Downloaded 09/25/21 11:27 PM UTC 3934 MONTHLY WEATHER REVIEW VOLUME 141 the 08C Isotherm (Between 4000 and 5000 M) Due to the Proﬁler Studies (E.G., May Et Al

NOVEMBER 2013 D O N A H E R E T A L . 3933

Wind Proﬁles in Tropical Cyclone Stratiform Rainbands over Land

SHAUNNA L. DONAHER,BRUCE A. ALBRECHT, AND MING FANG University of Miami, RSMAS, Miami, Florida

WILLIAM BROWN NCAR Earth Observing Laboratory, Boulder, Colorado

(Manuscript received 12 March 2013, in ﬁnal form 16 June 2013)

ABSTRACT

Observations of 14 stratiform periods in outer tropical cyclone rainbands are used to evaluate wind structure using a velocity–azimuth display (VAD) technique applied to KAMX (Miami) Weather Surveil- lance Radar-1988 Doppler (WSR-88D) data. These 14 cases occurred over land in southern Florida from Tropical Storm Fay and Hurricanes Gustav and Ike during 2008. Profiles show a maximum horizontal wind speed between 1000 and 1500 m in height, with occasional evidence of a secondary horizontal wind maximum near 3500–5000 m. Storm-relative wind components are calculated, and radial wind profiles show a mean transition from radial inflow at low levels to radial outflow around 2500–3000-m altitude. The radial inflow maximum is around 500 m, while maximum outflow is more variable. These profile characteristics are consistent with previous wind observations in rainbands over land and water. Changes in wind structure within one 4-h period are examined, with changes seen linked to the environmental influence on the rainband. All rainbands show a logarithmic wind speed decrease below 200 m. This layer is studied in detail using a log-wind fit method and a ratio method to calculate aerodynamic roughness length. Much lower ratios of surface to higher-level winds were found than in previous studies over open oceans. Another significant finding of this work is the lack of a constant aerodynamic roughness length despite similar storm wind profiles. These results are useful in broadening the understanding of low-level impacts of landfalling rainbands far from the storm center.

1. Introduction and Lorsolo et al. (2008), and those modeled by Foster (2005). Further, wind shear may play a role in promoting Rainbands are integral parts of tropical cyclones, tornado genesis over land (e.g., Novlan and Gray 1974; forming some of the main precipitation regions of the Gentry 1983; Verbout et al. 2007; Baker et al. 2009), system. Even if a tropical cyclone never makes landfall, which makes understanding the amount of low-level high winds and heavy rain from outer rainbands may still shear in landfalling storms important. Despite the im- affect land. Knowledge of the vertical variation of wind pact and importance of outer rainbands, observations of speed in these landfalling rainbands is critical in urban the wind structure in these bands over land are limited. areas where high-rise structures may experience winds Early work describing the structure of hurricane much stronger than those at the surface (Franklin et al. rainbands is based on data collected from research air- 2003). Wind gusts near the surface are also important, as craft flights in 1981 into storms over the water (e.g., they can cause significant damage near the shoreline, but Barnes et al. 1983; Barnes and Stossmeister 1986). Trop- can decrease rapidly inland due to surface friction ical cyclone rainbands are characterized by areas of en- (Powell et al. 1991). The vertical wind profile may be hanced reflectivity, or convective cells, embedded in a responsible for the generation of low-level roll struc- region of stratiform rainfall (e.g., Ishihara et al. 1986). tures such as those described in Morrison et al. (2005) These rainbands are shallow with radar echo tops around 9 km (e.g., Barnes et al. 1983; Powell 1990). Stratiform Corresponding author address: Dr. Shaunna Donaher, 4600 regions in rainbands can cover areas 10 times larger than Rickenbacker Causeway, Miami, FL 33149. the convective precipitation (Marks 2003) and are char- E-mail: [email protected] acterized by a distinct bright band just below the height of

DOI: 10.1175/MWR-D-13-00081.1

Ó 2013 American Meteorological Society Unauthenticated | Downloaded 09/25/21 11:27 PM UTC 3934 MONTHLY WEATHER REVIEW VOLUME 141 the 08C isotherm (between 4000 and 5000 m) due to the profiler studies (e.g., May et al. 1994), allowing for an lower reflectivity of ice crystals above the melting level improved understanding of the wind structure. (Marks 1985). Although the stratiform regions usually Recent observations of winds in landfalling rainbands contain less severe conditions than their convective often only use one or two instruments for data collection counterparts, wind maxima features are still present in (e.g., Knupp et al. 2006) or have limited vertical reso- wind profiles. lution (e.g., 105 m; Knupp et al. 2000) or specific heights The highest-resolution observations of wind speed (e.g., below 10 m; Masters et al. 2010) that they can an- profiles in hurricane rainbands come from dropsondes alyze. This study provides a unique opportunity to study (e.g., Franklin et al. 2003; Schwendike and Kepert 2008; multiple rainbands over land from three different trop- Zhang et al. 2011), which have become prevalent in ical systems using a variety of instrumentation. The oceanic flights in the past decade. These dropsondes are vertical structure of wind profiles from rainbands of released from aircraft typically flying at 700 hPa, and fall storms of different intensities and distances from the to the surface collecting measurements with a vertical center are compared. The high resolution of the VAD resolution of 5–7 m. Dropsondes allow for the study of dataset allows for examination of the log-wind regime in the low-level winds over the ocean with in situ mea- the hurricane boundary layer (HBL). In this work, the surements, something that is challenging to do with air- HBL is defined dynamically, following Kepert (2001), as craft as they cannot fly low enough. the shallow area closest to the surface that is influenced Observations of tropical cyclone rainbands over land by the frictional disruption of the gradient wind balance. used to be limited because of widely spaced coastal ra- The HBL has a logarithmic increase in wind speeds from dars, the inability to fly at low levels over land, and the the surface up to a wind maximum [usually located be- small possibility of a storm hitting a specific region in tween 500 and 1000 m above ground level (AGL); Smith a given year. With the installation of Next Generation and Montgomery 2010] that tops the HBL. The HBL is Weather Radar (NEXRAD) Weather Surveillance an area that is often unsampled during research flights Radar-1988 Doppler (WSR-88D), storms passing near but is crucial to understanding the impact of tropical cy- or over land can be studied (e.g., Skwira et al. 2005; clones on human activity and vegetation. Lorsolo et al. 2008), although the likelihood of a storm The objective of this study is to build upon previous passing close enough for rainband analysis is still lim- observations of wind profiles in stratiform rainbands, ited. Although several previous studies examined rain- with a focus on the low-level wind structure in land- bands using WSR-88Ds (e.g., Spratt et al. 1997; Blackwell falling rainbands far from the storm center. Individual 2000; Stewart and Lyons 1996) and through velocity– profiles of mean wind speed and direction, storm-relative azimuth display (VAD) techniques (e.g., Kim et al. 2009), wind components, and wind composites are presented this study evaluates the wind profiles in stratiform for 14 stratiform time periods in outer rainbands. Tem- rainbands at high resolution by combining WSR-88D poral variability within the longest rainband is examined level II data with the VAD technique, similar to Marks and the HBL is analyzed by calculating friction velocity et al. (1999) and Morrison et al. (2005). With this tech- and aerodynamic roughness length, calculating the ratios nique, the mean vertical resolution of the observations is of surface winds to those at the top of the HBL, and about 7 m, which is comparable to that of 5 m by Franklin comparing these ratios to previous studies over land et al. (2003) and 6 m by Schwendike and Kepert (2008) and water. and Zhang et al. (2011) using dropsondes. Although the VAD technique retrieves winds from a volume scan, the 2. Experiment description resulting horizontal winds are representative of a much larger area than a point measurement. In stratiform con- The data used in this work come from the Cloud ditions, this technique is preferable to observations from Precipitation Study (CPS), a collaborative field project dropsondes, which can drift significantly with the azi- involving the University of Miami Rosenstiel School of muthal wind in high wind regimes (Aberson 2008) as they Marine and Atmospheric Science (UM RSMAS), the fall for about 15 min. The areas covered by dropsondes National Center for Atmospheric Research Earth Ob- and a VAD scan are comparable in size, but the VAD serving Laboratory (NCAR EOL), and North Carolina scan is faster at 5–6 min. The increased resolution from State University (NC State). The field experiment took dropsondes and the VAD analysis is significantly higher place in south Florida over eight weeks in August and than the 500-m vertical resolution previously seen in September 2008, during which four tropical cyclones aircraft Doppler radar rainband studies over the ocean passed near the vicinity of the field site. The experiment (e.g., Eastin and Link 2009; Hence and Houze 2008) was based at the University of Miami South Campus and 200-m vertical resolution seen in landfalling wind [Center for Southeastern Tropical Advanced Remote

