High-Resolution Satellite Surface Latent Heat Fluxes in North Atlantic Hurricanes Jiping Liu, Judith A

Florida State University Libraries

Faculty Publications Center for Ocean-Atmospheric Prediction Studies (COAPS)

2011 High-resolution satellite surface latent heat fluxes in North Atlantic hurricanes Jiping Liu, Judith A. Curry, Carol Anne Clayson, and Mark Bourassa

Follow this and additional works at the FSU Digital Library. For more information, please contact [email protected] SEPTEMBER 2011 LIUETAL. 2735

High-Resolution Satellite Surface Latent Heat Fluxes in North Atlantic Hurricanes

JIPING LIU AND JUDITH A. CURRY School of Earth and Atmospheric Sciences, Georgia Institute of Technology, Atlanta, Georgia

CAROL ANNE CLAYSON Department of Earth, Ocean, and Atmospheric Science, The Florida State University, Tallahassee, Florida

MARK A. BOURASSA Department of Earth, Ocean, and Atmospheric Science, and Center for Ocean–Atmospheric Prediction Studies, The Florida State University, Tallahassee, Florida

(Manuscript received 24 June 2010, in ﬁnal form 6 March 2011)

ABSTRACT

This study presents a new high-resolution satellite-derived ocean surface flux product, XSeaFlux, which is evaluated for its potential use in hurricane studies. The XSeaFlux employs new satellite datasets using improved retrieval methods, and uses a new bulk flux algorithm formulated for high wind conditions. The XSeaFlux latent heat flux (LHF) performs much better than the existing numerical weather prediction reanalysis and satellite-derived flux products in a comparison with measurements from the Coupled Boundary Layer Air–Sea Transfer (CBLAST) field experiment. Also, the XSeaFlux shows well-organized LHF structure and large LHF values in response to hurricane conditions relative to the other flux products. The XSeaFlux dataset is used to interpret details of the ocean surface LHF for selected North Atlantic hurricanes. Analysis of the XSeaFlux dataset suggests that ocean waves, sea spray, and cold wake have substantial impacts on LHF associated with the hurricanes.

2 1. Introduction exceed 1000 W m 2, which is an order of magnitude larger than the summertime climatological latent heat The thermodynamic disequilibrium between the trop- flux (LHF) values (e.g., Trenberth and Fasullo 2007). To ical atmosphere and ocean provides an energy source for understand the dynamics of individual hurricanes as well tropical cyclones (e.g., Kleinschmidt 1951; Emanuel 1986), as the climate dynamics of hurricanes requires observa- arising primarily from the undersaturation of near-surface tions of the surface LHF on a scale of tens of kilometers air. The dependence of the air–sea energy transfer rate on and hours; regardless of the source of water vapor being wind has been hypothesized to be the principal feedback the LHF from below the storm or an influx of midlevel mechanism that allows hurricanes to develop (e.g., water vapor. Emanuel 2003). Hence, accurate knowledge of the la- The surface LHF in hurricanes is difficult to measure tent heat flux across the air–sea interface is required for and direct observations are rarely available outside of hurricane predictions, whether using a simple diagnostic a few dedicated field campaigns [e.g., the Coupled approach (e.g., Emanuel 1995) or a fully coupled dy- Boundary Layer Air–Sea Transfer (CBLAST; Black namical approach (e.g., Chen et al. 2007). The strong et al. 2007)]. In the absence of direct measurements, the surface winds associated with hurricanes increase surface LHF can be estimated using a bulk flux algorithm of the evaporation so that the latent heat loss by the ocean can following form:

LHF 5 r L C U(Q 2 Q ), (1) Corresponding author address: Jiping Liu, School of Earth and a y E s a Atmospheric Sciences, Georgia Institute of Technology, Atlanta, GA 30332-0340. where ra is the density of air, Ly is the latent heat of E-mail: [email protected] vaporization, CE is the turbulent exchange coefﬁcient, U

DOI: 10.1175/2011MWR3548.1

Ó 2011 American Meteorological Society 2736 MONTHLY WEATHER REVIEW VOLUME 139 is the near-surface wind speed, Qs is 98% of the satu- center. The European Centre for Medium-Range Weather rated speciﬁc humidity at the sea surface temperature Forecasts Interim Reanalysis (ERA-Interim; Simmons

