sea surface temperature variability in hurricanes: implications...

AUGUST 2003 1783C I O N E A N D U H L H O R N

q 2003 American Meteorological Society

Sea Surface Temperature Variability in Hurricanes: Implications with Respect toIntensity Change

JOSEPH J. CIONE

NOAA/AOML/Hurricane Research Division, Miami, Florida

ERIC W. UHLHORN

RSMAS/CIMAS, University of Miami, Miami, Florida

(Manuscript received 21 February 2002, in final form 17 January 2003)

ABSTRACT

Scientists at NOAA’s Hurricane Research Division recently analyzed the inner-core upper-ocean environmentfor 23 Atlantic, Gulf of Mexico, and Caribbean hurricanes between 1975 and 2002. The interstorm variabilityof sea surface temperature (SST) change between the hurricane inner-core environment and the ambient oceanenvironment ahead of the storm is documented using airborne expendable bathythermograph (AXBT) obser-vations and buoy-derived archived SST data. The authors demonstrate that differences between inner-core andambient SST are much less than poststorm, ‘‘cold wake’’ SST reductions typically observed (i.e., ;08–28Cversus 48–58C). These findings help define a realistic parameter space for storm-induced SST change within theimportant high-wind inner-core hurricane environment. Results from a recent observational study yielded esti-mates of upper-ocean heat content, upper-ocean energy extracted by the storm, and upper-ocean energy utilizationfor a wide range of tropical systems. Results from this analysis show that, under most circumstances, the energyavailable to the tropical cyclone is at least an order of magnitude greater than the energy extracted by the storm.This study also highlights the significant impact that changes in inner-core SST have on the magnitude of air–sea fluxes under high-wind conditions. Results from this study illustrate that relatively modest changes in inner-core SST (order 18C) can effectively alter maximum total enthalpy (sensible plus latent heat) flux by 40% or more.

The magnitude of SST change (ambient minus inner core) was statistically linked to subsequent changes instorm intensity for the 23 hurricanes included in this research. These findings suggest a relationship betweenreduced inner-core SST cooling (i.e., increased inner-core surface enthalpy flux) and tropical cyclone intensi-fication. Similar results were not found when changes in storm intensity were compared with ambient SST orupper-ocean heat content conditions ahead of the storm. Under certain circumstances, the variability associatedwith inner-core SST change appears to be an important factor directly linked to the intensity change process.

1. Introduction

The effect of the ocean on tropical cyclone (TC) gen-esis and maintenance has been well known for decades.The ocean provides the necessary energy to establishand maintain deep convection (Byers 1944; Palmen1948; Riehl 1954; Miller 1958; Malkus and Riehl 1960).Recent studies conducted by Shay et al. (2000) andBosart et al. (2000) have also shown that in some in-stances, warm upper-ocean features can significantly im-pact TC intensity. While findings from these case studiesare significant, it is still unclear how (and to what extent)variations in upper-ocean thermal structure directly im-pact changes in storm intensity. Tropical cyclone inten-sity change is a complex and interactive nonlinear pro-

Corresponding author address: Joseph J. Cione, Rosenstiel Schoolof Marine and Atmospheric Science, University of Miami, 4301 Rick-enbacker Cswy., Miami, FL 33149.E-mail: [email protected]

cess that often involves several competing or synergisticfactors (Riehl 1948; 1950; Miller 1958; Sadler 1978;Gray 1979; Holland and Merrill 1984; Emanuel 1986,1988; DeMaria and Pickle 1988; Molinari and Vollaro1990; Willoughby and Black 1996; Holland 1997;DeMaria and Kaplan 1994; 1999; Kaplan and DeMaria1999; Bosart et al. 2000; Shay et al. 2000). Identifyingthe quantitative impact a physical process has on inten-sity change is an arduous task and one that can only beattempted using controlled numerical methodology. On-going coupled ocean–atmosphere TC modeling effortsare works in progress; many of the numerical routinesand parameterizations (e.g., data initialization, grid res-olution, turbulent fluxes, atmospheric microphysics,etc.) used within the rarely observed high-wind stormenvironment still require significant improvement.

An additional stumbling block confronting TC mod-elers is data verification of the upper ocean and atmo-spheric boundary layer (ABL) hurricane environments.

1784 VOLUME 131M O N T H L Y W E A T H E R R E V I E W

Prior to 1997, accurate depictions of inner-core ABLthermodynamic and kinematic structure were essentiallyunknown. Also, since many of the earlier TC oceanresponse studies concentrated on studying the post-storm, cold wake region of the hurricane (Federov etal. 1977; Pudov et al. 1978; Pudov 1980; Price 1981),accurate depictions of the TC upper-ocean eyewall re-gion have been exceedingly rare over the past 30 years(Black 1983; Black et al. 1988; Shay et al. 1992; Blackand Holland 1995). As a result, numerical attempts toinitialize and verify this rarely observed ocean envi-ronment have relied on prestorm ambient sea surfacetemperatures (SST) ahead of the system and post-stormcold wake SSTs, typically valid several days after TCpassage (Bender et al. 1993; Bender et al. 2000). How-ever, recent experiments conducted during the HurricaneResearch Division’s (HRD) annual field program havehelped fill in these atmospheric and oceanic data voidregions by using Global Positioning System (GPS) drop-sondes and airborne expendable bathythermographs(AXBT).

To accurately document TC-induced SST change, thisstudy will use upper-ocean data obtained from AXBTsduring 1997–2002 HRD field experiments along withfixed and drifting buoy observations over the 1975–2002 period. It is believed that these (multistorm) ob-servations and analyses will improve the representationof SST cooling patterns typically observed in the high-wind hurricane environment. These rare observationswill serve as ‘‘ocean truth’’ for coupled modeling effortsattempting to simulate TC-induced SST change directlyunder the storm.

2. Research goals

Earlier observational studies have documented thesignificant impact hurricanes can have on the verticaland horizontal structure of the upper-ocean environment(Federov et al. 1977; Pudov et al. 1978; Pudov 1980;Price 1981; Black 1983; Shay et al. 1992). Figure 1illustrates this point and depicts ocean surface temper-atures 48–58C cooler in the post-Georges (1998) coldwake region (relative to the surrounding, ambient oceanenvironment). However, it should be recognized that thisocean response analysis was constructed seven days af-ter the storm’s passage and does not necessarily rep-resent typical SST cooling patterns observed directlyunder the hurricane. This is an important point sincemuch of the ocean-to-atmosphere exchange of energyin hurricanes occurs within a relatively limited area nearthe eyewall. As a result, obtaining accurate represen-tations of near-surface ocean temperatures within thiscritical region is of paramount importance. By usingbuoy observations with recent hurricane AXBT data,this study will quantify the relative magnitude and in-herent variability associated with differences betweenambient SSTs well ahead of the storm and SSTs ob-served near storm center. The authors will investigate

any potential relationships between these ‘‘SST differ-ences’’ and observed changes in storm intensity. In ad-dition, the potential impact SST variability has on air–sea fluxes within the high-wind hurricane environmentwill also be investigated. Finally, using observations andrecent findings from Cione et al. (2000), estimates forhurricane heat potential, energy extracted by the storm,and energy utilization will be made.