Unauthenticated | Downloaded 09/25/21 11:27 PM UTC NOVEMBER 2013 D O N A H E R E T A L . 3935

pattern formed by energy backscattered from the atmosphere is received by the remaining spatially sepa- rated receiving antennas. As air flows over the beams, atmospheric features are tracked across the receivers and the time lapse between spaced antennas gives wind measurements at each height. The MAPR has a time resolution of 30 s, which is considerably faster than the 10–15 min required by traditional Doppler beam swinging wind profilers (Cohn et al. 2001). During CPS, the MAPR operated with a variable range resolution between 100 and 200 m and collected data from 400 to 10 200 m AGL. Operating at a wavelength of 33 cm, the MAPR attenuation by rain is limited and allows for SNR, Doppler velocity, and turbulence intensity (spec- trum width) observations with height. Vertical profiles of the horizontal wind are retrieved from Level II KAMX data using a velocity–azimuth FIG. 1. Map of Florida and Cuba. The marker represents the display technique (Browning and Wexler 1968), thresh- location of vertically pointing instruments from the CPS experi- olded to remove all data points with SNR , 20 dB. The ment at the University of Miami South Campus (CSTARS). The radar measured radial velocities of many range gates at tracks of the three storms with rainbands sampled at CSTARS are also shown. Storm tracks are center positions from HURDAT. The the same distance are fit to a sine function against the symbols on the tracks are at 24-h intervals beginning at 0000 UTC azimuthal angle (Fang and Doviak 2008). The amplitude each day. and phase of the sine curve give the wind speed and direction (respectively) of the horizontal wind at the height of those range gates. This VAD analysis is ap- Sensing (CSTARS)] in southwest Miami, Florida, lo- plied to each range circle (250 m apart) between 5 and cated about 15 miles west (inland) of Biscayne Bay at 20 km from radar at each predefined elevation angle 25.618N, 80.398W (see Fig. 1). The terrain in south between 0.58 and 19.58. This provides reliable wind in- Florida is flat, with no natural locations higher than formation from 65 to 6550 m AGL. Although the S-band 10-m elevation. radar has a 18 beamwidth, limiting the radar resolution Multiple wavelength radars operated during this in the vertical to 87 m at 5 km from the radar and 349 m study, including a vertically pointing X-band radar from at 20 km from the radar, scanning at multiple angles RSMAS and a 915-MHz wind profiler [Multiple An- causes overlapping data points with height, so the mean tenna Profiling Radar (MAPR); for details, see Cohn vertical resolution of the dataset from VAD analysis has et al. (2001)] from the NCAR EOL. These remote sens- an average vertical resolution of about 7 m, which varies ing instruments provided high-resolution vertical map- with altitude from a minimum of 2 m at 65 m AGL to pings of signal-to-noise ratio (SNR), Doppler velocities, a maximum of 85 m at 6000 m AGL. It should be pointed and winds in the column above the CSTARS. Data from out that the vertical resolution for the dataset obtained KAMX, a scanning WSR-88D located 2.8 km from the from VAD is neither the radar resolution in the vertical field site, provided the vertical and horizontal structure direction nor the radar resolution in the radial direction, of winds and reflectivity. To complement the remote but the vertical distance between two adjacent range sensing datasets, rawinsonde launches occurred up to circles of VAD analysis that may not necessarily be at three times daily to provide thermodynamic and wind the same elevation angle. Complete volume scans took profiles. Surface precipitation drop size distributions about 5–6 min and covered up to a total volume of were collected with a high temporal resolution (10 s) ;4800 km3. Successful application of the VAD tech- Parsivel disdrometer provided by North Carolina State nique ensures that only wind structures comparable to University and a lower-resolution (1 min) Vaisala WXT and longer than the diameter of the range circle of the experimental capacity plate rain sensor from NCAR. WSR-88D are included in the dataset. Since this The two main observing instruments used in the wind technique involves averaging over a 3608 scan, it pro- analysis are the MAPR and Miami WSR-88D (KAMX). vides a smoothed vertical wind profile, thus removing The MAPR operates by using three (or four) vertically high-resolution horizontal structures, which makes it pointing beams arranged in a triangular (or square) pat- ideal for studying stratiform and homogenous pre- tern. A single vertical beam transmits, while the diffraction cipitation fields around the radar site. For this reason,

Unauthenticated | Downloaded 09/25/21 11:27 PM UTC 3936 MONTHLY WEATHER REVIEW VOLUME 141

FIG. 2. Mean profiles of band F3b (top left) horizontal wind speed (with inset showing log-wind profile in lowest 500 m), (top right) wind direction (degrees east of north), (bottom left) radial wind speed, and (bottom right) tangential wind speed during TS Fay. Profiles are produced using VAD technique on KAMX Level II data. this study focuses only on stratiform periods during the direction (direction of rainband approach), with veering rainbands. winds with height. The radial profile shows low-level Each complete 5–6-min WSR-88D scan creates a radial inflow below 500 m, a reversal to radial outflow at VAD profile, which are averaged together over strati- 3000 m, and a secondary radial inflow just above the form time periods for each rainband case and smoothed melting layer. The tangential wind profile is similar in in the vertical with a 5-point running mean prior to structure and strength to the mean wind profile, indicating analysis. An example of these smoothed wind profiles that the tangential component dominates the mean wind. for band F3b [Tropical Storm (TS) Fay, second strati- We intended to use observations from the vertically form period of third rainband] is shown in Fig. 2, and pointing X-band radar along with the WSR-88D VAD includes calculation of the mean radial and tangential and MAPR observations. Unfortunately, the radar re- wind profiles using the central storm location from the ceiver saturated under the rainband conditions, making closest time period in the National Hurricane Center’s the reflectivity data less useful. The Doppler velocities (NHC’s) North Atlantic Hurricane Database (HURDAT; from the X-band appear to be unaffected by the satu- http://www.nhc.noaa.gov/data/#hurdat), as discussed in ration and represent the sum of the precipitation and the section 5. In Fig. 2, the mean horizontal wind profile air velocity. In this study we use X-band velocity data to shows a log-wind regime below 500 m AGL (see inset) determine starting and ending times of the stratiform topped by a low-level wind maximum (LLWM) be- periods during rainbands and for establishing the level tween 1000 and 2000 m AGL. A secondary horizontal of the melting layer. wind maximum (SHWM) is also present around 4500 m Surface wind observations come from three heights AGL, which is the height of the melting level. The above the ground. The lowest wind speed information wind direction profile shows a general south-southeast comes from the NCAR Vaisala WXT-520 at about 3 m

Unauthenticated | Downloaded 09/25/21 11:27 PM UTC NOVEMBER 2013 D O N A H E R E T A L . 3937

