An Observational Study of Hurricane Boundary Layer Small-Scale Coherent Structures

Sylvie Lorsolo Atmospheric Science Group, Texas Tech University, Lubbock, Texas

Search for other papers by Sylvie Lorsolo in
Current site
Google Scholar
PubMed
Close
,
John L. Schroeder Atmospheric Science Group, Texas Tech University, Lubbock, Texas

Search for other papers by John L. Schroeder in
Current site
Google Scholar
PubMed
Close
,
Peter Dodge Hurricane Research Division, NOAA Atlantic Oceanographic and Meteorological Laboratory, Miami, Florida

Search for other papers by Peter Dodge in
Current site
Google Scholar
PubMed
Close
, and
Frank Marks Jr. Hurricane Research Division, NOAA Atlantic Oceanographic and Meteorological Laboratory, Miami, Florida

Search for other papers by Frank Marks Jr. in
Current site
Google Scholar
PubMed
Close
Full access

Abstract

Data with high temporal and spatial resolution from Hurricanes Isabel (2003) and Frances (2004) were analyzed to provide a detailed study of near-surface linear structures with subkilometer wavelengths of the hurricane boundary layer (HBL). The analysis showed that the features were omnipresent throughout the data collection, displayed a horizontal and vertical coherency, and maintained an average orientation of 7° left of the low-level wind. A unique objective wavelength analysis was conducted, where wavelength was defined as the distance between two wind maxima or minima perpendicular to the features’ long axis, and revealed that although wavelengths as large as 1400 m were observed, the majority of the features had wavelengths between 200 and 650 m. The assessed wavelengths differ from those documented in a recent observational study. To evaluate the correlation between the features and the underlying near-surface wind field, time and spectral analyses were completed and ground-relative frequency distributions of the features were retrieved. High-energy regions of the power spectral density (PSD) determined from near-surface data were collocated with the features’ ground-relative frequency, illustrating that the features have an influence on the near-surface wind field. The additional energy found in the low-frequency range of the PSDs was previously identified as characteristic of the hurricane surface flow, suggesting that the features are an integral component of the HBL flow.

Corresponding author address: Sylvie Lorsolo, HRD/NOAA/AOML, 4301 Rickenbacker Cswy., Miami, FL 33149-1097. Email: sylvie.lorsolo@noaa.gov

Abstract

Data with high temporal and spatial resolution from Hurricanes Isabel (2003) and Frances (2004) were analyzed to provide a detailed study of near-surface linear structures with subkilometer wavelengths of the hurricane boundary layer (HBL). The analysis showed that the features were omnipresent throughout the data collection, displayed a horizontal and vertical coherency, and maintained an average orientation of 7° left of the low-level wind. A unique objective wavelength analysis was conducted, where wavelength was defined as the distance between two wind maxima or minima perpendicular to the features’ long axis, and revealed that although wavelengths as large as 1400 m were observed, the majority of the features had wavelengths between 200 and 650 m. The assessed wavelengths differ from those documented in a recent observational study. To evaluate the correlation between the features and the underlying near-surface wind field, time and spectral analyses were completed and ground-relative frequency distributions of the features were retrieved. High-energy regions of the power spectral density (PSD) determined from near-surface data were collocated with the features’ ground-relative frequency, illustrating that the features have an influence on the near-surface wind field. The additional energy found in the low-frequency range of the PSDs was previously identified as characteristic of the hurricane surface flow, suggesting that the features are an integral component of the HBL flow.

Corresponding author address: Sylvie Lorsolo, HRD/NOAA/AOML, 4301 Rickenbacker Cswy., Miami, FL 33149-1097. Email: sylvie.lorsolo@noaa.gov

1. Introduction

Hurricanes are among the most destructive meteorological events that affect society. Proper forecasting of hurricane track and intensity is critical to minimizing a hurricane’s impact. However, an accurate forecast can only be expected if the kinematic and thermodynamic structure of the tropical cyclone is reasonably understood and modeled. A better knowledge of the hurricane boundary layer (HBL) is fundamental, as it would allow for the improved boundary layer (BL) parameterization in numerical weather prediction (NWP) models. Therefore, understanding smaller-scale coherent features embedded within the HBL has become an integral part of hurricane research. Both streaks and rolls are important features of the HBL since they are believed to be responsible for vertical heat, momentum, and moisture fluxes, which are related to hurricane intensity and wind irregularities (Morrison et al. 2005; Zhang et al. 2008). In their numerical study of Hurricane Andrew (1992), Yau et al. (2004) documented wind streaks with widths of ∼10 km. Streak formation was directly associated with latent heating and momentum transport.

