OceTHE OFFICIALa MAGAZINEnog OF THE OCEANOGRAPHYra SOCIETYphy

CITATION Doyle, J.D., R.M. Hodur, S. Chen, Y. Jin, J.R. Moskaitis, S. Wang, E.A. Hendricks, H. Jin, and T.A. Smith. 2014. Tropical cyclone prediction using COAMPS-TC. Oceanography 27(3):104–115, http://dx.doi.org/10.5670/oceanog.2014.72.

DOI http://dx.doi.org/10.5670/oceanog.2014.72

COPYRIGHT This article has been published inOceanography , Volume 27, Number 3, a quarterly journal of The Oceanography Society. Copyright 2014 by The Oceanography Society. All rights reserved.

USAGE Permission is granted to copy this article for use in teaching and research. Republication, systematic reproduction, or collective redistribution of any portion of this article by photocopy machine, reposting, or other means is permitted only with the approval of The Oceanography Society. Send all correspondence to: [email protected] or The Oceanography Society, PO Box 1931, Rockville, MD 20849-1931, USA.

DOWNLOADED FROM HTTP://WWW.TOS.ORG/OCEANOGRAPHY SPECIAL ISSUE ON NAVY OPERATIONAL MODELS

Tropical Cyclone Prediction Using COAMPS-TC

BY JAMES D. DOYLE, RICHARD M. HODUR, SUE CHEN, YI JIN, JONATHAN R. MOSKAITIS, SHOUPING WANG, ERIC A. HENDRICKS, HAO JIN, AND TRAVIS A. SMITH

Moderate Resolution Imaging Spectroradiometer (MODIS) image of Super Typhoon Jangmi, September 27, 2008. NASA image by Robert Simmon and Jesse Allen, based on MODIS data

ABSTRACT. A new version of the Coupled Ocean/ of COAMPS-TC to capture both intensity and multiscale Atmosphere Mesoscale Prediction System for Tropical Cyclones structure in agreement with observations. Results from the air- (COAMPS®-TC) has been developed for prediction of tropical ocean coupled COAMPS-TC simulations of Typhoon Fanapi cyclone track, structure, and intensity. The COAMPS-TC has (2010) and Super Typhoon Jangmi (2008) in the Western been tested in real time in both uncoupled and coupled modes Pacific indicate accurate predictions of the track and intensity, over the past several tropical cyclone seasons in the Western as well as the sea surface temperature cooling response to the Pacific and Atlantic basins at a horizontal resolution of 5 km. storm, in agreement with satellite measurements. The air- An evaluation of a large sample of forecasts in the Atlantic ocean-wave coupled simulations of the Atlantic Hurricane and Western Pacific basins reveals that the COAMPS-TC Frances (2004) highlight the capability of the COAMPS-TC intensity predictions are competitive with, and in some regards system to realistically capture not only sea surface temperature more accurate than, the other leading dynamical models, cooling following storms but also characteristics of ocean particularly for lead times beyond 36 hours. Recent real-time surface waves and their interactions with boundary layers forecasts of Hurricane Sandy (2012) highlight the capability above and below the ocean surface.

104 Oceanography | Vol. 27, No.3 INTRODUCTION track prediction through the use of characteristics (e.g., structure, morphol- The demand for more accurate forecasts global prediction models (e.g., Goerss, ogy, propagation), at least for continental of tropical cyclone track and intensity 2007; Hamill et al., 2011). A three-day locations (Fowle and Roebber, 2003; with longer lead times is greater than hurricane track forecast today is as Done et al., 2004). ever due to the enormous economic and skillful as a one-day forecast was The Coupled Boundary Layer Air–Sea societal impact. A dramatic example 30 years ago. Evacuating coastal areas Transfer (CBLAST) field program (Black occurred during October 2012 as before a hurricane is estimated to cost et al., 2007), conducted from 2002 to Hurricane Sandy threatened many com- $1 million for every mile of coastline 2004, provided important air–sea inter- munities along the US Eastern Seaboard. evacuated (e.g., Whitehead 2003). These action observations in hurricanes and Basic questions such as where Sandy dramatically improved track forecasts motivated new approaches to parame- would track and how strong it would have reduced the size of evacuation terization of these processes in tropical become had profound implications for areas and mitigated costs. However, cyclone models. Coupled air-ocean the millions of people in its path and intensity prediction remains a significant and air-ocean-wave tropical cyclone billions of dollars of vulnerable assets. challenge, and progress has been consid- modeling systems more realistically With an estimated total damage amount erably slower (DeMaria et al., 2005, 2014; represent these key air-sea interaction of $50 billion USD or more, Hurricane Rogers et al., 2006). The slow improve- processes (e.g., Bao et al., 2000; Bender Sandy is the second costliest hurricane ment in TC intensity and structure et al., 2007; Chen et al., 2007, 2010). The since 1900, and the deadliest hurricane forecasts has been attributed to a variety coupling to an ocean-circulation model to hit the northeastern United States of reasons, ranging from a lack of critical can improve the storm intensity forecast in four decades. observations in the TC inner core and through a more realistic representation The potential impact of tropical the surrounding environment to inaccu- of storm-induced cooling in the upper cyclones on military operations can rate representations of physical processes ocean and sea surface. Inclusion of also be enormous. Typhoon Cobra, also in numerical weather prediction (NWP) ocean waves and their feedback to the known as Halsey’s Typhoon after Admiral models. Marks and Shay (1998) note that atmosphere and ocean boundary layers William Halsey, struck the US Navy’s track prediction depends more on large- of a hurricane can yield more realistic Pacific Fleet in December 1944 during scale processes, while intensity depends momentum fluxes across the air-sea World War II. Three destroyers were lost, on both inner-core dynamics and its interface (e.g., Bao et al., 2000; Doyle, and a total of 790 sailors perished. More relationship to the environment. This 2002) and improve the model-predicted recently, during Sandy, the decision to motivates the requirement for accurate maximum wind speed–central pressure sortie Navy assets from Norfolk, VA, and representation of the key physical and relationship (e.g., Chen et al., 2007), other ports along the Eastern Seaboard dynamical processes within the storm which is an important aspect of the days in advance of the storm was criti- itself and in the larger-scale environ- hurricane intensity and structure cally dependent on forecasts of Sandy’s ment. The need to explicitly resolve the relationship. Inclusion of wave-current track, intensity (maximum sustained inner part of the storm, including the interaction through the use of an wind speed at the surface), and storm eye, the eyewall, and spiral rainbands, observation-based momentum drag structure (i.e., the size of the storm or has motivated modeling of the inner core formulation in the wave model, the wind radius of key wind speed thresholds). In at high horizontal resolution (e.g., Zhu source generation, and the total volumet- the western North Pacific basin, an area et al., 2004, 2006; Braun et al., 2006; ric dissipation is shown by Smith et al. of strategic importance for the US Navy, Chen et al., 2007; Davis et al., 2008). One (2013) to reduce coupled model forecast the Navy Pacific Fleet in the Philippine distinct advantage of applying models errors in significant wave height and Sea has been affected by numerous at high resolution (grid increments wave period for Hurricane Ivan (2004). storms, such as Typhoon Nanmadol of 5 km or less) is that convection Advances in high-resolution TC (2011) that exhibited erratic movement can be explicitly represented in the modeling and data assimilation are and was poorly forecasted. model, which precludes the need for a thought to be necessary in order to There has been remarkable convection parameterization and results significantly improve the intensity improvement in tropical cyclone (TC) in more accurately resolved convective and structure prediction. To this

