• Bender, M. A., , and Ginis I. , 2000: Real-case simulations of hurricane–ocean interaction using a high-resolution coupled model: Effects on hurricane intensity. Mon. Wea. Rev., 128, 917946.

    • Search Google Scholar
    • Export Citation
  • Bender, M. A., , Ginis I. , , Tuleya R. , , Thomas B. , , and Marchok T. , 2007: The operational GFDL coupled hurricane–ocean prediction system and a summary of its performance. Mon. Wea. Rev., 135, 39653989.

    • Search Google Scholar
    • Export Citation
  • Beven, J. L., , and Kimberlain T. B. , 2009: Tropical cyclone report: Hurricane Gustav (AL072008) 25 August–4 September 2009. National Hurricane Center, 38 pp. [Available online at http://www.nhc.noaa.gov/pdf/TCR-AL072008_Gustav.pdf.]

  • Bleck, R., 2002: An oceanic general circulation model framed in hybrid isopycnic-Cartesian coordinates. Ocean Modell., 4, 5588.

  • Bleck, R., , and Boudra D. , 1981: Initial testing of a numerical ocean circulation model using a hybrid (quasi-isopycnic) vertical coordinate. J. Phys. Oceanogr., 11, 755770.

    • Search Google Scholar
    • Export Citation
  • Blumberg, A. F., , and Mellor G. L. , 1987: A description of a three-dimensional coastal ocean circulation model. Three-Dimensional Coastal Ocean Models, Geophys. Monogr., Vol. 4, Amer. Geophys. Union, 1–16.

  • Booth, J. F., , Thompson L. A. , , Patoux J. , , and Kelly K. A. , 2012: Sensitivity of midlatitude storm intensification to perturbations in the sea surface temperature near the Gulf Stream. Mon. Wea. Rev., 140, 12411256.

    • Search Google Scholar
    • Export Citation
  • Brooks, A. D., 1983: The wake of Hurricane Allen in the western Gulf of Mexico. J. Phys. Oceanogr., 13, 117129.

  • Browning, G. L., , and Kreiss H.-O. , 1986: Scaling and computation of smooth atmospheric motions. Tellus, 38A, 295313.

  • Canuto, V. M., , Howard A. , , Cheng Y. , , and Dubovikov M. S. , 2001: Ocean turbulence. Part I: One-point closure model—Momentum and heat vertical diffusivities. J. Phys. Oceanogr., 31, 14131426.

    • Search Google Scholar
    • Export Citation
  • Canuto, V. M., , Howard A. , , Cheng Y. , , and Dubovikov M. S. , 2002: Ocean turbulence. Part II: Vertical diffusivities of momentum, heat, salt, mass, and passive scalars. J. Phys. Oceanogr., 32, 240264.

    • Search Google Scholar
    • Export Citation
  • Chelton, D. B., , DeSzoeke R. A. , , and Schlax M. G. , 1998: Geographical variability of the first baroclinic Rossby radius of deformation. J. Phys. Oceanogr., 28, 443460.

    • Search Google Scholar
    • Export Citation
  • Chen, S. S., , Price J. F. , , Zhao W. , , Doneland M. A. , , and Walsh E. J. , 2007: The CBLAST-Hurricane program and the next-generation fully coupled atmosphere–wave–ocean models for hurricane research and prediction. Bull. Amer. Meteor. Soc., 88, 311317.

    • Search Google Scholar
    • Export Citation
  • Cione, J. J., , and Uhlhorn E. W. , 2003: Sea surface temperature variability in hurricanes: Implications with respect to intensity change. Mon. Wea. Rev., 131, 17831796.

    • Search Google Scholar
    • Export Citation
  • Cooper, M., , and Haines K. , 1996: Altimetric assimilation with water property conservation. J. Geophys. Res., 101 (C1), 10591077.

  • Cuyers, Y., , Le Vailant X. , , Bouruet-Aubertot P. , , Vialard J. , , and McPhaden M. J. , 2013: Tropical storm-induced near-intertial waves during the Cirent experiment: Energy fluxes and impact on vertical mixing. J. Geophys. Res., 118, 358380, doi:10.1029/2012JC007881.

    • Search Google Scholar
    • Export Citation
  • Donlon, C. J., , Minnett P. J. , , Gentemann C. , , Nightingale T. J. , , Barton I. J. , , Ward B. , , and Murray M. J. , 2002: Toward improved validation of satellite sea surface skin temperature measurements for climate research. J. Climate, 15, 353369.

    • Search Google Scholar
    • Export Citation
  • Environmental Modeling Center, 2003: The GFS Atmospheric Model. NCEP Office Note 442, 14 pp. [Available online at http://www.emc.ncep.noaa.gov/officenotes/newernotes/on442.pdf.]

  • Falkovich, A. I., , Ginis I. , , and Lord S. , 2005: Ocean data assimilation and initialization procedure for the coupled GFDL/URI Hurricane Prediction System. J. Atmos. Oceanic Technol., 22, 19181932.

    • Search Google Scholar
    • Export Citation
  • Feliks, Y., 1985: On the Rossby radius of deformation in the ocean. J. Phys. Oceanogr., 15, 16051607.

  • Ferrier, B. S., 1994: A double-moment multiple-phase four-class bulk ice scheme. Part I: Description. J. Atmos. Sci., 51, 249280.

  • Ferrier, B. S., , Tao W. K. , , and Simpson J. , 1995: A double-moment multiple-phase four-class bulk ice scheme. Part II: Simulations of convective storms in different large-scale environments and comparisons with other bulk parameterizations. J. Atmos. Sci., 52, 10011033.

    • Search Google Scholar
    • Export Citation
  • Franklin, J. L., 2008: National Hurricane Center forecast verification. Preprints, 28th Conf. on Hurricanes and Tropical Meteorology, Orlando, FL, Amer. Meteor. Soc., 13A.1. [Available online at http://www.nhc.noaa.gov/verification.]

  • Franklin, J. L., 2009: 2008 National Hurricane Center forecast verification report. National Hurricane Center, 71 pp. [Available online at http://www.nhc.noaa.gov/verification/pdfs/Verification_2008.pdf.]

  • Gangopadhyay, A., , Robison A. R. , , and Arango H. G. , 1997: Circulation and dynamics of the western North Atlantic. Part I: Multiscale feature models. J. Atmos. Oceanic Technol., 14, 13141332.

    • Search Google Scholar
    • Export Citation
  • Gopalakrishnan, S., , Liu Q. , , Marchok T. , , Sheinin D. , , Surgi N. , , Tuleya R. , , Yablonsky R. , , and Zhang Z. , 2010: Hurricane Weather Research and Forecasting (HWRF) model scientific documentation. L. Bernardet, Ed., NOAA/Earth System Research Laboratory, 75 pp.

  • Jaimes, B., , and Shay L. K. , 2010: Near-inertial wave wake of hurricanes Katrina and Rita over mesoscale oceanic eddies. J. Phys. Oceanogr., 40, 1320–1337.

    • Search Google Scholar
    • Export Citation
  • Janjić, Z. I., 1994: The step-mountain coordinate model: Further developments of the convection, viscous sublayer and turbulence closure schemes. Mon. Wea. Rev., 122, 927945.

    • Search Google Scholar
    • Export Citation
  • Janjić, Z. I., , Gall R. , , and Pyle M. E. , 2010: Scientific documentation for the NMM solver. NCAR Tech. Note NCAR/TN-477+STR, 54 pp.

  • Kim, H.-S., , Gangopadhyay A. , , Rosenfeld L. K. , , and Bub F. , 2007: Developing a high-resolution climatology for the central California coastal region. Cont. Shelf Res., 27, 21352161.

    • Search Google Scholar
    • Export Citation
  • Kurihara, Y., , and Tuleya R. E. , 1974: Structure of a tropical cyclone developed in a three-dimensional numerical simulation model. J. Atmos. Sci., 31, 893919.

    • Search Google Scholar
    • Export Citation
  • Kurihara, Y., , Tuleya R. E. , , and Bender M. A. , 1998: The GFDL Hurricane Prediction System and its performance in the 1995 hurricane season. Mon. Wea. Rev., 126, 13061322.

    • Search Google Scholar
    • Export Citation
  • Landsea, C., , and Franklin J. , 2013: Atlantic hurricane database uncertainty and presentation of a new database format. Mon. Wea. Rev., 141, 3576–3592.

    • Search Google Scholar
    • Export Citation
  • Lee, C.-Y., , and Chen S. S. , 2012: Symmetric and asymmetric structures of hurricane boundary layer in coupled atmosphere–wave–ocean models and observations. J. Atmos. Sci., 69, 35763594.

    • Search Google Scholar
    • Export Citation
  • MacAfee, A. W., , and Pearson G. M. , 2006: Development and testing of tropical cyclone parametric wind models tailored for midlatitude application—Preliminary results. J. Appl. Meteor. Climatol., 45, 12441260.

    • Search Google Scholar
    • Export Citation
  • Marchok, T. P., 2002: How the NCEP tropical cyclone tracker works. Preprints, 25th Conf. on Hurricanes and Tropical Meteorology, San Diego, CA, Amer. Meteor. Soc., P1.13. [Available online at http://ams.confex.com/ams/pdfpapers/37628.pdf.]

  • Marks, F. D., Jr., 2007: Recent results from NOAA’s Hurricane Intensity Forecast Experiment (IFEX). Preprints, 11th Symp. on Integrated Observing and Assimilation Systems for the Atmosphere, Oceans, and Land Surface, San Antonio, TX, Amer. Meteor. Soc., 8.1. [Available online at https://ams.confex.com/ams/87ANNUAL/techprogram/paper_119567.htm.]