Drobinski and Foster (2003) defined streaks as quasi-two-dimensional structures located in the near-surface layer of sheared BLs. Streaks are described as elongated features, organized as areas of alternating high- and low-speed fluid. These features usually exist in the surface layer, but can also extend into the mixed layer (Moeng and Sullivan 1994). In a large-eddy simulation (LES) study Drobinski and Foster (2003) simulated a neutrally stratified dry atmosphere of the midlatitudes with a shear-dominated flow and obtained streaks with 380-m wavelength. The streaks were oriented at 20° from the geostrophic wind and aligned with the surface wind. Streaks were distinguished from BL rolls as features with shorter lifetimes (streak usually exist for tens of minutes whereas rolls could last for hours) and smaller vertical extent. Rolls are believed to be deeper structures of the BL, spanning the depth of the BL.

Khanna and Brasseur (1998) conducted an LES study to describe various features of the BL under different stability conditions. The authors tested three different BL stability states: a purely convective, a near-neutral, and a moderately convective BL, where a stability parameter (−zi/L) was defined in terms of zi, the height of the convective BL, and L, the Obukhov length. In the purely convective BL case (−zi/L = 730) small-scale thermal plumes near the ground grew into large cellular structures throughout the BL. The near-neutral stability case (−zi/L = 0.44) exhibited streaks that diminished in intensity with height. In the moderately convective BL cases (−zi/L = 3 and 8), both rolls and streaks were observed. The rolls appeared as small structures spanning the BL and aligned with the mean wind, whereas streaks were only present near the ground. The authors proposed that the formation of the rolls was a consequence of the streaks near the ground, which suggests that rolls and streaks can coexist and interact with each other.

Weckwerth et al. (1997) combined data from the Convection and Precipitation/Electrification (CAPE) project and from numerical simulations to investigate BL rolls. Observations revealed that a minimum wind of 5.5 m s−1 throughout the convective BL was required for roll occurrence. Wavelengths, defined as the distance between two wind maxima or minima perpendicular to the features long axis, ranged from approximately 0.7 to 6 km, and were found to be proportional to the convective BL depth with smaller wavelengths associated with shallower mixed layers. The observed rolls had aspect ratios ranging from ∼1.8 to ∼4.8, and were oriented close to the mean wind direction of the convective BL, as well as the low-level wind.

Although various numerical and observational studies have documented BL rolls and streaks, few of them have dealt with strong wind events such as tropical cyclones. The importance of the HBL small-scale features was noted in early studies. In an experiment using a rotating tank simulating an Ekman BL, Faller (1961) proposed that the Ekman layer instabilities induced rainbands. The author noted similarities with HBL processes, suggesting the possible effect of smaller-scale features on hurricane dynamics. Recently scientists have investigated HBL smaller-scale features using numerical models. Nolan (2005) analyzed a hurricane-like BL where small-scale instabilities with wavelengths between 3 and 5 km and phase velocity of 3–5 m s−1 resulted. However the documented structures appear to differ from HBL rolls or streaks in various respects. Another study by Foster (2005) focused on HBL rolls, and found subkilometer rolls within the radius of maximum wind (RMW), while wavelengths on the order of 1 to 2 km were determined outside the RMW. Roll orientation was between 10° and 20° left of the mean azimuthal wind direction above the BL and approximately 10° to the right of the surface wind. Foster speculated that roll formation could be favored by the presence of intense turbulence and other processes such as streaks. The author also proposed that in strong convective regions, the roll circulation would be disrupted.

Only two observational studies have documented and analyzed HBL smaller-scale features of wavelengths smaller than 2 km. Wurman and Winslow (1998) were the first to acquire high spatial resolution radar data from the HBL. The radar signature of the rolls was characterized by bands of alternating weak and strong winds superimposed on the general radial velocity field. Wavelengths of approximately 600 m and depths near 1000 m were found in the eyewall of Hurricane Fran (1996). The authors argued that the rolls transported higher momentum to the surface because of their counterrotating circulation and could be the cause of small-scale irregularities in damage patterns. They also suggested that perturbations as small as 100 m could coexist with the rolls.

This study contrasts in some respects with a recent study, which also documented BL rolls in Hurricane Fran. Morrison et al. (2005) analyzed Weather Surveillance Radar-1988 Doppler (WSR-88D) radar data from four landfalling hurricanes. The gate-to-gate spacing of the WSR-88D radar radial velocity data was 250 m. Therefore the authors concentrated on rolls parallel to the radar beam to avoid aliasing due to bin-spacing limitations. The wavelength distribution ranged from 500 m to 3 km. However, most of the wavelengths were greater than 1 km and the peak of the wavelength distribution was approximately 1400 m, which contrasts with the subkilometer average value qualitatively identified by Wurman and Winslow (1998). The depth of the HBL rolls also varies greatly, with values ranging from approximately 300 m to 1 km with a mean depth of 660 m. The rolls were oriented on average 4° to the left of the mean BL wind direction. The authors compared the HBL depth (estimated from the height of the inflection point in the radial wind profile) and the inflow layer depth and found that on average the rolls were deeper than the inflow layer and shallower than the HBL depth. They also found that with increasing radial distance, the depth of the rolls and the aspect ratio increased.