Oceanography | September 2014 105 end, the Naval Research Laboratory COAMPS-TC DESCRIPTION NAVDAS to represent the hurricane’s (NRL) in Monterey, CA, developed the The COAMPS-TC system is composed characteristics, and then blending Coupled Ocean/Atmosphere Mesoscale of data quality control, analysis, the synthetic observations with other Prediction System for Tropical Cyclones initialization, and forecast model available real-time observations that (COAMPS-TC), a new version of subcomponents (Doyle et al., 2011). The describe the larger-scale environment COAMPS designed specifically for system was transitioned to operations outside the TC circulation. The TC high-resolution tropical cyclone pre- at the Fleet Numerical Oceanography synthetics are necessary due insufficient diction (Doyle et al., 2011). The model and Meteorology Center (FNMOC) real-time in situ observations of tropical builds on the existing COAMPS infra- in 2012, and the first operational cyclones. Recently, a new method of structure and provides the framework to forecasts occurred in June 2013. The generating TC synthetics was introduced add new capabilities and advancements Navy Atmospheric Variational Data into COAMPS-TC that is based on in data assimilation, vortex initialization, Assimilation System (NAVDAS) is TC position, maximum winds, radius of maximum winds, mean radius of the 34 knot winds, and recent storm motion. The TC synthetics are generated AN EVALUATION OF A LARGE SAMPLE OF at one point in the center of each TC, FORECASTS IN THE ATLANTIC AND WESTERN PACIFIC and at eight points around each of nine BASINS REVEALS THAT THE COAMPS-TC INTENSITY concentric circles centered over the TC, for a total of 73 individual points. PREDICTIONS ARE COMPETITIVE WITH, AND IN “ At each of these points, profiles of the SOME REGARDS MORE ACCURATE THAN, THE OTHER u- and v-components are prescribed LEADING DYNAMICAL MODELS, PARTICULARLY FOR at 1000, 925, 850, 700, 600, 500, and LEAD TIMES BEYOND 36 HOURS. 400 hPa, along with geopotential height at 1,000 hPa. The winds’ horizontal structure is generated from a modified Rankine wind vortex that best fits the physical parameterization, and air-sea used to blend observations of winds, observed value of the maximum winds, coupling appropriate for high-resolution temperature, moisture, and pressure the radius of the maximum winds, tropical cyclone prediction. from” a plethora of sources such as radio- and the radius of the 34-knot winds. A This paper provides an overview of sondes, pilot balloons, satellites, surface Rankine vortex is a simplified model the main capabilities of COAMPS-TC measurements, ships, buoys, and aircraft of the radial structure of the tangential through examination of several different (Daley and Barker, 2000). As part of the wind field. The prescribed vertical representative tropical cyclone cases and TC analysis procedure, the pre-existing profile of the winds follows the method statistical analysis of real-time and ret- circulation in the COAMPS-TC first described by Liou and Sashegyi (2012). rospective forecasts that were conducted guess fields is relocated to allow for accu- The mean storm motion is added to the as part of a pre-operational evaluation rate representation of TC position for the Rankine-vortex winds. The 1,000 hPa of the system. We next provide a brief analysis background following Liou and geopotential heights are created by solv- description of the COAMPS-TC system, Sashegyi (2012). ing the equation for the Rankine vortex after which we address aspects of the Synthetic observations or profiles for the geopotential field. The innermost air-sea coupling, including boundary are used to incorporate TC structure ring of synthetics can be set to either layer and surface flux parameterizations and intensity into the initial conditions the reported radius of maximum wind and fully coupled air-ocean and air- based on National Hurricane Center (RMW) or to the location of the RMW ocean-wave options. The final section (NHC) and the Joint Typhoon Warning in the COAMPS-TC first-guess fields. provides an overview of the statistical Center (JTWC; Liou and Sashegyi, 2012) The latter is typically used to minimize performance of the system, and we specifications. This is accomplished by the size of the analysis increments. The conclude with a summary of the results. using these synthetic observations in remainder of the rings is evenly spaced