AGL. Although this is the standard height for surface observations, CSTARS was surrounded by trees about 6 m high, which reduced the wind speeds measured near the ground to levels much lower than typical open ex- posure over land. For this reason, wind observations ervation type are from RM Young sensors located at heights of 14.5 and distance to center Storm quadrant and Right front, 190 km Right front, 156 km Right front, 143 km Right front, 136 km Left rear, 426 km Right front, 453 km Right rear, 427 km 18mAGLonatowerattheCSTARSsiteprovide supplementary wind speed data at a 2-min temporal 1 1 1 1 1 1 1 1 2 2 2 2 2 is of the rainband and whether resolution. 2 2 2 , Right front, 191 km 8 8 8 8 8 8 8 8 3. Dataset 22 km h 13 km h 9kmh 9kmh 18 km h 3kmh 24 km h 17 km h The three storms with rainbands passing over the CPS site were Tropical Storm Fay (17–22 August), Hurricane Gustav (30–31 August), and Hurricane Ike (9–10 September). Although none of these storms’ eyewalls 1 1 1 1 1 1 1 1 passed over CSTARS, rainbands from each of the storms 2 2 2 2 2 2 2 2 did pass over the field site. WSR-88D loops from Miami 1002 mb26 m s 998 mb26 m s 995 mb26 m s 994 337 mb26 m s 989 mb 345 26 m s 993 mb 10 26 m s 948 mb 10 62 m s 965 mb 21 36 m s 270 314 290 and Key West were carefully analyzed to determine rainband occurrence, defined based on an obvious band sustained wind speed Storm motion structure in the radar loop. Vertically pointing X-band data confirmed the specific timing of bands based on Storm pressure and maximum edges of changing conditions. This coupled dataset t t t t t t V V V V t t V V consists of 20 rainbands during Fay, 2 during Gustav, V V and 2 during Ike, for a total of 24 distinct rainbands. The majority of the rainbands occurred in the right-front quadrants of storms moving toward the northwest or

north (see Table 1 for more detail on the eight rainbands Rainband containing stratiform periods). Because of this large observation type dataset, data came from storms of varying strength (tropical storm to category 4 hurricane), rainbands at distances from 135 to 730 km from the storm center, and rainbands approaching the site from a wide range of of the storm. Storm strength, location, and motion are from closest time period of NHC best track data. t Band heading directions. Rainbands passed over the ﬁeld site for V anywhere from 15 min to 5 h depending on whether they were observed cross-band or along-band (note that only 8 of the 24 rainbands are included in Table 1). Subjective examination of the radar returns for all 24 rainbands identiﬁed the stratiform portions of the rainbands based

on the existence of a bright band, which is an unam- Time of band biguous indicator of the presence of stratiform precip- observation (UTC) itation (Houze 1997). In these stratiform time periods an obvious bright band existed around 4500 m AGL, SNR remained relatively constant, and rainfall was light 2 (,10 mm h 1) and fairly steady. In total, 14 stratiform- type cases were found, all in outer rainbands (see Table 2), and encompass cases lasting from 6 to 165 min at distances of 135–452 km from storm center. Stratiform rainbands are named after the storm they

occurred in [Fay (F), Gustav (G), or Ike (I)], numbered 1. Timing and conditions observed in eight rainbands that contained a stratiform period lasting 30 min or longer. Rainband heading, quadrant, and obs after the rainband in which they are chronologically observed in each storm, and lettered to distinguish what ABLE T Storm Band No. Date (2008) FayFayFay 2Fay 3 18Fay 4 Aug 19Gustav 5 Aug 19Ike 1930–2020 6 Aug 1 19 0121–0436 Aug NNW 21 0500–0830 Aug 1 31 Aug N Crossband, normal to 0900–1130 N–NNE 1712–2105 9 Sep Along-band, 0124–0636 parallel Crossband, NNE to parallel to ESE Along-band, NW parallel to 1526–1854 Crossband, normal to Along-band, parallel to NW Along-band, parallel to estimated from WSR-88D loops. Observation type describesthe whether band the rainband is moved oriented over parallel the site to as or a normal slice through to (crossband) the or tangential along the winds major ax order they occurred within the rainband (if more than Fay 1 18 Aug 1640–1740 N–NNW Crossband, parallel to

Unauthenticated | Downloaded 09/25/21 11:27 PM UTC 3938 MONTHLY WEATHER REVIEW VOLUME 141

TABLE 2. Stratiform cases in rainbands observed during the eight the boundary layer of a hurricane has a logarithmic rainbands shown in Table 1. Average rainfall rate is from 1-min profile up to the base of the maximum wind speed level, Parsivel observations. Rainfall data are missing from bands F5, usually between 500 and 1800 m AGL. This wind max- F6a, and F6b. imum outside of the eyewall is normally found near the Average top of a turbulent boundary layer (or HBL; Franklin Stratiform Start time End time rainfall rate 2 et al. 2003). Above the level of horizontal wind maxi- band Date (UTC) (UTC) (mm h 1) mum, there typically is a decrease in wind speed with F1 18 Aug 2008 1640 1740 6.71 height due to a weakening of the radial pressure gradient F2 18 Aug 2008 1954 2000 5.92 in the warm core system, as well as a region of weak or F3a 19 Aug 2008 0206 0336 2.50 F3b 19 Aug 2008 0353 0411 3.18 no shear up to 2000 m. The Powell and Black (1990) F3c 19 Aug 2008 0418 0430 0.93 flight-observed profile is also similar to that observed F4a 19 Aug 2008 0509 0521 0.52 using GPS dropsondes by Franklin et al. (2003), where F4b 19 Aug 2008 0527 0548 1.98 the LLWM occurs around 500 m near the eyewall and F5 19 Aug 2008 0954 1006 N/A rises to about 1000 m AGL in outer vortex regions, with F6a 21 Aug 2008 1742 1830 N/A F6b 21 Aug 2008 2000 2042 N/A a logarithmic profile below 300 m. Although Powell and G1a 31 Aug 2008 0151 0436 0.63 Black (1990) construct wind speed profiles from flight- G1b 31 Aug 2008 0530 0600 1.59 level winds at 500, 1500, and 3000 m, Franklin et al. I1a 9 Sep 2008 1727 1738 1.62 (2003) use about 200 outer vortex dropsondes with 5-m I1b 9 Sep 2008 1751 1812 1.67 resolution to measure the wind speed profiles. These dropsondes can drift while falling, but since they are averaged they provide a useful high-resolution (5–7 m) one stratiform period occurred in a rainband). Strati- vertical profile dataset to compare with the 14 individual form cases could either consist of an entire rainband VAD wind speed profiles below 3000 m. (e.g., band F1), one stratiform period within a rainband The 14 VAD profiles of wind speed with height in the (e.g., band F5), or several stratiform cases within a rain- stratiform rainbands (Fig. 5) all show similar features to band (e.g., bands F3a, F3b, and F3c). An example illus- the over-ocean profiles described by Powell and Black trating the features of band G1 is shown in Fig. 3. Figure 3a (1990) and Franklin et al. (2003). The wind profiles in shows SNR and vertical velocity from the X band for the the nine cases during Fay show a horizontal wind speed entire 5 h of band G1, which includes the stratiform periods maximum between 1000 and 1500 m AGL. Bands F1, G1a (0151–0436 UTC) and G1b (0530–0600 UTC). F2, F3a, F3c, F4b, and F6b have a double wind maximum Figure 3b shows the same variables, but from the structure similar to case F3b (Fig. 2), although the SHWM MAPR dataset. These two datasets show excellent is not always as prominent. These secondary wind agreement with vertical velocity values, and show simi- maxima could be due to thermodynamically driven cir- lar structures in SNR intensity and melting layer height. culation at the melting level (Moon and Nolan 2010). Figure 3c shows a closeup of band G1b using MAPR Rainbands during Gustav and Ike also show the double data, and is an excellent example of the consistent bright wind maximum structure, but in these cases the maxi- band associated with the melting layer, as well as a fairly mum horizontal winds speeds are stronger in the upper uniform rain rate seen in the vertical velocities. Figure maximum, which is located around 4000 m AGL in band 3d shows a plan position indicator (PPI) from the Miami G1a and 3500 m AGL in the remaining three cases. WSR-88D at the start of band G1b, along with a black The wind directions from the VAD wind profiles (Fig. 5) arrow showing the mean rainband motion. Figure 4 show easterlies or southerlies since the majority of the shows the rainfall rates during G1b, from the Parsivel rainbands approached from the southeast. Bands F6a sensor, which shows variability ranging from 0.5 to and F6b have a westerly/southerly profile and were ob- 2 3.5 mm h 1. While this may appear to be significant var- served on the left side of TS Fay, while all other wind iability given the homogenous vertical velocity signal profiles were observed on the right side of their storms. shown in Fig. 3c, the difference in rain rate during this This could account for some of the differences in struc- 2 rainband is less than 3 mm h 1, showing that the rainfall ture seen in these two profiles, as the height and strength observed during this case is very light. of LLWM can vary in different quadrants of the storm (e.g., Kepert and Wang 2001). As seen in Fig. 5, the majority of the profiles show the 4. Mean horizontal wind most easterly wind components near the surface. Twelve Powell and Black (1990) indicate that the vertical of the 14 profiles show a nearly constant veering of wind profile of the horizontal component of the wind speed in direction up to the height of the wind maximum. In these

