detailed ocean response to Hurricane Gilbert as comparing to the relatively sparse observation. In addition to investigate the upper ocean thermal response, near-inertial and vertical response, the modification of WCE F by Hurricane Gilbert surface forcing will also studied.

3 The research described in this chapter is presented in four sections. Section 2 briefly describes the ocean model and numerical experiments. Section 3 discusses and summarizes the numerical results. Section 4 compares the model-simulated results with recent observation. Section 5 gives conclusions. 2 The ocean model and experiment design The ocean model used in this study is the Geophysical Fluid Dynamics Laboratory s modular ocean model version 2 (MOM2), which is a three-dimensional primitive equation model (Pacanowski 1996). The model has adapted the same ocean domain and resolutions as used in the study on the interaction between Hurricane Opal (1995) and a warm core ring in the Gulf of Mexico (Hong et al. 2000), since the two hurricanes had traversed in the Gulf of Mexico. The coefficients for the vertical mixing scheme such as maximum mixing coefficient, background diffusion coefficient and the viscosity coefficient are also same as in Hong et al. (2000). In lieu of an effective ocean data assimilation and prediction system, a 2-year spin-up ended at yearday 257 with the limited-area domain for the Gulf of Mexico provides an ocean state resembling the observed state prior to Hurricane Gilbert (Fig. 1). This realization is used as the initial condition in this study. In this realization there is a WCE with location and strength similar to the observed, located 200 km to the right of Gilbert s track. A prescribed surface wind stress field (Chang and Anthes 1978) for Hurricane Gilbert is used as a surface forcing. The maximum wind stress specified is 42 dyn cm 2 at the radius of maximum winds of 60 km. The translation direction and the speed of the stress field follow the best track of Hurricane Gilbert, as depicted in Fig. 1 and listed in Table 1 of Shay et al. (1992). Three numerical experiments are conducted for this study. Case C1 is a control experiment that simulates the response of the Loop Current and the WCE F to Hurricane Gilbert wind field. Case C2 studies the Gilbert-induced response of an initially motionless ocean without the WCE with uniformly distributed temperature. Case C3 is conducted without the hurricane forcing, in which the WCE is allowed to drift freely. All simulations start at 1200 UTC Sept 13. The simulation periods for Cases C1 and C3 are 45 days and for Case C2 is 10 days. These simulations encompass the periods prior to, during, one day after (Wake I), and three days after (Wake II Experiment in Shay et al. 1998) the passage of Gilbert. These results are compared to the observation (Shay et al. 1992; 1998). The 45-day experiments for Cases C1 and C3 extend to the yearday 302, which the analyses from observed WCE F are available for the comparison. 3 The ocean response in the Gulf of Mexico Since the WCE F is located to the right of the storm track at approximately 200 km, or about 3 radius of maximum wind (R max ), the Gilbert-induced ocean response is as expected, complicated due to this mesoscale variability. The horizontal extent of the cold wake region induced by the hurricane is limited by the WCE because it has a deeper mixed layer depth, so is the SST cooling. The induced divergent and convergent currents are modulated by the anticylonic current associated with the WCE.