106 Oceanography | Vol. 27, No.3 with increasing distance from the 40 vertical levels that extend from 10 m BOUNDARY LAYER AND AIR- RMW to the outermost ring at 600 km to approximately 30 km. The inner two SEA INTERACTION SENSITIVITY from the center. grid meshes follow the storm. Momentum exchange at the sea surface Sea surface temperature is analyzed The COAMPS-TC system has is dependent on the sea-state-dependent directly on the model computational the capability to operate in a fully drag coefficient, Cd. Prior to the past grid using the Navy Coastal Ocean coupled air-sea interaction mode decade, the characteristics of Cd had not Data Assimilation (NCODA) system, (Chen et al., 2010, 2011; Doyle, et al., been observed in a tropical cyclone and which makes use of all available satellite, 2011). The atmospheric module within were primarily based on extrapolations ship, float, and buoy observations COAMPS-TC is coupled to the Navy from field campaign measurements con- (Cummings, 2005). In coupled applica- Coastal Ocean Model (NCOM; Martin, ducted in much weaker wind conditions. tions, both the NCODA and NAVDAS 2000; Martin et al., 2006) to represent In a seminal study, Powell et al. (2003) systems are applied using a data assimi- air-ocean interaction processes. The analyzed data from GPS dropwind- lation cycle in which the first guess from COAMPS-TC system has an option to sondes deployed from aircraft into hur- the analysis is derived from the previous predict ocean surface waves and the ricanes; they found that the mean wind short-term forecast. interactions between the atmosphere, speed varied logarithmically with height The COAMPS-TC atmospheric model ocean circulation, and waves using in the lowest 200 m and was a maximum uses the nonhydrostatic and compressible either the Simulating WAves Nearshore at 500 m. They estimated the surface form of the dynamics and has prognostic (SWAN) or WAVEWATCH III models. stress, roughness length, and neutral sta- variables for the three components of the Wave and current interaction is param- bility Cd, and found a markedly reduced –1 wind (two horizontal wind components eterized through inclusion of Stokes Cd at wind speeds above 30 m s . Their and the vertical wind), the perturbation drift currents and Langmuir turbulence analysis showed a leveling off of the pressure, potential temperature, water (Kantha and Clayson, 2004; see Allard surface momentum flux as the winds vapor, cloud droplets, raindrops, ice et al., 2014, in this issue). The wind-wave increase above the hurricane threshold crystals, snowflakes, graupel (soft hail), interaction is represented through the and even a slight decrease of the Cd with and turbulent kinetic energy (Hodur, prediction of a sea-state-dependent further increases in wind speed. Donelan 1997). Physical parameterizations include Charnock parameter (Moon et al., 2004) et al. (2004) extended the Powell et al. representations of cloud microphys- that is a function of wave age and wind (2003) study through a series of wind- ical processes, convection, radiation, speed. A sea spray parameterization wave tank experiments and found that boundary layer processes, and surface (Fairall et al., 1994, 2009; Bao et al., Cd saturation occurs when wind speed layer fluxes. The COAMPS-TC model 2011) can be used to represent the injec- exceeds 33 m s–1. Beyond this wind contains a representation of dissipative tion of droplets into the atmospheric speed threshold, the surface roughness heating near the ocean surface, which boundary layer due to ocean surface no longer increases. Donelan et al. has been found to be important for wave breaking and shearing of the crest (2004) found a Cd saturation level of tropical cyclone intensity forecasts (Jin of breaking waves. The spray droplets 0.0025, similar to the saturation value of et al., 2007). The COAMPS-TC system impact the momentum and enthalpy 0.0026 found by Powell et al. (2003). uses a flexible nesting design, which has fluxes through increased mass loading, Both surface drag and sea spray proven useful when more than one storm air flow stratification, and evaporation processes play major roles in regu- is present in a basin at a given time, as and/or condensation. lating energy exchange at the air-sea well as special options for moving nested grid families that independently follow James D. Doyle ([email protected]) is a meteorologist and head of the individual tropical cyclone centers of Mesoscale Modeling Section in the Marine Meteorology Division, Naval Research interest. In the applications shown in Laboratory (NRL), Monterey, CA, USA. Richard M. Hodur is a scientist at Science this paper (unless otherwise noted), the Applications International Corporation, Monterey, CA, USA. Sue Chen, Yi Jin, atmospheric portion of COAMPS-TC Jonathan R. Moskaitis, Shouping Wang, Eric A. Hendricks, and Hao Jin are all meteorol- uses three nested grids of 45 km, 15 km, ogists in the Marine Meteorology Division, NRL, Monterey, CA, USA. Travis A. Smith is an and 5 km horizontal resolution and oceanographer in the Oceanography Division, NRL, Stennis, MS, USA.

Oceanography | September 2014 107 interface. An advanced version of a illustrates the sensitivity of TC intensity Figure 1a,b), however, attains a minimum sea spray parameterization developed forecasts to the Cd and the sea spray surface pressure that is 29 hPa higher by Fairall et al. (1994, 2009) and a parameterizations using COAMPS-TC and an intensity 20 m s–1 weaker than drag parameterization where Cd is simulations of Hurricane Isabel (2003). the best track analysis. In contrast, storm limited for wind speed above 30 m s–1 All of the simulations capture the initial intensity is increased in the limited

(hereafter referred to as the “limited Cd” rapid intensification apparent in the Cd experiment (Figure 1, magenta experiment) following Powell et al. best track data within the first 40 hours. line), likely a result of reduced surface (2003) and Donelan et al. (2004) are both The simulation with the standard friction. Including sea spray processes evaluated in COAMPS-TC. Figure 1 Charnock parameterization (blue line in substantially increases the surface latent heat fluxes, and the additional energy input at the air-sea interface further a Min. Sea Level Pressure b Max. Wind Speed enhances convection (not shown). The 1,020 80 simulation with both the limited Cd and the sea spray parameterization (red line) 70 1,000 attains a minimum pressure of 931 hPa and a 68 m s–1 maximum wind speed at 60 120 h, a significant improvement over the 980 standard Charnock parameterization. An ) –1 50 increase in the energy input, attributable 960 to the sea spray along with the higher 40 wind speeds due to the limited C at Speed (m s d Pressure (mb) high wind speeds, is clearly apparent in 940 30 Figure 1c,d, which compares the enthalpy

flux from the standard Charnock Cd Best Track 920 with that from the experiment that 20 Charnock Cd Limited Cd used the limited Cd and sea spray Limited C + Spray d parameterization. The magnitude of the 900 10 0 20 40 60 80 100 120 020406080100 120 enthalpy flux is nearly doubled due to the Time (hours) Time (hours) stronger wind speeds and evaporation of sea spray drops. The maximum enthalpy 450 c d –2 2,000 flux reaches 2,000 W m on the right side of the storm track where the wind 400 1,500 speed is stronger due, in part, to the

350 contribution by the translation speed km 1,000 of the storm. It should be noted that

300 500 the differences in the enthalpy fluxes in Figure 1c,d correspond to a time when 0 250 Charnock Cd Limited Cd + Spray the two simulations are quite different –1 250 300 350 400 450 250 300 350 400 450 (there is a ~ 20 m s difference in the km km maximum wind speed at 110 h). Thus, Figure 1. Impact of drag coefficient and sea-spray parameterizations on Hurricane Isabel (2003) the differences in the enthalpy fluxes are simulations. Time series of: (a) minimum sea level pressure, and (b) 10 m maximum wind speed. The 120 h Coupled Ocean/Atmosphere Mesoscale Prediction System for Tropical Cyclones mostly due to integrated differences in (COAMPS®-TC) simulation is initialized at 0000 UTC September 7, 2003. In these sensitivity tests, the development pathways of each simu-

Charnock Cd represents the standard Charnock relationship and “limited Cd” corresponds to a Cd lation. The impact of the limited Cd and (drag coefficient) that is constant at high wind speeds (above 30 m s–1). Enthalpy flux (W –2m ) sea spray is also manifested by the strong valid at 110 h for utilizing: (c) a standard Charnock Cd, and (d) limited Cd and a sea spray parame- terization. The enthalpy flux is displayed from the 5 km grid mesh. asymmetry shown by the enthalpy flux