  • Maximenko, N., , Niiler P. , , Centuioni L. , , Rio M.-H. , , Melnichenko O. , , Chambers D. , , Zlotnicki V. , , and Galperin B. , 2009: Mean dynamic topography of the ocean derived from satellite and drifting buoy data using three different techniques. J. Atmos. Oceanic Technol., 26, 19101919.

    • Search Google Scholar
    • Export Citation
  • Mehra, A., , and Rivin I. , 2010: A real time ocean forecast system for the North Atlantic Ocean. Terr. Atmos. Oceanic Sci., 21, 211228.

    • Search Google Scholar
    • Export Citation
  • Moon, I.-J., , Ginis I. , , Hara T. , , and Thomas B. , 2007: A physics-based parameterization of air–sea momentum flux at high wind speeds and its impact on hurricane intensity predictions. Mon. Wea. Rev., 135, 28692878.

    • Search Google Scholar
    • Export Citation
  • Paltoglou, G., , Beggs H. , , and Majewski L. , 2010: New Australian high resolution AVHRR SST products from the Integrated Marine Observing System. IMOS, 11 pp. [Available online at http://imos.org.au/fileadmin/user_upload/shared/SRS/Paltoglou_2010_IMOS_AVHRR_v1_0.pdf.]

  • Pan, H.-L., , and Wu W.-S. , 1995: Implementing a mass flux convection parameterization package for the NMC medium-range forecast model. NMC Office Note 409, 40 pp.

  • Price, J. F., , Sanford T. B. , , and Forristall G. Z. , 1994: Forced stage response to a moving hurricane. J. Phys. Oceanogr., 24, 233260.

    • Search Google Scholar
    • Export Citation
  • Rogers, R., and Coauthors, 2006: The intensity forecasting experiment: A NOAA multiyear field program for improving tropical cyclone intensity forecasts. Bull. Amer. Meteor. Soc., 87, 15231537.

    • Search Google Scholar
    • Export Citation
  • Sanford, T. B., , Price J. F. , , and Girton J. B. , 2011: Upper-ocean response to Hurricane Frances (2004) observed by profiling EM-APEX floats. J. Phys. Oceanogr., 41, 10411056.

    • Search Google Scholar
    • Export Citation
  • Schade, L. R., , and Emanuel K. A. , 1999: The ocean’s effect on the intensity of tropical cyclones: Results from a simple coupled atmosphere–ocean model. J. Atmos. Sci., 56, 642651.

    • Search Google Scholar
    • Export Citation
  • Shay, L. K., , Mariano A. J. , , Jacob S. D. , , and Ryan E. H. , 1998: Mean and near-inertial ocean current response to Hurricane Gilbert. J. Phys. Oceanogr., 28, 858889.

    • Search Google Scholar
    • Export Citation
  • Shay, L. K., , Goni G. J. , , and Black P. G. , 2000: Effects of a warm oceanic feature on Hurricane Opal. Mon. Wea. Rev., 128, 13661383.

  • Tallapragada, V., 2010: NCEP operational hurricane WRF (HWRF) modeling system. 11th WRF Users’ Workshop, Boulder, CO, NCAR. [Available online at http://www.mmm.ucar.edu/wrf/users/workshops/WS2010/presentations/session%202/2-2_vijay_wrf_users_workshop_2010.pdf.]

  • Teague, W. J., , Carron M. J. , , and Hogan P. J. , 1990: A comparison between the generalized digital environmental model and Levitus climatologies. J. Geophys. Res., 95 (C5), 71677183.

    • Search Google Scholar
    • Export Citation
  • Tuleya, R. E., 1994: Tropical storm development and decay: Sensitivity to surface boundary conditions. Mon. Wea. Rev., 122, 291304.

  • Wai, M., , and Stage S. A. , 1989: Dynamical analyses of marine atmospheric boundary layer structure near the Gulf Stream oceanic front. Quart. J. Roy. Meteor. Soc., 115, 2944.

    • Search Google Scholar
    • Export Citation
  • Warner, T. T., , Lakhtakia M. N. , , Doyle J. D. , , and Pearson R. A. , 1990: Marine atmospheric boundary layer circulations forced by Gulf Stream sea surface temperature gradients. Mon. Wea. Rev., 118, 309323.

    • Search Google Scholar
    • Export Citation
  • Williams, R. G., , Wilson C. , , and Hughues C. W. , 2007: Ocean and atmosphere storm tracks: The role of eddy vorticity forcing. J. Phys. Oceanogr., 37, 22672289.

    • Search Google Scholar
    • Export Citation
  • Xu, F., , and Ignatov A. , 2010: Evaluation of in situ surface temperatures for use in the calibration and validation of satellite retrievals. J. Geophys. Res., 115, C09022, doi:10.1029/2010JC006129.

    • Search Google Scholar
    • Export Citation
  • Yablonsky, R. M., , and Ginis I. , 2008: Improving the ocean initialization of coupled hurricane–ocean models using feature-based data assimilation. Mon. Wea. Rev., 136, 25922607.

    • Search Google Scholar
    • Export Citation
  • View in gallery

    The HWRF–HYCOM hurricane prediction system.

  • View in gallery

    Hurricane (inner) nested and (outer) RTOFS-Atlantic basin nesting domains.

  • View in gallery

    AXBT surveys (in color) and sample locations (o marks) superimposed on the NHC best track of Hurricane Gustav (07L).

  • View in gallery

    Prestorm survey, 28 Aug 2008: (a) navigation-corrected NOAA-18 AVHRR SST (color) superimposed with the survey track with probe locations (x and o) that are numbered at random and (b) comparison between collocated AXBT and AVHRR SST [circles in (a)] as a function of probe number.

  • View in gallery

    As in Fig. 4, but for a poststorm survey (8 Sep 2008).

  • View in gallery

    The (a) 2008 and (b) 2009 Atlantic TCs: the NHC best-track positions are shown at 6-h intervals with intensity at the Saffir–Simpson wind scale (solid circle).

  • View in gallery

    Comparisons of oceanic parameters for prestorm survey as a function of probe number: (a) SST from three sources and (b) MLD (blue) and Z26 (green) from simulations (thin) and observations (thick).

  • View in gallery

    Comparisons of (top) in-storm SST, (middle) MLD, and (bottom) Z26 between (left) HY09 simulations, (middle) AXBT observations, and (right) differences (of simulation from observation). Superimposed in each plot are the survey track (thin black lines) and the NHC best track (thick green line with + marks). The red–green–blue palette applies to both simulations and observations.

  • View in gallery

    Time evolution of footprint MLD (m) at six locations (inset) on the HY09 forecast track for Ike (09L) (IC, 0600 UTC 10 Sep 2008). Black open circle denotes the time of the storm pass at a given location.

  • View in gallery

    Comparison of track forecasts between HY09 (black) and H209 (gray) for 223 homogeneous cases. Absolute average error (lines) and standard deviation (vertical bars) are a function of forecast time and number of cases (parenthesis). Units are in nautical miles (n mi).

  • View in gallery

    Average track bias (nm) relative to the best-track position (0,0) for models (rows) and forecast hour (columns). The x axis and y axis of the individual boxes represent bias (nm) to the east and northdirections.

  • View in gallery

    Comparison of intensity forecast between HY09 (black) and H209 (gray) for 248 homogeneous cases showing (a) absolute mean error and standard deviation, and (b) bias. Number of samples (parenthesis) and forecast hour are shown on the x axis. Units are in knots (kt).

All Time Past Year Past 30 Days
Abstract Views 0 0 0
Full Text Views 117 117 17
PDF Downloads 99 99 12

Performance of Ocean Simulations in the Coupled HWRF–HYCOM Model

View More View Less
  • 1 * I. M. Systems Group Inc., and Marine Modeling and Analysis Branch, NOAA/NWS/NCEP/EMC, College Park, Maryland
  • | 2 Marine Modeling and Analysis Branch, NOAA/NWS/NCEP/EMC, College Park, Maryland
  • | 3 Hurricane Modeling, NOAA/NWS/NCEP/EMC, College Park, Maryland
  • | 4 I. M. Systems Group Inc., and Global Climate and Weather Modeling Branch, NOAA/NWS/NCEP/EMC, College Park, Maryland
  • | 5 Marine Modeling and Analysis Branch, NOAA/NWS/NCEP/EMC, College Park, and Operations and Requirements Division, NOAA/NWS/OCWWS, Silver Spring, Maryland
© Get Permissions
Full access

Abstract

This paper introduces a next-generation operational Hurricane Weather Research and Forecasting (HWRF) system that was developed at the U.S. National Centers for Environmental Prediction. The new system, HWRF–Hybrid Coordinate Ocean Model (HYCOM), retains the same atmospheric component of operational HWRF, but it replaces the feature-model-based Princeton Ocean Model (POM) with the eddy-resolving HYCOM. The primary motivation is to improve enthalpy fluxes in the air–sea interface, by providing the best possible estimates of the balanced oceanic states using data assimilated Real-Time Ocean Forecast System products as oceanic initial conditions (IC) and boundary conditions.