4 The structure and the drifting speed of the WCE F are also modified as Gilbert passes through the domain (Shay et al. 1992). 3.1 The upper ocean thermal response The simulated sea surface temperature (SST) distributions and changes after the storm at 1500 UTC 19 Sept 1988 are shown in Fig. 2a for Case C1 and in Fig. 2b for Case C2. In Case C1, the maximum simulated SST decrease is over 2.5 C and the pool of cold water (<26.0 C) extends to the western region of WCE F. The temperature gradient between the cool pool and WCE is about 2.0 C over a 100-km distance. This large horizontal temperature gradient induces advection that may reduce the temperature over the WCE. To the west, the decrease of SST is less than 1.0 C near the Mexican coast due to the shallow shelf. The cooler SST is caused by the mixing in strong nearinertial current shears across the base of the mixed layer (Shay et al. 1992). This is in contrast to the central WCE where the maximum surface cooling is <1 C during the storm as well as wake periods due to the deep warm layer. The pattern of upper-ocean thermal response differs greatly between Cases C1 and C2 as expected. The difference of temperature change between the two cases is due to the existence of WCE adjacent the storm track. The deeper and warmer layers associated with the WCE F reduce much of the storm-induced cooling. In Case C2, the temperature change is uniformly distributed between 1 and 2R max on the right side of the storm track and in the wake of the storm (Fig. 2b), similar to idealized studies of ocean response to hurricane (Chang and Anthes 1979, Price 1981). The uniform initial temperature field over the Gulf of Mexico is the main reason for this distribution. The SST decreases in C2 from above 28.5 C to below 25.0 C on the right side of the storm track, suggesting a maximum cooling of 3.5 C. 3.2 Near-inertial response The time series of normalized surface current speed of u- and v-components have been plotted for the cross-track distance at 2R max and for the along-track distance at 0.2k for Cases C1 and C2 (Fig. 3a d). The inertial wavelength k is ~600 km, given Gilbert s translation speed. The initial time is 0600 UTC 16 September to indicate the beginning of the storm period. The abscissa is scaled in inertial period (IP, about 30 h) relative to the initial time. Therefore, the along-track distance at 0.2k is closer to the location of the WCE than at 2R max.at2r max, the time evolution of the surface current patterns in the two cases is very similar (Fig. 3a, b) and is only slightly larger magnitude in C2. The surface velocity u- and v-components are in quadrature and have an asymmetric response skewed to the right of the storm track, as the results of Shay et al. (1998). However, the responses at the distance of 0.2k (120 km) from the two cases are significantly different. Although the near-inertial oscillation exists as in Shay et al. (1998) with the WCE F (Fig. 3c), the simulated surface velocity components are dominated by the WCE current speed. In Case C2 (Fig. 3d), the surface current response at 0.2k is similar to the idealized case in previous studies. 3.3 Vertical structures The depth-time cross-sections of the u-component for the cross-track distance at 2R max and for the along-track distance at 0.2k for C1 are shown at 2R max (Fig. 4a)

5 (a) With the WCE (1500 UTC 19 Sept) (a) (b) Without the WCE (1500 UTC 19 Sept) Fig. 2 Sea surface temperature SST (shaded) and the SST change (contour) for (a) Case C1, (b) Case C2 at 1500 UTC 19 Sept 1988 and at 0.2k (Fig. 4b). During the storm-forcing period (0 0.3 IP), the upper layer currents are accelerated to a maximum speed of 130 cm s 1 in the east-west direction and 180 cm s 1 in the north-south direction (Figure not shown). The maximum

6 (a) (b) (c) (d) Fig. 3 Normalized u- and v-components of surface current for Case C1 and Case C2 at 2R max and 0.2 k respectively, where the v m = 160 cm s 1. The abscissa is scaled in inertial period relative to the point where Hurricane Gilbert approached at 0600 UTC 16 Sept 1988 surface velocity in the C1 is larger than the observed maximum mixed layer velocities of 110 cm s 1 (Shay et al. 1998). The maximum current components decrease to 80 cm s 1 within the first 1.5 inertial periods. The current reversed direction below the thermocline, indicating a downward energy propagation of near-inertial waves (Shay et al. 1989). As shown in Fig. 3c, the near-inertial oscillation in the along-track direction weakens due to the presence of the anticyclonic circulation of the modulated WCE (Fig. 4b). The near-inertial oscillation is evident within the first 2 IP with a reversal of currents between the mixed layer and the thermocline. The current reversal also exists after 2 IP, but with decreasing amplitude. In the case without the WCE F (C2), the near-inertial oscillation and reversal current between the mixed

7 (a) (b) (c) Fig. 4 Depth-time cross section of u-component at (a) 2R max,(b) 0.2k for Case C1, (c) 0.2k for the Case C2. The abscissa is scaled as in Fig. 3 layer and the thermocline are comparatively more discernible after 2 IP (Fig. 4c), which is consistent with observations of Shay et al The modification of WCE F Hurricane Gilbert moved in a WNW direction and passed within 200 km south of WCE F (Fig. 1), it disturbs the WCE F more along the west side than along the east side of eddy. The time change of surface current at the center of WCE F along 24.3 N is shown in Fig. 5. The WCE F had a nearly asymmetric anticyclonic circulation before it was affected by Gilbert s wind stress. The induced current response first takes place along the east side of WCE when Hurricane Gilbert passed the Yucatan Peninsula and entered the western Gulf of Mexico, as indicated by the change of southward currents along the east side of eddy to the southwest. Both direction and speed of the current show significantly changes along the west side of eddy or west of 92 W after Sept 16. The simulated surface current speed increases from 60 to 140 cm s 1, a 133% increase under Gilbert s forcing. In the same period,