Unauthenticated | Downloaded 09/25/21 11:27 PM UTC NOVEMBER 2013 D O N A H E R E T A L . 3939

21 FIG. 3. (a) X-band (top) signal-to-noise ratio (SNR, in dB) and (bottom) vertical velocity (in m s ) for all of band G1, from Hurricane Gustav. Stratiform periods G1a and G1b are outlined by the dashed lines. (b) MAPR SNR and vertical velocity for all of band G1. Again stratiform periods G1a and G1b are outlined by the dashed lines. (c) WSR- 88D image from KAMX at 0531 UTC, corresponding to the start of band G1b. Black marker shows CSTARS location and arrow shows rainband motion. (d) MAPR SNR and vertical velocity during band G1b. It should be noted that SNR from the MAPR and X-band are uncalibrated and are shown to indicate overall rainband structure, not specific reflectivity values. Negative vertical velocities represent motion toward the radar and are due to a combination of rainfall rate and actual vertical motion. cases, the wind veers 208–308 over the lowest 1000– than the sharp veering at 3000 m AGL seen in band F6b 1500 m. In contrast, Marks et al. (1999) analyzed hourly that appears to be associated with a minimum in the composite profiles of VAD-derived wind direction wind speed profile. during the landfall of Hurricane Fran (1996) and found Figure 6 shows the composite horizontal wind speed little change in wind direction over the lowest 400 m, profile for each storm, and the rainbands from Fay show followed by a sharp veering in wind direction of 508 up to a general decrease in wind speed with height above the height of the wind speed maximum. Although their 2500 m while the Ike rainbands show a distinct narrow low-level wind direction is constant over 5 h of Fran’s wind maximum just above 3000 m. The rainbands ob- outside eyewall making landfall, the profiles shown in served during Gustav show a broad upper-level wind Fig. 5 are more consistent with that of a friction layer maximum from 3000 to 5000 m AGL. Bands during and show similar variations as those seen by Knupp et al. Gustav and Ike were observed farther from the storm (2006) using wind profiler measurements during the center (430 km for Ike and 450 km for Gustav versus an landfall of Tropical Storm Gabrielle (2001) in Florida. average distance of 162 km for Fay), which may help to The time progression of wind direction profiles during explain some of the variability in wind patterns. Al- TS Fay is seen in the upper right panel of Fig. 5. The though Gustav was the strongest storm at the time of legend lists the stratiform rainbands in order of occur- rainband observations, it had a small wind field (tropical rence over four days, and Fay’s path around the field site storm and hurricane force winds extending 370 and 130 km is evident. There is very little evidence of temporal from the center respectively; Beven and Kimberlain 2009), changes in wind direction with height during Fay, other which explains why the wind speeds in the VAD profile

Unauthenticated | Downloaded 09/25/21 11:27 PM UTC 3940 MONTHLY WEATHER REVIEW VOLUME 141

ground to outflow located around 3000 m AGL. Peak radial inflow at low levels (300–400 m) is obvious with the VAD but missed with the lowest range gate of the MAPR (at 400 m). Although the MAPR data show a second radial inflow above the melting level (4500 m), the VAD winds indicate outflow at all heights above 3000 m. The three left panels of Fig. 9 show individual profiles of radial wind by storm. In these profiles the lowest radial wind shifts occur during Ike at just above 1000 m, but the highest occur during Gustav at heights above 4000 m AGL. The majority of radial shifts from inflow to outflow during Fay occur between 2500 and 3000 m. This is in agreement with overwater observations from Barnes et al. (1983) where the radial shift occurred

21 between 2500 and 4000 m AGL at radial distances of FIG. 4. Parsivel rain rate (mm h ) during band G1b. 70–110 km from the storm center. The individual profiles of radial wind shown in Fig. 9 indicate a maximum are lower than the other two storms. Although the radial inflow below 500 m in every case except F6a and 2 center of Ike passed 430 km from the field site, its large F6b, which varies in strength between 3 and 8 m s 1. wind field (tropical storm and hurricane force winds This is in agreement with the peak radial inflow around extending 445 and 185 km respectively; Berg 2009) gave 500 m AGL observed by Marks et al. (1999) during the measured wind speeds comparable to those during Fay. landfall of Hurricane Fran (1996). Schwendike and Figure 7 shows the mean horizontal wind from 12 of Kepert (2008) report finding a shallower inflow layer the 14 VAD cases, along with variance bars representing toward the storm center over the ocean in Hurricane one standard deviation off of the mean. This averaging Danielle (1998), but evidence of such a layer is not seen does not include bands F6a and F6b, since they differ in this dataset. Maximum radial outflow is less consis- substantially from the other horizontal wind speed profile tent and exists at a wide variety of heights between 3000 2 structures. Many of the wind maxima features in this and 6000 m AGL and values between 0 and 4 m s 1. composite are smoothed by combining all of the profiles, Several of the Fay cases and both Ike cases show a sec- although the constant presence of friction decreasing ondary radial inflow, located at heights between 4500 and wind speed strength near the surface is evident, as is 5500 m AGL. Maximum values of radial outflow are less a decrease in wind speed above 3000 m. than maximum values of radial inflow. Nine out of the 14 stratiform rainbands show evidence of peak radial out- 2 flows of about 2 m s 1 located around 3500 m AGL, 5. Radial and tangential winds which corresponds to a clockwise shift in wind direction To discuss the wind profiles within the context of the at the same height (Fig. 5). storms it is useful to divide them into the radial and b. Tangential wind profiles tangential wind components. These wind profiles are calculated in a storm-relative manner, using the central The three right panels of Fig. 9 show individual pro- storm location and storm motion from the closest time files of mean tangential winds. All profiles are positive, period of HURDAT data. In this coordinate system, as expected from the tropical cyclone circulation in the negative radial wind speeds represent wind blowing to- Northern Hemisphere. These profiles show that a maxi- ward the center of the storm, positive radial winds are mum in the tangential wind is typically found between moving away from the eye of the storm, and positive 1000 and 3000 m AGL. Although most profiles decrease tangential winds represent cyclonic flow around the storm in speed with height above 3000 m, a few of the cases do center. show evidence of a second maximum in the tangential wind at about 4000 m, especially the Gustav and Ike cases a. Radial wind profiles where the upper maximum is stronger than the lower Composite profiles of radial winds during 11 of the maximum. These results are in agreement with the rain- stratiform cases from both the VAD and MAPR are band schematic from Barnes et al. (1983, their Fig. 9), shown in Fig. 8. The profiles from these two instruments which shows maximum tangential wind between 1000 agree nicely, with a switch from radial inflow near the and 2500 m AGL at radial distances of 50–110 km, with

Unauthenticated | Downloaded 09/25/21 11:27 PM UTC NOVEMBER 2013 D O N A H E R E T A L . 3941

