GEOPHYSICAL RESEARCH LETTERS, VOL. 31, L03110, doi:10.1029/2003GL018571, 2004

Simulation of extratropical Hurricane Gustav using a coupled atmosphere-ocean-sea spray model William Perrie,1 Xuejuan Ren,1,2 Weiqing Zhang,1,3 and Zhenxia Long1,3 Received 6 September 2003; revised 3 November 2003; accepted 2 January 2004; published 12 February 2004.

[1] Numerical simulations of extratropical Hurricane mixing. Storm-induced upwelling influences the mixed Gustav (2002) are performed using the MC2 (Mesoscale layer temperature by pumping deeper colder water and Compressible Community) atmospheric model, coupled to increasing the mixed layer depth. 1 the Princeton Ocean Model (POM), and a sea spray [3] When the wind speed is higher than 15 ms , parameterization. On one hand, the impact of coupling substantial sea spray is produced by breaking waves. POM to MC2 generates sea surface temperature (SST) Andreas and Emanuel [2001] suggest that sea spray cooling, through entrainment mixing at the bottom of the contributions to heat fluxes can provide a net air-sea mixed layer, with the passage of the storm. SST cooling enthalpy transfer in tropical cyclones. In this study, we reduces the sea surface heat fluxes compared to uncoupled use the spray parameterization of Andreas [2003] with the MC2 simulations, which have time-invariant SST. Reduced MC2 atmospheric model, coupled to POM to study extra- heat fluxes lead to reduced storm intensity. On the other tropical Hurricane Gustav. Tropical storms are not our hand, simulation of the heat and mass flux contributions of focus. Gustav was selected to study air-sea interaction sea spray enhances sea surface heat fluxes and slightly during extratropical hurricane development, where barocli- increases maximum storm intensity compared to coupled nicity and midlatitude system interactions are dominant MC2-POM simulations without spray. INDEX TERMS: factors. We show that although the impacts of ocean 0312 Atmospheric Composition and Structure: Air/sea constituent feedbacks are comparatively minor, they are easily seen fluxes (3339, 4504); 3339 Meteorology and Atmospheric in Gustav’s central sea level pressure (SLP) and 10 m Dynamics: Ocean/atmosphere interactions (0312, 4504); 4504 winds U10. Oceanography: Physical: Air/sea interactions (0312). Citation: Perrie, W., X. Ren, W. Zhang, and Z. Long (2004), Simulation of extratropical Hurricane Gustav using a coupled 2. Model Description atmosphere-ocean-sea spray model, Geophys. Res. Lett., 31, L03110, doi:10.1029/2003GL018571. [4] The MC2 model is a state-of-the-art, fully elastic, nonhydrostatic model solving the full Euler equations on a limited-area Cartesian domain. It is able to successfully 1. Introduction simulate midlatitude cyclones [Benoit et al., 1997]. Lateral [2] In tropical cyclones, variability in intensity can orig- boundary and initial conditions are taken from CMC (Ca- inate from the coupling between the storm and the under- nadian Meteorlogical Centre) analysis data. The model lying ocean. This coupling is dominated by momentum and domain is (79.5W–40.0W, 24.25N–56.25N), using a enthalpy exchanges at the air-sea interface occurring with latitude-longitude projection, 0.25 resolution, 30 vertical the passage of a storm. Tropical cyclones can impact the layers, and 600 s time steps. Over the sea, MC2’s interfacial upper ocean by causing SST cooling by as much as 6C, as fluxes of momentum, and sensible and latent heat are seen in satellite images [Lin et al., 2003], and related near- calculated using Monin-Obukhov theory. This theory leads inertial currents as large as 2 ms1 [Bender and Ginis, to a bulk turbulent flux formulation, based on turbulent 2000]. Because SST cooling can significantly decrease the transfer coefficients for these fluxes, which depend on air-sea heat exchange, it reduces intensity empirical similarity functions ym and yh and roughness [Schade and Emanuel, 1999; Bender and Ginis, 2000; Chan lengths for wind speed, temperature, and humidity, z0m,z0t, et al., 2001]. Ocean temperature and related mixed layer and z0q. heat and mass budgets are largely controlled by entrainment [5] The MC2 default parameterizations for ym, yh,z0m, mixing at the base of the mixed layer [Jacob et al., 2000]. In z0t, and z0q must be handled with care because Andreas storm conditions, vertical current shear at the mixed layer [2003] developed our bulk spray flux algorithm by sub- base is the dominant factor in determining entrainment tracting estimates of the interfacial heat fluxes from HEXOS measurements of the total heat fluxes. Thus the residual is the spray flux; consequently, his spray algorithm is tuned to a specific turbulent bulk flux algorithm, related to the 1Fisheries & Oceans Canada, Bedford Institute of Oceanography, COARE algorithm [Fairall et al., 1996]. The spray algo- Dartmouth, Canada. rithm is not valid if we use a different bulk flux algorithm. 2 Department of Atmospheric Sciences, Nanjing University, Nanjing, PR Therefore we ignore MC2’s default z0m,z0t,z0q, ym, China. 3 yh parameterizations and implement Andreas’ [2003] Department of Engineering Mathematics, Dalhousie University, parameterizations. Total momentum t , latent H ,and Halifax, Canada. T L,T sensible Hs,T fluxes, constituting the boundary conditions Copyright 2004 by the American Geophysical Union. at the lowest model level, are obtained by simply adding the 0094-8276/04/2003GL018571$05.00 corresponding bulk interfacial (t,HL,Hs) and spray fluxes

L03110 1of4 L03110 PERRIE ET AL.: SIMULATION OF EXTRATROPICAL HURRICANE GUSTAV L03110

ated, embedded within a southwesterly flow ahead of a mid- level trough centered on . Dominated by bar- oclinic processes, it intensified rapidly over 28oC Gulf Stream waters and began to merge with the associated midlatitude low. Gustav became a hurricane by 12 UTC, with maximum winds of 85 kt (43 ms1) near 18 UTC. It made landfall over Cape Breton with 80-kt winds, achieving minimum central sea level pressure (SLP) at 06 UTC 12 September. It transi- tioned to an , and turning north, made a second landfall over Newfoundland.

Figure 1. (a) Simulations of Gustav’s storm track from 18 4. Upper Ocean Impacts UTC 10 September to 12 UTC 13 September: Control MC2 (—6), MC2-POM (– –), MC2-POM-spray (—&), and [9] Figure 1a presents simulations of Gustav’s storm NHC analysis (—.). Six-hourly intervals shown. Surface track in comparison with the NHC (National Hurricane currents at 48 h in the simulations (! ms1), and SST (C) Center) analysis. The model integration is from 18 UTC at 48 h, minus the initial SST, from MC2-POM-spray 10 September to 12 UTC 13 September. Gustav moves model. (b) Difference in SST (C) distributions of the MC2- slowly during the initial portion of the simulation, and POM-spray minus MC2-POM simulations at 48 h. Storm accelerates during its extratropical phase as it moves past location is .. Buoy station 44011 is . . Compared to the NHC analysis, the simulation from MC2-POM-spray is slightly better than that of MC2- POM, which in turn is better than the uncoupled MC2. [10] Results for surface currents and SST are also shown (tsp,QL,sp,QS,sp), following Andreas and Emanuel [2001] and [Andreas, 2003], in Figures 1a and 1b. By 48 h in the MC2-POM-spray simulation, which is late in Gustav’s life cycle, widespread SST cooling has occurred on both sides of the storm track. t ¼ t þ t ð1Þ T sp A maximum of 8C cooling occurs along a narrow area to the right of the storm track because of rightward bias of the wind structure. This can be verified with satellite SST data H H Q 2 L;T ¼ L þ L;sp ð Þ from The Johns Hopkins University http://fermi.jhuapl.edu/ avhrr). Surface-currents induced by Gustav’s winds are also presented in Figure 1a: These achieve maximum speeds to Hs;T ¼ Hs þ QS;sp ð3Þ the right of the storm track. Maximum differences in surface currents and SST between MC2-POM-spray and MC2- 1 [6] The POM model [Mellor, 1998] is implemented for POM simulations, are 0.1 ms and 0.3C. the ocean domain (82W–40W, 20N–57.5N) on a [11] Figures 2a–2c show time–depth section profiles for latitude-longitude projection, with 0.1/6 horizontal resolu- ocean temperature To, horizontal U, and vertical w currents, tion, and 23 vertical layers, of which 8 are in the upper 80 m. Lateral prescribed barotropic transports are used to keep the Gulf Stream in the right position. Initial and boundary conditions for temperature and salinity are taken from monthly Naval Oceanographic Office Generalized Digital Environmental Model (GDEM) data [Bender and Ginis, 2000]. The model spin-up involves integration for one year, using monthly mean wind stress, heat flux, and fresh water flux data from the National Centers for Environmental Prediction (NCEP). The integration is continued for a second year using NCEP data for a given storm to provide a realistic pre-storm ocean representation. [7] The coupled model system exchanges information between the atmosphere and ocean at every coupling time step. Wind stress, sensible and latent heat fluxes, radiative flux and fresh water flux, as computed from MC2 (includ- ing sea spray), are transferred to POM, which is then integrated for two time-steps. The new POM-produced SST is then passed to MC2, which is then integrated forward for another five time-steps (50 minutes). Figure 2. Time-depth sections of (a) ocean temperature To (C), (b) horizontal current speed U (ms1), and (c) vertical velocity w (ms1). Location (65.34W, 41.33N) is the 3. Extratropical Hurricane Gustav (2002) position (Figure 1a) of greatest SST cooling. At 21 UTC 11 [8] Gustav was designated a tropical storm by 12 UTC on September, this location achieves its shortest distance to the 10 September, north of the Bahamas. Nearing storm center, as indicated .. Contour intervals are 1C, on 11 September, Gustav turned northeastward and acceler- 0.1 ms1 and 0.5 104 ms1, respectively.

2of4 L03110 PERRIE ET AL.: SIMULATION OF EXTRATROPICAL HURRICANE GUSTAV L03110

Figure 3. Comparison of uncoupled MC2 (– – –) and coupled models MC2-POM (—6), MC2-POM-spray (—) simulations, for with NHC (—) and CMC () results for (a) 1 central SLP (hPa), and (b) maximum U10 (ms ). for the location (shown in Figure 1a) of the strongest SST cooling (65.34W, 41.33N). Gustav’s passage deepens the thermocline base from 15 m to a final 30 m, cools the SST by 8C (Figure 2a), and generates horizontal and vertical currents of about 1.4 ms1 and 3.5 104 ms1, respec- tively. While the strongest horizontal current occurs at the surface and is highly sheared between the surface and the thermocline base, the strongest vertical current is at the base Figure 5. Comparisons of buoy 44011 (66.59W, of thermocline as part of the upwelling and entrainment 41.09N,) data (—), uncoupled MC2 (), and MC2- mixing that drives the SST cooling. POM-spray coupled simulations (– – –), for (a) SST (C), 1 (b) U10 (ms ), and (c) central SLP (hPa). CMC analysis 5. Impact on Gustav (—) is included in (d).