A proof-of-concept exercise of HWRF–HYCOM is conducted by validating ocean simulations, followed by the verification of hurricane forecasts. The ocean validation employs airborne expendable bathythermograph sampled during Hurricane Gustav (2008). Storm-driven sea surface temperature changes agree within 0.1° and 0.5°C of the mean and root-mean-square difference, respectively. In-storm deepening mixed layer and shoaling 26°C isotherm depth are similar to observations, but they are overpredicted at similar magnitudes of their ICs. The forecast verification for 10 Atlantic hurricanes in 2008 and 2009 shows that HWRF–HYCOM improves intensity by 13.8% and reduces positive bias by 43.9% over HWRF–POM. The HWRF–HYCOM track forecast is indifferent, except for days 4 and 5, when it shows better skill (8%) than HWRF–POM. While this study proves the concept and results in a better skillful hurricane forecast, one well-defined conclusion is to improve the estimates of IC, particularly the oceanic upper layer.

Corresponding author address: Hyun-Sook Kim, Marine Modeling and Analysis Branch, NOAA/NWS/NCEP/EMC, 5830 University Research Court, College Park, MD 20740. E-mail: hyun.sook.kim@noaa.gov

Abstract

This paper introduces a next-generation operational Hurricane Weather Research and Forecasting (HWRF) system that was developed at the U.S. National Centers for Environmental Prediction. The new system, HWRF–Hybrid Coordinate Ocean Model (HYCOM), retains the same atmospheric component of operational HWRF, but it replaces the feature-model-based Princeton Ocean Model (POM) with the eddy-resolving HYCOM. The primary motivation is to improve enthalpy fluxes in the air–sea interface, by providing the best possible estimates of the balanced oceanic states using data assimilated Real-Time Ocean Forecast System products as oceanic initial conditions (IC) and boundary conditions.

A proof-of-concept exercise of HWRF–HYCOM is conducted by validating ocean simulations, followed by the verification of hurricane forecasts. The ocean validation employs airborne expendable bathythermograph sampled during Hurricane Gustav (2008). Storm-driven sea surface temperature changes agree within 0.1° and 0.5°C of the mean and root-mean-square difference, respectively. In-storm deepening mixed layer and shoaling 26°C isotherm depth are similar to observations, but they are overpredicted at similar magnitudes of their ICs. The forecast verification for 10 Atlantic hurricanes in 2008 and 2009 shows that HWRF–HYCOM improves intensity by 13.8% and reduces positive bias by 43.9% over HWRF–POM. The HWRF–HYCOM track forecast is indifferent, except for days 4 and 5, when it shows better skill (8%) than HWRF–POM. While this study proves the concept and results in a better skillful hurricane forecast, one well-defined conclusion is to improve the estimates of IC, particularly the oceanic upper layer.

Corresponding author address: Hyun-Sook Kim, Marine Modeling and Analysis Branch, NOAA/NWS/NCEP/EMC, 5830 University Research Court, College Park, MD 20740. E-mail: hyun.sook.kim@noaa.gov

1. Introduction

Traditionally, short-range weather prediction produced at the U.S. National Weather Service (NWS) is conducted with fixed-in-time distributions of sea surface temperature (SST), assuming its temporal variability has little effect within the forecast range. However, coastal upwelling, boundary currents, eddies, and ocean responses to strong winds significantly affect SST temporal variability and its gradient over a large area. They can also affect atmospheric motions, engaging at least the planetary boundary layer (Wai and Stage 1989; Warner et al. 1990), for example, the development and evolution of atmospheric vortices sustained by enthalpy fluxes from the ocean in the storm corridor in the midlatitude (Booth et al. 2012) or in the hurricane and typhoon corridors in the tropics. Storm-driven SST changes are driven by multiscale physical processes, including turbulent air–sea flux modified by wave and air entrainment, wind-stress-induced mixing, and vertical advection. The changes can be substantial on the storm footprint (defined as an area within a radius from the storm center; Price et al. 1994; Sanford et al. 2011). Such variability influences air–sea enthalpy fluxes, and it eventually affects the storm path and intensity (Schade and Emanuel 1999; Chen et al. 2007; Lee and Chen 2012). If oceanic mesoscale features exist on the storm path, the interaction can be more complex (Jaimes and Shay 2010; Shay et al. 2000). Chen et al. (2007) and Lee and Chen (2012) have demonstrated in their numerical studies that two-way interaction must be included, in order to improve atmosphere and ocean forecast skill. Based on the analysis of the 11-yr storm guidance performance of the operational Geophysical Fluid Dynamics Laboratory (GFDL) hurricane–ocean model, Bender et al. (2007) have concluded that skill improvement in hurricane intensity forecast can be also be improved by accurate initialization of ocean mesoscale features.

Toward having a next-generation nonhydrostatic mesoscale atmospheric model that can be operated at a much higher resolution, the Hurricane Weather Research and Forecasting (HWRF) system has been developed at the National Centers for Environmental Prediction (NCEP)’s Environmental Modeling Center (EMC), in close collaboration with scientists from GFDL and the University of Rhode Island. HWRF became operational in 2007 and has been providing useful guidance at a similar forecast skill (Franklin 2009) as the operational GFDL model (Bender and Ginis 2000; Bender et al. 2007). The original intention of HWRF development was to couple to the Hybrid Coordinate Ocean Model (HYCOM; Bleck and Boudra 1981; Bleck 2002), and to replace the GFDL model. However, at present both HWRF and GFDL are running in operations, being coupled to the Princeton Ocean Model (POM; Blumberg and Mellor 1987).

POM is computationally efficient. However, the initialization is nontrivial. Each cycle of the ocean model starts from ½° monthly generalized digital environmental model (GDEM) climatology (Teague et al. 1990). The model is then integrated for 0.5 and 2 months in two separate phases in order to establish dynamically consistent monthly currents, temperature, and salinity (Falkovich et al. 2005). The final initial conditions are produced by two numerical integrations: a 72-h spinup using assimilation of daily NCEP SST combined with feature models (e.g., Gangopadhyay et al. 1997 and Kim et al. 2007) that project the surface information to the upper layer, followed by a 48-h forward computation forced by the National Hurricane Center (NHC) forecast winds along the storm track in order to simulate a cold wake (Yablonsky and Ginis 2008). The routine is repeated for each cycle run.

In contrast, the HWRF–HYCOM system adopts a rather simple and yet comprehensive initial condition (IC) and boundary condition (BC) procedure. Currently, two sets of ocean model runs are available at NCEP from Real-Time Ocean Forecast Systems (RTOFSs): one is the Atlantic basin version (RTOFS-Atlantic) that became operational in 2005 (Mehra and Rivin 2010) and the other is a global version (RTOFS-Global) that became operational in 2011. Both RTOFSs produce data-assimilation-based ocean nowcast using HYCOM at eddy-resolving resolution. These models provide a 3D estimate of the balanced ocean state with special attention given to hurricane forcing, thus removing the need for separate ocean initialization for individual hurricane simulations. RTOFS–HYCOM that is developed and maintained at NCEP in collaboration with the U.S. Navy and several other partners provides a unified operational environment that could help to leverage resources required for development of the coupled hurricane model.

This work introduces a capability of a complex hurricane forecast system, such as HWRF, in examining the effects of one ocean component against another on hurricane predictions. The initial development of the oceanic component, reported here, was based on RTOFS-Atlantic, limiting the applicability to the Atlantic basin only. Our primary focus of this study is to examine the quality of simulated ocean states in the context of air–sea interactions under hurricane conditions. This includes validation of the ocean simulations, by investigating ocean conditions critical to the atmosphere–ocean interaction such as SST, mixed layer depth (MLD), and the depth of 26°C isotherm (Z26). It is followed by the verification of HWRF–HYCOM forecasts for 10 Atlantic tropical storms in 2008 and 2009, and the comparison of the HWRF hurricane forecasts. To attain an equitable comparison to the operational HWRF, we use the same configuration of the 2009 operational version in HWRF–HYCOM. Calibration of all the components of HWRF–HYCOM is outside the scope of this study.

Section 2 introduces the HWRF–HYCOM system and a summary of the ocean model. Section 3 documents the observations employed, including airborne expendable bathythermograph (AXBT) surveys, AXBT data quality assurance, the NHC-issued best-track data, and synopsis of Hurricane Gustav (07L) 2008. Section 4 presents results from validation of the ocean simulation, followed by the performance of HWRF–HYCOM for 10 Atlantic hurricane forecasts. The final section, section 5, presents a discussion and conclusions.

2. The coupled HWRF–HYCOM model

Figure 1 shows the components and interconnections of HWRF and HYCOM. The core of the system (solid ellipse) consists of the HWRF atmospheric model, the HYCOM ocean model, and a coupler (dashed box). The atmospheric component is one-way nested to the Global Forecast System (GFS; Environmental Modeling Center 2003); the ocean component is one-way nested to RTOFS-Atlantic. The coupler exchanges SST (A in Fig. 1) and atmospheric forcing between the atmospheric and ocean model at a given numerical step.

Fig. 1.
Fig. 1.

The HWRF–HYCOM hurricane prediction system.

Citation: Journal of Atmospheric and Oceanic Technology 31, 2; 10.1175/JTECH-D-13-00013.1

For each 6-h cycle, the ocean component carries out a nowcast procedure (dashed ellipse in Fig. 1) to update ocean IC for the ellipse. The nowcast is realized as follows: The previous ocean ICs are employed to perform a 6-h ocean spinup forced by GFS modified fluxes. The modification is conducted by melding GFS winds with fine-resolution parametric winds (MacAfee and Pearson 2006) using the storm position and maximum winds from the NHC Tropical Cyclone Vitals Database (TCVitals). BCs for the ocean component are prepared from 6-hourly RTOFS-Atlantic fields. HWRF is initialized by interpolating the global analysis fields of GFS onto the HWRF grid, followed by a vortex relocation using the location, intensity, and structure from the NHC TCVitals. If the 6-h forecast from the previous cycle is available, the vortex initialization is replaced by modifying the HWRF vortex. Otherwise, a bogus two-dimensional (2D) axisymmetric vortex is employed. HWRF utilizes boundary conditions from the GFS forecasts at 6-h intervals.