FIG. 5. (left) Mean VAD profiles of horizontal wind speed with height during 14 stratiform rainband cases: (top) 10 cases from Fay, (middle) 2 cases from Gustav, and (bottom) 2 cases from Ike. (right) As in (left), but for mean VAD profiles of horizontal wind direction with height, in degrees east of north. an occasional secondary peak around 4000 m AGL de- of the HBL, but the tangential wind speed drops off pending on the flight location relative to the cells moving more sharply in these cases than in the other Fay, along the rainband. Gustav, and Ike cases. This could be related to the di- Individual profiles of tangential wind are similar in rection of rainband approach, as these two cases are in magnitude and structure to the individual rainband westerly flow, indicating a longer fetch over land before profiles of horizontal wind (Fig. 5), indicating that as reaching Miami. expected the tangential wind component is the most significant to the overall wind speed. As in Fig. 5, bands 6. Temporal variability F6a and F6b differ greatly in strength and structure from the other Fay cases. Like the other profiles, they have The temporal variability in the profiles is examined a low-level tangential wind maximum just above the top using band G1 (shown in Fig. 3). Band G1 is broken down

Unauthenticated | Downloaded 09/25/21 11:27 PM UTC 3942 MONTHLY WEATHER REVIEW VOLUME 141

FIG. 6. Compositing of VAD mean horizontal wind Vh profiles FIG. 7. Mean VAD horizontal wind speed during all stratiform by storm. The solid black line represents the mean of rainbands cases (thick black line) except for bands F6a and F6b, with one observed when Fay had a maximum sustained wind speed of 2 standard deviation (gray bars). 26 m s 1 in the eyewall. The solid gray line represents rainbands observed when Ike had a maximum sustained wind speed of 21 36 m s in the eyewall. The dashed black line is the mean of during band G1a. The bottom panel of Fig. 10 shows the rainbands observed when Gustav had a maximum sustained wind 2 speed of 62 m s 1 in the eyewall. Bands F6a and F6b have been time progression of the tangential wind component. The excluded from the composites because of their large differences in patterns seen here mimic those of the mean wind tem- wind speeds from the other cases. poral variability, as the tangential wind component is significantly larger than the radial wind component. A careful study of WSR-88D loops during and after into two stratiform regions: band G1a, which represents bands G1a and G1b indicates that while there are no the longest period of stratiform conditions, followed by distinct changes in the structural appearance of the an hour-long gap where conditions are more convective rainband, as the rainband moves north-northwest over than stratiform, and then a second short stratiform pe- the CSTARS location the tail end of the rainband is riod (band G1b). Figure 10 uses half-hour averaged approached. Shortly after the end of band G1b, the VAD profiles to illustrate the changes in mean, radial, and tangential wind over 4 h as the band moves over site along its major axis. Mean wind speed profiles (top panel) show that while winds in the HBL remained relatively constant over the time period, the wind speeds near the low-level (1000 m) wind maximum increased steadily. Knupp et al. (2006) also observed an ascending and strengthening LLWM during an hour time progression of stratiform conditions in TS Gabrielle (2001). During the half-hour gap between bands G1a and G1b this wind maximum disappears and the wind speed drops to just below the initial wind speed at the start of band G1a. The upper wind maximum between 3000 and 5000 m AGL exists for all time periods during band G1a, but has also disappeared by band G1b. The middle panel of Fig. 10 shows the time progression of radial wind speed. The low-level radial inflow decreases in strength with time during band G1a, FIG. 8. Mean radial wind profiles with height from VAD (solid with a larger decrease during band G1b. Although the line) and MAPR (black dashed line) during 11 stratiform cases. height of reversal from radial inflow to outflow shows no Bands F5, F6a, and F6b are excluded from this averaging because of their differences in wind speed and structure. Both profiles are trend throughout band G1a, it is about 500 m in altitude smoothed with a 5-point running mean. The dashed gray line marks lower in band G1b. Band G1b also has a much lower and the transition from radial inflow (negative values) to radial outflow stronger secondary radial inflow than any of the profiles (positive values).

Unauthenticated | Downloaded 09/25/21 11:27 PM UTC NOVEMBER 2013 D O N A H E R E T A L . 3943

FIG. 9. (left) Mean VAD radial wind profiles with height by storm. Light gray line marks transition from radial inflow (negative values) to outflow (positive values). (right) Mean VAD tangential wind profiles with height by storm. Band F5 is excluded from this plot as it lies outside the bounds of one standard deviation of the mean. entire rainband begins to dissipate, and the line structure decrease in horizontal wind speed from 500 m to the of the rainband weakens. This postband environment surface (see Fig. 11 for band F3c), similar to the typical could potentially begin affecting the rainband conditions logarithmic variation of wind speed with height in during band G1b, which could explain the sharp changes a neutral surface layer (Stull 1988). To investigate this in profiles seen between the two stratiform cases. decay rate, statically neutral conditions were assumed, which is expected for stratiform conditions and supported by the linear fit shown in Fig. 12. For each strati- 7. Low-level mean wind form case, wind speed was plotted against the natural The individual plots of mean wind speed versus height logarithm of height minus a 6-m displacement distance for each stratiform rainband all show an exponential using the log-wind equation (e.g., Stull 1988, p. 377)

Unauthenticated | Downloaded 09/25/21 11:27 PM UTC 3944 MONTHLY WEATHER REVIEW VOLUME 141

FIG. 11. Similar to upper-left panel in Fig. 2. Close-up of lowest level of the HBL in band F3c (thick line), along with log-wind proﬁle calculated using Stull (1988), extrapolated to lower levels (dashed line). Data points from the CSTARS tower at 18 and 14.5 m are shown as black stars.

u (z 2 d) U 5 * ln , (1) k z0 2 5 k 1 ln(z d) U ln(z0). (2) u*

A best linear ﬁt was made for each of the wind proﬁles using VAD values from 65 to 120 m AGL to represent the surface layer. The slope and intercept of each line are used to solve for u* and z0 by case following Eq. (2). The results are shown in Table 3 and show u varying 2 * from about 0.4 to 1.4 m s 1 with a mean of about 21 1.0 m s . The z0 values vary from about 0.1 to 3.9 m, although an outlier of 9 m was found for band G1b. As expected, these roughness heights are substantially

FIG. 10. Averaged half-hour VAD proﬁles of (top) mean wind larger than would be expected over the ocean (z0 of speed, (middle) radial wind, and (bottom) tangential wind during about 0.001 m for a disturbed sea state; Stull 1988). We bands G1a and G1b. expect the aerodynamic roughness length for the south Florida terrain to range from 0.7 m for large towns and small cities to 2.5 m for centers of large cities with tall shown in Eq. (1), where U is the average rainband wind buildings. With the environment around the ﬁeld site 21 speed; u* is the friction velocity (m s ); k is the von consisting of trees and urban housing, bands F5, G1b, Karman constant, chosen to be 0.35; d is the displace- and I1a are well above the expected range for z0, while ment distance in meters (chosen to be 6 m as approx- bands F1 and G1a are more representative of the out- imate height of surrounding trees); z is height above skirts of towns. Unlike previous studies (e.g., Colin and ground in meters; and z0 is aerodynamic roughness Faivre 2010; Powell et al. 2003; Vickery et al. 2009) there length, also in meters. Equation (1) is written in the is no set value for z0 based on the 14 stratiform cases. form of y 5 mx 1 b in Eq. (2), where ln(z 2 d)isused There is also not an obvious relationship between z0 and as the y axis, and the wind speed is on the x axis in fetch, wind speed, or direction of rainband approach

Fig. 12. (not shown), although higher values of z0 are associated

Unauthenticated | Downloaded 09/25/21 11:27 PM UTC NOVEMBER 2013 D O N A H E R E T A L . 3945

log-wind ﬁt was varied by 50 m up to a height of 200 m. The outcome was surprising, with only 60% of the

height ranges resulting in z0 values that were less than 4 m. Thus, the fact that a consistent range of heights for the best-ﬁt line was chosen in an attempt to calculate

a constant z0 factor rather than choosing ranges that would optimize the goodness of ﬁt could inﬂuence the

z0 values. Additionally, ﬁtting winds only up to 120 m AGL (100 m above radar level) should help keep z0 values lower, as Masters et al. (2010) showed that including higher measurements could increase roughness lengths. By removing the two cases with less than a 91% correlation in the linear ﬁt (bands F6a and G1b) from 21 the averaging, z0 and u* become 1.54 m and 0.98 m s , respectively.