3. Methodology and data used

Early ocean-response studies concentrated on ana-lyzing the horizontal and vertical thermal structure ofthe upper-ocean within the TC-modified, cold wake en-vironment. In addition to these poststorm wake studies,Black (1983) conducted analyses of the upper-oceanthermal structure for a number of Atlantic, Gulf of Mex-ico, and eastern Pacific hurricanes between 1971 and1980. Due to difficult observing conditions and limitedopportunities, there have only been a few case studiesthat have attempted to document the hurricane high-wind upper-ocean environment (Black et al. 1988; Shayet al. 1992; Black and Holland 1995). As such, quan-tifying the interstorm variability associated with upper-ocean thermal conditions near/within the hurricane eye-wall has been difficult. A primary goal of this study isto improve SST cooling estimates near the hurricaneeyewall using upper-ocean measurements from many(23) hurricanes. To accomplish this, observations fromthe Tropical Cyclone Buoy Database (TCBD) (Cione etal. 2000) are used. The TCBD includes information onhurricane position and intensity as well as near-surfacemeteorological and oceanographic data from fixed anddrifting platforms. The TCBD incorporates observationsfrom over 40 hurricanes between 1975 and 2002. Mostof the TCBD observations were acquired from the Na-tional Data Buoy Center quality controlled buoy archive(Gilhousen 1988, 1998).1 In addition to the TCBD data,upper-ocean AXBT observations from nine HRD fieldexperiments were also used. From both buoy and AXBTdata, 33 along-track SST transects were obtained for 23hurricanes (Table 1).

To document TC-induced SST change, clear defini-tions for both ambient and inner-core SSTs must beestablished. For this study, ambient SSTs (SSTA) arelocated well ahead of the storm center (.28 latitude) ineither the right front or left front quadrant (defined ina storm-relative coordinate system). The location ofSSTA is defined as the point where SST initially de-creases. Inner-core SST (SSTIC) is defined as the min-imum SST within a 60-km radius of the analyzed TCcenter. Since it is a primary goal of this research toinvestigate the magnitude and variability of SST change

1 Detailed information on platform locations and configurations,sensor descriptions and levels of accuracy, data acquisition, aver-aging, quality control, and archival techniques can be found onlineat http://www.noaa.ndbc.gov/.


FIG. 1. Advanced Very High Resolution Radiometer (AVHRR) image (courtesy of Johns Hopkins University Applied Physics Laboratory)depicting the Gulf of Mexico SST distribution averaged over a 7-day period ending 1 Oct 1998 2303 UTC. The 38–58C storm-induced coldwake is visible to the right of Hurricane Georges’s track.

for mature tropical systems, only observations from hur-ricanes are included in this analysis. All storms are pre-landfall, located south of 368N, and attain hurricaneintensity (i.e., maximum surface wind speed . 32 ms21) at some point during the period of observation.Whenever possible, estimates for inner-core wake SST(SSTICW) are also included. SSTICW is defined as theminimum SST observed in either the right rear or leftrear quadrant of the storm. However, unlike typical TC-induced cold wake SST fields (Fig. 1), SSTICW obser-vations are located ,200 km from the storm center. Assuch, SSTICW observations are often located in areas ofmoderate surface wind (typically . 15 m s21). Due to

various factors such as instrument failure, systems mak-ing landfall, and experimental design limitations, 26 ofthe 33 horizontal SST transects listed in Table 1 includeestimates for DSSTICW.

4. Horizontal variability of SST in hurricanes

The 33 (26) DSSTIC (DSSTICW) estimates shown inTable 2 represent the difference between SSTIC(SSTICW)and SSTA ahead of the storm. Table 2 is stratified byDSSTIC with associated DSSTICW values listed wheneveravailable. Figure 2 illustrates the geographic location


TABLE 1. Summary of along-track sea surface temperature transects. In all, 33 transects from 23 hurricanes are included.

Hurricane Year Data source (Buoy; AXBT)Total number of

transectsNumber of DSSTICW

estimates

LiliIsidoreBretDennisIrene

20022002199919991999

42001 AXBTAXBTAXBTAXBTAXBT

21131

21110

GeorgesEarlBonnieDanielleErika

19981998199819981997

42040 SANF1 DRYF1 SMKF142003 42040 AXBTAXBTAXBTAXBT

43111

42101

EdouardAllisonLuisOpalEmily

19961995199519951993

41611 4161442003415194200144019

21111

21011

AndrewBobElenaAliciaFrederic

19921991198519831979

FWYF1 42003DSLN420074200842003

21111

11111

AnitaBelleEloise

197719761975

420024100242001

111

111

TABLE 2. Magnitude of SST change by storm. A total of 33 inner-core SST change (DSSTIC, ambient minus inner core) and 26 inner-core wake SST change (DSSTICW, ambient minus inner-core wake)estimates are included.

Hurricane YearObservingplatform

DSSTIC (8C)SSTIC-SSTA

DSSTICW (8C)SSTICW-SSTA

AliciaGeorgesEmilyErikaBelle

19831998199319971976

420084204044019AXBT41002

21.821.521.521.521.2

22.723.421.521.521.5

ElenaEarlLuisGeorgesDennis

19851998199519981999

420074200341519SANF1AXBT

21.020.920.920.820.8

21.7——

21.4—

GeorgesDennisLiliGeorgesDanielle

19981999200219981998

DRYF1AXBT42001SMKF1AXBT

20.820.820.720.720.7

20.9—

21.921.2

—AndrewAllisonBobAnitaIrene

19921995199119771999

FWYF142003DSLN42002AXBT

20.720.620.620.620.6

—20.620.621.4

—FredericDennisEarlEloiseBonnie

19791999199819751998

42003AXBT4204042001AXBT

20.520.520.420.320.3

20.820.520.520.820.3

BretLiliOpalEarlIsidore

19992002199519982002

AXBTAXBT42001AXBTAXBT

20.320.320.120.120.1

20.420.720.320.820.2

EdouardEdouardAndrew

199619961992

416114161442003

0.00.00.0

0.020.220.1

and magnitude for all 33 DSSTIC estimates listed inTable 2.