The atmospheric model component employed in operations evolves each year with updates to model physics, model algorithms, and resolution. The atmospheric component also employs novel vortex initialization, and other innovations resulting from extensive preimplementing testing and evaluation. This proof-of-concept work is based on the 2009 operational version of HWRF (Tallapragada 2010). In the following, we present a brief introduction to the atmospheric and oceanic component.

a. The atmospheric component—HWRF 2009

HWRF is a regional model configured with two-way interaction between an inner domain (7.2° × 6.0°) and an outer domain (77.5° × 77.5°). While the latter is quasi stationary at 27-km resolution, the former follows a storm at a 9-km resolution. Both domains maintain a 1:3 ratio in the horizontal resolution and the numerical integration step but use the same vertical resolution. The HWRF model is built upon the WRF Nonhydrostatic Mesoscale Model (NMM) dynamic core to solve the 3D primitive equations (Janjić et al. 2010). Solutions are computed on a longitude–latitude Arakawa E grid and 42 hybrid (pressure and sigma) layers, with a time step of 18 s for the inner domain and 54 s for the outer domain, both of which should be the integer divisor of the coupler time step (e.g., 540 s).

HWRF employs a suite of physics modules specifically modified for the hurricane prediction problem. They are listed in Table 1. More details are presented in Gopalakrishnan et al. (2010).

Table 1.

Summary table of the HWRF physics modules.

Table 1.

b. The ocean model component—HYCOM 2009

1) Model configuration

The hurricane–ocean domain covers sections of the Atlantic warm pool and the hurricane corridor. The model component is RTOFS–HYCOM (Mehra and Rivin 2010), which is based on HYCOM (Bleck and Boudra 1981; Bleck 2002). Solutions to the governing 3D primitive equations are sought on an Arakawa C grid, using the time step that is leapfrog for the internal mode and a predictor corrector for the external mode. The advection scheme is fourth order, and the gradient operator is numerically curl free. HYCOM supports several vertical mixing and diffusion schemes. RTOFS-Atlantic employs the Goddard Institute for Space Studies (GISS) scheme (Canuto et al. 2001, 2002). The nested domain configured in this demonstration of concept is shown in Fig. 2. While both nested and nesting domains have the same vertical coordinate, the horizontal grid points of the former are collocated with those of the latter, but with half resolution. The nesting grid is curvilinear with its size varying smoothly from ~9 km in the Gulf of Mexico (GOM) to ~34 km in the eastern North Atlantic.

Fig. 2.
Fig. 2.

Hurricane (inner) nested and (outer) RTOFS-Atlantic basin nesting domains.

Citation: Journal of Atmospheric and Oceanic Technology 31, 2; 10.1175/JTECH-D-13-00013.1

2) Initial and boundary conditions

ICs are estimated from RTOFS-Atlantic 24-h nowcasts [see details in section 2b(3)]. Because of a coarser resolution of the nested domain, the ICs are prepared by upscaling (a simple volume-weighted spatial average).

BCs are prescribed with upscaled values of RTOFS-Atlantic forecasts over a 144-h period. The values required for each barotropic and baroclinic time step are obtained by linear interpolation of hourly surface and 6-hourly volume data, respectively. For the barotropic mode, the tangential velocity component is prescribed, while the normal component and pressure field are obtained from the well-posed BC model of Browning and Kreiss (1986). The temperature and salinity from the RTOFS potential temperature, potential density, and layer thickness are relaxed within a neighborhood of the open boundary (40 grid points), and the baroclinic fluxes are extrapolated from the interior at the open boundary.

3) Preparation of RTOFS-Atlantic nowcasts

Nowcasts are estimated daily, employing the assimilation of SST and sea surface height (SSH). The SST assimilation is performed by nudging the model SST to an estimate obtained by linear interpolation of two SST analyses, each of which is produced using 2D variational data assimilation (2DVAR). The initial guess for the analysis is the model SST, and the observations include remotely sensed [Advanced Very High Resolution Radiometer (AVHRR), Geostationary Operational Environmental Satellite (GOES)] and in situ SST observations transmitted to the Global Telecommunication System. The SSH assimilation is performed in two stages: The first stage is to produce an analysis of the absolute SSH valid at the initial time of the nowcast using 2DVAR from SSH anomalies obtained from altimeters [Jason-1, Jason-2, Environmental Satellite (Envisat)] and an estimate of the mean dynamic topography (Maximenko et al. 2009). The second stage includes the preparation of the background with the 24-h mean model SSH, followed by adding the misfit of the analysis and the background to the model SSH. The latter is done by rearranging the thickness of model layers, employing a procedure similar to Cooper and Haines (1996).

3. Observations

This section describes AXBT observations and data quality evaluation, followed by the synoptic history of Hurricane Gustav (07L) in 2008, and the NHC-observed best-track data of 10 Atlantic tropical cyclones in 2008 and 2009.

a. AXBT

As part of the multiyear Intensity Forecast Experiment project (Rogers et al. 2006; Marks 2007), the upper water temperatures in the Gulf of Mexico and Caribbean Sea for the 2008 season were monitored by AXBT from seven field campaigns (Fig. 3). A total of 136 profiles were available for a period from 28 August to 8 September 2008, including 50 samples from the prestorm survey (1 in Fig. 3), and 47 from the poststorm survey (7 in Fig. 3). In-storm periods consisted of five surveys (2–6 in Fig. 3), covering different stages of the storm—a category 1 hurricane for survey 2, intensification from a category 2 to category 3 hurricane for survey 3, de-intensification from a category 4 hurricane to a category 3 hurricane for survey 4, a category 3 hurricane for survey 5, and de-intensification from a category 3 hurricane to a category 2 hurricane for survey 6.

Fig. 3.
Fig. 3.

AXBT surveys (in color) and sample locations (o marks) superimposed on the NHC best track of Hurricane Gustav (07L).

Citation: Journal of Atmospheric and Oceanic Technology 31, 2; 10.1175/JTECH-D-13-00013.1

Historical AXBT data have been exploited in the past, for example, Cione and Uhlhorn (2003). In this section, nonetheless, quantitative evaluation was conducted exclusively on the 2008 data to ensure the accuracy, using independently available remote sensing observations from AVHRR.

AXBT data processing included filtering, despiking, bias correction, and interpolation, followed by extrapolation to the sea surface from the first true depth (2 m). The typical measurement error was on the order of 0.2°C (E. W. Uhlhorn 2008, personal communication). The final processed profiles were available at 1-m intervals to a nominal depth of approximately 400 m. For quality assurance of the measurements and later use for validation, we chose temperature at 2-m depth, SST2m.

The remote sensing data were 1-km-resolution retrievals from National Oceanic and Atmospheric Administration Satellite 18 (NOAA-18), with corrections to skin temperatures by adding 0.17°C (SSTAVHRR; Donlon et al. 2002; Paltoglou et al. 2010). A matchup pair was obtained by averaging pixels existing in the 40-km radius (the internal Rossby radius; Feliks 1985; Chelton et al. 1998) of circle centered at a given AXBT site, after choosing the satellite pass coincident with a survey of interest.

A total of 26 collocated pairs identified for the prestorm survey (1600–2100 UTC 28 August 2008) (Fig. 4) showed a mean and root-mean-square (RMS) difference of −0.02° and 0.30°C, respectively. This was the same magnitude of the SSTAVHRR accuracy reported in Xu and Ignatov (2010). Both measurements similarly varied between 29.2° and 30.5°C for SST2m, and from 29.2° to 30.3°C for SSTAVHRR, and had the same mean value (29.8°C) (Fig. 4b). The largest difference (0.67°C), however, existed at 25.49°N, 88.49°W, which might be due to the combined effect between the relatively large distance (23.4 km) to the matchup SSTAVHRR and large variability near the Loop Current (LC) front. Both datasets exhibited warm water (≥29.4°C) in the central region of the GOM and relatively cool water (≤29.5°C) toward the northern coastal area (Fig. 4b).

Fig. 4.
Fig. 4.

Prestorm survey, 28 Aug 2008: (a) navigation-corrected NOAA-18 AVHRR SST (color) superimposed with the survey track with probe locations (x and o) that are numbered at random and (b) comparison between collocated AXBT and AVHRR SST [circles in (a)] as a function of probe number.

Citation: Journal of Atmospheric and Oceanic Technology 31, 2; 10.1175/JTECH-D-13-00013.1

The scarce coverage of NOAA-18 for a poststorm period (1600–2100 UTC 8 September 2008) resulted in 19 matchups situated in the eastern and northern GOM (Fig. 5). They showed 0.14°C for the mean difference and 0.27°C for the RMS difference, with large differences in the northeast region of the LC (between 0.35° and 0.65°C). Both SST2m and SSTAVHRR had a similar range of variation: 28.3°–29.6° and 28.7°–29.7°C, respectively (Fig. 5b). Compared to the prestorm temperature range (Fig. 4), the remnant of a cold wake after Gustav was apparent in the eastern GOM.