(SST), and Qa is the near-surface air specific humidity. et al. 2007) also shows a very weak LHF response to the Surface analyses from numerical weather prediction hurricanes (Manning and Hart 2007). (NWP) models provide one source of ocean surface tur- In this paper, we use a new high-resolution satellite- bulent fluxes, which are calculated from (1) using model- derived ocean surface turbulent flux dataset (called predicted variables. However, there are a number of XSeaFlux) that is designed to resolve the extreme surface established problems with NWP fluxes, including vari- latent heat flux in hurricanes. Recent new satellite-derived ables that are more dependent on physical parameteri- surface and near-surface parameters with improved re- zations used in the particular model and not strongly trieval methods (see section 2 for details) are applied in constrained by observations. Satellite-derived observa- this study, in addition to a new bulk flux model that in- tions of ocean surface turbulent fluxes (for a summary see cludes careful consideration of the flux transfer at high Curry et al. 2004) provide an alternative to the NWP winds (see section 2 for details). The XSeaFlux LHF is fluxes, but uncertainties in the retrievals of near-surface evaluated and interpreted using field observations. Ap- values of air humidity and temperature continue to pla- plications of the XSeaFlux data are made to interpret gue the flux calculations. Current ocean surface flux details of ocean surface latent heat fluxes for Hurricanes products derived from satellites and NWP reanalyses Fabian and Isabel. typically have resolutions of ;100 km (or coarser) and a temporal resolution of 6 h to 1 day. Particular difficul- ties in calculating the LHF in hurricanes from bulk flux 2. Data and methods models in the form of (1) include determination of the The datasets employed in this study include the fol- turbulent exchange coefficient under conditions of high lowing: winds and a highly disturbed sea state and understanding the effects of sea spray. Surface momentum exchange co- d the established ocean surface LHF datasets, includ- efficients increase with wind speed and are enhanced by ing the NCEP2 (NWP reanalysis; Kanamitsu et al. fetch-limited waves or opposing swell, but level off or de- 2002), the OAFlux (a blended product using in situ ob- 2 crease above an extreme high wind threshold (.45 m s 1; servations, NWP reanalysis, and satellite product; Yu e.g., Powell et al. 2003; Black et al. 2007). and Weller 2007), the HOAPS3 (satellite-derived prod- An example of the deficiencies in current NWP and uct; Andersson et al. 2007), and the ERA-Interim (NWP satellite-derived flux products is illustrated in Fig. 1 by the reanalysis; Simmons et al. 2007); spatial distribution of the LHF for Hurricanes Fabian d the new satellite-derived XSeaFlux dataset that pro- (1800 UTC 2 September 2003) and Isabel (1800 UTC vides ocean surface LHF as well as surface and near- 13 September 2003). Compared to the National Oceanic surface parameters on a spatial scale of 0.258 and and Atmospheric Administration-12 (NOAA-12) Advanced a temporal scale of 6 h; Very High Resolution Radiometer (AVHRR) images of d in situ airborne observations of the LHF from the Hurricanes Fabian (2000 UTC 2 September) and Isabel CBLAST field experiment (Drennan et al. 2007) for (2100 UTC 13 September), none of the available NWP Hurricanes Fabian and Isabel (2003); and satellite-derived flux products reproduce a well- d North Atlantic hurricane data obtained from the organized and distinct structure of the LHF associated Extended Best Tracks dataset (Demuth et al. 2006). with the hurricanes. Specifically, the National Centers The satellite-derived XSeaFlux LHF has been de- for Environmental Prediction reanalysis II (NCEP2; veloped to take advantage of new developments of the Kanamitsu et al. 2002) identifies some LHF signals in SEAFLUX project (e.g., Curry et al. 2004). The high- response to the hurricanes, but is located too far north resolution XSeaFlux dataset (0.258 and 6-hourly) used for Hurricane Fabian and too far northeast for Hurricane here incorporates the following satellite-derived input Isabel as compared to the observed position of the hurri- variables: canes. The Objectively Analyzed air–sea Fluxes (OAFlux; Yu and Weller 2007) does not have discernible LHF d SST: A new high-resolution SST analysis (version 2) has signals associated with the hurricanes. The Hamburg been developed based upon the operational SST prod- Ocean–Atmosphere Parameters and Fluxes version 3 uct produced by NOAA (Reynolds et al. 2007) using (HOAPS3; Andersson et al. 2007) seems to have the optimum interpolation (OI), which has a spatial reso- hurricane-related LHF variations at correct locations, lution of 0.258 and temporal resolution of 1 day. The but the magnitude of the LHF is weak, and the majority XSeaFlux product includes a parameterization and of the data is missing within ;100–200 km of the storm additional satellite products to allow for determination SEPTEMBER 2011 LIUETAL. 2737

22 FIG. 1. Spatial distribution of LHF (W m ) for (top) Hurricane Fabian (1800 UTC 2 Sep 2003) and (bottom) Hurricane Isabel (1800 UTC 13 Sep 2003) for the NWP and satellite-derived ﬂux products (NCEP2, OAFlux, HOAPS3, and ERA-Interim). The contours are the NCEP2 surface pressure. NOAA-12 AVHRR image of Hurricane Fabian (2000 UTC 2 Sep 2003) and Isabel (2100 UTC 13 Sep 2003) were obtained online at http://meto.umd.edu/;stevenb/hurr/03/03.html. 2738 MONTHLY WEATHER REVIEW VOLUME 139