An alternative way of estimating z0 is to use the ra- FIG. 12. Low-level VAD winds from band F3c plotted against tios of wind speeds near the surface and at the top of natural log (height minus displacement distance) in black. The the log-wind layer. This is done by comparing mean best-fit line (gray dashes) is calculated using data between 65 and winds at 18 m (high in situ measurement from tower) to 120 m (black stars). mean winds at the top of the log-wind layer (inflection point in VAD curve). The top of the log-wind layer, which is defined as the height at which a logarithmic with higher values of u* and the maximum wind speed in profile no longer fits the VAD data, varies between 472 the lowest 2000 m. and 950 m AGL (see Table 3), with VAD data over Overall the goodness of fit r between the linear fit and a 30-m range around the inflection point used to cal- log-wind wind speeds is extremely high (see Table 3), as culate the average wind speed. Starting with the log- demonstrated in the example shown in Fig. 12. It ap- wind equation [Eq. (1)], the ratio of the wind between pears that z0 depends on the heights used in calculating 18 m and the top of the log-wind layer can be written as the best-fit line, since a small shift in included heights Eq. (3), where the subscript 1 denotes a height of 18 m can change z0 by over 5 m. A sensitivity study was AGL and the subscript 2 represents the height of the conducted for z0, where the height range used in the inflection point.

TABLE 3. Observed variability of friction velocity u* and aerodynamic roughness length z0 by using a logarithmic best-ﬁt line to each stratiform case. Ratios are calculated comparing mean wind speed at 18 m to these values. The ﬁnal column shows z0 calculated from the ratios using Eq. (4). All heights are given in altitude above ground level (AGL).

Vh at top of Ratio of 18-m wind u z0 (m) from Goodness of Top of log-wind log-wind layer to top of log-wind z0 (m) from ratio *2 2 Band (m s 1) log-wind ﬁt log-wind ﬁt r layer (m) (m s 1) layer wind calculation F1 0.72 0.44 0.99 950 18.79 0.24 3.50 F2 0.63 0.06 0.99 510 19.73 0.36 1.34 F3a 1.18 1.04 0.90 502 21.97 0.40 0.89 F3b 1.22 1.15 0.99 491 22.33 0.38 1.10 F3c 1.31 1.68 0.99 472 21.93 0.35 1.47 F4a 1.32 1.69 0.97 485 21.48 0.41 0.80 F4b 1.19 1.34 0.98 510 21.10 0.46 0.43 F5 1.44 3.86 0.97 493 21.69 0.41 0.80 F6a 0.66 1.02 0.89 553 11.70 0.36 1.34 F6b 0.65 1.48 0.95 788 12.18 0.27 2.84 G1a 0.40 0.24 0.97 765 17.56 0.31 2.08 G1b 1.04 9.05 0.90 523 16.03 0.39 0.99 I1a 0.98 3.21 0.96 485 19.28 0.33 1.76 I1b 0.74 2.32 0.99 858 20.72 0.31 2.08 Mean 0.96 2.04 0.97 599 19.04 0.36 1.53 Std dev 0.32 2.28 0.03 164 3.51 0.06 0.85

Unauthenticated | Downloaded 09/25/21 11:27 PM UTC 3946 MONTHLY WEATHER REVIEW VOLUME 141 2 TABLE 4. Mean wind speeds from the VAD technique at 65 m z1 d ln 2 2 AGL (a) compared with wind speeds calculated from Eq. (1) using z0 ln(z1 d) ln(z0) R 5U /U 5 5 , (3) (b) z0 by band from the log-wind fit, (c) z0 by band from the ratio fit, 12 1 2 z 2 d ln(z 2 d) 2 ln(z ) ln 2 2 0 (d) the mean z0 from the log-wind fit, and (e) the mean z0 from the z ratio fit. For (b)–(e), the u value for each band from the log-wind 0 * 2 fit is used. All wind speeds are in m s 1. The RMS error using ln(z 2 d) 2 R ln(z 2 d) methods (b)–(e) is also shown. z 5 1 12 2 ln( 0) 2 . (4) 1 R12 (a) (b) (c) (d) (e) Since the ratio in Eq. (3) is independent of the F1 9.4 10.1 5.8 6.9 7.5 F2 11.8 12.4 6.81 6.1 6.56 strength of the wind, this equation can be solved for z0 F3a 12.6 13.6 14.1 11.3 12.3 [Eq. (4)], and the roughness length for each rainband F3b 12.6 12.7 13.9 11.7 12.6 can be calculated. The ratios used for each rainband and F3c 12.1 13.3 13.8 12.6 13.7 F4a 12.6 13.4 16.2 12.7 13.8 the new calculated z0 values are shown in Table 3. The average ratio of the 18-m winds to the inflection F4b 12.0 12.9 16.7 11.4 12.4 F5 10.3 11.2 17.7 13.8 15.0 point winds is 0.36, which is significantly lower than those F6a 7.3 7.7 7.1 6.3 6.9 previously reported over land and over water, even though F6b 6.2 6.8 5.6 6.2 6.8 the near-surface ratios were calculated to the inflection G1a 5.7 6.3 3.8 3.8 4.2 point wind speed, which was lower than the maximum G1b 4.9 5.6 12.1 10.0 10.9 wind speed found in the VAD profiles at heights between I1a 6.8 8.2 9.8 9.4 10.2 I1b 5.9 6.8 7.1 7.1 7.7 2 1300 and 3300 m AGL. Previous work found over land RMS (m s 1) — 0.9 3.8 2.6 2.8 ratios ranging from 0.55 to 0.85 (e.g., Powell and Black 1990; Blackwell 2000) and overwater ratios close to 0.90 2 (e.g., Powell et al. 2009; Dunion et al. 2003; Franklin et al. wind speed, the RMS error is higher at 3.8 m s 1. Using

2003). A direct comparison between this work and pre- the mean z0 from Table 3 for the log-wind method vious work is challenging since the height of the upper (2.04 m) and ratio method (1.53 m) produces 65-m winds 2 wind used in the ratio varied between aircraft at a ﬁxed that have RMS errors of 2.6 and 2.8 m s 1, respectively. height [e.g., 500–3000 m in Powell and Black (1990)] or Unfortunately, even larger errors are generated if pressure [e.g., 850 hPa in Blackwell (2000)]. The compar- a mean z0 value between 1.5 and 2.0 m is used to cal- ison is further complicated when considering the surface culate wind speeds elsewhere in the surface layer (not wind data, which were typically obtained at 10 m AGL but shown). Whether this z0 is derived from the log-wind observed from a variety of ocean buoy, dropsonde and method or the ratio method, it underestimates wind land-based data collection methods. speeds by 30%–40% in the leading rainband cases, and