Fig. 5.
Fig. 5.

As in Fig. 4, but for a poststorm survey (8 Sep 2008).

Citation: Journal of Atmospheric and Oceanic Technology 31, 2; 10.1175/JTECH-D-13-00013.1

b. Synopsis of Hurricane Gustav 2008

Hurricane Gustav (07L) has 11 days of life span from 25 August to 4 September 2008 (Fig. 6a). It originally formed on the morning of 25 August, about 410 km north of Caracas, Venezuela. From its genesis to the final landfall, the storm experienced intensification 3 times and landfall 5 times (Beven and Kimberlain 2009). The general route of the storm was northwestward from the Caribbean Sea to GOM, except for one turn to the southwest after the first landfall in Haiti (Fig. 6a). The first intensification to a category 1 hurricane (80 kt, where 1 kt = 0.51 m s−1; 981 mbar) took place 24 h after the genesis. The strength degraded to a tropical storm (TS) after the first landfall, and it kept the status for ~60 h until rapid re-intensification from a category 1 to a category 4 hurricane (130 kt and 941 mbar) within 24 h south of Cuba. After the second and third landfalls on Cuba, the storm weakens from the category 4 strength to category 2 strength (95 kt and 955 mbar). Despite the storm passing over the warm LC, it continued weakening, influenced by mid- and upper-level dry air intrusion. When the storm made the final landfall on Louisiana later on 1 September, the category 2 intensity was further degraded to 90 kt and 954 mbar. Its unusually large size—TS force wind extended ~200 nm and hurricane force winds extended 70 nm—brought substantial damage and casualties over Louisiana. The demise of the storm took place 3 days after the landfall.

Fig. 6.
Fig. 6.

The (a) 2008 and (b) 2009 Atlantic TCs: the NHC best-track positions are shown at 6-h intervals with intensity at the Saffir–Simpson wind scale (solid circle).

Citation: Journal of Atmospheric and Oceanic Technology 31, 2; 10.1175/JTECH-D-13-00013.1

c. NHC best-track data

Verification of the HWRF–HYCOM tropical cyclone (TC) forecasts is conducted for 264 cases using the NHC verification software (Franklin 2008). The cases chosen for the study include four TCs in 2008 (155 cases; Fig. 6a) and six TCs in 2009 (109 cases; Fig. 6b) covering a variety of intensities and tracks (Table 2). The NHC verification is conducted against the NHC best-track data of nominal 6-hourly maximum wind, central pressure, track positions, and wind radii, stratified by tropical storm, category 1 and 2 hurricanes, and major hurricanes. A recent report (Landsea and Franklin 2013) shows that uncertainties associated with the NHC best-track data depends on the availability of satellite remote sensing and airborne data, and also the NHC hurricane specialists. For instance, the track uncertainty with respect to the absolute value ranges from 10% for category 1 and 2 hurricanes, and major hurricanes to 12.5% for tropical storms; relative uncertainties for the absolute intensity (maximum wind) value are 25%, ~15%, and ~10% for tropical storms, category 1 and 2 hurricanes, and major hurricanes, respectively.

Table 2.

TC, period, maximum sustained wind speed (kt), minimum central pressure (mbar), and the Saffir–Simpson wind scale.

Table 2.

4. Results

This section consists of validation of ocean simulations and verification of hurricane forecasts of HWRF–HYCOM. The ocean validation is conducted for prestorm (~1600–2130 UTC 28 August) and in-storm periods (29 August–1 September) separately. The model and observations are compared by constructing averages of model values within a 32-km radius around observations. For in-storm comparison, concurrent model data are prepared with 24-h forecast from the cycle simulation that was initialized 24 h prior to the individual surveys. The hurricane forecast verification is performed on a set of model intensities and tracks produced by the NCEP tracker (Marchok 2002), using the NHC verification toolbox and the NHC best-track data as reference. Comparison between HWRF–HYCOM (HY09) and the 2009 operational HWRF (H209) is also included in the verification.

a. Validation of HYCOM ocean simulations

1) Prestorm survey

Figure 7 shows comparisons of SST (Fig. 7a), MLD (Fig. 7b), and Z26 (Fig. 7b) between 45 identified matchups. LC is evident in HY09 SST (Fig. 7a) by the typical convex shape of pattern for probes 1–7, 8–14, and 15–20, whereas it is less clear in AXBT, showing large variation between probes. HY09 SST linearly decreases over probes 25–34, but in situ SST varies significantly at a scale of ~50–150 km. This accounts for a difference of ~0.6°C. For probes 35–44, a trend of warm SST followed by cold temperature is similar for both model and data, except for probes 38–44, where HY09 SST is relatively warmer than AXBT. A total of 45 model–data pairs show <0.1° and ~0.3°C for the mean and RMS difference, respectively. The discrepancy existing in the LC region (probes 1–20) is a major contribution (0.1°C) to the differences. Twenty-nine concurrent AVHRR observations (Fig. 7a) exhibit better agreement with a model by the mean difference of −0.1°C than AXBT observations by −0.2°C. In particular, the trend and variability of AVHRR in the northern survey area agree better with HY09 than AXBT.

Fig. 7.
Fig. 7.

Comparisons of oceanic parameters for prestorm survey as a function of probe number: (a) SST from three sources and (b) MLD (blue) and Z26 (green) from simulations (thin) and observations (thick).

Citation: Journal of Atmospheric and Oceanic Technology 31, 2; 10.1175/JTECH-D-13-00013.1

The prestorm HY09 MLD (Fig. 7b) delineates the deep MLD LC (probes 2–6) and MLD in the Gulf Common Water (probes 7–14). However, HY09 MLD for these locations is 11.4 m deeper than AXBT MLD (Fig. 7b). HY09 presents a dome-shaped pattern over probes 15–20, opposing AXBT. From probe 25–45, both HY09 and AXBT MLD show the same deepening trend. However, HY09 MLD varies at a similar degree as AXBT but shows the opposite of AXBT at four locations—probe 26 (Vernon Basin), probes 29–31 (Lund), probes 36–37 (Walker Ridge), and probes 41–44 (Mississippi Canyon). Overall, the prestorm HY09 MLD is 4.6 m deeper than the observed and having an RMS difference of 15.5 m, which is mostly explained by the HY09 MLD variation.

Models Z26 (HY09-Z26 in Fig. 7b) and AXBT Z26 (AXBT-Z26 in Fig. 7b) exhibit a thick layer of the LC warm water for probes 2–6 and 15–20. But, differences exist. For example, HY09 simulates deep Z26 by <60 m for probes 1–7, and by <11 m for probes 15–20, compared to the observed. Simulations for probes 25–31 and 41–44 indicate an anomalous thick warm layer, forming a convex-shaped pattern. On the other hand, observations show a dome-shaped pattern, and they illustrate thin layers at the both locations. HY09 Z26 is shallower east (probes 22–24) and west of LC (probes 35–40) than AXBT 26. HY09 and AXBT present a similar pattern over probes 8–11, but the latter exhibits deep Z26 instead. In general, HY09 Z26 is deeper by an average of 27.4 m than AXBT, and it has an RMS difference of 30.9 m.

2) In-storm survey

Figure 8 shows comparisons of SST, MLD, and Z26 between 50 model–data pairs. Simulated SST in the survey 2 area varies from 27.6° to 29.8°C (Fig. 8a). They are in good agreement with observed SST (Fig. 8b) within a mean difference less than 0.2°C (Fig. 8c). But, HY09 simulates a colder cold wake at the rear-right quadrant (27.6°C) by 1.4°C than AXBT (29.0°C). Since HY09 MLDs (Fig. 8d) are comparable to the observation (Fig. 8e), a cause of the overcooling probably is due to overpredicted intensity, 73 kt, compared to the best-track data (65 ± 5 kt). Model SSTs for survey 3 match AXBT SSTs quite well, showing the spatial mean of 29.4° and 29.2°C for simulations and measurements, respectively. It is, however, noticed that HY09 SSTs at the storm centers are consistently warmer by 0.37°–0.70°C (Fig. 8c) than the counterparts. This discrepancy may be due to uncertainty existing at comparison between instantaneous (HY09) versus synoptic (AXBT) scales. SST simulations for survey 4 are very consistent with the observations. For the total of 14 pairs covering the LC, HY09 SST ranges between 28.5° and 29.7°C (Fig. 8a), similar to in situ data (28.1°–29.6°C; Fig. 8b). The highest model SST (29.7°C; Fig. 8c) exists at the storm center and ~270 km in front of the storm center, whereas the observed only matches the latter. Both model and AXBT show the same SST cooling, O(28.2°C), at the rear-right quadrant. During survey 5, simulated SST along the storm track is approximately the same as the observed SST, but HY09 SST on the cross track is predominantly warmer (0.53°C; Fig. 8c) than AXBT. The warm bias is particularly large (1°C; Fig. 8c) for an AXBT sampled at the LC frontal eddy. HY09 exhibits skillful eddy simulation, but the difference is largely due to the offset predicted location (≥32 km eastward). For survey 6, HY09 SSTs substantially agree with observations, particularly in the left quadrant (28.4°–29.5°C; Fig. 8a). However, at three locations along the track and one location east off the storm center model estimates are 1.4°C colder and 1.6°C warmer than observed (Fig. 8c), respectively.

Fig. 8.
Fig. 8.