None of the previous observational studies used data from experiments specifically designed to document the HBL smaller-scale features. Hence, it was essential to obtain additional data with higher spatial and/or temporal resolution to allow for a more in-depth analysis of these features and help understand the discrepancies between the past studies. These features were identified as HBL rolls in previous observational studies (Wurman and Winslow 1998; Morrison et al. 2005); however, because of the disparities between the studies and the fact that the features identified in this study also share similarities with streaks, they will be referred to as small-scale features until documentation of a full set of characteristics (wavelength, cross- and along-feature circulation, depth, etc.) allows for a formal identification.

HBL features are believed to be responsible for transport of higher momentum to the surface resulting in higher surface wind speeds (Wurman and Winslow 1998; Gall et al. 1998; Morrison et al. 2005). Damage surveys conducted after Hurricane Andrew’s landfall revealed discrepancies between the overall mean wind of the tropical cyclones and the resulting damage patterns (Wakimoto and Black 1994; Fujita 1992). Although possible causes for such discrepancies were suggested, small-scale features could be considered. Thus, the role of the small-scale linear coherent structures on near-surface wind fields and damage is a high priority. Powell et al. (1992) conducted a spectral analysis of radar reflectivity and wind speed data from Hurricane Bob (1991). The spectrum indicated the presence of a convective peak, representative of the contribution of convective scales of motion to the wind flow during the passage of rainbands. Powell et al. (1996) reviewed spectral analyses of past hurricanes and found that turbulent and convective mesogamma (10–100 h−1, 0.6–6 min) scales contributed to variations in the wind speed time histories. Spectra from Hurricane Bob displayed high energy in the mesogamma and mesobeta (2–10 h−1, 6–30 min) range, which could be due to rainbands or BL features.

Similar results were found for Hurricane Bonnie (1998) (see Schroeder and Smith 2003). In this study, the authors generated power spectrum density (PSD) estimates of wind speed for a 2-h period near the peak intensity of the storm and found a high-energy area in the low-frequency range (<0.01 Hz), characteristic of the mesogamma scales of motion previously mentioned by Powell et al. (1996). The high-energy area in the low-frequency range of Hurricane Bonnie’s PSD did not match any energy present when compared with universal spectra (Geurts 1997). Universal spectra represent the atmospheric surface layer under conditions such as a flow over a perturbed terrain, over a flat and smooth uniform surface, or under conditions described in the Kaimal model (Kaimal and Finnigan 1994). This result suggests that this energy could be due to the storm environment. Although these studies were able to identify energy characteristic of the HBL flow, they could not tie the excess energy to a specific type of phenomena.

The primary goal of the present study was to analyze the small-scale features found in the HBL of Hurricanes Isabel and Frances and thoroughly document their vertical and horizontal structure and orientation. Previous observational studies assessed the scales of motion associated with these features using subjective analysis. Given the noted discrepancies between the results, a second goal was to develop a robust methodology to objectively determine the wavelength of the features. The third goal of this research effort was to establish a correlation between the small-scale features’ kinematic signature and the underlying near-surface wind field.

2. Experiment

Two field experiments were conducted to acquire high-resolution data of the HBL and document the embedded coherent features. The first experiment took place in North Carolina during the landfall of Hurricane Isabel in 2003, and the second occurred in Florida during the landfall of Hurricane Frances in 2004. The platforms used during each experiment were two mobile Doppler radars and three instrumented towers. The Shared Mobile Atmospheric Research and Teaching (SMART) radars (SR1 and SR2), which are two C-band radars (5 cm) mounted on trucks with a half-power beamwidth of 1.5° (Biggerstaff et al. 2005), were employed. During the deployments the gate-to-gate spacing was set at 66.7 m to better resolve the small-scale features under investigation. The 10-m towers were instrumented with an anemometer (10 m AGL), along with a temperature, humidity, and barometric pressure sensor. The sampling rates for towers were 5 and 1 Hz for hurricane Isabel and Frances, respectively.

a. Hurricane Isabel deployment

Hurricane Isabel made landfall at 1700 UTC 18 September 2003 near Drum Inlet, North Carolina (Lawrence et al. 2005). Texas Tech University (TTU) and the University of Oklahoma (OU) collaborated to deploy and operate the observational assets. TTU’s goal was to acquire data to better characterize turbulence present in the HBL. TTU was responsible for deploying SR2 and three instrumented towers at the Craven Regional Airport in New Bern, North Carolina. The towers were positioned at close range from the radar (<1.5 km from SR2) in order to facilitate comparison between the radar-identified HBL small-scale features and the tower-observed surface wind field. The deployment site was approximately 60 km from the coastline, and the center of Hurricane Isabel passed approximately 47 km east of the deployment site. This configuration placed the deployment site in offshore flow throughout the experiment (Fig. 1).