22 FIG. 2. Time series of the LHF (W m ) for Hurricanes Fabian (2–4 Sep 2003) and Isabel (12–14 Sep 2003) from the CBLAST, and NWP and satellite-derived ﬂux products (NCEP2, OAFlux, HOAPS3, and ERA-Interim).

of the diurnal cycle of SST (following Clayson and water contents (Roberts et al. 2010). Comparisons of

Weitlich 2007) to produce a 6-hourly SST dataset. the Qa and Ta retrievals from this methodology to d U: A blended ocean surface wind speed (at 10 m above a database of over 200 000 ship and buoy observations sea level) analysis has been produced by the National demonstrated that the root-mean-square (rms) error 21 Climatic Data Center (Zhang et al. 2006), which has of this algorithm for Qa is found to be about 1.3 g kg 2 a spatial spacing of 0.258 and temporal spacing of 6 h. A accompanied with a small positive bias of 0.16 g kg 1,

spatial-temporally weighted interpolation is used to while for Ta the rms error is 1.38C with a bias of less integrate wind speed derived from multiple satellites, than 0.18C. The data have a spatial resolution of 0.258 including the Special Sensor Microwave Imager (SSM/ and temporal resolution of 6 h. For missing data I) on board the Defense Meteorological Satellite points, an optimal interpolation scheme is used in Program, the Tropical Rainfall Measuring Mission both space and time (e.g., Zhang et al. 2006). Microwave Imager (TMI), the Quick Scatterometer (QuikSCAT), and the Advanced Microwave Scanning The XSeaFlux LHF is calculated from the satellite-

Radiometer for Earth Observing System (AMSR-E). derived input values of SST, U, Qa,andTa using a bulk Note: wind speed for multiple satellites is obtained flux model in (1). The bulk flux model employed here is from the Remote Sensing Systems for the uniformity based upon the surface renewal algorithm of Clayson of the retrieval algorithms. The blended multiple- et al. (1996), which has been shown by Brunke et al. (2003) satellite observations provide improved space and to perform very well against eddy correlation observa- time coverage, and reduce the subsampling aliases tions of surface latent heat fluxes in the tropical oceans. and random errors. Brunke et al. (2003) found that bulk flux models begin to d Qa and Ta: A new method for retrievals of Qa and Ta diverge rapidly from the verification data at wind speeds 2 over the oceans has been developed by using the SSM/I exceeding 15 m s 1. To account for the impact of high data with a first-guess SST and a neural network winds and a highly disturbed sea state on the turbulent technique, in addition to improvements gained by exchange coefficients, we employ the adjustment de- properly accounting for the effects of high cloud liquid scribed by Bourassa (2006) to account for the influence

TABLE 1. Summary of input meteorological state variables. Values in parentheses are the weighting assigned to different sources of input variables.

NCEP2 OAFlux HOAPS3 ERA-Interim Horizontal resolution ;1.98 18 18 18 Temporal resolution 6-hourly Daily 12-hourly 6-hourly SST NCEP2 NCEP1 (1), OI (2) AVHRR ERA-Interim U NCEP2 NCEP1 (1), NCEP2 (1), SSM/I (4), QSCAT (4), AMSR-E (4) SSM/I ERA-Interim Qa NCEP2 NCEP (1), NCEP2 (1) SSM/I ERA-Interim SEPTEMBER 2011 LIUETAL. 2739

FIG. 3. As in Fig. 2, but for the CBLAST and XSeaFlux. of ocean waves and Andreas et al. (2008) to account There are several sea spray models in the literature for the influence of sea spray. (e.g., Fairall et al. 1994; Andreas and Emanuel 2001), To account for the effects of ocean waves, following that rely on assumptions on the size distribution of Bourassa (2006), we replace U in the bulk flux model by spray droplets and how the droplets evaporate. Here, the magnitude of the vector difference jU 2 Uwj,where we use a spray flux model based on Andreas et al. Uw is substituted by 80% of the orbital velocity at the (2008), whereby the LHF is a combination of the in- crest of the dominant waves. This substitution is justified terfacial latent heat flux (LHFi), which is computed in Bourassa (2006) for determining surface stress. The with the bulk interfacial flux algorithm of the afore- magnitude of the orbital velocity (Uorb) is determined mentioned flux model, and the ‘‘nominal’’ spray flux based on the dominant waves, and is in the mean direction that is computed with the fast microphysical spray pa- of the propagating waves. This approach to modeling rameterization (LHFs). Specifically, Andreas et al. wave influences works by modification of the vertical wind (2008) hypothesize that 50-mm droplets are good in- shear rather than modification of Charnock’s parameter. dicators of the total spray latent heat flux and model One advantage of this approach is that the same value it as LHF 5 r L f1 2 [r(t )/50mm]3gV (u ), where s s y f ,50 L * of Charnock’s parameter is a good fit to wind waves tf ,50 is the approximate residence time for droplets with 5 and swell. The orbital velocity is approximated as Uorb an initial radius of 50-mm, r(tf,50) is the radius these pHs/Tp,whereHs is significant wave height and Tp is the droplets have when they fall back into the sea, and the dominant wave period. Thus, the dependence of CE on 50 mm reiterates the hypothesis that 50-mm droplets wind speed is atypical in that we replace the traditional lead the spray latent heat flux. Here V (u ) is the wind L * earth-relative wind speed with jU 2 U j.TheU in- function, which is expressed as V (u ) 5 1:10 3 1027u2:22 w w L * * creases as wave slope increases, resulting in a reduced based on the analyses of both the Humidity Exchange value of the latent heat flux over steep waves, relative to over the Sea (HEXOS) experiment and the Fronts and the latent heat flux for the same 10-m wind speed and Atlantic Storm Track Experiment (FASTEX). a nonmoving surface with the same roughness. This result is similar to using the 10-m wind without considering U , w 3. Evaluation of latent heat fluxes for Hurricanes and reducing C as waves become steeper. The wave in- E Fabian and Isabel formation is obtained from the outputs of the new generation wave model (WAVEWATCH III) developed at Hurricanes Fabian and Isabel (September 2003) are NOAA/NCEP, which has a spatial resolution of 1831.258 selected for intensive analysis and evaluation because of and a temporal resolution of 6 h. the availability of field measurements from the CBLAST

TABLE 2. Comparison of LHF between the NWP and satellite-derived datasets and CBLAST.

NCEP2 OAFlux HOAPS3 ERA-Interim XSeaFlux 2 Bias (W m 2) 299.5 2150.0 272.3 2123.2 26.9 Correlation 0.51 0.46 0.48 0.58 0.91 2740 MONTHLY WEATHER REVIEW VOLUME 139

21 21 FIG. 4. Time series of (a) SST (8C), (b) U (m s ), and (c) Qa (g kg ) from the CBLAST, XSeaFlux, and ERA-Interim.

Hurricane Program (Black et al. 2007; Drennan et al. Isabel on 12–14 September 2003. The CBLAST obser- 2007; Uhlhorn et al. 2007). As part of CBLAST, a NOAA vations were collected between 1800 and 2200 UTC WP-3D research aircraft was instrumented to conduct di- along the flight track. rect turbulent flux measurements in the high wind bound- To obtain the parameters (LHF, SST, U, and Qa) from ary layer of Hurricanes Fabian and Isabel. Data used in the aforementioned NWP and satellite-derived products this study include flux flights made during Hurricane at the CBLAST positions, we extract the parameters that Fabian on 2–4 September 2003, and during Hurricane have the smallest distance and time difference relative to SEPTEMBER 2011 LIUETAL. 2741

TABLE 3. Absolute percentage error (APE) of SST, U, and Qa between XSeaFlux and CBLAST, and induced APE in LHF.

SST U Qa Input parameters 1.18% 1.47% 5.32% Resultant LHF 8.68% 1.04% 20.64%

the CBLAST observations (e.g., falling within half grid spacing and temporal scale of different products). Figure 2 compares time series of the NWP and satellite-derived LHF with the CBLAST LHF. None of the current flux products captures the extremely large LHF values observed on 3 September associated with Hurricane Fabian. Furthermore, the NCEP2 LHF values are biased low for much of the time relative to the CBLAST. The OAFlux LHF values are substantially smaller than the CBLAST, and do not capture the observed variations. The HOAPS3 has large negative biases on 14 September. The ERA- Interim substantially underestimates the LHF associated with Hurricane Isabel. Thus, current NWP and satellite- derived flux products do not reproduce the LHF variations associated with Hurricanes Isabel and Fabian. Also, there are large discrepancies among these flux products arising from their determinationofLHFthroughtheuse of different sources of the input meteorological state variables (see Table 1) and the use of different bulk flux algorithms (e.g., Liu and Curry 2006). Figure 3 compares time series of the XSeaFlux LHF with the CBLAST LHF (Note: the XSeaFlux LHF is not available for much of 3 September as a result of the missing input wind speed values owing to heavy pre- cipitation and sampling problems.) The XSeaFlux shows a much better agreement with the CBLAST. For the times when the XSeaFlux LHF values are available, the XSea- 2 Flux LHF values have a bias of 26.9 W m 2, which is a factor of 3–6 smaller than the biases of the NCEP2, OAFlux, HOAPS3, and ERA-Interim (Table 2). Moreover, the XSeaFlux LHF values are more strongly correlated with the CBLAST as compared to the NCEP2, OAFlux, HOAPS3, and ERA-Interim (Ta- ble 2). Figure 4 compares time series of input variables (SST,

U, and Qa) for the CBLAST, XSeaFlux, and ERA-In- terim, which is the highest-resolution NWP analysis that is available. The CBLAST-observed SST varies from ;268 to ;308C for Hurricanes Fabian and Isabel. Comparisons suggest that the XSeaFlux SST capture the observed variations during the hurricanes, although sometimes they show cold biases on the magnitude of ; 8 8 FIG. 5. Spatial distribution of (a),(b),(g),(h) SST (8C); 0.5 –1 C (Fig. 4a). The ERA-Interim SST does not 21 21 (c),(d),(i),(j) U (m s ); and (e),(f),(k),(l) Qa (g kg ) for (top) capture the observed variations, since it uses weekly Hurricane Fabian (1800 UTC 2 Sep 2003) and (bottom) Hurricane speciﬁed SST. Also, the XSeaFlux SST shows somewhat Isabel (1800 UTC 13 Sep 2003). localized warming in the domain of hurricanes, which is 2742 MONTHLY WEATHER REVIEW VOLUME 139

FIG. 6. As in Fig. 1, but for the XSeaFlux. absent in the ERA-Interim SST (Fig. 5). The observed U shows a well-organized and distinct structure of the LHF 2 varies from 16 to 28 m s 1 forHurricanesFabianand associated with the hurricanes, and the largest LHF ex- 2 Isabel. Comparisons show that the XSeaFlux U tracks the ceeds 1000 W m 2 near the storm center, and decreases CBLAST observations very closely. By contrast, the outwards. This result illustrates the importance of spatial ERA-Interim U does not capture the observed variations resolution in the estimation of heat ﬂuxes, and suggests (Fig. 4b), and greatly underestimates wind speed in the that further improvements could be made with better domain of the hurricanes as compared to the XSeaFlux U resolution in both space and time.

(Fig. 5). The observed Qa varies between 15 and 18.5 g 2 kg 1 for Hurricane Fabian and ﬂuctuates around 19 g 2 4. Analysis of hurricane energetics using XSeaFlux kg 1 for Hurricane Isabel. Comparisons indicate that Q a data from both the XSeaFlux and ERA-interim overestimates

CBLAST Qa observations for Hurricane Fabian, but The high-resolution LHF field available from the show good agreement with the CBLAST observations XSeaFlux dataset enables a broad range of scientific for Hurricane Isabel (Fig. 4c). However, the ERA-In- questions to be addressed regarding hurricane ener- terim Qa has a more distinct spatial structure and larger getics. values of Qa in the domain of the hurricanes as compared Possible impacts of ocean waves on near-surface wind to the XSeaFlux Qa (Fig. 5). speed and drag coefficient have been discussed in the- The impact of the error of each input variable (SST, oretical studies and for various regions (e.g., Bourassa

U, and Qa) on the resultant error of the XSeaFlux LHF 2006; Kara et al. 2007), but there is little quantitative for hurricane conditions is estimated. Speciﬁcally, the understanding of the impact of spatial and temporal averaged absolute percentage error (APE) of each input variability of ocean waves on the LHF associated with variable relative to the CBLAST observations is calcu- hurricanes. Figures 7a,b shows the spatial distribution lated. Then the APE of each variable is added to the of the LHF in Hurricane Fabian in response to ocean time series of that variable as inputs for the new bulk ﬂux waves. Ocean waves produce an asymmetric LHF re- model to estimate the resultant APE of LHF. As shown sponse (e.g., enhancing the LHF in the forward half in Table 3, the APE of Qa contributes the largest error to of the hurricane relative to the storm motion), of the 2 the LHF, whereas the APE of U has the least impact. magnitude of ;10–30 (;10–40) W m 2 for Hurricane The temporal variability is clearly dominated by the Fabian (Isabel), and reducing the LHF mainly in the changes in wind speed. right back quadrant of the storm, of the magnitude of 2 To further interpret how the XSeaFlux dataset repre- ;10–30 (;10) W m 2 for Hurricane Fabian (Isabel). sents LHF spatial variations in the domain of the hurri- Note that ocean waves also enhance the LHF in the canes, Fig. 6 shows the spatial distribution of the direct backward of Hurricane Isabel. This asymmetry XSeaFlux LHF at 1800 UTC 2 September for Hurricane depends on the translational speed of the hurricane. 2 Fabian and at 1800 UTC 13 September for Hurricane For wind speeds less than roughly 20 m s 1, the turbu- Isabel. In contrast to Fig. 1, the high-resolution XSeaFlux lent heat transfer occurs almost exclusively at the air–sea SEPTEMBER 2011 LIUETAL. 2743

22 FIG. 7. Differences in the LHF (W m ) with and without parameterizations of (a),(b) ocean waves and (c),(d) sea sprays for (left) Hurricane Fabian (1800 UTC 2 Sep 2003) and (right) Hurricane Isabel (1800 UTC 13 Sep 2003). interface. But with increasing wind speed, sea spray would offset the impact of ocean waves for high wind production increases, and heat and moisture transfer conditions. Farther away from the high winds, the impact also occurs at the surface of the spray droplets. When of sea spray is comparable to or smaller than that of the spray droplets leave the sea surface no heat is removed ocean waves. For Hurricane Isabel (Fig. 8, dotted gray in contrast to evaporation. The cooling is due to the line), the spray-enhanced LHF averaged within the radius 2 droplets’ heat loss in the air by evaporation and sensible of 34-kt wind (;17.5 m s 1) fluctuates mainly between 10 2 heat loss. The droplets eventually fall down, producing and 100 W m 2 on 8–14 September, then increases to 2 surface cooling. There has been considerable debate on ;280 W m 2 as Hurricane Isabel changed its path to the effects of sea spray on heat transfer. Fairall et al. north-northwestward on 15–16 September, and decreases 22 (1994) suggested a doubling of CE due to spray effects to ;100–200 W m during landfall on 18 September. 2 for U at 20 m s 1, whereas Makin (1998) predicts a 20% To explore the potential role of hurricanes in the cli- 21 increase of CE for U at 30 m s . The results of Andreas mate system, Trenberth and Fasullo (2007) estimated and Emanuel (2001) lie in between. As shown in Figs. the enthalpy loss by the tropical ocean due to hurricanes 7c,d, including a spray flux parameterization (Andreas within 400 km of the storm center based on a simple et al. 2008) results in a substantial increase of the LHF relationship between surface enthalpy fluxes and wind 2 for high wind regime (greater than ;18 m s 1), of the speed. Here we provide direct estimations from the 2 magnitude of ;60–150 W m 2. The impact of sea spray XSeaFlux data of the surface energy loss by the ocean as is a factor of 4–5 larger than that of ocean waves for high LHF for Hurricane Isabel since its LHF has the least winds. Thus, small error in sea spray parameterization missing values. Specifically, the LHF values are calculated 2744 MONTHLY WEATHER REVIEW VOLUME 139

12 22 FIG. 8. Time series of the surface energy loss as the LHF (310 Wm , dotted black line), 2 and difference in the LHF (W m 2) with and without parameterization of sea spray (dotted 2 gray line) within radius of 34-kt wind (;17.5 m s 1) of the storm center, and the category for Hurricane Isabel (bar). based on the average of the LHF within the radius of 34-kt in the evolution of the hurricane, which is partly due to 2 wind (;17.5 m s 1), which is obtained from the tropi- the increase of the hurricane size. The latent heat loss by cal cyclone extended best-track dataset (Demuth et al. the ocean from the time that Hurricane Isabel changed 2006). its path to north-northwestward (on 15–16 September) As shown in Fig. 8 (dotted black line), Hurricane to when it made landfall (on 18 September) is about Isabel pumps a considerable amount of heat out of the a factor of 2–3 larger than the heat loss when Isabel was ocean, and that amount is generally increasing with time major hurricanes (categories 4 and 5, see bars in Fig. 8),

22 FIG. 9. Spatial distribution of (a),(b) LHF (W m ) and (c),(d) SST (8C) for Hurricane Isabel (left) 1800 UTC 13 Sep 2003 and (right) 1800 UTC 17 Sep 2003. SEPTEMBER 2011 LIUETAL. 2745