8 Fig. 5 Time change of surface current at the center of the WCE F along 24.3 N for Case C1 the current direction changes from northward to northwestward then westward. Subsequently, as Gilbert moved on, the induced current backs from eastward to westward. The isotach patterns in Fig. 5 delineated the inertial oscillation in the surface current. The current speeds in general decrease with time after Sept 17 as Gilbert continued to more away from the eddy. The current east of the eddy was only increased slightly over 60 cm s 1. The current speed in the center of WCE F has a corresponding increase and decrease pattern due to the induced oscillation. The westward movement of WCE F over the 8-day period is evident in Fig. 5. The shape and the position of WCE F were affected by Hurricane Gilbert too. Results from the two 45-day simulations with (C1) and without (C3) Gilbert s forcing are compared in Fig. 6, where the 0 and 30 or 40 cm surface height contours have been plotted for C1 (solid lines) and C3 (dotted lines) at various days from the simulations. Both simulations start from Sept 14 with the same initial condition with

9 Fig. 6 The contour lines of 0 and 30/40 cm surface height for Cases C1 (solid lines) and C3 (dotted lines) at the various days into the simulations the WCE as shown in Fig. 6a. After Hurricane Gilbert passes the WCE F (Fig. 6b), the eddy center location moves up more to the northwest in comparison to C3. These are caused by the strong hurricane-induced northward current along the south side of eddy. The shape and the translation speed induced differ when the WCE F propagates towards the Mexican coast (Fig. 6c f). The moving direction changes from southwest to west in C1 by the Hurricane Gilbert effect. By Oct 26, i.e.,

10 Fig. 7 Time series of surface temperature from (a) buoys for Fabian (Niiler, personal communication), (b) model simulated for Gilbert. Letters left and right indicate the temperature changes are collected at the left and right side of the storm yearday 300 (Fig. 6f), the WCE F locates at 24.4 N and 93.6 W in C1, and at 23.7 N and 93.3 W in C3. The central location of WCE F at the yearday 300 from the observation given by Shay et al. (1998) is at 24.5 N and W. The distances from the observed location to the model locations of C1 and C3 are 72 km and 138 km. The location in C1 is about 8.6 south of the observed location, but the location in C3 is 40 south of the observed location. There is about 42% more accurate in distance and 78% more accurate in direction for C1. Hence, the drift of

11 the warm core eddy is more accurately reproduced in the simulation with Gilbert s forcing (C1) than that without (C3). 4 Comparison with recent observation Observational campaign in the Hurricane Component of the coupled boundary layers air-sea transfer (CBLAST) has collected unprecedented rich data sets of ocean response, among them SST and temperature profiles during and after Hurricane Fabian (Sept 2 4, 2003) and Isabel (Sept 12 14, 2003) by drifting buoys and subsurface floats (Black, 2004). Time series of observed surface temperature for Fabian (Niiler, personal communication), and our Gilbert simulation (C1) are shown on Fig. 7. All the temperature variations feature a rapid decrease for 1.5 day after the storm passage, and oscillated in the wake. The inertial period is about 24 h for Fabian and 30 h for Gilbert. Since some buoys were located at the left side of the storm and others at the right side of the storm, the decrease of temperature and the inertial oscillations had different magnitude and amplitude indicating a strong right bias. Larger temperature decrease and oscillation amplitude are shown for the right side. This right bias of ocean response to a storm is consistent with previous observation (Price 1981; Sanford et al. 1987; Shay et al. 1992). There are noticeable differences of surface and subsurface temperature changes between the observed in Fabian and simulated for Gilbert. These can be attributed to the differences in different mixed layer depth and structure in different basins, and the model initial state and imperfections, such as a lack of recovering mechanism. Temperature profiles from model simulation for Gilbert (Case C1) during the storm and after the passage (Fig. 8a) are compared with those collected from CTD float in Fabian (Terrill, personal communication, Fig. 8b). There are general similarities in the observed and simulated mixed layer structures after the storm passage. Generally speaking, temperature decreases at the surface and upper mixed layer, and increases below the thermocline. The model is obviously incomparable to depict the sharp gradient at the bottom of the mixed layer due to limited vertical resolution. In late time, the variation of deep layer temperature is complicated by upwelling and horizontal advection in both the observation and simulation. 5 Conclusions The ocean response to Hurricane Gilbert has been studied using the GFDL MOM2 with an initial ocean state including the Loop Current and the WCE F similar to the observed features and locations prior to Gilbert. The results from the numerical simulation were able to conform to the observation about the modification of the Loop Current WCE F by Hurricane Gilbert and to provide us a more detailed ocean features in three-dimensional and over a wider region than observation. The maximum upper ocean temperature cooling due to the storm-induced mixing and upwelling was over 2.5 C. The existence of Warm Core Eddy adjacent to the right side of Gilbert s path reduced the decrease of temperature caused by the storm due to the fact of deep warm mixed layer inside the eddy. In a sensitivity test where