Two scanning strategies were implemented to document the HBL small-scale features. The scanning strategies were performed in an alternating manner, each lasting 30 min. Table 1 describes the scanning strategies and their associated goals. The VERT strategy was designed to document the HBL structure while the TEMP strategy was used to study the features temporal evolution. Overall, the towers recorded over 47 h of data while SR2 recorded over 6 h of data using these scanning strategies. At the time of landfall (1700 UTC 18 September 2003) Hurricane Isabel had a minimum pressure of 960 hPa and a maximum wind of 90 kt (Lawrence et al. 2005). Table 2 gives some characteristics of the surface flow at the New Bern deployment site at landfall at 1700 UTC. Figure 2 presents tangential and radial wind profiles derived from the velocity–azimuth display (VAD) processing at 1700 UTC. At this time, the top of the HBL, defined as the height of the inflection point in the radial wind profile, was at 1.4 km AGL. Caution should be applied if comparing the wind profiles and the tower data as data included in the wind profiles extend past the area close to the tower.

b. Hurricane Frances deployment

Hurricane Frances made its first landfall on the east coast of Florida, near Hutchinson Island, Florida, around 0430 UTC 5 September 2004 (Franklin et al. 2006). TTU deployed observational assets to acquire dual-Doppler radar data targeting the coastal wind field transition zone. Single Doppler radar data was also acquired to support further documentation of the HBL small-scale features. For this experiment, SR1 and SR2 both collected data from the HBL. Both radars were positioned much closer to the coastline than in Hurricane Isabel. SR1 was deployed at the Merrit Island Airport located 8 km from the open ocean. SR2 was deployed at the Space Coast Regional Airport in Titusville, Florida, located 24 km from the open ocean and 21.9 km north-northwest of SR1. One 10-m tower was also located at the Space Coast Regional Airport, 392 m from SR2. Figure 3 is a map of the deployment area that includes the track of Hurricane Frances. The center of Hurricane Frances passed approximately 100 and 105 km south of SR1 and SR2, respectively. In this case, the deployment sites were subject to onshore flow during the entire data collection effort.

Similar scanning strategies as those used in Hurricane Isabel were implemented to document the HBL’s small-scale features. The azimuthal extent used in both strategies was increased to allow for a larger analysis domain. The vertical extent of the VERT scanning strategy was increased to provide data higher in the HBL. The TEMP strategy was limited to only one elevation angle (3°) to provide high temporal resolution. The TR20DD strategy was implemented during the landfall of Hurricane Frances for other scientific objectives but was useful for detailed description of the features because of its vertical oversampling. The instrumented tower recorded over 70 h of data, while SR1 and SR2 acquired data for approximately 20 h. At the time of landfall (0430 UTC 5 September 2004) Hurricane Frances had a minimum pressure of 957 hPa, a maximum wind of 90 kt (Franklin et al. 2006). Table 3 gives some characteristics of the surface flow at the Titusville site at 0000 UTC, a few hours before landfall, during the data collection using the TEMP scanning strategy. The tower was located with the radar executing the TEMP scanning strategy (one low elevation), which precluded VAD profiles for this time period. Fig. 4 presents tangential and radial wind profiles at the Merrit Island, Florida, site at 0400 UTC. The HBL depth retrieved from the radial wind profile was 1.1 km.