108 Oceanography | Vol. 27, No.3 pattern in Figure 1d, which further sug- against best-track values. In addition, as on September 28. Jangmi’s intensity gests that these processes impact not only Figure 2b shows, the radii of the 34 kt change is highlighted here to show the the TC’s intensity but also its structure. In wind speed mean absolute error and impact of the evolving ocean during general, the current state-of-the-science mean error decrease for all lead times. the forecast through comparisons of sea spray parameterizations are still lim- In particular, at 96 h, the mean absolute high-resolution coupled and uncoupled ited by numerous uncertainties, in part error is reduced by 20% and the mean simulations. The NCOM ocean model due to the lack of reliable and accurate error almost by half. is employed, with a single domain at measurements of sea spray (e.g., Andreas As a demonstration of the 15 km resolution and 36 vertical levels et al., 2008; Bao et al., 2011). It remains COAMPS-TC air-ocean coupled from the ocean surface to ~ 4 km depth. an outstanding challenge to develop a capability, Figure 3 shows results from First guess and boundary conditions physically based and observationally an air-ocean coupled simulation of Super are from the US Navy Operational verifiable sea spray parameterization that Typhoon Jangmi, a 2008 category-5 Global Atmospheric Prediction System is fully interactive with ocean waves and storm in the western North Pacific basin (NOGAPS) for the atmosphere and the atmospheric boundary layer. that occurred during the THe Observing global NCOM for the ocean for a series The sensitivity of the predicted system Research and Predictability of five-day forecasts; here, we highlight track, intensity, and structure of EXperiment (THORPEX) Pacific Asian the forecast initialized at 0000 UTC tropical cyclones to the representation Regional Campaign (T-PARC) and September 25, 2008. The same initial sea of the atmospheric planetary boundary the Office of Naval Research’s (ONR’s) surface temperature (SST) is used for the layer (PBL) is illustrated through a Tropical Cyclone Structure-08 (TCS-08) coupled and uncoupled forecasts, but SST comparison of two different versions experiments. During its lifetime from is unchanged during the uncoupled run, of the turbulence parameterization September 24 to October 2, Jangmi and the ocean model predicts SST during in COAMPS-TC. One version of the formed in a region of deep ocean mixed the coupled run. PBL parameterization makes use of layer northeast of the island of Yap, Both the coupled and uncoupled a buoyancy-based nonlocal mixing then moved northwestward, crossing simulations accurately predicted Jangmi’s length to represent turbulent mixing over several warm eddies followed track for the four days prior to landfall (e.g., Bougeault and André, 1986; Grinier by a series of cold eddies, and then (Figure 3a). However, the intensity is and Bretherton, 2002). This approach made landfall over northern Taiwan considerably different between the two has significantly improved COAMPS-TC intensity forecasts due to its suitability for turbulent mixing in deep convection. abIntensity accuracy and bias R34 accuracy and bias

The radius of the 34 kt winds, however, (276) (270) (228) (192) (156) (129) (992) (864) (736) (596) (480) 60 is too large, a result of the excessive 15 mixing in the boundary layer. In order 50 to improve the prediction of the TC 10 40 wind field within the boundary layer, the 5 Bougeault and André mixing length is 30 0 replaced with a Mellor-Yamada mixing 20 length representation (Mellor and −5 MAE (kt, solid), ME dashed) PBL b 10 PBL b MAE (nm, solid), ME dashed) Yamada, 1982) in the lowest 3 km, where abPBL a PBL a 0 wind shear dominates in the turbulence 0 24 48 72 96 120 0 24 48 72 96 120 Lead time (h) Lead time (h) production, while maintaining the Figure 2. Comparison of two planetary boundary layer (PBL) parameterizations for: (a) forecast Bougeault and André mixing length intensity, and (b) forecast radius of the 34 kt wind. The mean absolute errors (MAE, solid) and above the boundary layer. This new mean errors (ME, dashed) are shown for a large sample of western Atlantic Ocean storms. PBL-a method leads to a reduction in the inten- uses a Bougeault and André (1986) mixing length only, while PBL-b features a combination of Bougeault and André (1986) and Mellor and Yamada (1982) mixing lengths. The number of sity bias (Figure 2a) in a large sample samples for each forecast lead time is shown along the top of the plot, with a greater number of of Atlantic basin TC forecasts verified samples in (b) because the evaluation is performed for each quadrant.

Oceanography | September 2014 109 simulations. The coupled simulation with the best track than the uncoupled During September–November 2010, reduces the intensification rate during simulation. This more rapid decay prior the air-ocean coupled version of the second day of the simulation, and has to landfall in the coupled simulation is COAMPS-TC was used to provide an improved intensity forecast relative associated with a considerably reduced real-time forecasts in support of the ONR to the uncoupled simulation (Figure 3b). surface enthalpy flux associated with up Impact of Typhoons on the Ocean in the After reaching peak intensity, the storm to 4°C cooling of the SSTs that occurred Pacific (ITOP) field program (D’Asaro in the coupled simulation decays faster as Jangmi passed over the series of cold et al., 2011), which was focused on after day four and compares better eddies (Figure 3a). gaining a new understanding of tropical cyclone forced cold wakes, surface fluxes, and the dynamical interaction 19W Tracks from 0000 UTC 25 Sep 2008 between tropical cyclones and the 30°N 31 a ocean. The COAMPS-TC forecasts 28°N 30 were successfully used during ITOP to guide the observing strategy in both 26°N 29 the atmosphere and ocean. A storm 24°N of particular focus during ITOP was 28 Typhoon Fanapi, which formed southeast 22°N of Taiwan in mid-September, intensified 27 20°N to a category-3 storm, and then made 26 landfall in Taiwan on September 19. 18°N The predicted storm track is in excellent 25 16°N agreement with the best track, as Figure 4 shows for a forecast initialized at 24 14°N 0000 UTC September 15, 2010. There is

12°N 23 general agreement in SST distribution as well, derived from satellite observations 10°N 22 (Figure 4a) and the forecast (Figure 4b). 114°E 117°E 120°E 123°E 126°E 129°E 132°E 135°E 138°E Both predicted and satellite-observed 1000 cold wakes form on the right side of the b 26/00Z 27/00Z 28/00Z 29/00Z 990 track, with a similar amount of cooling 980 (~ 3°C) relative to SST prior to the storm. 970 960 The most intense cooling occurs on the 950 eastern flank of the wake, in both the 940 observations and the forecast, due to the 930 920 initial slow movement of the storm in the

Minimum Pressure (hPa ) 910 September 15–17, 2010, time period. It 900 is noteworthy that cold water is advected 890 01224364860728496 108 120 to the north at both the western and the Simulation Time (h) eastern ends of the wake, surrounding JMA JTWC Coupled Uncoupled the warm water near 126°E and 24.5°N. Figure 3. Simulations of Super-Typhoon Jangmi corresponding to the Note that cooling in the wake evolves coupled and uncoupled COAMPS-TC experiments initialized at 0000 UTC during the forecast and is not present in September 25, 2008, for a five-day forecast displaying (a) track and sea surface temperature (°C) valid at 96 h (0000 UTC September 29), and (b) minimum the ocean initial conditions. central pressure (hPa) for the five day forecast. The best track from the Joint The capabilities of the air-ocean- Typhoon Warning Center (JTWC) is displayed in (a), and the JTWC (gray) wave coupled COAMPS-TC system and Japan Meteorological Agency (black) estimates of the minimum central pressure are shown in (b). are highlighted for Atlantic Hurricane