The DSSTIC and DSSTICW values depicted in Table 2are much less than the 48–58C TC-induced cold wakeSST reductions often observed 1–2 days after a hurricanepassage (Price 1981; Mayer et al. 1981; Black 1983). Incomparison, average DSSTIC and DSSTICW are 20.78Cand 21.08C, respectively. These values represent ;15%–25% of the 48–58C cold wake reductions shown in Fig.1 and are in reasonable agreement with early mixed layerTC–ocean modeling studies (Elsberry et al. 1976; Changand Anthes 1978) as well as inner-core observations fromBlack (1983). Cumulative distributions for both DSSTIC

and DSSTICW are given in Fig. 3. These results show thatboth SSTIC and SSTICW are significantly warmer thanpoststorm, TC cold wake SSTs. Figure 4 captures thealong-track variability of SST change as a function ofradial distance (RD) from the storm center for all 22 (19)DSSTIC (DSSTICW) buoy transects listed in Table 2(AXBT-derived DSSTIC and DSSTICW observations arenot included in Fig. 4 due to horizontal and temporalsampling limitations associated with these data). SST re-ductions of 1.58C or less were noted for 21 of 22 (15 of19) SSTIC(SSTICW) buoy transects.

The summary shown in Table 3 includes both buoyand AXBT-derived data. Stratifying results by DSSTIC,upper and lower 50th percentile statistical summariesare illustrated for DSSTIC, DSSTICW, SSTA, SSTIC,SSTICW, the radial distance at which SSTA initially de-creased (RDA), the radial distance where SSTIC was ob-served (RDIC), the radial distance where SSTICW wasmeasured (RDICW), and storm-specific parameters suchas storm latitude (TCLAT), storm intensity (TCIWIND;


FIG. 2. The geographic locations for the 33 DSSTIC values listed in Table 2. The estimateshave been divided into two groups, DSST . 20.78C (less SST cooling, open triangles, 17 events)and DSSTIC , 20.78C (more SST cooling, filled triangles, 16 events).

TCIBAR), and storm speed (TCSPEED). Tests for statisticalsignificance between upper and lower percentile meanswere conducted. Table 3 depicts differences betweenupper and lower 50th percentile means for SSTIC,SSTICW, TCLAT, and RDA (bolded values). It should benoted that statistically significant findings were notfound for SSTA. In fact, ambient SST showed little var-iability with respect to inner-core SST change. Resultsfrom Table 3 show that lower (higher) latitude events,on average, exhibited less (more) inner-core SST cool-ing. This result may in part be explained by climatology.It stands to reason that, on average, lower latitude eventswould encounter deeper, warm water upper-ocean en-vironments. These events (assuming all other factors tobe equal) should experience less upper-ocean cooling.In addition, for a given wind speed, the onset of SSTcooling for the lower 50% sample would tend to occurcloser to the storm center due to the presence of (rel-atively) deeper warm water. This may partly explainwhy average RDA (i.e., the point at which SSTA firstdecreases) was ;120 km closer to the storm center forthe lower 50th percentile group of observations. Whiledifferences between upper and lower sample meanswere not found for storm speed or storm initial intensity,results from Table 3 suggest that faster moving, initiallyweaker storms may be more likely to experience reducedDSST values. These trends, while statistically incon-clusive, are in agreement with earlier results (Black1983; Bender et al. 2000; Chan et al. 2001) and arephysically consistent given the reduced level of upper-ocean turbulent mixing one would expect from quicker-moving, relatively weaker tropical systems.

5. Hurricane heat potential and energy extractedby the storm

Leiper and Volgenau (1972) first defined the term‘‘hurricane heat potential’’ as the integrated verticaltemperature from the sea surface to the depth of the268C isotherm. Hurricane heat potential (also known as‘‘upper-ocean heat content’’) is denoted by QH and isdefined as

0

Q (x, y, t) 5 rc DT(x, y, z, t) dz, (1)H p Ez(T526)

where cp is the specific heat of water at constant pressure(4178 J kg21 K21), r is the average density of the upperocean (1026 kg m23), and DT is the difference betweenT(z) and 268C over the depth interval dz. The units forQH are given in kJ cm22 (5107 J m22), as is commonin the literature.

Since much of the ocean-to-atmosphere exchange ofenergy in tropical cyclones occurs within the high-windinner core, analyses and estimates presented here willfocus on conditions potentially present within this im-portant region. Exactly how much of the available QH

is extracted by the storm within the inner core is verydifficult to quantify without highly accurate, direct, andcontinuous measurement. However, given the stormspeed, storm initial intensity, and upper-ocean thermalprofile ahead of the storm, it is possible to estimate theamount of upper-ocean heat content extracted by thestorm. In order to maintain consistency with earlier def-initions, the inner-core upper-ocean environment is de-fined to be 0–60 km from the storm center. For these


FIG. 3. The cumulative distribution for ‘‘core minus ambient’’ SSTchange. All 33 DSSTIC and 26 DSSTICW values listed in Table 2 areincluded. DSSTIC and DSSTICW are given in 8C.

FIG. 4. Individual horizontal SST profiles for the 22 buoy-derivedSST transects listed in Table 1 (note: No AXBT profiles are illus-trated). Here 21 of the 22 transects extend 200 km to the rear of thestorm center. In addition to the individual transects, the mean DSSTprofile (heavy horizontal line) and standard deviation estimates (ver-tical heavy lines) are also illustrated. Dashed individual profiles de-note along-track profiles that exhibit DSSTICW cooling values in excessof 1.58C. DSST is measured in 8C and distance is measured in degreeslatitude from the storm center. Negative radial distance indicates ob-servations that were obtained in either the right rear or left rearquadrant of the storm.

calculations, the initial upper-ocean temperature profileused to calculate QH is located 60 km ahead of the stormalong the storm track. Due to the relatively short anal-ysis period (3–13 h), coupled with the fact that the initialprofile 60 km ahead of the storm is already well mixed,inner-core QH estimates are assumed to remain constantfor this analysis. By utilizing various storm speeds andintensities, estimates for upper-ocean heat content ex-tracted by the storm (QHpext) and estimates for ‘‘upper-ocean energy utilization’’ (QHputil) can be constructed.

First, the inner-core enthalpy (sensible plus latentheat) flux (Hcore) is computed using the standard bulkaerodynamic formulas:

H 5 H 1 Hcore S L

5 rU{c C (SST 2 T ) 1 L C (q 2 q )}, (2)p h A y e SST A

where r is the density of air, TA is the air temperatureat 10 m, cp is the specific heat of air at constant pressure,and Ly is the latent heat of vaporization at a given TA.Here, U represents the 1-min wind speed at 10 m, whileqSST and qA are the saturation mixing ratio at the SSTand the actual mixing ratio of the air at 10 m, respec-tively; Ch and Ce are the dimensionless coefficients ofheat exchange and moisture exchange at 10 m; Hcore hasunits of W m22 (51027 kJ cm22 s21). Then QHpext andQHputil are determined from Hcore by

Q 5 H TC , (3)Hpext core transit-time

Q 5 Q /Q , (4)Hputil Hpext H

where TCtransit-time is the time in seconds for a storm totravel the inner-core diameter (120 km). To first order,

QHputil estimates how much of the upper-ocean heat con-tent available to the system is actually extracted (i.e.,‘‘utilized’’) by the storm. Similar to QH, estimates forQHpext are given in kJ cm22. For this particular analysis,Hcore serves as a proxy for storm intensity since it hasa strong wind speed dependence.