The z0 values for each rainband calculated using the overestimates wind speeds by up to 50% in the later ratio method are more realistic than the z0 values cal- cases where the bands are from stronger storms but are culated using the automated log-wind fit. The average z0 much farther from the storm center. from the ratio method for all 14 stratiform cases is 1.53 m, which is what is expected of an urban area 8. Summary and discussion without tall buildings. Although the rainband cases with high z0 values with the log-wind fit method do not cor- Stratiform periods during 14 rainbands over land in respond to the higher z0 values calculated with the ratio southern Florida are analyzed using a VAD technique method, the deviation between rainband z0 values has with Level II KAMX radar data to study vertical wind decreased by over half with the ratio method. structure. Results of these overland profiles are similar When the calculated u* and z0 values are put back into to previous observations over water, showing a LLWM Eq. (1) to solve for the wind speed at 65 m AGL and between 1000 and 1500 m AGL. Several cases show compared to VAD wind speeds at that height, an in- evidence of a SHWM around 3500–5000 m AGL, with teresting comparison arises, as shown in Table 4. As wind profiles from the stronger tropical systems having expected, using the individual z0 for each rainband cal- stronger SHWM than LLWM. Powell et al. (1991) ob- culated from the log-wind method provides wind speeds served a similar SHWM during a rawinsonde launched that are closest to the VAD winds, since 65 m is in the at Charleston, South Carolina, when Hurricane Hugo height range used to calculate each u and z0. The root- (1989) was 170 km from making landfall, further sup- * 2 mean-square (RMS) error in this case is 0.9 m s 1,or porting that stronger storms may have a secondary wind within 10% of the VAD winds. If the individual z0 values maximum near the melting layer. It has been suggested from the ratio method are used to solve for the mean that well-organized, intense rainbands that have secondary

Unauthenticated | Downloaded 09/25/21 11:27 PM UTC NOVEMBER 2013 D O N A H E R E T A L . 3947 wind maxima weaken storms overall by acting as a bar- previous observations. The second method uses wind rier to inflow (Barnes et al. 1983), which is contrary to ratios, which provides more realistic z0 values with less this work. However, these rainbands occurred at such variability, but far smaller ratios of surface to higher- large distances from their storm centers that there may level wind than what has been estimated over the ocean. be little relevance in their ability to weaken the storm The much lower ratios calculated here may partially be overall. due to the friction effects of being over land, the high Observed storm-relative wind components are found trees surrounding the field site and thus limiting the to be in close agreement with previous observations. 18-m wind, and the fact that the observations are located Radial wind profiles transition from inflow at low levels far from the storm center, as Kepert and Wang (2001) to outflow around 2500–3000 m. The radial inflow max- state that surface wind reduction factors are lower out- imum is around 500 m, while maximum outflow is more side of RMW. variable and weaker in strength. As expected, tangential Based on this dataset, it appears that the log-wind fit wind profiles by rainband closely match the mean hori- method is better for predicting wind speeds in the sur- zontal wind in strength and structure, as they make up face layer than the ratio method, although the variability the largest component of total horizontal winds, al- in calculated z0 values is concerning for both methods. though changes in wind directions appear to be associ- Unlike previous studies, a mean z0 was not obvious, ated with changes in the radial wind. which has implications for building codes and modeling Examination of temporal variability within one 4-h of HBL winds over land. One possible explanation could 2 period shows an ascending and strengthening low-level be that observed wind speeds were all below 23 m s 1 in wind maximum, along with a decrease in the low-level the lowest 500 m, which is lower than the majority of radial inflow over time. Analysis of plan position in- wind profiles used by previous studies (e.g., Powell et al. dicator (PPI) radar imagery indicates that band G1a was 2003). Given that wind speeds are low for all 14 cases the sampled as a diagonal cross section through the early differences between calculated and actual winds may part of the rainband, and then an hour-long gap of more not make a significant impact when winds are at or be- convection conditions occurred before the shorter band low tropical storm strength. G1b at the end of the rainband. The drastic differences Rainband characteristics can change drastically de- in wind structure between the 30-min averages during pending on storm intensity, storm-relative location, and bands G1a and G1b could possibly be explained by en- rainband stage, so it is important to remember that even vironmental conditions impacting the band during G1b, with several case profiles, the observations shown here and further help to illustrate that while both cases are are still only point profiles in a much larger tropical stratiform periods, the part of the rainband sampled can system. While many of the observed wind structure cause the observed conditions to vary widely. components agree with previous observations, the high Although not shown in log-wind form, all 14 of the vertical resolution of this dataset and wider variability in VAD profiles have a logarithmic decay of wind speed in observing instruments allows for the study of the vari- the HBL. In every case, this log-wind regime extends up ability in vertical structure in greater depth. Although to a height of about 200 m AGL. The log-wind profile not all of these observations agree with previously still exists from 200 m to the height of the wind maxi- published work, having several cases for analysis helps mum (top of the HBL), but it has a slope that differs to solidify the belief that the differences shown could from that in the near-surface (below 200 m) portion of be partially due to the location over land, which makes the HBL. This is similar to the composite profile of outer this dataset a valuable comparison to data collected vortex dropsondes by Franklin et al. (2003), where the over the ocean, particularly at the lowest levels where it depth of the entire log-wind region exists up to 700– is nearly impossible to collect data with aircraft. It is 800 m, but the exponential slope changes in the lowest important to study and understand the layer of the 200 m. A change in slope on a logarithmic plot may in- atmosphere closest to the surface in order to better be dicate the presence of an internal boundary layer re- able to predict storm conditions and their impact on sulting from a change in surface roughness (Stull 1988). life and property. In addition to a detailed study of the Two methods are used to describe winds in the lowest HBL, this provides further evidence of the location levels of the HBL and calculate friction velocity and and strength of wind speed maxima in tropical cy- aerodynamic roughness length. The first method fits clone rainbands over land. Even in conditions that are a log-wind profile with a linear fit. The calculated fric- thought of as stratiform, wind speed maxima are ev- tion velocity based on this method is, as expected, larger ident and understanding where they are located and than that observed over water. However, the aerodynamic how they change is important current and ongoing roughness length is found to be much more variable than work.

Unauthenticated | Downloaded 09/25/21 11:27 PM UTC 3948 MONTHLY WEATHER REVIEW VOLUME 141