110 Oceanography | Vol. 27, No.3 Frances (2004), which was observed error is about 5 m s–1 (not shown). The COAMPS-TC mean upper 50 m and during the ONR CBLAST field campaign COAMPS-TC minimum SST in the 50–100 m ocean temperature biases are near the northern shore of the Bahamas wake is 25.5°C, which is 3.7°C colder –0.07°C and 1.83°C, respectively. The Archipelago. The three-way coupled than the model initial SST. The forecast maximum significant wave height is COAMPS model configuration includes SST is in good agreement with the located in the front right quadrant of the the three grid meshes (45/15/5 km), a 229 ocean temperature profiles sampled storm ahead of the trailing cold wake single 3 km horizontal resolution ocean by nine in situ CBLAST floats during on the right side of the model track. The mesh, and a single 10 km horizontal res- a three-day forecast (D’Asaro et al., forecast mean significant wave height olution wave model. In this application, 2007; not shown). The quite reasonable error on September 1 is about 3.5 m the atmospheric model has 60 vertical sigma levels and the ocean model has 30°N 49 vertical levels with 35 sigma layers. 31 The wave model, SWAN, has 36 discrete 28°N 30 directional and 33 frequency bands. For the forecast highlighted here, the coupled 26°N 29 COAMPS-TC is initialized on 1200 UTC 24°N 28 August 31, 2004. The initial and bound- ary conditions for the atmospheric and 22°N 27 ocean models are provided by NOGAPS 26 20°N and global NCOM, respectively. 25 Figure 5 shows the air-ocean-wave 18°N 120°E 122°E 124°E 126°E 128°E 130°E 120°E 122°E 124°E 126°E 128°E 130°E 132°E coupled COAMPS-TC forecast SST, Figure 4. The cold wake generated by Typhoon Fanapi (2010) as shown by the: (a) GHRSST significant wave height, 10 m wind, total (Group of High Resolution Sea Surface Temperature) and (b) coupled COAMPS-TC 96 h surface currents, and Stokes currents forecast during the 2010 Impact of Typhoons on the Ocean in the Pacific (ITOP) field campaign at the 48 h time. Comparisons with the (initialized at 0000 UTC September 15, 2010). Color shading denotes the daily mean SST on September 19, 2010; the solid black line corresponds to the storm track with its positions marked best track show that the 48 h forecast by black dots every 12 h starting at the initial time and arrows displaying the mean surface track error is 100 nm and the intensity current vectors. The observed track is shown in (a) and the COAMPS-TC forecast track in (b).

ab 32 27.5°N 16 29°N

28°N 27.0°N 14 31 27°N 26.5°N 12 30 26°N 26.0°N 10 25°N 29 25.5°N 24°N 8 25.0°N 23°N 28 24.5°N 6 22°N 27 24.0°N 4 21°N 23.5°N 36.7 m s–1 20°N 26 3.2 m s–1 2 23.0°N 0.6 m s–1 19°N 25 0 78°W 80°W 82°W 84°W 86°W 88°W 90°W 92°W 83.5°W 84.5°W 85.5°W 86.5°W 87.5°W 88.5°W

Figure 5. Air-ocean-wave coupled COAMPS-TC 48 h forecast of (a) sea surface temperature (°C) and (b) significant wave height (m; shading), wind (black arrows), surface current (blue arrows), and surface Stokes current (magenta arrows) vectors for Hurricane Frances (2004). COAMPS-TC is initial- ized on 1200 UTC August 31, 2004. The unit vectors for the wind, surface current, and surface Stokes current vectors are shown in (b). A subset of the domain is shown in (b) as well for clarity purposes. The model and best track are by the black and red lines and squares, respectively (every 6 h).

Oceanography | September 2014 111 as verified using the Scanning Radar assimilation used beyond) and lateral validate the model for as many samples Altimeter on the NOAA-P3 aircraft. boundary conditions for these tests used as possible in order to ensure robust Alignment of the 10 m wind and Stokes the Global Forecast System (GFS). The performance for the variety of different surface current vectors are seen in the COAMPS-TC forecasts for a particular TC forecast scenarios seen in nature. cold wake (not shown), suggestive of TC were produced every 6 h while it Figure 6 summarizes the statistics an increase in Langmuir turbulence was defined to exist according to the comparing the performance of (Van Roekel et al., 2012). operational assessment of NHC (for COAMPS-TC forecasts and real-time Atlantic TCs) or JTWC (for western forecasts from other operational regional PREDICTION OF TRACK, North Pacific TCs) (typically numbered dynamical tropical cyclone models. INTENSITY, AND STRUCTURE or named storms). In the Atlantic, we The operational model forecasts were We validated COAMPS-TC predictions produced retrospective forecasts for sourced from the Automated Tropical of track and intensity over large samples 41 tropical cyclones that occurred in Cyclone Forecast system (Sampson and of cases in both the Atlantic and the 2010–2012, yielding nearly 800 forecasts. Schrader, 2000) archive, along with the western North Pacific tropical cyclone In the western North Pacific, we pro- “best-track” analyses of TC position and basins. For these tests, COAMPS-TC was duced real-time forecasts for 65 tropical intensity used as verification. Following run in uncoupled mode for efficiency cyclones from 2010–2012, yielding just conventional TC forecast validation reasons. The background field for the over 1,100 forecasts. While it is computa- procedures, a forecast case is only NAVDAS analysis corresponding to the tionally costly to build such large sample included in the validation sample if the first forecast for a storm (cycling data of COAMPS-TC forecasts, it is vital to best-track indicates the storm is a TC at the forecast initial time and valid time, and also that all models in the compar- Track accuracy Intensity accuracy and bias ison made a forecast (i.e., the sample is (798) (776) (624) (510) (400) (306) (811) (789) (633) (515) (401) (306) 300 20 a b homogeneous). For the Atlantic sample, 250 15 COAMPS-TC forecasts are compared against those from the Geophysical Fluid 200 10 Dynamics Laboratory model (GFDL) 5 150 and the Hurricane Weather Research and MAE (nm) 100 0 Forecasting model (HWRF), both run by COAMPS−TC COAMPS−TC the National Centers for Environmental 50 −5

GFDL MAE (kt, solid), ME dashed) GFDL HWRF HWRF Prediction. Figure 6a displays track −10 0 0 24 48 72 96 120 0 24 48 72 96 120 accuracy results and Figure 6b shows Lead time (h) Lead time (h) intensity accuracy (solid lines) and bias Intensity accuracy and bias (dashed lines) results. The COAMPS-TC 30 (1118)(941) (744) (561) (399) (272) Figure 6. Summary statistics for homogeneous mean absolute error for track is slightly 25 c comparisons of COAMPS-TC and operational regional dynamical tropical cyclone (TC) model 20 higher than the corresponding values forecasts. (a) Track mean absolute error for the for the operational models. However, 15 Atlantic basin sample described in the text, for 10 COAMPS-TC retrospective forecasts and real- the COAMPS-TC intensity forecasts are 5 time Geophysical Fluid Dynamics Laboratory more skillful than HWRF and GFDL (GFDL) and Hurricane Weather Research for all lead times beyond 24 h. The 0 and Forecasting model (HWRF) forecasts. −5 (b) Intensity mean absolute error (solid) and mean intensity error for COAMPS-TC