Recent findings from Cione et al. (2000) helped betterdefine near-surface atmospheric thermodynamic con-ditions typically observed within the hurricane inner-core environment. Using these estimates and assuminga fairly typical tropical Atlantic summer season QH val-ue of 75 kJ cm22 (57.5 3 108 J m22), estimates forQHputil were constructed.2 A primary objective of thisstudy is to establish a reasonable ‘‘parameter space’’ forQHputil within the well-mixed hurricane inner core overa wide range of possible storm speeds and intensities.Upper-ocean energy utilization as a function of stormspeed and total surface enthalpy flux is illustrated inFig. 5. Figure 5 includes a wide array of potential stormspeeds (2.5–10 m s21) and inner-core surface flux val-ues/intensities (650–2600 W m22). Applying bulk aero-

2 The spatial and seasonal variability of QH for the North Atlanticcan be found at http://www.aoml.noaa.gov/phod/cyclone/data/2002/map.html.


TABLE 3. Statistical summary of along-track SST transects (sorted by DSSTIC). Bold values indicate statistical significance at the 95%level (or higher) between upper and lower 50th percentile means. The events that cooled the most (least) are in the upper (lower) 50thpercentile sample.

Statisticalsummary

DSSTIC

(8C)DSSTICW

(8C)SSTA

(8C)SSTIC

(8C)SSTICW

(8C)RDA

(8lat)RDIC

(8lat)RDICW

(8lat)TCLAT

(8N)TCIWIND

(ms21)TCIBAR

(mb)TCSPEED

(ms21)

All transectsMinMaxMeanMedianStd devCount

21.80.0

20.720.6

0.533

23.40.0

2120.8

0.826

26.730.228.9290.9

33

2629.928.228.6

133

24.729.827.928.1

1.226

2.04.52.82.60.9

33

00.60.30.20.2

33

0.11.91.21.30.6

26

14.135.925.825.8

4.733

29.069.643.741.210.333

9221000972.497619.233

1.210.8

5.35.12.4

33

Upper 50th percentileMinMaxMeanMedianStd devCountSig level %

21.820.72120.9

0.416

N/A

23.420.921.821.5

0.710

N/A

26.730.228.828.8

1.116

2629.527.827.8

1.11699

24.728.627.026.8

1.31099

2.04.53.43.50.9

1699

00.60.30.20.2

16

0.21.91.31.30.6

10

23.535.927.325.6

3.61696

29.169.646.546.310.016

9221000968.696519.016

1.810.8

5.24.02.9

16

Lower 50th percentileMinMaxMeanMedianStd devCount

20.60.0

20.320.3

0.217

21.40.0

20.520.5

0.316

27.130.029.029.1

0.817

27.129.928.728.7

0.717

26.929.828.528.5

0.816

2.042.32.00.7

17

0.10.60.30.20.2

17

0.11.91.21.30.6

16

14.135.224.324.8

5.217

29.159.341.236.110.217

935993976.198319.217

1.27.85.45.71.8

15

dynamic formulas [Eq. (2)] with exchange coefficientsas determined by Garratt (1977), composite analysesfrom Cione et al. (2000) showed that the average inner-core total surface enthalpy flux was approximately 1300W m22 for a category 1 hurricane (i.e., maximum sur-face wind speed between 33 and 43 m s21). The fullrange of energy utilization for this analysis is estimatedto be between 1.0% (for a tropical storm moving at 10m s21) and 16.6% (major hurricane moving at 2.5 ms21). It should be noted that these energy utilizationestimates assume that the upper-ocean profile never en-ters the storm’s eye. If the profile were to temporarilyexperience reduced surface winds within the eye, theenergy utilization estimates presented in Fig. 5 (andestimated above) would be reduced.

Even though these findings do not take into accountall the physical processes and/or situations that couldpotentially come into play (such as a stationary systemor warm/cold water advection near a highly baroclinicocean front), the results nevertheless illustrate the vastenergy resources available to most tropical cyclones un-der most storm conditions. These results suggest thatfor the large majority of propagating systems, the mag-nitude of upper-ocean heat content (QH) should not bea limiting factor affecting storm maintenance and/or in-tensification.

6. The impact of SST change on inner-coresurface enthalpy flux

Results previously illustrated in Table 3 depict anaverage difference in inner-core SST of 0.78C between

the upper and lower 50th percentile samples. This rel-atively small difference in SST can potentially have asignificant impact on the resulting surface enthalpy fluxto the storm within the high-wind inner core. Changesto qA, U, and/or TA (or modifications to the exchangecoefficient expressions) will also significantly impactthe magnitude of Hcore. However, a primary objectiveof this research is to isolate the impact that SST-de-pendent variables (SST and qSST) potentially have onthe storm’s ability to extract energy from the inner-coreupper-ocean environment.

Figure 6 illustrates the percent change in upper-ocean energy extracted by the storm (DQHpext ) as afunction of inner-core SST change [where non-SST-dependent variables in Eq. (2) remain constant]. Theinitial values for SSTIC, storm speed (TCspeed ), surfaceair temperature (TA), relative humidity (RH), surfacewind speed (U ), and minimum sea level pressure (P)are shown at the top of the illustration and representbulk mean values obtained from Cione et al. (2000).Changes in inner-core total surface heat flux (relativeto the initial 1300 W m22) are also illustrated withinthe body of the figure.

Figure 6 shows that, all other factors being equal,relatively small variations in inner-core SST can dra-matically impact inner-core surface heat flux and, assuch, the magnitude of upper-ocean energy extracted bythe storm (QHpext). Figure 6 illustrates that a 10.78Cdifference in inner-core SST (i.e., average DSSTIC be-tween upper and lower 50th percentile samples shownin Table 3) results in a 30% increase in the amount of


FIG. 5. The percentage of available upper-ocean energy extractedby the storm or ‘‘energy utilization’’ (QHputil) as a function of inner-core surface heat flux (W m22 5 1027 kJ cm22 s21), and storm speed(in m s21). In this illustration, total surface heat flux within the innercore was held constant at 1300 W m22 as storm speed varied from2.5 to 10 m s21. Similarly, storm speed was held constant at 5 m s21

as inner-core surface heat flux varied between 50% and 200% of theinitial 1300 W m22 value (i.e., 650–2600 W m22). The dashed linedenotes the 1300 W m22 surface enthalpy flux and 5 m s21 stormspeed values that were initially used. In all cases, hurricane heatpotential is 75 kJ cm21.

FIG. 6. The percent change in upper-ocean energy extracted by thestorm (DQHpext), as a function of inner-core SST change, DSSTIC (rel-ative to the initial inner-core SST of 288C). Changes in total surfaceheat flux (relative to the initial 1300 W m22 value) are also shownand range between 2839 and 1035 W m22. Values for storm speed(TCspeed), hurricane heat potential (QH), surface air temperature (TA),relative humidity (RH), surface wind speed (U), and minimum sealevel pressure (P) are given at the top of the illustration.

upper-ocean energy extracted by the storm (QHpext), fora category 1 hurricane. This represents a 26% increasein HL(270 W m22) and a 33% change in HS(80 W m22).On the other hand, it should be noted that total enthalpyflux changes of this magnitude (i.e., 350 W m22) wouldnot impact the total amount of upper-ocean heat content(QH) available to a propagating system by more than1%–2% (Fig. 5).