3. Processing and methodology

The current study is based on the analysis of 3 h of data from Hurricane Isabel and 5 h of data from Hurricane Frances. The raw data were processed using the Radar Data Software Library (RSL) developed by the National Aeronautics and Space Administration (NASA) Tropical Rainfall Measuring Mission (TRMM) satellite validation office (see online at http://trmm-fc.gsfc.nasa.gov/trmm_gv/software/rsl). The editing phase of the radar data (unfolding, removal of bad data due to blockage/ground clutter) was completed using the National Center for Atmospheric Research (NCAR) software package SOLOII (Nettleton et al. 1993). Once editing was complete, scripts were developed and implemented to convert the radar data from Doppler Radar Data Exchange (DORADE) format (Lee et al. 1994) into ASCII format. The VAD method (Browning and Wexler 1968) was used to retrieve the mean wind profile. The mean wind was then subtracted from the radial velocity field in order to obtain the residual radial velocity field. The residual velocities depict in detail the features under investigation. Figure 5a displays a sector scan of radial velocity acquired from Hurricane Frances before the VAD technique was performed. The HBL’s small-scale features can be identified in the radar radial velocity field as striations superimposed on the mean wind field, but details are difficult to distinguish. Figure 5b represents the associated residual radial velocity field after the removal of the VAD-retrieved mean wind. The residual velocity field results in a much more defined depiction of the features.

To obtain a complete representation of the three-dimensional structure of the HBL features, the radar data were converted into a Cartesian framework using a Cressman scheme (Cressman 1959). A radius of influence of 150 m was applied and the gridded data had a horizontal and vertical resolution of 50 and 25 m, respectively. It was also crucial that the analysis be performed without inclusion of aliased data and a predefined “nonaliased” domain was determined using the SMART radar specifications (bin spacing and beam spreading). It was required that the selected domain ensure wavelengths as small as 300 m could be resolved at the maximum range. The final chosen domain was limited to a range of 5.7 km from the radar.

4. General feature characteristics

a. Physical representation

Residual radial velocities displayed via plan position indicators (PPIs) were used to verify feature presence and organization. Figures 5b and 6 display residual velocity images from Hurricanes Frances (3° elevation) and Isabel (3.2° elevation), respectively. Small-scale features appear as coherent linear structures in both cases and are approximately aligned with the mean near-surface wind direction as illustrated in the images. The analysis of multiple time periods revealed that the features were always present, regardless of the location of the deployment sites relative to the vortex. A comparison between the two hurricanes suggests that the structure of the features was more defined in Hurricane Frances relative to Hurricane Isabel. In Hurricane Isabel, the features take on a more cellular appearance. This cellular appearance may be due to the fact that the radar was also deployed farther inland and situated in offshore flow in Hurricane Isabel compared to the onshore near-coastal flow in Hurricane Frances. In Hurricane Isabel, the deployment would thus have been prone to increased frictional effects that might have altered the physical characteristics of the features.

The magnitude of the residual velocities are similar in both hurricanes, ranging from approximately −6 to +6 m s−1. This range does not seem to vary significantly throughout the analyzed periods. In all the examined data, the residual velocity magnitudes did not exceed 10 m s−1. The documented magnitudes are weaker than those documented in Morrison et al. (2005), where magnitudes reached values of 13 m s−1.

Examination of PPIs (Figs. 5b and 6) suggests a change of feature orientation and an increase in feature wavelength with height. To aid interpretation, horizontal and vertical cross sections were plotted. Figure 7 is a horizontal cross section taken at 300 m AGL for Hurricane Frances. This representation is less noisy than the original scans because of the smoothing effect of the interpolation. A qualitative assessment of the features indicates subkilometer wavelengths as stated in Wurman and Winslow (1998), with average wavelengths near 500 m.

Figure 8 presents horizontal cross sections of the gridded residual velocity data displayed at multiple heights. Examination of the image reveals a definite vertical coherency of the features. If one looks at these results through 800 m AGL, a slight veering of the features’ orientation with height is apparent between the lowest and the highest horizontal cross sections. In Hurricane Isabel, the veering of the features is less subtle (not shown here), but stays minimal. This difference between the two storms could be due to stronger radial shear in the BL of Hurricane Isabel.

Figure 9a presents a vertical cross section taken perpendicular to the features’ orientation, approximately 2 km in range from the radar. In both hurricanes the features appear well defined, exhibiting a strong vertical coherency. Globally, the features extend uninterrupted from the ground to the top of the analyzed depth with stronger amplitudes of the residual velocity data near the ground. Close examination of the cross sections (Fig. 9b) indicates that the features exhibit only a minimal amount of tilt with height.

The features tend to lose definition near the top of the analyzed depth. This loss of definition could be due to the reduced effect of friction higher in the HBL. However, assuming a three-dimensional circulation such as the one speculated in Morrison et al. (2005), this loss of definition could also be because the observed flow may change directions to include cross-feature flow.

b. Orientation

A qualitative examination of various horizontal cross sections and sector scans indicated that the HBL small-scale features in both Hurricane Isabel and Frances are approximately oriented with the mean near-surface wind direction. However, a more accurate evaluation of the orientation was required for subsequent analysis steps and therefore a quantitative analysis of the features’ orientation was completed using a method of minimum variance. It was assumed that along the direction parallel to the orientation of the long axis of the features, the variance of the residual velocity data would be minimized. The method is illustrated in Fig. 10. For a considered image, the residual velocity field was rotated every 0.5°. At each angle, the variance along 100 cross sections was calculated. Each cross section contained the same sample size so that the variance would not be biased. For each rotation angle, the assigned variance was computed as the mean of the individual cross-section variances. The rotation angle associated with the minimum variance was then retrieved and related to the features’ long axis orientation.