MAE (kt, solid), ME dashed) −10 COAMPS−TC mean error (dashed), for the same sample and is generally closer to zero than for the GFDN models as in (a). (c) Intensity mean absolute −15 operational models, indicating that 0 24 48 72 96 120 error (solid) and mean error (dashed) for the Lead time (h) western North Pacific basin sample described COAMPS-TC also performs better in in the text, for COAMPS-TC and GFDN (GFDL- terms of intensity bias as well as intensity Navy model) real-time forecasts. In all panels, the sample size is shown as a function of lead accuracy. Figure 6c shows intensity time via the numbers at the top of the plot. error and bias results for the western

112 Oceanography | Vol. 27, No.3 North Pacific sample of COAMPS-TC highlights a particularly accurate forecast, also had a reasonable intensity forecast forecasts, in comparison with the initialized at 12 UTC October 25, 2012. (Figure 7b), but perhaps more relevant to GFDN model, the Navy’s version of the Figure 7a shows the 120 h COAMPS-TC the prediction of coastal impacts of such Geophysical Fluid Dynamics Laboratory track forecast for this case alongside a storm (waves, surge), the model made model. At lead times from 0 to 48 h, the the observed track (from the National an accurate prediction of Sandy’s surface GFDN mean absolute error is slightly Hurricane Center “best-track”). The fore- wind field. Figure 7c shows the 84 h lower than that of COAMPS-TC, but cast shows landfall close to the observed lead time COAMPS-TC forecast of the for later lead times, COAMPS-TC has a location along the New Jersey coast 10 m wind field, which can be compared considerably lower mean absolute error between four and five days lead time, to the OSCAT (Oceansat-2 scatterom- than GFDN. Encouraging results for capturing the remarkably unusual “hard eter) ocean surface wind observations such large samples of cases provided left turn” caused by the mutual interac- (Figure 7d) collected near the forecast the necessary evidence to facilitate tion of Sandy with a deep mid-latitude valid time. The COAMPS-TC forecast the transition of COAMPS-TC into trough over the East Coast. The model captures the large size of the wind field operations at FNMOC. In spite of the generally successful performance of COAMPS-TC, the sta- a c COAMPS−TC 40°N 40°N tistical evaluation points to several issues in which the forecast model requires additional improvement. One of the most obvious issues impacting the inten- 35°N 35°N sity predictions is the rapid development of a weak maximum wind speed bias in 60 the first 12 h of the forecasts, apparent 50 in Figure 6b,c and Figure 2a. This bias 30°N 30°N 00 h 12 h results from an adjustment process due 40 24 h to an unbalanced initial state within the 36 h 48 h TC vortex. The imbalance arises from 72 h 30 the NAVDAS analysis, which is not 96 h d 25°N 120 h 40°N able to accurately represent the wind 20 and mass field correlations within the 10 vortex. Examination of new methods for initializing the vortex and minimizing 78°W 75°W 72°W 35°N 0 the adjustments during the early stages 90 of the forecasts is underway. b 80 Perhaps the most well-known tropical 70 30°N cyclone from the validation samples 60 described above is Hurricane Sandy Intensity (kt) 50 Best track 40 COAMPS−TC (2012). Sandy had a very atypical track 0 24 48 72 96 120 80°W 75°W 70°W 65°W and was an unusually large tropical Lead time (h) cyclone due to the interaction of the Figure 7. (a) COAMPS-TC real-time forecast track (blue) for Hurricane Sandy, initialized at TC with multiple mid-latitude weather 1200 UTC October 25, 2012, and corresponding best-track (black). The multicolored circles show the TC forecast position at different lead times and the corresponding best-track systems. COAMPS-TC real-time position valid at those times. (b) COAMPS-TC intensity forecast (blue) and best-track forecasts of Sandy handled these complex intensity (black). (c) COAMPS-TC forecast 10 m wind field at 84 h lead time, valid at 0000 UTC interactions quite well, producing October 29, 2012. (d) OSCAT (Indian Space Research Organization’s OceanSat-2 satellite) ocean surface wind observations collected near 0000 UTC October 29, 2012. Image courtesy of accurate simulations of the storm’s NASA In panels (c) and (d), the wind speed in miles per hour (mph) is shown in color and the track, intensity, and structure. Figure 7 vectors show wind direction.