The large values of total surface enthalpy flux thatresult from relatively modest changes in inner-core SST,coupled with the realization that most hurricanes utilizeless than 10% of the upper-ocean energy available tothem, accentuates the need for a shift in focus fromanalyzing upper-ocean heat content to accurately ob-serving (and predicting) the short-term variability ofhurricane inner-core SST conditions.

7. Linkages between SST change and TC intensitychange

Earlier studies have had difficulty linking environ-mental SST ahead of the storm (SSTA) with subsequent

changes in TC intensity. Much of this inability can belinked to the fact that SST measurements, especiallysatellite-derived skin temperatures, do not adequatelydepict thermal conditions below the surface (Reynolds1988; Reynolds and Smith 1994). Results in Fig. 7a alsodepict this trend, illustrating little or no relationshipbetween SSTA and subsequent TC intensity change (de-fined in all cases as the 24-h change in maximum surfacewind speed centered at the time SSTIC was recorded).Figure 7b also depicts little or no relationship betweenhurricane heat potential (QH) ahead of the storm andintensity change. In contrast, the regression resultsshown in Figs. 7c and 7d illustrate clear relationshipsbetween SST change (DSSTIC and DSSTICW) and sub-sequent changes in storm intensity. The 4.4% and 4.2%explained variances (i.e., 100r2) shown in Figs. 7a and7b, increase to 33.4% and 42.1% in Figs. 7c and 7d,respectively, when DSSTIC and DSSTICW are used. Thismarked increase in explained variance suggests that nor-malized differences between SST within and ahead ofthe storm may, under certain circumstances, be closelytied to observed changes in hurricane intensity. The re-lationship between reduced inner-core SST cooling andsubsequent intensification is plausible since relativelysmall changes in inner-core SST can significantly altersurface energy fluxes within the high-wind hurricane


environment (Fig. 6). Storms experiencing reduced in-ner-core SST cooling would have larger surface fluxesand, as a result, would be more likely to experienceenhanced intensification (assuming all other factors po-tentially impacting hurricane intensity change to beequal).

This hypothesis suggesting a quantifiable relationshipbetween SST change and subsequent changes in TCintensity is tested. Table 4a is a statistical summary ofresults found when observations were sorted by TC in-tensity change (DTCI). Similar to the definition used inFigs. 7a–d, TC intensity change is defined as the best-track-derived (Neumann et al. 1993) 24-h change inmaximum surface wind speed centered at the time SSTIC

was obtained. Since the average transit time betweenSSTICW and SSTA was found to be ;23 h and closelymatched the 24-h period of intensification used in thisanalysis (vs. ;13-h transit time between SSTIC andSSTA), only intensity change events that also includedcorresponding DSSTICW values were used. Results from24 DTCI events are included in Table 4a. In additionto statistics on DTCI, Table 4a includes upper and lower50th percentile summary statistics for many of the var-iables listed in Table 3 including SSTA, SSTIC, SSTICW,DSSTIC, DSSTICW, TCLAT, TCIWIND, and TCSPEED. Table4a also includes satellite-derived estimates for upper-ocean heat content ahead of the storm (QH) as well asthe following synoptic-scale atmospheric parameters:200-hPa level zonal wind (U 200), 200-hPa air temper-ature (T 200), 850–200-hPa wind shear (S 850-200), 200-hPa divergence (D 200), and 200-hPa eddy flux of an-gular momentum (E 200). These atmospheric parame-ters, in addition to the weekly averaged climatologicalSST (SSTA-CLIM), were obtained from the StatisticalHurricane Intensity Prediction Scheme (SHIPS) data-base (DeMaria and Kaplan 1994, 1999) and were av-eraged over a TC-centered area of 500 km radius. Usinga standard Student’s t-test, statistically significant dif-ferences between upper and lower 50th percentile meanswere found at the 95% level (or higher) for DSSTIC andDSSTICW (Table 4a, bold values). For the storms in-cluded in this research, inner-core SST change is linkedto TC intensity change. Statistically significant differ-ences between upper and lower 50th percentile meanswere not found for any other variable shown in Table4a. However, while the findings are not statistically sig-nificant, results illustrated in Table 4a suggest thatquick-moving, low-latitude, relatively weak storms maybe more likely to intensify when compared to slow-moving, high-latitude, strong systems. Similar (nonsta-tistically significant) trends were also found in Table 3between SST change and storm latitude, storm speed,and initial intensity.

It has been well documented that the magnitude ofatmospheric shear can dramatically impact the intensityof tropical systems (Gray 1968; Merrill 1988). However,for the sample of storms used in this research, the dif-ference in shear (S850-200) between the group that no-

ticeably intensified (upper 50th percentile) and the groupthat did not (lower 50th percentile) was not found to bestatistically significant. It is probable that a number ofphysical processes impacted the rate of intensity changefor the 24 events included in Table 4a. Nevertheless,the results presented in this analysis strongly suggestthat the magnitude of inner-core SST change can, undercertain circumstances, significantly impact the physicalprocesses controlling storm maintenance and intensitychange.

8. Maximum potential intensity and TC intensitychange

Table 4b is a statistical summary of maximum po-tential intensity (MPI) as a function of TC intensitychange (DTCI). The SST-dependent MPI formulationused in this analysis was obtained from DeMaria andKaplan (1994) and is given by

C(T2T )0MPI 5 A 1 Be , (5)

where T is the SST, T0 is a specified reference temper-ature, and A, B, and C are constants. Letting T0 5 30.08Cand using a least squares fit, DeMaria and Kaplan (1994)determined the constants to be A 5 34.21, B 5 55.80,and C 5 0.1813 after accounting for storm translationspeed. Potential intensity (POTA) is also illustrated inTable 4b and is defined as the difference between MPIand the observed TC intensity, prior to intensification.An additional parameter listed in Table 4b is DMPIA-ICW.It is defined as the difference in MPI using SSTICW andSSTA. All potential intensity terms listed in Table 4b aregiven in m s21.