Comparisons of (top) in-storm SST, (middle) MLD, and (bottom) Z26 between (left) HY09 simulations, (middle) AXBT observations, and (right) differences (of simulation from observation). Superimposed in each plot are the survey track (thin black lines) and the NHC best track (thick green line with + marks). The red–green–blue palette applies to both simulations and observations.

Citation: Journal of Atmospheric and Oceanic Technology 31, 2; 10.1175/JTECH-D-13-00013.1

For survey 2, HY09 MLD (Fig. 8d) is deeper than the observed (Fig. 8e) by 11.6 m. The simulation has a windward trend that gradually shoals from 80.5 to 51.9 m, as opposed to the measurement that varies from 36.0 to 60.0 m; this results in the largest difference (44.5 m; Fig. 8f), ~85 km, ahead of the storm center, which is more than twice the mean bias (20.7 m) of the simulation. HY09 Z26 (Fig. 8g) shows a similar trend and deep bias (~60 m) as seen with MLD. AXBT Z26 exhibits a deepening trend to the wind direction, similarly seen in AXBT MLD. Despite the initial overestimate of MLD and Z26, the deepening upper layer is properly simulated as the response to a category 2 hurricane wind.

HY09 for survey 3 overpredicts MLD (Fig. 8d) by 9.6 m on average (Fig. 8f). AXBT observes deep MLD at the storm center rather than at its surrounding, forming a bowl-shaped pattern. It is, however, noticed that during the three samples at the storm center over a ~2 h period, the observed center MLD progressively ascends from 84.0 to 56.0 m. On the other hand, HY09 simulates a pattern that has shallow MLD in the rear-right quadrant (52.6–70.6 m) and deep in the front-left side (67.9–80.4 m). HY09 Z26 (Fig. 8g) varies between 156.6 and 207.0 m. These values are overpredicted by 44.9 m on average (Fig. 8i) than AXBT Z26 (100.0–190.0 m; Fig. 8h), while HY09 and AXBT show the same trend of Z26 in the rear-right being shallower than one in the front-left section.

For survey 4, HY09 (Fig. 8d) simulates MLD similarly to AXBT (Fig. 8e), having shallow (≥27.3 m) ahead of Gustav and deep (≤89.4 m) beneath the category 3 hurricane wind, O(108 kt). The largest MLD difference (31.4 m) exists at the storm center (Fig. 8f), where HY09 MLD is 89.4 m and AXBT MLD is 58.0 m. HY09 Z26 (Fig. 8g) is deep in the LC (209.4 and 233.0 m), so is AXBT Z26 (152.0 and 188.0 m; Fig. 8h). However, the simulated LC Z26 is deeper by an average of 14.1 m (Fig. 8i), and the areal variation is 27.9 m larger than the observed.

HY09 MLD over the survey 5 area (Fig. 8d) simulates little fluctuation (±6 m) from the mean (39.9 m), in contrast to AXBT MLD (Fig. 8e). The observations vary between 24.0 and 152.0 m, and trend leeward along the track and eastward along the cross track (Fig. 8e). HY09 MLD in the rear-right quadrant agrees relatively well with AXBT, but overprediction exists in the front-left quadrant (≤33.0 m; Fig. 8f). The LC Z26 for the survey area is simulated in a range of 145.0–177.6 m, relatively close to the observed (148.9–152.0 m; Fig. 8h). But, HY09 Z26 for the rest of area overpredicts, except for a sample ~56.7 km west of the frontal eddy, 55.0 (Fig. 8g) versus 86.0 m (Fig. 8h).

Both models MLD and AXBT MLD for survey 6 show a deeper layer at the left quadrant O(53 m) than the rest. The spatial mean of HY09 MLD is deeper by ~9.0 m than the observed, which is primarily contributed by the difference of 42 m in that quadrant (Fig. 8f). The HY09 Z26 variation (Fig. 8g) is very similar to the HY09 MLD (Fig. 8d), and AXBT Z26 (Fig. 8h) also shows the same similarity as AXBT MLD. HY09 Z26 varies between 22.2 and 124.7 m (Fig. 8g), and AXBT Z26 ranges from 26.0 to 58.0 m (Fig. 8h). The largest difference (78.7 m) exists near a cold wake.

3) Mixed layer response

Figure 9 shows the temporal evolution of footprint HY09 MLD at six locations along the Hurricane Ike (09L) forecast track. The MLD at each location behaves similarly—rising and falling at a period of ~30 h. The undulation dissipates with the increase in time. It is noticed that the initial shoaling is already in progress at least 6 h before the storm. This is probably the response to the approaching storm. MLD reaches the shallowest depth at 12–18 h after the storm. This probably explains the coldest SST of cold wakes, and the periodic MLD undulations partially illustrate the wake waves of the near-inertial period (Brooks 1983; Shay et al. 1998; Cuyers et al. 2013).

Fig. 9.
Fig. 9.

Time evolution of footprint MLD (m) at six locations (inset) on the HY09 forecast track for Ike (09L) (IC, 0600 UTC 10 Sep 2008). Black open circle denotes the time of the storm pass at a given location.

Citation: Journal of Atmospheric and Oceanic Technology 31, 2; 10.1175/JTECH-D-13-00013.1

In summary, the study of storm-driven SST change suggests that HY09 reasonably simulates SST of ~0.1° and 0.5°C of the mean and RMS difference, respectively, except for a few disagreements that mostly occur at the storm center. Predicted poststorm MLD undulating is similar to observations in the past. The point-to-point comparison to observations exhibits overforecasts in MLD and Z26 before and during the storm passage. However, the in-storm MLD and Z26 response deepens and shoals, respectively. Compared to the initial MLD, in-storm MLD deepens by ~30%, and the agreement to observed MLD is degraded by 90% compared to the prestorm condition. On the other hand, the initial Z26 overestimate in GOM still exists for in-storm cases, but the in-storm Z26 simulation is shallower overall and improved by reducing the difference from observation by 38%.

b. Verification of HWRF–HYCOM hurricane predictions

1) Track prediction

Figure 10 shows the results of the homogeneous verification of hurricane tracks for all 223 cases—the average track forecast errors (lines) and the standard deviation (vertical bars) for HY09 (new model; black) and H209 (operational HWRF; gray). Both HY09 and H209 show a similar trend, such that the error gradually increases from ~9 to ~260 nm, with the increase in forecast hour. There is no difference between HY09 and H209, except the forecast hours are longer than 96 h, showing an overall improvement of 3.7% for HY09. At days 4 and 5, HY09 is more skillful than H209 by ≤8.0%.

Fig. 10.
Fig. 10.

Comparison of track forecasts between HY09 (black) and H209 (gray) for 223 homogeneous cases. Absolute average error (lines) and standard deviation (vertical bars) are a function of forecast time and number of cases (parenthesis). Units are in nautical miles (n mi).

Citation: Journal of Atmospheric and Oceanic Technology 31, 2; 10.1175/JTECH-D-13-00013.1

The error standard deviation of the HY09 track forecast increases from 7.5 to 48.8 nm for the first 36 h. The rate of increase levels off to 51.7 nm at 48 h, followed by a steady increase in the standard deviation from 67.7 nm at 72 h to 96.4 nm at 96 h. The largest deviation (189.9 nm) is at 120 h. Compared to HY09, H209 shows a similar trend, except for the relatively larger bias of 8.3 nm at 96 h and of 2.9 nm at 120 h.

Figure 11 shows the track bias at the given forecast hours (x axis). The vector length represents the bias magnitude, and the orientation denotes the directional bias to the north (y axis) and the east (x axis) from the observed storm position (origin). The HY09 track is predominantly northwest biased, but the trend abruptly changes from west to north for lead hours between 72 and 96. A similar direction change exists with H209, but the bias magnitude is relatively smaller at the 12–72-h period than for HY09. The trend change in the later forecast hours is more substantial for H209 than for HY09, and the absolute bias of H209 at 96 h is about 1.3 times larger than for HY09.

Fig. 11.
Fig. 11.

Average track bias (nm) relative to the best-track position (0,0) for models (rows) and forecast hour (columns). The x axis and y axis of the individual boxes represent bias (nm) to the east and northdirections.

Citation: Journal of Atmospheric and Oceanic Technology 31, 2; 10.1175/JTECH-D-13-00013.1

2) Intensity prediction

The intensity forecast is evaluated by comparing the absolute average error, standard deviation, and bias for 248 homogeneous samples. The average intensity error of HY09 (black in Fig. 12a) gradually increases from 2.5 to 19.9 kt for the 120-h forecast period, having a maximum increase between 0 and 12 h and little change after 48 h. Compared with H209 (gray), the HY09 error is consistently smaller (by up to 4.1 kt), showing a maximum of 20.6% improvement at 120 h and an overall improvement of 13.8% over H209.

Fig. 12.
Fig. 12.

Comparison of intensity forecast between HY09 (black) and H209 (gray) for 248 homogeneous cases showing (a) absolute mean error and standard deviation, and (b) bias. Number of samples (parenthesis) and forecast hour are shown on the x axis. Units are in knots (kt).

Citation: Journal of Atmospheric and Oceanic Technology 31, 2; 10.1175/JTECH-D-13-00013.1

The standard deviation of HY09 (black) varies from 2.8 kt at 0 h to 17.1 kt at 120 h (Fig. 12a). Compared to H209 (gray), HY09 consistently shows better performance for most of the forecast hours, particularly at 96 h. But, HY09 is worse at 12 and 72 h, by 0.4 kt, respectively, than H209.