Table 4 displays results of the orientation relative to true north at approximately 1500 UTC for Hurricane Isabel and 0000 UTC for Hurricane Frances, using 177 sector scans. The range of orientations is relatively narrow, ranging from −5° to 15° for Hurricane Isabel and from 35° to 65° for Hurricane Frances. Negative values represent orientations that are left of flow when looking downwind. The mean orientation determined for Hurricane Isabel was ∼5° with a variance of ∼6°2. For Hurricane Frances, the mean orientation was 53° with a variance of ∼9°2. Since the orientation of the HBL small-scale features changes with time, it was more informative to classify their orientation with respect to the mean near-surface wind.

The mean near-surface wind direction (defined as the mean wind between 0 and 250 m AGL) was 358° and 45° for Hurricanes Isabel and Frances, respectively (Table 5). The difference between the average feature orientation and the near-surface wind direction for Hurricane Isabel was between −26° and 4°. In Hurricane Frances, this difference was between −20° and 10°. Negative values represent an orientation to the left of the near-surface wind direction. Figure 11 is a histogram summarizing the feature orientation with respect to the mean near-surface wind direction for both hurricanes. In total, 65% of the feature orientations were within 10° of the mean near-surface wind direction, and over 90% were within 20°. Although there were isolated occurrences for which the orientation was up to 30° from the mean near-surface wind direction, the features’ orientation was on average 7° to the left of the near-surface wind direction.

Overall, these results agree with those documented by Morrison et al. (2005). Morrison et al. (2005) indicated orientations between −50° and 50° mainly to the left of the mean HBL wind direction. In numerical studies, LES simulations conducted by Drobinski and Foster (2003) provided streaks aligned with the surface wind field, which is slightly different from the results found in the present study. However, in Foster (2005), revealed instabilities aligned at ∼10° to the right of the surface wind.

c. Wavelength

One of the main goals of this study was to evaluate the wavelength distribution of the HBL small-scale features using an objective method. To obtain a relevant distribution of the feature wavelength, it was important to include a substantial amount of data in the analysis. The following is a detailed description of the method used to determine the wavelength distribution.

  1. Gridded data were segregated into 30-min periods to prevent the effect of any significant changes in orientation due to storm translation. Because the features were approximately aligned with the mean wind direction, which was assumed to be quasi-stationary over 30 min, it was hypothesized that the orientation of the features would also be quasi-stationary during the same period. This assumption was corroborated by a comparison between the orientation found from the first and the last volume scans of the considered time period.

  2. Vertical cross sections were taken normal to the orientation of the features. To examine the horizontal coherency of the feature wavelength, several cross sections were obtained at various ranges from the radar.

  3. Residual velocities along the vertical cross sections at specific heights were then obtained. This method allowed for an assessment of the wavelength coherency in the vertical direction. Figure 12 is an example of residual radial velocity data at 250 m AGL extracted from a vertical cross section. Although the shape of the signal is strongly sinusoidal, the amplitudes of the maxima and the minima are highly variable.

  4. Individual wavelengths were determined by measuring the distance between a peak (local maximum) and an adjacent trough (local minimum) using a peak-to-trough method, and multiplying the result by 2 to obtain the full wavelength. Since missing data due to editing or bad data could bias the results, only consecutive minima and maxima (with no missing data in between) were considered.

The wavelength analysis consisted of 465 vertical cross sections for Hurricane Isabel and 336 for Hurricane Frances. A great number of qualitative verifications were conducted to confirm that the automated algorithm was providing correct results. These results were also in accordance with the visual assessment of the wavelengths. Figure 13 represents the overall wavelength distribution with height for Hurricane Frances. The majority of the determined wavelengths were found in a range from 200 to 650 m, with a mean around 350–500 m. The figure illustrates the vertical coherency of the wavelength as the distribution remains coherent from the surface up to 600 m AGL. There is a slight widening of the distribution with height toward larger wavelengths. This widening might be due to the fact that the features of interest are losing definition, and/or that larger wavelengths become more dominant aloft. Similar results were found for Hurricane Isabel. Few wavelengths larger than 1200 m were found using the employed method. The subkilometer scale was found whether the features were located close to (as in Hurricane Isabel) or far from (as in Hurricane Frances) the eyewall.