Similar to results illustrated in Table 4a for DSSTIC

and DSSTICW, statistically significant differences be-tween upper and lower 50th percentile means forDMPIIC and DMPIICW were found. (These results are tobe expected since MPI is a function of SST.) Compar-isons between upper and lower 50th percentile DMPImeans (given in m s21) and observed changes in hur-ricane intensity between the two groups (also in m s21)were conducted. As was done previously in the DSSTanalysis, only DMPIICW values were used since mean23-h TC transit times from SSTICW to SSTA more closelyapproximate the 24-h period of intensity change utilizedin this study (relative to the ;13-h transit time betweenSSTIC and SSTA). The statistical summary presented inTable 4b illustrates that average MPIICW reductions re-sulting from DSSTICW ranged from 24.1 m s21 (for theupper 50th percentile) to 29.7 m s21 (for the lower 50thpercentile). These results demonstrate the capacity ofthe upper ocean to potentially limit the magnitude ofTC intensification while highlighting the variable natureof this ‘‘braking’’ process. The 5.6 m s21 difference inDMPI found between the upper and lower sample meansrepresents 46% of the total 12.2 m s21 difference in TCintensity change found between the upper and lower50th percentile means shown in Table 4b (i.e., 14.5


FIG. 7. (a) Scatterplot of TC intensity change (DTCI) as a function of SSTA. The resultinglinear best-fit equation is illustrated. For this fit, 4.4% of the variance is explained (i.e., r2 50.044). The units for DTCI are m s21 24 h21. SST is measured in 8C. (b) As in (a) except thathurricane heat potential (QH) ahead of the storm (kJ cm22) was used instead of SSTA. Here, 4.2%of the variance is explained. (c) As in (a) except inner-core SST minus SSTA(DSSTIC) was usedinstead of SSTA. 33.4% of the variance is explained. (d) As in (a) except inner-core wake SSTminus SSTA(DSSTICW) was used instead of SSTA. Here, 42.1% of the variance is explained.

m s21–2.3 m s21). These results suggest that the ‘‘brake’’on TC intensification may have been more ‘‘readily en-gaged’’ for the group of storms that intensified the least(i.e., lower 50th percentile sample).

These results also suggest that ambient MPI estimatesmay not always give an accurate account of MPI con-ditions found within the important hurricane inner-coreenvironment. Results illustrated in Table 4b show thatMPIIC values were on average relatively close to MPIA

values for the events that significantly intensified (i.e.,the upper 50% sample). However, by simply looking atmean MPIA, one might have expected to find no sig-nificant difference in intensification between the upperand lower 50% samples given the fact that average ini-tial MPIA values were quite similar for both groups.These findings highlight the importance of obtainingaccurate observations of SST within the hurricane inner

core. In many cases MPIA does not give an accuratemeasure of how close a system is to its maximum in-tensity where it matters most, within the hurricane high-wind environment. This is a significant point since po-tential intensity (i.e., MPI minus initial storm intensity)is an important and proven predictor used in the SHIPSforecast model (DeMaria and Kaplan 1994, 1999).

9. Summary

It is well accepted in the operational and researchcommunities that the upper ocean can have a significantimpact on maintaining and/or modifying TC structureand intensity. However, exactly how and to what extentvariations in upper-ocean thermal structure directly im-pact local convective tendencies, overall TC structure,and ultimately, changes in storm intensity is still not


TA

BL

E4a

.S

tati

stic

alsu

mm

ary

fact

ors

link

edto

trop

ical

cycl

one

inte

nsit

ych

ange

(sor

ted

byD

TC

I).

Bol

dva

lues

indi

cate

stat

isti

cal

sign

ifica

nce

at(o

rab

ove)

the

95%

leve

lbe

twee

nup

per

and

low

er50

thpe

rcen

tile

mea

ns.

The

even

tsth

atin

tens

ified

the

mos

t(l

east

)ar

ein

the

uppe

r(l

ower

)50

thpe

rcen

tile

sam

ple.

Sum

mar

y

DT

CI

(ms2

1

day2

1)

SS

TA

(8C

)S

ST

IC

(8C

)S

ST

ICW

(8C

)D

SS

TIC

(8C

)D

SS

TIC

W

(8C

)

OH

AH

EA

D(k

Jcm

23)

TC

LA

T

(8N

)T

CI W

IND

(ms2

1)

TC

SP

EE

D

(ms2

1)

SS

TA

SH

IPS

(8C

)

U20

0

SH

IPS

(ms2

1)

T20

0

SH

IPS

(8C

)

S85

0-20

0

SH

IPS

(ms2

1)

D20

0

SH

IPS

(s2

110

27)

E20

0S

HIP

S(m

s21

day2

1)

All

data

Min

Max

Mea

nM

edia

nS

tdde

vC

ount

28.

922

.3 7.9

6.7

7.5

24

27.1

30.0

29.0

29.2 0.8

24

26.1

29.9

28.3

28.6 1.0

24

24.7

29.8

28.0

28.2 1.2

24

21.

80.

02

0.6

20.

60.

524

23.

40.

02

1.0

20.

80.

824

6.8

135.

072

.476

.037

.621

14.1

35.9

25.4

25.3 5.2

24

29.0

59.3

44.8

46.3 9.3

24

2.2

8.2

5.3

5.4

1.8

24

27.0

30.3

28.9

29.2 1.0

19

27.

811

.3 0.6

20.

65.

619

255

.82

50.7

254

.32

54.8 1.5

19

0.8

13.0 5.6

4.8

3.2

19

7.5

78.0

31.9

31.0

18.8

19

24.

010

.0 2.4

1.0

3.7

19

Upp

er50

thpe

rcen

tile

Min

Max

Mea

nM

edia

nS

tdde

vC

ount

Sig

leve

l%

7.6

22.3

14.5

15.6 4.2

11 N/A

27.1

30.0

28.9

29.0 1.0

11

27.1

29.9

28.6

28.6 0.9

11

26.9

29.8

28.4

28.2 0.9

11

20.

60.

02

0.3

20.

30.

211 99

21.

40.

02

0.5

20.

40.

411 99

20.0

135.

077

.581

.033

.411

14.1

35.2

23.5

24.2 5.9

11

29.0

56.7

42.1

41.2

10.5

11

3.9

7.8

5.7

5.7

1.2

11

27.0

30.0

28.8

28.9 1.2

8

27.

87.

22

0.1

22.

95.

98

255

.82

51.9

254

.62

55.6 1.4

8

0.8

13.0 5.0

2.9

3.8

8

10.0

78.0

38.1

32.5

19.4 8

0.0

6.5

2.0

0.5

2.3

8

Low

er50

thpe

rcen

tile

Min

Max

Mea

nM

edia

nS

tdde

vC

ount

28.

96.

72.

32.

74.

413

27.3

29.9

29.0

29.3 0.8

13

26.1

29.3

28.1

28.6 1.0

13

24.7

29.2

27.7

28.0 1.4

13

21.

80.

02

0.9

20.

80.

613

23.

42

0.1

21.

42

1.4

0.9

13

6.8

110.

666

.768

.549

.110

20.1

36.5

27.0

26.0 4.2

13

31.3

59.3

47.1

46.3 7.8

13

2.2

8.2

4.9

4.8

2.1

13

27.4

30.3

29.0

29.2 0.8

11

27.

511

.3 1.1

22.

05.

611

255

.72

50.7

254

.12

54.3 1.6

11

1.5

12.2 6.0

5.4

2.8

11

7.5

58.0

26.7

18.0

17.3

11

24.