12 Fig. 8 Temperature profiles from (a) CTD float (Terrill, personal communication) obtained every 4 h for Fabian, (b) model simulation for Gilbert

13 the initial condition in the Gulf of Mexico is quiescent, the maximum induced SST decreased 3.5 C. Gilbert modified the WCE F more in the west side than in the east side because of its track. The maximum surface current speed of the eddy under Gilbert s forcing increased by 133% of its pre-storm value with the current direction changed from northward to northwestward then westward periodically with inertial period. In the vertical direction, the near-inertial oscillation propagated downward, however, in the case with the WCE the vertical current reversal between the mixed layer and the thermocline is reduced after 2 IP due to the well-balanced eddy fields. The drift speed and location of WCE F in its translation to the west of the Gulf of Mexico was closer to the observed with Gilbert s forcing. The simulated ocean response to Hurricane Gilbert is compared with recent CBLAST observations in Hurricane Fabian. There are general agreements in terms of rapid cooling during the storm period, inertial oscillation during the wake period, and evolution of the mixed layer structure. However, due to the limited vertical resolution, the model is incomparable to depict the sharp gradient at the bottom of the mixed layer. Furthermore, the model imperfections, such as a lack of recovering mechanism could be one of the reasons that results surface temperature decrease during the inertial oscillation (Fig. 7). The Gulf of Mexico is characterized by the Loop Current, episodic shedding of warm core eddies and their westward propagation. These mesoscale activities transport the warm and salty Caribbean subtropical underwater (SUW) into the Gulf. The life cycle of the eddy have a profound effect on the circulation in the Gulf of Mexico and greatly affect the local fishing, offshore drilling and other industries and can impact the behavior of atmospheric disturbances such as Hurricane Opal (Hong et al. 2000; Shay et al. 2000; Bosart et al. 2000). Therefore, accurate analysis and forecast of eddy propagation will be important not only for hurricane forecasting but also for many local industries. Since there is about 47% chance every year for intense hurricane cross the Gulf of Mexico, it will be necessary to consider the influence of hurricane in the studies of Loop Current WCE propagation. Acknowledgements This research was supported by NRL Basic Research Program PE601153N and the ONR 6.2 program PE602435N. The computations were performed at the Naval Oceanographic Office (NAVOCEANO) of the Department of Defense Major Shared Resource Center. References Black PG (2004) An overview of CBLAST flights into Hurricane Fabian and Isabel (2003), 26th Conference on hurricane and tropical meteorology, AMS, 3 7 May, Miami Beach, FL Bosart LF, Velden CS, Bracken WE, Molinari J, Black PG (2000) Environmental influences on the rapid intensification of Hurricane Opal (1995) over the Gulf of Mexico. Mon Wea Rev 128: Chang SW, Anthes RA (1978) Numerical simulations of the ocean s nonlinear baroclinic response to translating hurricanes. J Phys Oceanogr 8: Chang SW, Anthes RA (1979) The mutual response of the tropical cyclone and the ocean. J Phys Oceanogr 9: Hong X, Chang SW, Raman S, Shay LK, Hodur RM (2000) The interaction between Hurricane Opal (1995) and a warm core ring in the Gulf of Mexico. Mon Wea Rev 128:

14 Pacanowski RC (1996) MOM 2 (version 2.0) documentation user s guide and reference manual. GFDL Ocean Tech Rep 32, pp 329 Price JF (1981) Upper ocean response to a hurricane. J Phys Oceanogr 11: Sanford TB, Black PG, Haustein JR, Feeney JW, Forristall GZ, Price JF (1987) Ocean response to a hurricane. Part: I observations. J Phys Oceanogr 17: Shay LK, Elsberry RL, Glack PG (1989) Vertical structure of the ocean response to a hurricane. J Phys Oceanogr 19: Shay LK, Black PG, Mariano AJ, Hawkins JD, Elsberry RL (1992) Upper ocean response to Hurricane Gilbert. J Geophy Res 97: Shay LK, Mariano AJ, Jacob SD, Ryan EH (1998) Mean and near-inertial ocean current response to Hurricane Gilbert. J Phys Oceanogr 28: Shay LK, Goni GJ, Black PG (2000) Effects of a warm oceanic feature on Hurricane Opal. Mon Wea Rev 125: