Simulating the formation of Hurricane Katrina (2005)

Transcription

1 GEOPHYSICAL RESEARCH LETTERS, VOL. 35, L11802, doi: /2008gl033168, 2008 Simulating the formation of Hurricane Katrina (2005) Yi Jin, 1 Melinda S. Peng, 1 and Hao Jin 2 Received 2 January 2008; revised 28 March 2008; accepted 29 April 2008; published 3 June [1] The formation of Hurricane Katrina (2005) is simulated using COAMPS 1 with high resolution (27 km, 9 km and 3 km grid spacing). Whereas most Atlantic hurricanes form in conjunction with easterly waves, no clear synoptic-scale tropical disturbance was evident as a precursor for Katrina. The triply-nested simulation, initialized 66 h before Katrina was identified as a tropical depression, successfully predicts the location and timing of Katrina s formation in the 3 km domain. The organization of the cyclone is associated with concentrated precipitation resolved with microphysics in the southeast part of the circulation. The storm does not form in the 9 km domain when the Kain-Fritsch cumulus scheme and the microphysics are used. However, a separate simulation using microphysics alone in the 9 km domain captures Katrina s formation, though with less organized structure. This study suggests that explicitly resolving clouds in high-resolution models holds promise for predicting tropical cyclone (TC) formation. Citation: Jin, Y., M. S. Peng, and H. Jin (2008), Simulating the formation of Hurricane Katrina (2005), Geophys. Res. Lett., 35, L11802, doi: /2008gl Introduction [2] Tropical cyclone (TC) formation (or genesis) has been the focus of many theoretical and numerical modeling studies, yet accurately predicting TC formation remains a great challenge. Advances in physics parameterization as well as increased resolution have improved the ability of global models to simulate TC genesis. Shen et al. [2006] indicated that improving the representation of cloud-scale interactions between the Madden-Julian Oscillation (MJO) and TCs played a crucial role in predicting TC formation. Using 1/8-degree equivalent grid spacing in a global model, B.-W. Shen et al. (Forecasts of tropical cyclogenesis with a global mesoscale model: Preliminary results with six tropical cyclones, submitted to Geophysical Research Letters, 2007) successfully simulated the genesis of six TCs in May 2002 two to three days in advance. [3] Additionally, improved TC intensity forecasts in high-resolution (1 3 km) models have been demonstrated [Gopalakrishnan et al., 2002; Yau et al., 2004; Braun et al., 2006; Chen et al., 2007]. The development of TC Diana (1984) was simulated by Powers and Davis [2004] using a mesoscale regional model with 1.2 km grid spacing. With a synoptic-scale disturbance in the initial condition, a mesoscale convective system first developed along a remnant 1 Marine Meteorology Division, Naval Research Laboratory, Monterey, California, USA. 2 Science and Applications International Corporation, Monterey, California, USA. Copyright 2008 by the American Geophysical Union /08/2008GL frontal zone that yielded a mesoscale vortex. After a period of quiescence, banded convection organized about the vortex to from isolated cells that were explicitly resolved. The system then acquired a warm core and intensified into a tropical storm. In another simulation of TC Diana at 3 km grid spacing, Hendricks et al. [2004] suggested that vortical hot towers with strong vertical motions played the most important role in the formation of the storm. Tory et al. [2006] examined the mechanism of TC formation using the Australian Bureau of Meteorology s Tropical Cyclone Limited Area Prediction System and found that 0.15-degree grid spacing adequately represented the mean vertical motions in real TC formation convective regions, and suggested that perhaps explicitly resolving individual convective cells might not be necessary for qualitative TC formation forecasts. [4] The objective of this study is to investigate the performance of a high-resolution (3 km) mesoscale model in predicting the formation of Hurricane Katrina (2005). Katrina was the costliest TC in the US history and resulted in more than 3000 human deaths. While most hurricanes in the Atlantic form in conjunction with easterly waves, very little evidence of a synoptic-scale tropical disturbance was apparent prior to Katrina s formation. Our investigation seeks to determine if a high-resolution model can predict tropical cyclogenesis from such a weak initial disturbance. Additional experiments will demonstrate that the use of microphysics instead of cumulus parameterization holds the key to the success for this particular case. 2. Model and Experiment Setup [5] The model used for this study is the Coupled Ocean/ Atmosphere Mesoscale Prediction System (COAMPS 1,a trademark of the Naval Research Laboratory) [Hodur, 1997]. COAMPS 1 has been used for studies of various weather events at a wide range of scales from meso-alpha down to several hundred meters [Doyle et al., 2005; Thompson et al., 2007]. COAMPS 1 is a fully compressible model using terrain-following coordinates. The model physics include the Kain-Fritsch cumulus scheme [Kain and Fritsch, 1993]; the explicit microphysics based on Rutledge and Hobbs [1984]; the radiative transfer parameterization [Harshvardhan et al., 1987]; the surface layer parameterization [Louis, 1979; Fairall et al., 1996] with modifications incorporating effect of high winds over water; the 2.5 closure scheme for planetary boundary layer parameterization [Mellor and Yamada, 1982], and a physically consistent method of including dissipative heating [Jin et al., 2007]. [6] The model is initialized at 0000 UTC 21 August 2005 using the NOGAPS [Hogan and Rosmond, 1991] one-degree global analysis as the first guess field and NOGAPS forecast fields for lateral boundary conditions during the 72 h simulation. Figure 1 shows the three nested domains designed for this study with horizontal grid spacing of 27 km, 9 km and 3 km L of6

2 Figure 1. COAMPS initial wind vectors at 850 hpa from the Katrina simulation in domain 1 valid for 0000 UTC August 21. Domains D1, D2, and D3 have grid spacing of 27 km, 9 km, and 3 km, respectively. The inner dashed box indicates the display area in Figures 2 and 3. The three black dots indicate the first three observed locations of Katrina at 1800 UTC 23 August, 0000 UTC and 0600 UTC 24 August from east to west. for the first, second and third domain, respectively. These domains are centered near the genesis location of Katrina at 1800 UTC 23 August. There are 40 vertical levels with the highest resolutions near the surface. In the control run, the Kain-Fritsch cumulus parameterization is used in the outer two domains, but not in the inner-most domain. A separate simulation is performed using the Kain-Fritsch cumulus scheme in the 27-km domain only. Note that microphysics is present in all three domains for both simulations, but its effect is very small when the cumulus parameterization is activated (see discussions below). 3. Results 3.1. Domain 3 (3 km) Results [7] At the initial time of the simulation (0000 UTC 21 August), the southern tip of a mid-latitude trough extends into the northern part of the outer-most domain (Figure 1). The wind distribution in the southern part of the domain indicates the presence of easterly waves in that region. Meanwhile, the 3 km domain is dominated by easterly flow with weak wave signals in the wind fields to the south. After 12 h of simulation, pronounced southeasterly flow moves into the southeastern portion of the 3 km domain, accompanied by light precipitation (not shown). Wide-spread precipitation develops in the subsequent 12 h of the simulation with a concentrated precipitation maximum (70 80 mm in 12 h) between 21 23N and 68 70W (Figure 2a). The wind fields display a wave-like structure with a wavelength of about 1000 km with some small-scale disturbances imbedded within it. At this time, a noticeable change occurs in the 850-mb winds over southern Bahamas. The easterly flow weakens and westerly flow develops near the southeastern coast of Cuba, indicating that a closed circulation is about to form. 2of6

3 Figure 2. COAMPS simulated 12 h accumulated precipitation (mm) and 850 hpa wind vectors at 3 km grid spacing: (a) 24-h forecast valid for 0000 UTC 22 August; (b) 48-h forecast valid for 0000 UTC 23 August; (c) 60-h forecast valid for 1200 UTC 23 August; and (d) 72-h forecast valid for 0000 UTC 24 August. The black line in Figure 2c indicates the location of the vertical cross section shown in Figure 4a. [8] In the subsequent 12 hours (from h), the precipitation decreases and the wind weakens slightly while the scale of the northwestward moving wave contracts. Precipitation on the east side increases significantly by 48 h within the convergent region of the large-scale southwesterly and southeasterly flow. Wind speeds also increase as the circulation becomes more organized (Figure 2b). By 60 h (1200 UTC 23 August), a well-defined cyclonic circulation develops and contracts in scale (Figure 2c). As the circulation develops, the precipitation remains concentrated on the southeast side of the circulation. The 12 h- accumulated precipitation at 60 h is slightly less than that at 48 h, but banded structures begin to form. The more organized banded structure of the precipitation continues through the final 12-h of the simulation. By the end of the simulation, precipitation extends to the south and wraps around the storm center (Figure 2d). The winds have also become quite symmetric by this stage. The organization in the wind and precipitation fields suggests a correlation between them that is under further investigation. [9] The simulated maximum surface wind speed of 13 m s 1 and minimum sea level pressure (SLP) of 1008 hpa at 1800 UTC 23 August (the time of the initial formation of the observed tropical depression) compare well with the best track data (15 m s 1 for the maximum wind speed and 1008 hpa for the minimum SLP; pdf/tcr-al122005_katrina.pdf). The simulated circulation center at this time is located at 24.3N, 74.7W, about 139 km away from the observed location (23.1N, 75.1W). This result is considered encouraging given that it is 60 h into the simulation and no initial cyclonic disturbance is used. At the end of the 72-h simulation, the vorticity center 3of6

4 Figure 3. COAMPS simulated 12 h accumulated precipitation (mm) and 850 hpa wind vectors in the 9 km domain with the cumulus parameterization turned on: (a) 24-h forecast valid for 0000 UTC 22 August; (b) 60-h forecast valid for 1200 UTC 23 August; (c) as in Figure 3a, but for the run without using the cumulus scheme; and (d) as in Figure 3b, but for the run without the cumulus scheme. The black lines in Figures 3b and 3d indicate the locations of the vertical cross sections shown in Figures 4b and 4c, respectively. is located at 24.5N, 75.0W, about 140 km from the observed location (23.4N, 75.7W) of Katrina Domain 2 (9 km) Results [10] The simulation in the 9 km domain is displayed in Figure 3 over a sub-region that corresponds to the area covered by the 3 km domain. At 9 km, both the Kain- Fritsch cumulus parameterization and the microphysics scheme are active. The first 12 h of the simulation in domain 2 are very similar to those in domain 3, with light precipitation over the southeastern corner of the region (not shown). However, significant differences arise at subsequent times. While a wave structure forms in the 9 km domain, the precipitation is largely scattered at 24 h (Figure 2a). There is little precipitation on the east and southeast side in the convergent region where the easterly flow and southerly flow merge, and furthermore, there is no mesoscale circulation. The 12-h accumulated precipitation in domain 2 during the next 48 h remains very light and scattered and the wave takes on a more north-south orientation (Figure 3b). Again, there is little precipitation near the confluent region of the easterly and southerly flow. Due to the lack of intense and concentrated precipitation, the simulation fails to produce a closed cyclonic circulation in the region where Katrina s formation took place. Although the microphysics is activated the microphysics contributes less than 0.2 mm to the 12-h accumulated precipitation. This value is two orders of magnitude smaller than that from the cumulus parameterization (10 20 mm). Clearly, the contribution by the explicit microphysics is insignificant when the cumulus scheme is turned on. 4of6