010

.0 2.6

1.0

4.6

11


TABLE 4b. Statistical summary of maximum potential intensity and potential intensity (sorted by DTCI). Bold values indicate statisticalsignificance at (or above) the 95% level between upper and lower 50th percentile means. The events that intensified the most (least) are inthe upper (lower) 50th percentile sample.

Summary

DTCI(m s21

day21)POTA

(m s21)MPIA

(m s21)MPIIC

(m s21)MPIICW

(m s21)DMPIIC

(m s21)DMPIICW

(m s21)

DTCI SHIPS(m s21

day21)

POTA

SHIPS(m s21)

MPIA

SHIPS(m s21)

All dataMinMaxMeanMedianStd devCount

28.922.3

7.96.77.5

24

10.561.136.138.412.624

67.290.180.982.5

6.624

61.789.076.077.5

6.924

55.688.073.774.5

7.924

215.30.0

24.924.4

3.924

221.20.0

27.226.0

3.924

28.922.3

7.56.77.7

19

9.961.136.336.213.919

66.693.180.782.5

7.819

Upper 50th percentileMinMaxMeanMedianStd devCountSig level %

7.622.314.515.6

4.211

N/A

10.561.138.031.114.511

67.290.180.180.8

7.411

67.289.077.677.5

6.811

66.088.076.074.5

6.811

25.20.0

22.622.6

1.91199

211.20.0

24.123.8

3.11199

7.622.314.414.5

5.08

N/A

9.961.137.531.215.8

8

66.690.179.780.0

9.18

Lower 50th percentileMinMaxMeanMedianStd devCount

28.96.72.32.74.4

13

14.352.134.538.011.213

68.489.081.583.4

6.113

61.783.474.777.5

7.113

55.684.571.873.0

8.513

215.30.0

26.826.7

4.113

221.220.929.7

210.15.6

13

28.96.72.42.74.7

13

18.853.035.436.213.111

68.893.181.582.5

7.011

well understood. This is partly attributed to the fact thatongoing upper-ocean and atmospheric modeling effortsare still lacking in several key areas (e.g., insufficienthorizontal/vertical grid resolution, flawed/incompleteparameterization schemes, incomplete/crude physicalrepresentations of the atmospheric and oceanic bound-ary layers, etc). A likely explanation as to why changesin upper-ocean thermal structure have never been di-rectly and quantitatively linked to changes in storm in-tensity is due to a limited number of observations. Bothfrom an oceanic and atmospheric standpoint, the innercore is the most difficult region to routinely and accu-rately observe within the hurricane environment. Issuessuch as nearly continuous cloud cover, 10–20-m oceanwaves, and wind speeds in excess of 50 m s21 makethis region difficult for in situ platforms to survive, dan-gerous to traverse/circumnavigate, and nearly impos-sible for remote satellites to fully document. As a result,high-resolution, accurate depictions of the atmosphericand oceanic boundary layers within the TC inner corewere rare prior to 1997. Since 1997, however, the useof highly durable and accurate GPS dropwindsondes andavailable AXBTs has enabled NOAA’s Hurricane Re-search Division to penetrate and observe the TC innerand outer core boundary layer environments in severalAtlantic hurricanes. By using the AXBT data from thesestorms in conjunction with archived SST data obtainedduring several ‘‘TC–buoy encounters,’’ the storm-to-storm variability associated with differences in SST be-

tween the hurricane inner core and the ocean environ-ment ahead of the storm have been well documented.

Results from this research suggest that differencesbetween inner-core and ambient SST are significantlyless than horizontal SST changes typically observed inthe post storm, TC cold wake environment (i.e., ;08–28C vs. 48–58C). This finding should prove useful tomodeling efforts attempting to verify the critically im-portant (but infrequently observed) inner-core SST/mixed layer temperature. Estimates of upper-ocean heatcontent, energy extracted by the storm, and energy uti-lization were made. Findings from this analysis suggestthat under most conditions, the upper-ocean heat contentis an order of magnitude or more greater than the energyextracted by the storm. Results also show that relativelymodest changes in inner-core SST can dramatically alterair–sea fluxes within the high-wind inner-core stormenvironment. Initial estimates show that SST changeson the order of 18C lead to surface enthalpy flux changesof 40% or more.

For the subset of observations used in this study, itwas shown that the magnitude of SST change (innercore minus ambient) was statistically linked to changesin TC intensity. These results suggest that storms ex-periencing reduced levels of inner-core SST coolinglikely experience an increase in surface enthalpy flux,and as a result, are more likely to intensify. AmbientSST and upper-ocean heat content ahead of the stormwere not associated with observed changes in storm


intensity. Besides SST change, no other variable wasstatistically linked to changes in storm intensity (for the24-event sample used in this study). Tropical cycloneintensity change is a complex, nonlinear process thatoften involves several competing or synergistic factors.Nevertheless, the results presented in this researchstrongly suggest that the (often ‘‘unseen’’ and numeri-cally unaccounted for) variability associated with SSTcooling within the hurricane inner core can, under cer-tain circumstances, significantly impact the physicalprocesses controlling storm maintenance and intensitychange.

Acknowledgments. The authors wish to thank JohnKaplan (AOML/HRD) for providing SHIPS data andDr. Gustavo Goni (AOML/PhoD) for computing satel-lite-derived upper-ocean heat content estimates. Wethank Dr. Frank Marks and Dr. Chris Landsea (AOML/HRD), Dr. Gary Barnes (University of Hawaii), and twoanonymous editors for their thoughtful insights and re-view of the initial manuscript. The authors would alsolike to thank Joseph Cione Sr. for his editorial assistance.Finally, the authors wish to acknowledge the employeesat the NOAA Aircraft Operations Center. Without theirexpertise and assistance this research would not havebeen possible.

REFERENCES

Bender, M. A., I. Ginnis, and Y. Kurihara, 1993: Numerical simulationof tropical cyclone–ocean interaction with a high-resolutionmodel version 4 (MM4). J. Geophys. Res., 98, 23 245–23 263.

——, ——, and Y. Kurihara, 2000: Real case simulation of hurricane–ocean interaction using a high-resolution coupled model: Effectson hurricane intensity. Mon. Wea. Rev., 128, 917–946.

Black, P. G., 1983: Ocean temperature changes induced by tropicalcyclones. Ph. D. dissertation, The Pennsylvania State University,278 pp.

——, and G. J. Holland, 1995: The boundary layer of Tropical Cy-clone Kerry (1979). Mon. Wea. Rev., 123, 2007–2028.

——, R. L. Elsberry, L. K. Shay, R. P. Partridge, and J. F. Hawkins,1988: Atmospheric and oceanic mixed layer observations in Hur-ricane Josephine obtained from air-deployed drifting buoys andresearch aircraft. J. Atmos. Oceanic Technol., 5, 683–698.