22 FIG. 10. Spatial distribution of the annual mean LHF (W m ) in 2003 for (a) NCEP2, (b) OAFlux, (c) HOAPS3, (d) ERA-Interim, and (e) XSeaFlux. suggesting that there is no simple relationship between Prestorm (undisturbed) LHF has a range of ;100– 2 2 surface energy loss and hurricane intensity (category). 200 W m 2,whichisreducedto;20–80 W m 2 after the The time-integrated surface energy loss contributed by passage of the storm. Using measurements from an array of Hurricane Isabel is of the magnitude of 1.1 3 1020 J, air-deployed floats and surface drifters, D’Asaro et al. which is mixed deeper into ocean and that then must be (2007) also suggested a ;77% (61%) reduction of heat flux compensated for by ocean heat transports. due to the effect of SST cooling when averaged over a The high-resolution XSeaFlux dataset enables resolution 100-km (300 km) radius for Hurricane Frances in 2004. of the cold wake behind the hurricane. Figure 9 shows Thus, large reduction in heat flux could result in reduced spatial distribution of the XSeaFlux SST and LHF at 1800 intensity of the storm, if the cold wake is more toward the UTC 13 and 17 September for Hurricane Isabel. In the storm center, which might occur for a slowly moving storm. wake of the hurricane, the SST is cooled by up to ;28–38C with the largest cooling occurring in a narrow band in the 5. Conclusions back quadrant of the storm and trailing effects of the cold wake can be seen along the previous track of the hurricane. In this paper, we have investigated the ocean surface The cold wake has substantial impacts on air–sea fluxes. latent heat flux associated with two North Atlantic 2746 MONTHLY WEATHER REVIEW VOLUME 139 hurricanes. A new satellite-derived ocean surface flux accuracy of the satellite-derived latent heat fluxes in dataset is described, XSeaFlux, which combines new hurricanes. satellite-derived input variables using improved retrieval methods and a new bulk flux algorithm that in- Acknowledgments. We thank for Jun Zhang providing cludes parameterizations for ocean waves and sea spray. the CBLAST data. This research was supported by The XSeaFlux and existing NWP reanalysis and satel- NASA NEWS and NSF Polar Programs (0838920). lite-derived flux products are evaluated against the CBLAST field measurements during Hurricanes Fabian and Isabel in 2003. Comparison results suggest REFERENCES that the LHF of the newly developed high-resolution 8 Andersson, A., S. Bakan, K. Fennig, H. Grassl, C.-P. Klepp, and XSeaFlux (0.25 and 6 hourly) is closest to the CBLAST J. Schulz, 2007: Hamburg Ocean Atmosphere Parameters and observations. For hurricane domain comparisons, the Fluxes from Satellite Data—HOAPS-3—Twice daily composite. XSeaFlux shows well-organized LHF structures and World Data Center for Climate, digital media, doi:10.1594/WDCC/ large LHF values in response to the hurricanes. By HOAPS3-DAILY. [Available online at http://cera-www.dkrz.de/ 5 contrast, the other flux products (NCEP2, OAFlux, WDCC/ui/Compact.jsp?acronym HOAPS3_DAILY.] Andreas, E. L, and K. A. Emanuel, 2001: Effects of sea spray on HOAPS3, and ERA-Interim) produce no discernible or tropical cyclone intensity. J. Atmos. Sci., 58, 3741–3751. weak LHF signals, and no distinct structure in response ——, P. O. G. Persson, and J. E. Hare, 2008: A bulk turbulent air– to the hurricanes. sea flux algorithm for high-wind, spray conditions. J. Phys. Applications of the XSeaFlux data have been made to Oceanogr., 38, 1581–1596. interpret details of the ocean surface latent heat flux for Black, P. G., and Coauthors, 2007: Air–sea exchange in hurricanes: Synthesis of observations from the Coupled Boundary Layer Hurricanes Fabian and Isabel in 2003. Preliminary Air–Sea Transfer Experiment. Bull. Amer. Meteor. Soc., 88, analysis suggests that ocean waves, sea spray, and cold 357–374. wake have substantial impacts on latent heat flux asso- Bourassa, M. A., 2006: Satellite-based observations of surface ciated with the hurricanes. This new satellite-derived turbulent stress during severe weather. Atmosphere–Ocean hurricane latent heat flux dataset has the potential to Interactions, Vol. 2, W. Perrie, Ed., Advanced Fluid Me- chanics Series, Vol. 39, Wessex Institute of Technology, 35–52. provide important new insights into hurricane thermo- Brunke, M. A., C. W. Fairall, X. Zeng, L. Eymard, and J. A. Curry, dynamics and dynamics, particularly when coupled with 2003: Which bulk aerodynamic algorithms are least problem- observations of the structure of latent heat release within atic in computing ocean surface turbulent fluxes? J. Climate, a hurricane (Guimond et al. 2010). Figure 10 shows the LHF 16, 619–635. averaged over January–December 2003 for the NCEP2, Chen,S.S.,J.F.Price,W.Zhao,M.A.Donelan,andE.J.Walsh, 2007: The CBLAST-Hurricane Program and the next generation OAFlux, HOASP3, ERA-Interim, and XSeaFlux. The fully coupled atmosphere–wave–ocean models for hurricane general pattern of the XSeaFlux LHF resembles those research and prediction. Bull. Amer. Meteor. Soc., 88, 311–317. of the existing NWP reanalysis and satellite-derived flux Clayson, C. A., and D. Weitlich, 2007: Variability of tropical di- products (e.g., the largest LHF is found in the trade wind urnal sea surface temperature. J. Climate, 20, 334–352. belts of both hemispheres and in the western boundary ——, C. W. Fairall, and J. A. Curry, 1996: Evaluation of turbulent fluxes at the ocean surface using surface renewal theory. current regions: the Kuroshio and Gulf Stream) due to J. Geophys. Res., 101, 28 515–28 528. strong winds coupling with large sea–air humidity Curry, J. A., and Coauthors, 2004: SEAFLUX. Bull. Amer. Meteor. difference in these regions, and the smallest LHF is Soc., 85, 409–424. found in the eastern equatorial Pacific and Atlantic D’Asaro, E. A., T. B. Sanford, P. P. Niiler, and E. J. Terrill, 2007: due to upwelling induced cold SST associated with Cold wake of Hurricane Frances. Geophys. Res. Lett., 34, L15609, doi:10.1029/2007GL030160. weak winds, and in the high latitudes due to the Demuth, J., M. DeMaria, and J. A. Knaff, 2006: Improvement of poleward decrease of SST. Thus, this new satellite- Advanced Microwave Sounder Unit tropical cyclone intensity derived ocean latent heat flux dataset is not only and size estimation algorithms. J. Appl. Meteor. Climatol., 45, better for hurricane-related studies, but also good for 1573–1581. other applications. However, we note that these prod- Drennan, W. M., J. A. Zhang, J. R. French, C. McCormick, and P. G. Black, 2007: Turbulent fluxes in the hurricane boundary ucts differ considerably in magnitude, with the ERA-In- layer. II. Latent heat fluxes. J. Atmos. Sci., 64, 1103–1115. terim being the strongest and the OAFlux being the Emanuel, K., 1986: An air-sea interaction theory for tropical cy- weakest. clones. Part I: Steady-state maintenance. J. Atmos. Sci., 43, Further efforts to improve the retrievals and accuracy 585–604. of wind and specific humidity for hurricane conditions ——, 1995: Sensitivity of tropical cyclones to surface exchange coefficients and a revised steady-state model incorporating are needed. Efforts to make in situ flux measurements in eye dynamics. J. Atmos. Sci., 52, 3969–3976. hurricanes are needed to support continuing bulk flux ——, 2003: Tropical cyclones. Annu. Rev. Earth Planet. Sci., 31, algorithm improvements, and assess and improve the 75–104. SEPTEMBER 2011 LIUETAL. 2747