Oceanography | September 2014 113 and shows a core of strong winds near capability of the COAMPS-TC system Hurricane Forecast Improvement Project the center surrounded by an east-west to realistically capture not only the SST (HFIP). We also appreciate support elongated area of relatively light winds, cooling in the wake of the storm, but for computational resources through a qualitatively similar to the observations. also the characteristics of the ocean grant of Department of Defense High However, the OSCAT data do indicate surface waves, such as the significant Performance Computing time from the stronger winds than the model along the wave height maximum in the front right DoD Supercomputing Resource Center coast north of 35°N. quadrant of the storm. at Stennis, MS, and Vicksburg, MS. While COAMPS-TC has accurately SUMMARY AND CONCLUSIONS predicted the evolution of Sandy, COAMPS® is a registered trademark of Prediction of tropical cyclone track and, Fanapi, Frances, and Jangmi, as well the Naval Research Laboratory. particularly, intensity and structure, as other tropical cyclones (not shown) remains among the greatest challenges in real time and in retrospective cases, REFERENCES facing meteorologists today. The results there are other examples that were not Allard, R., E. Rogers, P. Martin, T. Jensen, P. Chu, T. Campbell, J. Dykes, T. Smith, of this research highlight the promise predicted as well. These storms, and the J. Choi, and U. Gravois. 2014. The US Navy of a new high-resolution capability for data collected during their life cycles, coupled ocean-wave prediction system. Oceanography 27(3):92–103, http://dx.doi.org/​ tropical cyclone prediction using the provide great opportunities to study 10.5670/oceanog.2014.71. Navy’s COAMPS-TC. During the past and obtain a greater appreciation of the Andreas, E.L., P.O.G. Persson, and J.E. Hare. 2008. A bulk turbulent air-sea flux algorithm several tropical cyclone seasons in the complex physical interactions that occur for high-wind, spray conditions. Journal Western Pacific and Atlantic basins, in these systems, and to use this infor- of Physical Oceanography 38:1,581–1,596, http://dx.doi.org/10.1175/2007JPO3813.1. COAMPS-TC has been tested in real mation to improve our COAMPS-TC Bao, J.-W., C.W. Fairall, S.A. Michelson, and time and for a series of retrospective modeling system. L. Bianco. 2011. Parameterizations of sea-spray impact on the air-sea momen- cases in both coupled and uncoupled This research will lead to new tum and heat fluxes. Monthly Weather modes at high horizontal resolution. capabilities in the form of mesoscale TC Review 139:3,781–3,797, http://dx.doi.org/​ 10.1175/MWR-D-11-00007.1. An evaluation of a large sample of ensemble forecasts, providing the Navy Bao, J.-W., J.M. Wilczak, and J.-K. Choi. 2000. forecasts for 2010–2012 in the Atlantic with probabilistic forecasts of tropical Numerical simulations of air–sea interaction under high wind conditions using a coupled and Western Pacific basins reveals that cyclone intensity and structure for the model: A study of hurricane development. the COAMPS-TC intensity predictions first time. It is also expected that this Monthly Weather Review 128:2,190–2,210, http://dx.doi.org/10.1175/1520-0493(2000)​ are competitive with, and in some research will help motivate new field 128<2190:NSOASI>2.0.CO;2. regards more accurate than, the other campaigns that focus on the key mea- Bender, M.A., I. Ginis, R. Tuleya, B. Thomas, and T. Marchok. 2007. The operational leading dynamical models, particularly surements needed to further advance our GFDL Coupled Hurricane–Ocean Prediction for lead times beyond 36 h. Recent real- understanding of the convective structure System and a summary of its performance. Monthly Weather Review 135:3,965–3,989, time forecasts of Hurricane Sandy (2012) and dynamics of these systems and also http://dx.doi.org/10.1175/2007MWR2032.1. illustrate the capability of COAMPS-TC provide forecast validation. The flexibility Black, P.G., E.A. D’Asaro, T.B. Sanford, W.M. Drennan, J.A. Zhang, J.R. French, to capture both the intensity and the of the COAMPS-TC design will also P.P. Niiler, E.J. Terrill, and E.J. Walsh. 2007. fine-scale structure in agreement with allow us to test more advanced physics Air–sea exchange in hurricanes: Synthesis of observations from the Coupled Boundary Layer observations. Typhoon Fanapi (2010) and numerical methods in an effort to Air–Sea Transfer Experiment. Bulletin of the and Super Typhoon Jangmi (2008) gain a better physical understanding of American Meteorological Society 88:357–374, http://dx.doi.org/10.1175/BAMS-88-3-357. in the Western Pacific are accurately the model’s intensity forecast skill. Bougeault, P.H., and J.-C. André. 1986. On predicted using the air-ocean coupled the stability of the third-order turbulence closure for the modeling of the strato- COAMPS-TC, with regard to not only ACKNOWLEDGEMENTS cumulus-topped boundary layer. Journal of the track and intensification but also the We acknowledge the support of the the Atmospheric Sciences 43:1,574–1,581, http://dx.doi.org/10.1175/1520-0469(1986)​ sea surface cooling induced through the Office of Naval Research Program 043<1574:OTSOTT>2.0.CO;2. mixing and upwelling in agreement with Element (PE) 0602435N and PMW- Braun, S.A., M.T. Montgomery, and X. Pu. 2006. High-resolution simulation of Hurricane satellite measurements. The air-ocean- 120 PE 0603207N, as well as the Bonnie (1998). Part I: The organization of eye- wave coupled simulations of the Atlantic National Oceanic and Atmospheric wall vertical motion. Journal of the Atmospheric Sciences 63:19–42, http://dx.doi.org/10.1175/ Hurricane Frances (2004) highlight the Administration (NOAA) sponsored JAS3598.1.