5 Figure 4. Vertical cross sections of absolute vorticity (10 5 s 1 ) at 60 h of simulation valid for 1200 UTC 23 August (a) along the black line shown in Figure 2c; (b) along the black line shown in Figure 3c, and (c) along the black line shown in Figure 3d. (d) The GOES-12 IR image of Hurricane Katrina from the 0353 UTC overpass on 24 August The black dot indicates the location of Katrina formation at 1800 UTC 23 August [11] A TC intensity study of Katrina by Shen et al. [2006], with model equivalent grid spacing of degree, similar to our 9 km grid, indicated that deactivating the cumulus parameterization in lieu of the resolved microphysics produced a more realistic intensification rate for Katrina over the 5-day period from 26 to 31 August. We conducted a similar sensitivity test by deactivating the Kain- Fritsch scheme in the 9 km domain. The 12-h accumulated precipitation and 850 hpa-level winds from this test show much improvement (Figures 3c and 3d). The precipitation in the 9 km domain is now more intense than in the 3 km domain at 24 h (see Figure 3a) and a closed circulation has already formed at this early stage. This experiment indicates that convection generated by Kain-Fritsch cumulus parameterization scheme failed to organize the circulation into a strong mesoscale system in the preferred formation area. [12] The maximum precipitation in the microphysicsonly 9 km simulation is located on the eastern side of the system where the easterly flow and the southerly flow converge. As noted above, the precipitation maximum in domain 3 is locked in a similar position. However, the 3 km simulation produces more fine-scale banded structure with greater spatial coverage, especially during the last day of the simulation (see Figures 2d and 3d). Note the well-organized rain band wrapping around the east and southeast side of the circulation. Furthermore, the circulation in the 9 km domain becomes more elongated in the east-west direction after 60 h with weaker northerly flow on the western portions of the circulation. Therefore, the circulation is not as symmetric as seen in the 3 km domain during the last day of the simulation, although the maximum surface winds and the minimal SLP are about the same. [13] The development of the cyclone is also illustrated in the vertical distribution of vorticity at 60 h across the center of the vortex/disturbance (Figure 4) along 24N latitude (see Figures 2c, 3b, and 3d). In the 3 km domain, the vorticity field shows a symmetric pattern with respect to the vortex center, with a maximum located at 800 to 850 hpa (Figure 4a). The vorticity increases steadily through the development of the storm in this domain. In contrast, the vorticity profile in 5of6

6 the 9 km domain with the Kain-Fritsch cumulus parameterization shows very weak vorticity near 500 hpa and no development of vorticity at lower levels (Figure 4b). Note that the small-scale vorticity near 200 hpa is not related to the low-level disturbance of interest. When the cumulus parameterization is turned off, using only microphysics in the 9 km domain, strong vorticity develops to the east of the cyclone at lower levels (Figure 4c). As expected, the scale of the vorticity field in the 9 km domain is greater than in the 3 km simulation. The vorticity differences associated with the varying rainfall amounts indicate a possible role that convective precipitation plays in organizing and strengthening the weak disturbance. 4. Summary and Conclusions [14] COAMPS 1 is used to simulate the formation of Hurricane Katrina using triply nested domains with 27 km, 9 km and 3 km grid spacing. The 3 km simulation with explicitly resolved microphysics closely predicts the timing and location of the formation of Katrina. The formation of Katrina is associated with concentrated precipitation that evolves into a banded structure that crudely resembles the satellite image at the same time (Figure 4d). In contrast, the 9 km simulation with the Kain-Fritsch scheme produces only light, scattered precipitation with no TC formation. However, when the Kain-Fritsch scheme is deactivated, the 9 km simulation also captures the formation of Katrina fairly well. [15] Both the 3 km domain results and the differences between the 9 km domain results with and without the cumulus parameterization indicate a close relationship between the explicitly resolved convection and the organization of the wind circulation. For cases where TCs form within well organized easterly waves, explicit microphysics may be less critical, though additional research is needed to assess this possibility. Furthermore, the 3 km simulation is still superior to the 9 km simulation using the same microphysics. This study demonstrates that fine horizontal resolution (3 km) with explicit