Both HY09 and H209 show positive intensity bias (Fig. 12b). HY09 varies from −0.6 kt at 0 h to 1.0 kt at 24 h, and gradually increases before reaching a peak of 10.9 kt at 120 h. In contrast, H209 diverges from HY09 from 0 h, and the diverging rate increases with time, having a maximum difference of 6.8 kt at 120 h. Hence, HY09 reduces the H209 positive bias by an average of 43.9%.

5. Discussion and conclusions

This study demonstrates the skill of the HYCOM ocean model in coupled atmosphere–ocean simulations in hurricane environments. The HWRF–HYCOM system is well integrated into the NCEP operational eddy-resolving RTOFS ocean models, ensuring a seamless acquisition of all the improvements to these operational models and presenting a challenge to the operational models to have the best representations of the ocean under hurricane conditions.

The validation study illustrates a marked agreement between simulated and observed SST, showing the mean and RMS difference of 0.1° and 0.5°C, respectively. It is particularly found that simulated SST changes are comparable to those observed, and also the variability of mesoscale features such as the LC and eddies are well represented. Deepening MLD and shoaling Z26 in response to the storm forcing are similar to observations; yet improvements are necessary to compensate overpredicted MLD and Z26 in some areas. This can be achieved by mitigating the initial conditions, for example, through better resolution of the upper layer.

Mixing and its parameterization can be also accountable for the MLD and Z26 overprediction. They are a well-known issue in both oceanic and atmospheric modeling. Regardless, relatively few studies have been done especially in the hurricane environment with dynamic two-way interaction. While HYCOM employs the same GISS mixing for the coupled and noncoupled configurations, the model is forced by regional, high-resolution (27 and 9 km) HWRF and global, coarse-resolution (½°) GFS, respectively. Different degrees in episodic cooling during storm passage as well as domainwide variations of SST have been observed in between the coupled and noncoupled simulations. This suggests that relevant representation of oceanic mixing and the associated parameterization should be sought.

The simulations of storm-driven oceanic variations are implicitly realized in the forecast skill. Coupling HWRF to HYCOM without altering the configuration of operational HWRF–POM has improved the intensity forecast skill by as much 20.6% and on average by 13.8%, over the 2009 operational HWRF. In the configuration, HWRF–HYCOM reduces the positive intensity bias of the operational model by 43.9%. The typical ratio of spatial and temporal scales of oceanic response to atmospheric fluctuation is on the order of (e.g., Williams et al. 2007), implying the maximum return is 10%. Considering this, it can be stated that the improvement is above the noise level, which can be attributed to the eddy-resolving ocean model along with realistic SST initialization.

The estimates of surface flux exchange are highly uncertain, in particular at low (<5 m) and high wind (>30 m) regimes. The approach employed in the HWRF system is the use of empirical formulas of the bulk transfer coefficients, which depend among other things upon the wind speed. Given the same air–sea parameterization, having HYCOM in the 2009 configuration might exchange the turbulent fluxes at a different rate from the one for POM. This is due to not only the physical processes simulated based on different oceanic turbulent mixing, the ocean model grid spacing, the upper vertical layer discretization, and the initial oceanic thermal state but also the air–sea flux formulations, which also depend upon the atmospheric model grid resolution. To have an equitable comparison, this study has used the same configuration as the operational HWRF–POM, but it is anticipated that improvements in the representation of turbulent fluxes under hurricane conditions are likely to yield further improvements in the hurricane forecasts.

This work proves the concept of HWRF–HYCOM and results in skillful hurricane forecast. One well-defined conclusion from the study, however, is to improve the estimates of IC in real time, which leads to our one current effort that explores the use of RTOFS-Global products combined with data assimilation in regional domains.

Acknowledgments

The authors thank the members of three teams in EMC involved in the hurricane forecast system development in different aspects: L. Liu from the Marine Modeling and Analysis Branch; Y. Kwon, S. Trahan, Z. Zhang, and J. O’Connor of the Hurricane Weather Research and Forecasting model group; and Q. Liu of the Global Climate and Weather Branch. We also thank E. Uhlhorn of the Hurricane Research Division for the useful information on AXBT quality, data collection, and processing; and L. Shay and M. Patrick of RSMAS/UM for the additional data collection and for providing quality-controlled AXBT data. Finally, we would like to express special gratitude to three reviewers for their involvement in improving the manuscript.

REFERENCES

  • Bender, M. A., , and Ginis I. , 2000: Real-case simulations of hurricane–ocean interaction using a high-resolution coupled model: Effects on hurricane intensity. Mon. Wea. Rev., 128, 917946.

    • Search Google Scholar
    • Export Citation
  • Bender, M. A., , Ginis I. , , Tuleya R. , , Thomas B. , , and Marchok T. , 2007: The operational GFDL coupled hurricane–ocean prediction system and a summary of its performance. Mon. Wea. Rev., 135, 39653989.

    • Search Google Scholar
    • Export Citation
  • Beven, J. L., , and Kimberlain T. B. , 2009: Tropical cyclone report: Hurricane Gustav (AL072008) 25 August–4 September 2009. National Hurricane Center, 38 pp. [Available online at http://www.nhc.noaa.gov/pdf/TCR-AL072008_Gustav.pdf.]

  • Bleck, R., 2002: An oceanic general circulation model framed in hybrid isopycnic-Cartesian coordinates. Ocean Modell., 4, 5588.

  • Bleck, R., , and Boudra D. , 1981: Initial testing of a numerical ocean circulation model using a hybrid (quasi-isopycnic) vertical coordinate. J. Phys. Oceanogr., 11, 755770.

    • Search Google Scholar
    • Export Citation
  • Blumberg, A. F., , and Mellor G. L. , 1987: A description of a three-dimensional coastal ocean circulation model. Three-Dimensional Coastal Ocean Models, Geophys. Monogr., Vol. 4, Amer. Geophys. Union, 1–16.

  • Booth, J. F., , Thompson L. A. , , Patoux J. , , and Kelly K. A. , 2012: Sensitivity of midlatitude storm intensification to perturbations in the sea surface temperature near the Gulf Stream. Mon. Wea. Rev., 140, 12411256.

    • Search Google Scholar
    • Export Citation
  • Brooks, A. D., 1983: The wake of Hurricane Allen in the western Gulf of Mexico. J. Phys. Oceanogr., 13, 117129.

  • Browning, G. L., , and Kreiss H.-O. , 1986: Scaling and computation of smooth atmospheric motions. Tellus, 38A, 295313.

  • Canuto, V. M., , Howard A. , , Cheng Y. , , and Dubovikov M. S. , 2001: Ocean turbulence. Part I: One-point closure model—Momentum and heat vertical diffusivities. J. Phys. Oceanogr., 31, 14131426.

    • Search Google Scholar
    • Export Citation
  • Canuto, V. M., , Howard A. , , Cheng Y. , , and Dubovikov M. S. , 2002: Ocean turbulence. Part II: Vertical diffusivities of momentum, heat, salt, mass, and passive scalars. J. Phys. Oceanogr., 32, 240264.

    • Search Google Scholar
    • Export Citation
  • Chelton, D. B., , DeSzoeke R. A. , , and Schlax M. G. , 1998: Geographical variability of the first baroclinic Rossby radius of deformation. J. Phys. Oceanogr., 28, 443460.

    • Search Google Scholar
    • Export Citation
  • Chen, S. S., , Price J. F. , , Zhao W. , , Doneland M. A. , , and Walsh E. J. , 2007: The CBLAST-Hurricane program and the next-generation fully coupled atmosphere–wave–ocean models for hurricane research and prediction. Bull. Amer. Meteor. Soc., 88, 311317.

    • Search Google Scholar
    • Export Citation
  • Cione, J. J., , and Uhlhorn E. W. , 2003: Sea surface temperature variability in hurricanes: Implications with respect to intensity change. Mon. Wea. Rev., 131, 17831796.

    • Search Google Scholar
    • Export Citation
  • Cooper, M., , and Haines K. , 1996: Altimetric assimilation with water property conservation. J. Geophys. Res., 101 (C1), 10591077.

  • Cuyers, Y., , Le Vailant X. , , Bouruet-Aubertot P. , , Vialard J. , , and McPhaden M. J. , 2013: Tropical storm-induced near-intertial waves during the Cirent experiment: Energy fluxes and impact on vertical mixing. J. Geophys. Res., 118, 358380, doi:10.1029/2012JC007881.

    • Search Google Scholar
    • Export Citation
  • Donlon, C. J., , Minnett P. J. , , Gentemann C. , , Nightingale T. J. , , Barton I. J. , , Ward B. , , and Murray M. J. , 2002: Toward improved validation of satellite sea surface skin temperature measurements for climate research. J. Climate, 15, 353369.

    • Search Google Scholar
    • Export Citation
  • Environmental Modeling Center, 2003: The GFS Atmospheric Model. NCEP Office Note 442, 14 pp. [Available online at http://www.emc.ncep.noaa.gov/officenotes/newernotes/on442.pdf.]

  • Falkovich, A. I., , Ginis I. , , and Lord S. , 2005: Ocean data assimilation and initialization procedure for the coupled GFDL/URI Hurricane Prediction System. J. Atmos. Oceanic Technol., 22, 19181932.

    • Search Google Scholar
    • Export Citation
  • Feliks, Y., 1985: On the Rossby radius of deformation in the ocean. J. Phys. Oceanogr., 15, 16051607.

  • Ferrier, B. S., 1994: A double-moment multiple-phase four-class bulk ice scheme. Part I: Description. J. Atmos. Sci., 51, 249280.