Fairall, C. W., J. D. Kepert, and G. J. Holland, 1994: The effect of Reynolds, R. W., T. M. Smith, C. Liu, D. B. Chelton, K. S. Casey, sea spray on surface energy transports over the ocean. Global and M. G. Schlax, 2007: Daily high-resolution-blended Atmos. Ocean Syst., 2, 121–142. analyses for sea surface temperature. J. Climate, 20, 5473– Guimond, S. R., G. M. Heymsfield, and F. J. Turk, 2010: Multiscale 5496. observations of Hurricane Dennis (2005): The effects of hot Roberts, J. B., C. A. Clayson, F. R. Robertson, and D. L. Jackson, towers on rapid intensification. J. Atmos. Sci., 67, 633–654. 2010: Predicting near-surface atmospheric variables from Kanamitsu, M., W. Ebisuzaki, J. Woolen, J. Potter, and M. Fiorino, Special Sensor Microwave/Imager using neural networks with a 2002: NCEP/DOE AMIP-II Reanalysis (R-2). Bull. Amer. first-guess approach. J. Geophys. Res., 115, D19113, doi:10.1029/ Meteor. Soc., 83, 1631–1643. 2009JD013099. Kara, A. B., E. J. Metzger, and M. A. Bourassa, 2007: Ocean cur- Simmons, A., S. Uppala, D. Dee, and S. Kobayashi, 2007: ERA- rent and wave effects on wind stress drag coefficient over the Interim: New ECMWF reanalysis products from 1989 on- global ocean. Geophys. Res. Lett., 34, L01604, doi:10.1029/ wards. ECMWF Newsletter, No. 110, ECMWF, Reading, 2006GL027849. United Kingdom, 25–35. Kleinschmidt, E., 1951: Grundlagen einer theorie der tropischen Trenberth, K. E., and J. Fasullo, 2007: Water and energy budgets of zyklonen. Arch. Meteor. Geophys. Bioklimatol., 4A, 53–72. hurricanes and implications for climate change. J. Geophys. Liu, J., and J. A. Curry, 2006: Variability of the tropical and Res., 112, D23107, doi:10.1029/2006JD008304. subtropical ocean surface latent heat flux during 1989-2000. Uhlhorn, E. W., P. G. Black, J. L. Franklin, M. Goodberlet, Geophys. Res. Lett., 33, L05706, doi:10.1029/2005GL024809. J. Carswell, and A. S. Goldstein, 2007: Hurricane surface wind Makin, V. K., 1998: Air–sea exchange of heat in the presence of measurements from an operational stepped frequency mi- wind waves and spray. J. Geophys. Res., 103, 1137–1152. crowave radiometer. Mon. Wea. Rev., 135, 3070–3085. Manning, D. M., and R. E. Hart, 2007: Evolution of North Atlantic Yu, L., and R. A. Weller, 2007: Objectively analyzed air–sea heat ERA40 tropical cyclone representation. Geophys. Res. Lett., fluxes for the global ice-free oceans (1981–2005). Bull. Amer. 34, L05705, doi:10.1029/2006GL028266. Meteor. Soc., 88, 527–539. Powell, M. D., P. J. Vickery, and T. A. Reinhold, 2003: Reduced Zhang, H.-M., J. J. Bates, and R. W. Reynolds, 2006: Assessment of drag coefficient for high wind speeds in tropical cyclones. composite global sampling: Sea surface wind speed. Geophys. Nature, 422, 279–283. Res. Lett., 33, L17714, doi:10.1029/2006GL027086.