Bosart, L. F., W. E. Bracken, J. Molinari, C. S. Velden, and P. G.Black, 2000: Environmental influences on the rapid intensifi-cation of Hurricane Opal (1995) over the Gulf of Mexico. Mon.Wea. Rev., 128, 322–352.

Byers, H. R., 1944: General Meteorology. McGraw-Hill, 645 pp.Chan, J. C., Y. Duan, and L. K. Shay, 2001: Tropical cyclone intensity

change from a simple ocean–atmosphere coupled model. J. At-mos. Sci., 58, 154–172.

Chang, S. W., and R. A. Anthes, 1978: Numerical simulations of theocean’s nonlinear baroclinic response to translating hurricanes.J. Phys. Oceanogr., 8, 324–336.

Cione, J. J., P. G. Black, and S. H. Houston, 2000: Surface obser-vations in the hurricane environment. Mon. Wea. Rev., 128,1550–1568.

DeMaria, M., and J. D. Pickle, 1988: A simplified system of equationsfor the simulation of tropical cyclone. J. Atmos. Sci., 45, 1542–1554.

——, and J. Kaplan, 1994: A Statistical Hurricane Intensity PredictionScheme (SHIPS) for the Atlantic Basin. Wea. Forecasting, 9,209–220.

——, ——, 1999: An updated Statistical Hurricane Intensity Predic-tion Scheme (SHIPS) for the Atlantic and eastern North Pacificbasins. Wea. Forecasting, 14, 326–337.

Elsberry, R. L., T. Fraim, and R. Trapnell, 1976: A mixed layer modelof the ocean thermal response to hurricanes. J. Geophys. Res.,81, 1153–1162.

Emanuel, K. A., 1986: An air–sea interaction theory for tropicalcyclones. Part I: Steady-state maintenance. J. Atmos. Sci., 43,585–604.

——, 1988: The maximum intensity of hurricanes. J. Atmos. Sci.,45, 1143–1155.

Federov, K. N., A. A. Varfolomeyev, A. J. Ginzberg, A. G. Zatsepin,A. Yu. Krasnopevtsev, A. G. Ostrovskiy, and V. Ye. Sklyarov,1977: Thermal response of the ocean to the passage of HurricaneElla. Oceanology, 19, 656–661.

Garratt, J. R., 1977: Review of drag coefficients over oceans andcontinents. Mon. Wea. Rev., 105, 915–929.

Gilhousen, D. B., 1988: Quality control of meteorological data fromautomated marine stations. Preprints, Fourth Int. Conf. on In-teractive Information and Processing Systems for Meteorology,Oceanography, and Hydrology, Miami, FL, Amer. Meteor. Soc.,113–117.

——, 1998: Improved real-time quality control of NDBC measure-ments. Preprints, 10th Symp. on Meteorological Observationsand Instrumentation, Phoenix, AZ, Amer. Meteor. Soc., 363–366.

Gray, W. M., 1968: Global view of the origin of tropical disturbancesand storms. Mon. Wea. Rev., 96, 669–700.

——, 1979: Hurricanes: Their formation, structure, and likely rolein the tropical circulation. Meteorology over the TropicalOceans, D. B. Shaw, Ed., Royal Meteorological Society, 155–218.

Holland, G. J. 1997: The maximum potential intensity of tropicalcyclones. J. Atmos. Sci., 54, 2519–2541.

——, R. T. Merrill, 1984: On the dynamics of tropical cyclone struc-ture changes. Quart. J. Roy. Meteor. Soc., 110, 723–745.

Kaplan, J., and M. DeMaria, 1999: Climatological and synoptic char-acteristics of rapidly intensifying tropical cyclones in the NorthAtlantic basin. Preprints, 23d Conf. on Hurricanes and TropicalMeteorology, Dallas, TX, Amer. Meteor. Soc., 592–597.

Leiper, D., and D. Volgenau, 1972: Hurricane heat potential of theGulf of Mexico. J. Phys. Oceanogr., 2, 218–224.

Malkus, J. S., and H. Riehl, 1960: On the dynamics and energytransformations in steady-state hurricanes. Tellus, 12, 1–20.

Mayer, D. H., H. O. Mofjeld, and K. D. Leaman, 1981: Near-inertialwaves on the outer shelf in the Middle Atlantic Bight in thewake of Hurricane Belle. J. Phys. Oceanogr., 11, 86–106.

Merrill, R. T., 1988: Environmental influences on hurricane intensi-fication. J. Atmos. Sci., 45, 1678–1687.

Miller, B. I., 1958: On the maximum intensity of hurricanes. J. Me-teor., 15, 184–195.

Molinari, J., and D. Vollaro, 1990: External influences on hurricaneintensity. Part II: Vertical structure and response of the hurricanevortex. J. Atmos. Sci., 47, 1902–1918.

Neumann, C. J., B. R. Jarvinen, C. J. McAdie, and J. D. Elms, 1993:Tropical Cyclones of the North Atlantic Ocean, 1871–1992. His-torical Climate Series, Vol. 6-2, National Climatic Data Center,193 pp.

Palmen, E., 1948: On the formation and structure of tropical cyclones.Geophysica, 3, 26–38.

Pudov, V. D., 1980: Mesostructure of the temperature and currentvelocity fields of a baroclinic ocean layer in the wake of typhoonVirginia. Oceanology, 20, 141–146.

——, A. A. Varfolomeyev, and K. N. Federov, 1978: Vertical struc-ture of the wake of a typhoon in the upper-ocean. Oceanology,18, 141–146.

Price, J. F., 1981: Upper-ocean response to a hurricane. J. Phys.Oceanogr., 11, 153–175.

Reynolds, R. W., 1988: A real-time global sea surface temperatureanalysis. J. Climate, 1, 75–86.


——, and T. M. Smith, 1994: Improved global sea surface temperatureanalyses using optimum interpolation. J. Climate., 7, 929–948.

Riehl, H. 1948: On the formation of west Atlantic hurricanes. Studiesof upper-air conditions in low latitudes. Rep. 24, University ofChicago, 67 pp. [Available from University of Chicago, Chicago,IL 60637.]

——, 1950: A model of hurricane formation. J. Appl. Phys., 21, 917–925.

——, 1954: Tropical Meteorology. McGraw-Hill, 392 pp.Sadler, J. C., 1978: Mid-season typhoon development and intensity

changes and the tropical upper tropospheric trough. Mon. Wea.Rev., 106, 1137–1152.

Shay, L. K., P. G. Black, A. J. Mariano, J. D. Hawkins, and R. L.Elsberry, 1992: Upper-ocean response to Hurricane Gilbert. J.Geophys. Res., 97, 20 227–20 248.

——, G. J. Goni, and P. G. Black, 2000: Role of a warm ocean featureon Hurricane Opal. Mon. Wea. Rev., 128, 1366–1383.

Willoughby, H. E., and P. G. Black, 1996: Hurricane Andrew inFlorida: Dynamics of a disaster. Bull. Amer. Meteor. Soc., 77,543–549.

sea surface temperature variability in hurricanes: implications...

Documents