Acknowledgments. We thank Sandra Yuter and the Franklin, J. L., M. L. Black, and K. Valde, 2003: GPS dropwind- North Carolina State Cloud and Precipitation Processes sonde wind profiles in hurricanes and their operational im- and Patterns research group for providing the Parsivel plications. Wea. Forecasting, 18, 32–44. Gentry, R. C., 1983: Genesis of tornadoes associated with hurri- data used in Fig. 4 and Table 2. We would also like to canes. Mon. Wea. Rev., 111, 1793–1805. thank Tom Snowdon for his assistance in radar opera- Hence, D. A., and R. A. Houze Jr., 2008: Kinematic structure of tions, and Brianne Winkler, Xue Zheng, and Virendra convective-scale elements in the rainbands of Hurricanes Ghate for their assistance in the field operations. Special Katrina and Rita (2005). J. Geophys. Res., 113, D15108, thanks go to Frank Marks for his helpful comments on doi:10.1029/2007JD009429. Houze, R. A., 1997: Stratiform precipitation in regions of convec- this manuscript. This work was supported by NSF Grant tion: A meteorological paradox? Bull. Amer. Meteor. Soc., 78, ATM0715235 and ONR Grant N000140810465. 2179–2196. Ishihara, M., Z. Yanagisawa, H. Sakakibara, K. Matsuura, and J. Aoyagi, 1986: Structure of a typhoon rainband observed by two Doppler radars. J. Meteor. Soc. Japan, 64, 923–939. REFERENCES Kepert, J. D., 2001: The dynamics of boundary layer jets within the Aberson, S. D., 2008: Large forecast degradations due to synoptic tropical cyclone core. Part I: Linear theory. J. Atmos. Sci., 58, surveillance during the 2004 and 2005 hurricane seasons. Mon. 2469–2484. Wea. Rev., 136, 3138–3150. ——, and Y. Wang, 2001: The dynamics of boundary layer jets Baker, A. K., M. D. Parker, and M. D. Eastin, 2009: Environmental within the tropical cyclone core. Part II: Nonlinear enhance- ingredients for supercells and tornadoes within Hurricane ment. J. Atmos. Sci., 58, 2485–2501. Ivan. Wea. Forecasting, 24, 223–244. Kim, D.-K., K. R. Knupp, and C. R. Willams, 2009: Airflow and Barnes, G. M., and G. J. Stossmeister, 1986: The structure and precipitation properties within the stratiform region of Trop- decay of a rainband in Hurricane Irene (1981). Mon. Wea. ical Storm Gabrielle during landfall. Mon. Wea. Rev., 137, Rev., 114, 2590–2601. 1954–1971. ——, E. J. Zipser, D. Jorgensen, and F. Marks, 1983: Mesoscale and Knupp, K. R., J. Walters, and E. W. McCaul Jr., 2000: Doppler convective structure of a hurricane rainband. J. Atmos. Sci., profiler observations of Hurricane Georges at landfall. Geo- 40, 2125–2137. phys. Res. Lett., 27, 3361–3364. Berg, R., 2009: Tropical Cyclone Report: Hurricane Ike. National ——, ——, and M. Biggerstaff, 2006: Doppler profiler and radar Hurricane Center, 55 pp. [Available online at http://www.nhc. observations of boundary layer variability during the landfall noaa.gov/pdf/TCR-AL092008_Ike_3May10.pdf.] of Tropical Storm Gabrielle. J. Atmos. Sci., 63, 234–251. Beven, J. L., and T. B. Kimberlain, 2009: Tropical Cyclone Report: Lorsolo, S., J. L. Schroeder, P. Dodge, and F. Marks, 2008: An Hurricane Gustav. National Hurricane Center, 38 pp. [Available observational study of hurricane boundary layer small-scale online at http://www.nhc.noaa.gov/pdf/TCR-AL072008_Gustav. coherent structures. Mon. Wea. Rev., 136, 2871–2893. pdf.] Marks, F. D., 1985: Evolution of the structure of precipitation in Blackwell, K. G., 2000: The evolution of Hurricane Danny (1997) Hurricane Allen (1980). Mon. Wea. Rev., 113, 909–930. at landfall: Doppler-observed eyewall replacement, vortex ——, 2003: State of the science: Radar view of tropical cyclones. contraction/intensification, and low-level wind maxima. Mon. Radar and Atmospheric Science: A Collection of Essays in Wea. Rev., 128, 4002–4016. Honor of David Atlas, Meteor. Monogr., No. 30, Amer. Me- Browning, K. A., and R. Wexler, 1968: The determination of teor. Soc., 33–74. kinematic properties of a wind field using Doppler radar. ——, P. Dodge, and C. Sandin, 1999: WSR-88D observations of J. Appl. Meteor., 7, 105–113. hurricane atmospheric boundary layer structure at landfall. Cohn, S. A., W. O. J. Brown, C. L. Martin, M. E. Susedik, Preprints, 23rd Conf. on Hurricanes and Tropical Meteorol- G. Maclean, and D. B. Parsons, 2001: Clear air boundary layer ogy, Dallas, TX, Amer. Meteor. Soc., 1051–1054. spaced antenna wind measurement with the multiple antenna Masters, F. J., H. W. Tieleman, and J. A. Balderrama, 2010: Surface profiler (MAPR). Ann. Geophys., 19, 845–854. wind measurements in three Gulf Coast hurricanes of 2005. Colin, J., and R. Faivre, 2010: Aerodynamic roughness length es- J. Wind Eng. Ind. Aerodyn., 98, 533–547. timation from very high-resolution imaging lidar observations May, P. T., G. J. Holland, and W. L. Ecklund, 1994: Wind profiler over the Heihe Basin in China. Hydrol. Earth Syst. Sci., 12, observations of Tropical Storm Flo at Saipan. Wea. Fore- 2661–2669. casting, 9, 410–426. Dunion, J. P., C. W. Landsea, S. H. Houston, and M. D. Powell, Moon, Y., and D. S. Nolan, 2010: The dynamic response of the 2003: A reanalysis of the surface winds for Hurricane Donna hurricane wind field to spiral rainband heating. J. Atmos. Sci., of 1960. Mon. Wea. Rev., 131, 1992–2011. 67, 1779–1805. Eastin, M. D., and M. C. Link, 2009: Miniature supercells in an Morrison, I., S. Businger, F. Marks, P. Dodge, and J. A. Businger, offshore outer rainband of Hurricane Ivan (2004). Mon. Wea. 2005: An observational case for the prevalence of roll vortices Rev., 137, 2081–2104. in the hurricane boundary layer. J. Atmos. Sci., 62, 2662–2673. Fang, M., and R. J. Doviak, 2008: WSR-88D observed spatial Novlan, D. J., and W. M. Gray, 1974: Hurricane-spawned torna- spectra of turbulence in precipitation. Preprints, 13th Conf. on does. Mon. Wea. Rev., 102, 476–488. Aviation, Range and Aerospace Meteor., New Orleans, LA, Powell, M. D., 1990: Boundary layer structure and dynamics in Amer. Meteor. Soc., 12.6. [Available online at https://ams. outer hurricane rainbands. Part I: Mesoscale rainfall and confex.com/ams/pdfpapers/129142.pdf.] kinematic structure. Mon. Wea. Rev., 118, 891–917. Foster, R. C., 2005: Why rolls are prevalent in the hurricane ——, and P. G. Black, 1990: The relationship of hurricane recon- boundary layer. J. Atmos. Sci., 62, 2647–2661. naissance flight-level wind measurements to winds measured

Unauthenticated | Downloaded 09/25/21 11:27 PM UTC NOVEMBER 2013 D O N A H E R E T A L . 3949

by NOAA’s oceanic platforms. J. Wind Eng. Ind. Aerodyn., Spratt, S. M., D. W. Sharp, P. Welsh, A. Sandrik, F. Alsheimer, and 36, 381–392. C. Pacton, 1997: A WSR-88D assessment of tropical cyclone ——, P. P. Dodge, and M. L. Black, 1991: The landfall of Hurricane outer rainband tornadoes. Wea. Forecasting, 12, 479–501. Hugo in the Carolinas: Surface wind distribution. Wea. Fore- Stewart, S. R., and S. W. Lyons, 1996: A WSR-88D radar view of casting, 6, 379–399. Tropical Cyclone Ed. Wea. Forecasting, 11, 115–135. ——, P. J. Vickery, and T. A. Reinhold, 2003: Reduced drag co- Stull, R. B., 1988: An Introduction to Boundary Layer Meteorology. efﬁcient for high wind speeds in tropical cyclones. Nature, 422, 279–283. Kluwer Academic, 666 pp. ——, E. W. Uhlorn, and J. D. Kepert, 2009: Estimating maximum Verbout, S. M., D. M. Schultz, L. M. Leslie, H. E. Brooks, surface winds from hurricane reconnaissance measurements. D. J. Karoly, and K. L. Elmore, 2007: Tornado outbreaks as- Wea. Forecasting, 24, 868–883. sociated with landfalling hurricanes in the North Atlantic Schwendike, J., and J. D. Kepert, 2008: The boundary layer winds basin: 1954–2004. Meteor. Atmos. Phys., 97, 255–271. in Hurricanes Danielle (1998) and Isabel (2003). Mon. Wea. Vickery, P. J., D. Wadhera, M. D. Powell, and Y. Chen, 2009: A Rev., 136, 3168–3192. hurricane boundary layer and wind ﬁeld model for use in en- Skwira, G. D., J. L. Schroeder, and R. E. Peterson, 2005: Surface observations of landfalling hurricane rainbands. Mon. Wea. gineering applications. J. Appl. Meteor., 48, 381–405. Rev., 133, 454–465. Zhang, J. A., R. F. Rogers, D. S. Nolan, and F. D. Marks, 2011: On Smith, R. K., and M. T. Montgomery, 2010: Hurricane boundary the characteristic height scales of the hurricane boundary layer theory. Quart. J. Roy. Meteor. Soc., 136A, 1665–1670. layer. Mon. Wea. Rev., 139, 2523–2535.

Unauthenticated | Downloaded 09/25/21 11:27 PM UTC