[12] SST cooling can lead to reduced air-sea fluxes and decreased storm intensity. For example, idealized studies [Chan et al., 2001] found that 3.6C SST cooling corre- propagated faster than a typical tropical cyclone, it produced sponds to a central SLP increase of about 30 hPa, depending strong SST cooling. This impacts the storm intensity in on the ocean mixed layer and the atmospheric system. In terms of central SLP and wind speed (Figures 3a and 3b). real storm studies, Bender and Ginis [2000] obtained Comparing the uncoupled MC2 simulation to the MC2- substantial improvements in storm intensity predictions by POM and MC2-POM-spray simulations, the biggest differ- accounting for SST cooling. These results were obtained for ences are 14 hPa and 11 hPa, respectively. These simula- slowly moving tropical storms. For rapidly moving storms tions are less intense than the NHC analysis, in terms of over deep oceanic mixed layers, SST cooling has little central SLP and winds, reflecting the relatively less intense impact [Schade and Emanuel, 1999]. CMC analysis data used as driving fields. The MC2-POM [13] In late summer and early autumn, the mixed layer is and MC2-POM-spray central SLP results are in better shallow in the northwest Atlantic. Thus, although Gustav agreement with NHC results than the uncoupled MC2 results in the latter half of the simulation period, after Gustav’s minimum central SLP. The largest SLP difference between MC2-POM and MC2-POM-spray simulations is 1 3 hPa at 18 UTC 12 September, and in U10, 3ms at 00 UTC 12 September, demonstrating the role of sea spray. [14] Figures 4a and 4b present maximum air-sea surface latent and sensible heat fluxes from the uncoupled and coupled simulations. These estimates are area-averages on 2 2 centered on the maxima of the latent and sensible heat fluxes, respectively, following the storm’s trajectory. The dominance of latent heat flux over sensible heat flux is evident. Cumulatively, for the time series presented in Figures 4a and 4b, the MC2-POM simulation delivers 24% less total (latent + sensible) heat than simulations by the uncoupled MC2 model, 10% less than the MC2-POM- spray model. The peak latent heat flux, and also the peak Figure 4. Comparison of uncoupled MC2 (– – –), MC2- sensible heat flux, for the uncoupled MC2 simulation, POM (—6), and MC2-POM-spray (—)simulations exceeds that of the MC2-POM simulation by about for sea surface fluxes of maximum (a) latent heat, and 100 W m2. This reflects greater SST cooling and less (b) sensible heat, area-averages on 2 2 centered on the storm intensity in the MC2-POM simulation. The peak total maxima of the latent and sensible heat fluxes, respectively, heat fluxes from the MC2-POM-spray simulation also following the storm’s trajectory. Units are W m2. exceeds that of the MC2-POM simulation by about

3of4 L03110 PERRIE ET AL.: SIMULATION OF EXTRATROPICAL HURRICANE GUSTAV L03110

uncoupled MC2 simulation. The latter experiences no storm-induced SST depression and produces excessive heat fluxes to the atmosphere, compared to the coupled simu- lations, resulting in overestimates in storm intensity. On one hand, the coupled model simulations show storm-induced SST cooling, resulting from strong upwelling and entrain- ment mixing below the mixed layer. On the other hand, the contribution of sea spray to heat and momentum fluxes is notable under high-wind conditions. At the peak of the storm, spray enhances the winds by as much as 3.5 ms1 compared to the MC2-POM coupled simulation without spray.