  • Ferrier, B. S., , Tao W. K. , , and Simpson J. , 1995: A double-moment multiple-phase four-class bulk ice scheme. Part II: Simulations of convective storms in different large-scale environments and comparisons with other bulk parameterizations. J. Atmos. Sci., 52, 10011033.

    • Search Google Scholar
    • Export Citation
  • Franklin, J. L., 2008: National Hurricane Center forecast verification. Preprints, 28th Conf. on Hurricanes and Tropical Meteorology, Orlando, FL, Amer. Meteor. Soc., 13A.1. [Available online at http://www.nhc.noaa.gov/verification.]

  • Franklin, J. L., 2009: 2008 National Hurricane Center forecast verification report. National Hurricane Center, 71 pp. [Available online at http://www.nhc.noaa.gov/verification/pdfs/Verification_2008.pdf.]

  • Gangopadhyay, A., , Robison A. R. , , and Arango H. G. , 1997: Circulation and dynamics of the western North Atlantic. Part I: Multiscale feature models. J. Atmos. Oceanic Technol., 14, 13141332.

    • Search Google Scholar
    • Export Citation
  • Gopalakrishnan, S., , Liu Q. , , Marchok T. , , Sheinin D. , , Surgi N. , , Tuleya R. , , Yablonsky R. , , and Zhang Z. , 2010: Hurricane Weather Research and Forecasting (HWRF) model scientific documentation. L. Bernardet, Ed., NOAA/Earth System Research Laboratory, 75 pp.

  • Jaimes, B., , and Shay L. K. , 2010: Near-inertial wave wake of hurricanes Katrina and Rita over mesoscale oceanic eddies. J. Phys. Oceanogr., 40, 1320–1337.

    • Search Google Scholar
    • Export Citation
  • Janjić, Z. I., 1994: The step-mountain coordinate model: Further developments of the convection, viscous sublayer and turbulence closure schemes. Mon. Wea. Rev., 122, 927945.

    • Search Google Scholar
    • Export Citation
  • Janjić, Z. I., , Gall R. , , and Pyle M. E. , 2010: Scientific documentation for the NMM solver. NCAR Tech. Note NCAR/TN-477+STR, 54 pp.

  • Kim, H.-S., , Gangopadhyay A. , , Rosenfeld L. K. , , and Bub F. , 2007: Developing a high-resolution climatology for the central California coastal region. Cont. Shelf Res., 27, 21352161.

    • Search Google Scholar
    • Export Citation
  • Kurihara, Y., , and Tuleya R. E. , 1974: Structure of a tropical cyclone developed in a three-dimensional numerical simulation model. J. Atmos. Sci., 31, 893919.

    • Search Google Scholar
    • Export Citation
  • Kurihara, Y., , Tuleya R. E. , , and Bender M. A. , 1998: The GFDL Hurricane Prediction System and its performance in the 1995 hurricane season. Mon. Wea. Rev., 126, 13061322.

    • Search Google Scholar
    • Export Citation
  • Landsea, C., , and Franklin J. , 2013: Atlantic hurricane database uncertainty and presentation of a new database format. Mon. Wea. Rev., 141, 3576–3592.

    • Search Google Scholar
    • Export Citation
  • Lee, C.-Y., , and Chen S. S. , 2012: Symmetric and asymmetric structures of hurricane boundary layer in coupled atmosphere–wave–ocean models and observations. J. Atmos. Sci., 69, 35763594.

    • Search Google Scholar
    • Export Citation
  • MacAfee, A. W., , and Pearson G. M. , 2006: Development and testing of tropical cyclone parametric wind models tailored for midlatitude application—Preliminary results. J. Appl. Meteor. Climatol., 45, 12441260.

    • Search Google Scholar
    • Export Citation
  • Marchok, T. P., 2002: How the NCEP tropical cyclone tracker works. Preprints, 25th Conf. on Hurricanes and Tropical Meteorology, San Diego, CA, Amer. Meteor. Soc., P1.13. [Available online at http://ams.confex.com/ams/pdfpapers/37628.pdf.]

  • Marks, F. D., Jr., 2007: Recent results from NOAA’s Hurricane Intensity Forecast Experiment (IFEX). Preprints, 11th Symp. on Integrated Observing and Assimilation Systems for the Atmosphere, Oceans, and Land Surface, San Antonio, TX, Amer. Meteor. Soc., 8.1. [Available online at https://ams.confex.com/ams/87ANNUAL/techprogram/paper_119567.htm.]

  • Maximenko, N., , Niiler P. , , Centuioni L. , , Rio M.-H. , , Melnichenko O. , , Chambers D. , , Zlotnicki V. , , and Galperin B. , 2009: Mean dynamic topography of the ocean derived from satellite and drifting buoy data using three different techniques. J. Atmos. Oceanic Technol., 26, 19101919.

    • Search Google Scholar
    • Export Citation
  • Mehra, A., , and Rivin I. , 2010: A real time ocean forecast system for the North Atlantic Ocean. Terr. Atmos. Oceanic Sci., 21, 211228.

    • Search Google Scholar
    • Export Citation
  • Moon, I.-J., , Ginis I. , , Hara T. , , and Thomas B. , 2007: A physics-based parameterization of air–sea momentum flux at high wind speeds and its impact on hurricane intensity predictions. Mon. Wea. Rev., 135, 28692878.

    • Search Google Scholar
    • Export Citation
  • Paltoglou, G., , Beggs H. , , and Majewski L. , 2010: New Australian high resolution AVHRR SST products from the Integrated Marine Observing System. IMOS, 11 pp. [Available online at http://imos.org.au/fileadmin/user_upload/shared/SRS/Paltoglou_2010_IMOS_AVHRR_v1_0.pdf.]

  • Pan, H.-L., , and Wu W.-S. , 1995: Implementing a mass flux convection parameterization package for the NMC medium-range forecast model. NMC Office Note 409, 40 pp.

  • Price, J. F., , Sanford T. B. , , and Forristall G. Z. , 1994: Forced stage response to a moving hurricane. J. Phys. Oceanogr., 24, 233260.

    • Search Google Scholar
    • Export Citation
  • Rogers, R., and Coauthors, 2006: The intensity forecasting experiment: A NOAA multiyear field program for improving tropical cyclone intensity forecasts. Bull. Amer. Meteor. Soc., 87, 15231537.

    • Search Google Scholar
    • Export Citation
  • Sanford, T. B., , Price J. F. , , and Girton J. B. , 2011: Upper-ocean response to Hurricane Frances (2004) observed by profiling EM-APEX floats. J. Phys. Oceanogr., 41, 10411056.

    • Search Google Scholar
    • Export Citation
  • Schade, L. R., , and Emanuel K. A. , 1999: The ocean’s effect on the intensity of tropical cyclones: Results from a simple coupled atmosphere–ocean model. J. Atmos. Sci., 56, 642651.

    • Search Google Scholar
    • Export Citation
  • Shay, L. K., , Mariano A. J. , , Jacob S. D. , , and Ryan E. H. , 1998: Mean and near-inertial ocean current response to Hurricane Gilbert. J. Phys. Oceanogr., 28, 858889.

    • Search Google Scholar
    • Export Citation
  • Shay, L. K., , Goni G. J. , , and Black P. G. , 2000: Effects of a warm oceanic feature on Hurricane Opal. Mon. Wea. Rev., 128, 13661383.

  • Tallapragada, V., 2010: NCEP operational hurricane WRF (HWRF) modeling system. 11th WRF Users’ Workshop, Boulder, CO, NCAR. [Available online at http://www.mmm.ucar.edu/wrf/users/workshops/WS2010/presentations/session%202/2-2_vijay_wrf_users_workshop_2010.pdf.]

  • Teague, W. J., , Carron M. J. , , and Hogan P. J. , 1990: A comparison between the generalized digital environmental model and Levitus climatologies. J. Geophys. Res., 95 (C5), 71677183.

    • Search Google Scholar
    • Export Citation
  • Tuleya, R. E., 1994: Tropical storm development and decay: Sensitivity to surface boundary conditions. Mon. Wea. Rev., 122, 291304.

  • Wai, M., , and Stage S. A. , 1989: Dynamical analyses of marine atmospheric boundary layer structure near the Gulf Stream oceanic front. Quart. J. Roy. Meteor. Soc., 115, 2944.

    • Search Google Scholar
    • Export Citation
  • Warner, T. T., , Lakhtakia M. N. , , Doyle J. D. , , and Pearson R. A. , 1990: Marine atmospheric boundary layer circulations forced by Gulf Stream sea surface temperature gradients. Mon. Wea. Rev., 118, 309323.

    • Search Google Scholar
    • Export Citation
  • Williams, R. G., , Wilson C. , , and Hughues C. W. , 2007: Ocean and atmosphere storm tracks: The role of eddy vorticity forcing. J. Phys. Oceanogr., 37, 22672289.

    • Search Google Scholar
    • Export Citation
  • Xu, F., , and Ignatov A. , 2010: Evaluation of in situ surface temperatures for use in the calibration and validation of satellite retrievals. J. Geophys. Res., 115, C09022, doi:10.1029/2010JC006129.

    • Search Google Scholar
    • Export Citation
  • Yablonsky, R. M., , and Ginis I. , 2008: Improving the ocean initialization of coupled hurricane–ocean models using feature-based data assimilation. Mon. Wea. Rev., 136, 25922607.

    • Search Google Scholar
    • Export Citation
Save