To study the horizontal coherency of the wavelength distribution and to verify that there was no aliasing effects with increasing range from the radar, histograms of data taken at various ranges from the radar were produced. Figures 14 and 15 present wavelength histograms versus height at ranges of 1 and 3 km from the radar. Although the peak wavelength seems to have shifted slightly, both figures present the same overall characteristics with wavelengths ranging from 200 to 650 m, indicating that there is no significant change in wavelength with increasing distance from the radar within the chosen domain. The histograms are noisier than in Fig. 13 since only a small portion of the data is represented.

d. Discussion

The subkilometer wavelength distributions found during the current observational investigation contrast with the results presented in many of the previous HBL-focused studies. Morrison et al. (2005) identified rolls with wavelengths as small as 500 m, but they were not commonly found in their dataset. Foster (2005) found wavelengths in the subkilometer range of the hurricane profile he investigated, but only inside the RMW. Only Wurman and Winslow (1998) noted average subkilometer wavelengths of 600 m, which agree with those found in this study. The wavelength differences between the various observational studies could be explained by several factors, including the differences in the analysis technique, height range over which the analysis was performed, aliasing (due to the differing spatial resolutions of the data), thresholding and the inherent superposition of various active scales of motion.

The wavelength analysis conducted in the present study differs from the technique employed by Morrison et al. (2005). The method employed in the current study was completely quantitative and did not rely on subjective estimations. The analysis also accounts for changes in wavelength with height or distance from the radar. Moreover, given the relatively small gate-to-gate range resolution of the SMART radar, the analysis was not limited to data taken from beams parallel to the long axis of the features. Instead, data were extracted on planes normal to the long axis of the features. In general, this method was best suited to identify smaller scales of motion that were not retrieved by Morrison et al. (2005). However, overall examination of vertical cross sections allowed for the identification of a larger scale of motion, superimposed on the small-scale features.

Examination of the residual radial velocities indicates superposition of different scales of motion. If one considers a vertical cross section (Fig. 16a), which yields the 350 m AGL residual velocity data represented by the solid line in Fig. 16b, the record shows the presence of the small oscillatory characteristic of the small-scale features evident in the PPIs. However, a close examination of the data reveals that larger scales of motion are also present in the signal (dashed line). If one bases the wavelength analysis only on absolute maximum and minimum residual velocities, it would bias the determined wavelengths to larger scales. A close examination of Fig. 16a shows some of the smaller features that are well defined at lower altitudes would be missed above ∼350 m AGL. Wavelengths determined above this height would result in an overestimation of the wavelength of the small-scale features. The presence of superposed scales of motion illustrates the complexity of the HBL structure. In addition to analyzing the characteristics of specific scales of motion, it appears that discriminating the different scales of motion is also important. The coexistence of various scales of motion was mentioned by Wurman and Winslow (1998), suggesting that smaller perturbations were embedded with the rolls. Likewise, Foster (2005) suggested that smallerscale vortices could interact with roll instabilities.

Given this superposition of scales, performing a thresholding technique on the residual velocity data prior to assessing the wavelength can significantly alter the results of the wavelength analysis. Examination of the data indicated that larger scales of motion could effectively mask the smaller-scale contribution (Fig. 16). Therefore, if the wavelength retrieval only considered features characterized by adjacent positive and negative residual velocity values, the small-scale features might not be properly characterized. Applying a thresholding technique such as only considering residual velocities of ±3 m s−1 would result in a very different wavelength distribution. It is therefore important to analyze the effect of any previously employed thresholding technique.

Because of the discrepancies found between previous wavelength studies, WSR-88D data acquired during Hurricane Fran (1996), as used by Morrison et al. (2005), were analyzed using the objective method employed in this study. A qualitative examination of vertical cross sections indicates wavelengths of approximately 500 m near the ground (not shown here). The wavelengths appear smaller near the ground, with broader features above. Overall, the wavelengths appear to be subkilometer.

A quantitative assessment of the wavelength was then completed using the peak-to-trough method employed in this study. It provided similar results to those found for Hurricanes Isabel and Frances. Figure 17a is a histogram presenting the wavelength distribution for the considered volume. The dominant wavelengths were found between 300–700 m with a mean value near 500 m, with some wavelengths greater than 1 km. These results are similar to those found in Wurman and Winslow (1998) for the same storm. The vertical coherency is verified, as the histogram displays a consistent distribution at each height from 100 m to 1 km m AGL. In contrast, Fig. 17b shows the wavelength distribution when a ±3 m s−1 threshold technique was applied to the same data prior to using the peak-to-trough method. In this case, the dominant wavelengths were approximately 700 m below 100 m AGL and between 900 and 1500 m up to 1 km AGL. These results are closer to those qualitatively determined by Morrison et al. (2005) and indicate how thresholding can alter the results. It should be noted that the number of individual wavelength estimates included in the analysis is much smaller when the thresholding is applied (45 versus 314).