[17] Acknowledgments. Funding was from the Panel on Energy Research and Development (PERD) of Canada. We thank John Gyakum and Ron McTaggart-Cowan for setting up the MC2 atmospheric model, and Ed Andreas for the sea spray parameterization and for comments on this manuscript. We thank Prof. Issac Ginis for the URI version of POM. Figure 6. Distribution of the difference in U10 fields of MC2-POM-spray minus MC2-POM, at 00 UTC 12 References September, when coupled simulations have peak surface Andreas, E. L. (2003), An algorithm to predict the turbulent air-sea fluxes wind speed. Storm center is ., and storm track from the in high-wind, spray conditions. 12th Conf. Interaction of the Sea and MC2-POM simulation are shown. Atmosphere, Long Beach, CA, Am. Meteorol. Soc., 7 pp. Andreas, E. L., and J. DeCosmo (2002), The signature of sea spray in the HEXOS turbulent heat flux data, Bound. Layer Meteorol., 103, 303–333. 100 W m2. Thus, heat fluxes are reduced by SST cooling, Andreas, E. L., and K. A. Emanuel (2001), Effects of sea spray on tropical cyclone intensity, J. Atmos. Sci., 58, 3741–3751. but enhanced by spray, particularly in high winds and large Bender, M. A., and I. Ginis (2000), Real-case simulations of hurricane- air-sea temperature differences. ocean interaction using a high-resolution coupled model: Effects on hur- [15] Figures 5a–5c compare observed SST, U ,and ricane intensity, Mon. Wea. Rev., 128, 917–946. 10 Benoit, R., M. Desgagne, P. Pellerin, Y. Chartier, and S. Desjardins (1997), central SLP data from National Data Buoy Center (NDBC) The Canadian MC2: A semi-implicit semi-Lagrangian wide-band atmo- buoy 44011 at (66.59W, 41.09N) with uncoupled and spheric model suited for fine-scale process studies and simulation, Mon. coupled simulations. SST cooling is evident in the NDBC Wea. Rev., 125, 2382–2415. Chan, J. C. L., Y. Duan, and L. K. Shay (2001), Tropical cyclone intensity data and in the MC2-POM-spray simulation but not in the change from a simple ocean-atmosphere couple model, J. Atmos. Sci., 58, uncoupled MC2 simulation, because it assumes time-invari- 154–172. ant SSTs. The MC2 and MC2-POM-spray simulations Fairall, C. W., E. F. Bradley, D. P. Rogers, J. B. Edson, and G. S. Young compare well to U and central SLP buoy data in Figures (1996), Bulk parameterization of air-sea fluxes for Tropical Ocean-Global 10 Atmosphere Coupled-Ocean Atmosphere Response Experiment, J. Geo- 5a and 5b, and to CMC analysis in Figure 5c. The phys. Res., 101(C2), 3747–3764. atmosphere-ocean coupling in MC2-POM-spray simula- Jacob, S. D., L. K. Shay, and A. J. Mariano (2000), the 3D oceanic mixed tions depress the SST and wind fields relative to those of layer response to hurricane Gilbert, J. Phys. Ocean., 30, 1407–1429. Lin, I.-I., W. T. Liu, C.-C. Wu, C. H. Chiang, and C.-H. Sui (2003), Satellite MC2. But the effect is slight; spray and ocean coupling have observations of modulation of surface winds by typhoon-induced upper little impact on the storm at this phase of its life cycle. ocean cooling, Geophy.Res.Lett., 30(3), 1131, doi:10.1029/ However, at the peak of the storm (Figure 6), the maximum 2002GL015674. impact of spray is about 3.5 ms1 and covers a large area. Mellor, G. L. (1998), User’s guide for a three-dimensional primitive equa- tion numerical ocean model. Program in Atmospheric and Oceanic Sciences, Princeton Univ., 35 pp. Schade, L. R., and K. A. Emanuel (1999), The ocean’s effect on 6. Summary the intensity of tropical cyclones: Results from a simple coupled atmosphere-ocean model, J. Atmos. Sci., 56, 642–651. [16] Hurricane Gustav is simulated with a coupled atmo- sphere-ocean model. The model consists of the MC2 atmospheric model, the POM ocean model, and a recent sea spray model of [Andreas, 2003; Andreas and DeCosmo, Z. Long, W. Perrie, X. Ren, and W. Zhang, Fisheries & Oceans Canada, 2002]. The MC2-POM and MC2-POM-spray coupled sim- Bedford Institute of Oceanography, Dartmouth, Canada. (perriew@ ulations achieve better storm intensity estimates than the dfo-mpo.gc.ca)

4of4