114 Oceanography | Vol. 27, No.3 Chen, S., T.J. Campbell, H. Jin, S. Gaberšek, Doyle, J.D. 2002. Coupled atmosphere-ocean Martin, P.J., J.W. Book, and J.D. Doyle. 2006. R.M. Hodur, and P. Martin. 2010. Effect wave simulations under high wind conditions. Simulation of the northern Adriatic circulation of two-way air–sea coupling in high and Monthly Weather Review 130:3,087–3,099, during winter 2003. Journal of Geophysical low wind speed regimes. Monthly Weather http://dx.doi.org/10.1175/1520-0493(2002)​ Research 111, C03S12, http://dx.doi.org/​ Review 138:3,579–3,602, http://dx.doi.org/​ 130<3087:CAOWSU>2.0.CO;2. 10.1029/2006JC003511. 10.1175/2009MWR3119.1. Doyle, J.D., Y. Jin, R. Hodur, S. Chen. H. Jin, Mellor, G.L., and T. Yamada. 1982. Development of Chen, S., T.J. Campbell, S. Gabersek, H. Jin, and J. Moskaitis, A. Reinecke, P. Black, J. Cummings, a turbulence closure model for geophysical fluid R.M. Hodur. 2011. Next generation air–ocean– E. Hendricks, and others. 2011. Real time problems. Reviews of Geophysics 20:851–875, wave Coupled Ocean/Atmosphere Mesoscale tropical cyclone prediction using COAMPS-TC. http://dx.doi.org/10.1029/RG020i004p00851. Prediction System (COAMPS). NRL Review, Pp. 15–28 in Advances in Geosciences, vol. 28. Moon, I.-J., T. Hara, I. Ginis, S.E. Belcher, and 9 pp, http://www.nrl.navy.mil/content_images/ C.-C. Wu and J. Gan, eds, World Scientific H. Tolman. 2004. Effect of surface waves Atmospheric_2011.pdf. Publishing Company, Singapore. on air-sea momentum exchange. Part I: Chen, S.S., J.F. Price, W. Zhao, M.A. Donelan, Fairall, C.W., M.L. Banner, W.L. Peirson, W. Asher, Effect of mature and growing seas. Journal and E.J. Walsh. 2007. The CBLAST-Hurricane and R.P. Morison. 2009. Investigation of of the Atmospheric Sciences 61:2,321–2,333, Program and the next-generation fully coupled the physical scaling of sea spray spume http://dx.doi.org/10.1175/1520-0469(2004)​ atmosphere–wave–ocean models for hurricane droplet production. Journal of Geophysical 061<2321:EOSWOA>2.0.CO;2. research and prediction. Bulletin of the Research 114, C10001, http://dx.doi.org/​ Powell, M.D., P.J. Vickery, and T.A. Reinhold. 2003. American Meteorological Society 88:311–317, 10.1029/2008JC004918. Reduced drag coefficient for high wind speeds http://dx.doi.org/10.1175/BAMS-88-3-311. Fairall, C.W., J.D. Kepert, and G.J. Holland. 1994. in tropical cyclones. Nature 422:279–283, Cummings, J.A. 2005. Operational multivariate The effect of sea spray on surface energy trans- http://dx.doi.org/10.1038/nature01481. ocean data assimilation. Quarterly Journal of ports over the ocean. Global Atmosphere-Ocean Rogers, R., S. Aberson, M. Black, P. Black, J. Cione, the Royal Meteorological Society 131:583–604, System 2:121–142. P. Dodge, J. Gamache, J. Kaplan, M. Powell, http://dx.doi.org/10.1256/qj.05.105. Fowle, M.A., and P.J. Roebber. 2003. Short- J. Dunion, and others. 2006. The Intensity Daley, R., and E. Barker. 2000. NAVDAS: range (0–48 h) numerical prediction Forecasting Experiment: A NOAA multiyear Formulation and diagnostics. Monthly Weather of convective occurrence, mode, and field program for improving tropical cyclone Review 129:869–883, http://dx.doi.org/10.1175/​ location. Weather and Forecasting 18:782–794, intensity forecasts. Bulletin of the American 1520-0493(2001)129<0869:NFAD>2.0.CO;2. http://dx.doi.org/10.1175/1520-0434(2003)​ Meteorological Society 87:1,523–1,537, D’Asaro, E.A., T.B. Sanford, P.P. Niiler, and 018<0782:SHNPOC>2.0.CO;2. http://dx.doi.org/10.1175/BAMS-87-11-1523. E.J. Terrill. 2007. Cold wake of Hurricane Goerss, J.S. 2007. Prediction of consensus tropical Sampson, C.R., and A.J. Schrader. 2000. The Frances. Geophysical Research Letters 34, cyclone track forecast error. Monthly Weather Automated Tropical Cyclone Forecasting L15609, http://dx.doi.org/10.1029/​ Review 135:1,985–1,993, http://dx.doi.org/​ System (Version 3.2). Bulletin of the American 2007GL030160. 10.1175/MWR3390.1. Meteorological Society 81:1,231–1,240, D’Asaro, E., P. Black, L. Centurioni, P. Harr, Hamill, T.M., J.S. Whitaker, D.T. Kleist, M. Fiorino, http://dx.doi.org/10.1175/1520-0477(2000)081​ S. Jayne, I.-I. Lin, C. Lee, J. Morzel, and S.G. Benjamin. 2011. Predictions of <1231:TATCFS>2.3.CO;2. R. Mrvaljevic, P. P. Niiler, and others. 2010’s tropical cyclones using the GFS and Smith, T.A., S. Chen, T. Campbell, E. Rogers, 2011. Typhoon-ocean interaction in ensemble-based data assimilation methods. S. Gabersek, D. Wang, S. Carroll, and the western North Pacific: Part 1. Monthly Weather Review 139:3,243–3,247, R. Allard. 2013. Ocean-wave coupled Oceanography 24(4):24–31, http://dx.doi.org/​ http://dx.doi.org/10.1175/MWR-D-11-00079.1. modeling in COAMPS-TC: A study of 10.5670/oceanog.2011.91. Hodur, R.M. 1997. The Naval Research Hurricane Ivan (2004). Ocean Modelling Davis, C., W. Wang, S.S. Chen, Y. Chen, Laboratory’s Coupled Ocean/Atmosphere 69:181–194, http://dx.doi.org/10.1016/​ K. Corbosiero, M. DeMaria, J. Dudhia, Mesoscale Prediction System (COAMPS). j.ocemod.2013.06.003. G. Holland, J. Klemp, J. Michalakes, and others. Monthly Weather Review 125:1,414–1,430, Van Roekel, L.P., B. Fox-Kemper, P.P. Sullivan, 2008. Prediction of landfalling hurricanes http://dx.doi.org/10.1175/1520-0493(1997)​ P.E. Hamlington, and S.R. Haney. 2012. with the advanced hurricane WRF model. 125<1414:TNRLSC>2.0.CO;2. The form and orientation of Langmuir cells Monthly Weather Review 136:1,990–2,005, Jin, Y., W.T. Thompson, S. Wang, and C.-S. Liou. for misaligned winds and waves. Journal http://dx.doi.org/10.1175/2007MWR2085.1. 2007. A numerical study of the effect of of Geophysical Research 117, C05001, DeMaria, M., M. Mainelli, L.K. Shay, J.A. Knaff, dissipative heating on tropical cyclone http://dx.doi.org/10.1029/2011JC007516. and J. Kaplan. 2005. Further improvements intensity. Weather and Forecasting 22:950–966, Whitehead, J.C. 2003. One million dollars per mile? to the Statistical Hurricane Intensity http://dx.doi.org/10.1175/WAF1028.1. The opportunity costs of hurricane evacuation. Prediction Scheme (SHIPS). Weather and Kantha, L.H., and C.A. Clayson. 2004. On Ocean & Coastal Management 46:1,069–1,083, Forecasting 20:531–543, http://dx.doi.org/​ the effect of surface gravity waves on http://dx.doi.org/10.1016/j.ocecoaman.2003.​ 10.1175/WAF862.1. mixing in the oceanic mixed layer. Ocean 11.001. DeMaria, M., C.R. Sampson, J.A. Knaff, and Modelling 6:101–124, http://dx.doi.org/10.1016/ Zhu, T., and D.-L. Zhang. 2006. Numerical K.D. Musgrave. 2014. Is tropical cyclone S1463-5003(02)00062-8. simulation of Hurricane Bonnie (1998). intensity guidance improving? Bulletin of the Liou, C.-S., and K.D. Sashegyi. 2012. On the Part II: Sensitivity to varying cloud micro- American Meteorological Society 95:387–398, initialization of tropical cyclones with a physical processes. Journal of the Atmospheric http://dx.doi.org/10.1175/BAMS-D-12-00240.1. three-dimensional variational analysis. Natural Sciences 63:109–126, http://dx.doi.org/10.1175/ Donelan, M.A., B.K. Haus, N. Reul, W.J. Plant, Hazards 63:1,375–1,391, http://dx.doi.org/​ JAS3599.1. M. Stianssnie, H.C. Graber, O.B. Brown, and 10.1007/s11069-011-9838-0. Zhu, T., D.-L. Zhang, and F. Weng. 2004. E.S. Saltzman. 2004. On the limiting aerody- Marks, F.D., and L.K. Shay. 1998. Landfalling trop- Numerical simulation of Hurricane Bonnie namic roughness of the ocean in very strong ical cyclones: Forecast problems and associated (1998). Part I: Eyewall evolution and intensity winds. Geophysical Research Letters 31, L18306, research opportunities. Bulletin of the American changes. Monthly Weather Review 132:225–241, http://dx.doi.org/10.1029/2004GL019460. Meteorological Society 79:305–323. http://dx.doi.org/10.1175/1520-0493(2004)​ Done, J., C. Davis, and M. Weisman. 2004. The Martin, P.J. 2000. Description of the NAVY Coastal 132<0225:NSOHBP>2.0.CO;2. next generation of NWP: Explicit forecasts Ocean Model Version 1.0. NRL Report NRL/ of convection using the Weather Research FR/7322-00-9962, 42 pp. and Forecast (WRF) Model. Atmospheric Science Letters 5:110–117, http://dx.doi.org/​ 10.1002/asl.72.

Oceanography | September 2014 115