microphysics has the potential to predict TC formation and resolve mesoscale structures of TCs in their early stages. More detailed analysis of the simulations will be reported in the future. We plan to use high-resolution COAMPS 1 forecasts to support field experiments dedicated to TC formations in the western North Pacific during the 2008 season. [16] Acknowledgments. This research is sponsored by the Office of Naval Research under program N. The authors are thankful for the insightful discussions with Jason Nachamkin, James Doyle and William T. Thompson at the Naval Research Laboratory (NRL). The constructive comments by Bowen Shen at NASA Goddard Space Flight Center and the anonymous reviewer helped improve the quality of this work. References Braun, S. A., M. T. Montgomery, and Z. Pu (2006), High-resolution simulation of Hurricane Bonnie (1998). Part I: The organization of eyewall vertical motion, J. Atmos. Sci., 60, Chen, S. S., J. F. Price, W. Zhao, M. A. Donelan, and E. J. Walsh (2007), The CBLAST-Hurricane program and the next-generation fully coupled atmosphere wave ocean models for hurricane research and prediction, Bull. Am. Meteorol. Soc., 88, Doyle, J. D., M. A. Shapiro, Q. Jiang, and D. L. Bartels (2005), Largeamplitude mountain wave breaking over Greenland, J. Atmos. Sci., 62, Fairall, C., F. Bradley, D. P. Rogers, J. B. Edson, and G. S. Young (1996), Bulk parameterization of air-sea fluxes for Tropical Ocean Global Atmosphere Coupled Ocean Atmosphere Response Experiment, J. Geophys. Res., 101, Gopalakrishnan, S. G., et al. (2002), An operational multiscale hurricane forecasting system, Mon. Weather Rev., 130, Harshvardhan,, R. Davies, D. Randal, and T. Corsetti (1987), A fast radiation parameterization for atmospheric circulation models, J. Geophys. Res., 92, Hendricks, E. A., M. T. Montgomery, and C. A. Davis (2004), The role of vertical hot towers in the formation of tropical cyclone Diana (1984), J. Atmos. Sci., 61, Hodur, R. M. (1997), The Naval Research Laboratory s Coupled Ocean/ Atmosphere Mesoscale Prediction System (COAMPS), Mon. Weather Rev., 125, Hogan, T. F., and T. E. Rosmond (1991), The description of the Navy Operational Global Atmospheric Prediction System s spectral forecast model, Mon. Weather Rev., 119, Jin, Y., W. T. Thompson, S. Wang, and C.-S. Liou (2007), A numerical study of the effect of dissipative heating on tropical cyclone intensity, Weather Forecasting, 22, Kain, J. S., and J. M. Fritsch (1993), Convective parameterization for mesoscale models: The Kain-Fritsch scheme. The representation of cumulus convection in numerical models, Meteorol Monogr., 46, Louis, J.-F. (1979), A parametric model of vertical eddy fluxes in the atmosphere, Boundary Layer Meteorol., 17, Mellor, G. L., and T. Yamada (1982), Development of a turbulence closure for geophysical fluid problems, Rev. Geophys., 20, Powers, J. G., and C. A. Davis (2004), A cloud-resolving regional simulation of tropical cyclone formation, Atmos. Sci. Lett., 3(1), 15 24, 18 March. Rutledge, S. A., and P. V. Hobbs (1984), The mesoscale and microscale structure and organization of clouds and precipitation in midlatitude cyclones. XII: A diagnostic modeling study of precipitation development in narrow cold-frontal rainbands, J. Atmos. Sci., 41, Shen, B.-W., R. Atlas, O. Reale, S.-J. Lin, J.-D. Chern, J. Chang, C. Henze, and J.-L. Li (2006), Hurricane forecasts with a global mesoscale-resolving model: Preliminary results with Hurricane Katrina (2005), Geophys. Res. Lett., 33, L13813, doi: /2006gl Thompson, W. T., T. Holt, and J. Pullen (2007), Investigation of a sea breeze front in an urban environment, Q. J. R. Meteorol. Soc., 133, Tory, K. J., M. T. Montgomery, and N. E. Davidson (2006), Prediction and diagnosis of Tropical Cyclone Formation in an NWP System. Part I: The critical role of vortex enhancement in deep convection, J. Atmos. Sci., 63, Yau, M. K., Y. Liu, D.-L. Zhang, and Y. Chen (2004), A multiscale numerical study of Hurricane Andrew (1992). Part VI: Small-scale inner-core structures and wind streaks, Mon. Weather Rev., 132, H. Jin, Science and Applications International Corporation, 550 Camino El Estero Suite 103, Monterey, CA 93940, USA. Y. Jin and M. S. Peng, Naval Research Laboratory, 7 Grace Hopper Avenue, Stop 2, Monterey, CA , USA. (yi.jin@nrlmry.navy.mil) 6of6