The effect of aliasing must also be considered. Although Morrison et al. (2005) analyzed data taken when the beam was directed parallel to the long axis of the features; it is possible that the data could have been subject to aliasing. Since the amplitude of features with wavelengths ranging from 400 to 600 m could not be fully resolved by the WSR-88D radar, resulting wavelength distributions obtained when only considering adjacent positive and negative residual velocities could have represented aliased energy.

As previously documented, the current study analyzed data located between the surface and 450 m AGL for Hurricane Isabel and 600 m AGL for Hurricane Frances, whereas the Morrison et al. (2005) study considered data between ∼100 m and 3.5 km AGL. It is not clear to what degree these differences in analysis height affected the resulting wavelength distributions. However, it can be speculated that since a qualitative analysis we conducted indicated that larger scales of motion were more dominant in the upper levels of the HBL, the Morrison et al. (2005) wavelength distribution would result in larger scales. These results seem to imply that the features considered in Morrison et al. (2005) and those considered in the present study represent different scales of motion superimposed on each other. This superposition is furthermore illustrated, as the aliased signal from the WSR-88D radar presented greater magnitude that the signal from the SMART radar. Thus, it appears that a large spectrum of scales of motion exist in the HBL. This result leads to a more fundamental question of the importance of each of these features in the HBL, in terms of energy fluxes and correlation with the near-surface wind field.

5. Correlation with the near-surface wind field

Several studies (Wakimoto and Black 1994; Fujita 1992) have inferred irregularities in the surface wind field based on the resulting damage patterns. However, it is not clear what phenomena caused these irregularities. To better understand the interaction of the HBL small-scale features and the near-surface wind field, radar, and tower data were compared, providing a simultaneous view of the near-surface wind field.

a. Comparison of time histories

The most straightforward method used to compare the impact of the HBL small-scale features on the near-surface wind field was the direct comparison of the tower wind speed and the radar radial velocity time histories. Because Doppler radars only measure radial velocity, the tower data had to be projected onto the radar-tower axis to derive the equivalent radial velocity to compare with the Doppler radar data. The radar data used for comparison were collected from above the nearby towers at 30.4 m AGL for Hurricane Isabel and 20.5 m for Hurricane Frances. Only results for one bin of data above the tower are presented; however, the analysis was also completed for five–nine bins located over the tower deployment site. These additional comparisons yielded no noticeable changes relative to the one-bin approach. Various averaging times were employed on the tower data to maximize the correlation between the two signals. The best comparison was obtained using 30-s windows centered on the time of the radar acquisition. Some of the results of this one-to-one comparison are presented in Fig. 18. The radar and the tower time histories are similar. The individual values of the radar radial velocity data are in general greater than the values obtained from the tower, which is partly due to the increase of the wind speed with height in the HBL. The comparison is not as straightforward since the two types of measurement are inherently different, with tower measurements being point measurements and radar data representing volumetric measurements.

The next step of the analysis was to investigate a possible link between the HBL small-scale features with the near-surface wind field. Comparisons between pronounced peaks and troughs of the tower time series and images of radar residual velocity were made. As expected, the radial wind speed peaks were associated with positive residual velocities while the troughs were associated with negative residual velocities. However, the highest (lowest) tower-derived radial velocity amplitudes did not necessarily correspond to the highest (lowest) residual velocity amplitudes. It is likely that some of the peaks are not associated with the small-scale features under investigation, but rather to other scales superimposed in the signal. These results suggest a link between the HBL small-scale features and the near-surface flow, but it is impossible to discriminate the effect of each scale of motion and link a specific peak or trough in the time histories to a particular scale of motion, or to assess how much each scale of motion contributes to a specific peak or trough in the wind speed time history.

b. Comparison in the frequency domain

A frequency analysis was used to retrieve the ground-relative frequency range of the small-scale features. The PSD of the surface wind field was then examined for the potential presence of a significant amount of energy at the identified range. Obtaining PSDs of the surface wind field was straightforward since PSDs could easily be computed from the high-resolution tower data using a nonparametric method (Bendat and Piersol 1986). On the other hand, although the temporal resolution of the radar data was high enough to follow the evolution of the features, it was not high enough to determine relevant PSDs. The frequency of the features was therefore determined using an indirect method. The following are steps taken to determine the ground-relative frequency of the features:

  1. The translation vector of the features was estimated.

  2. The features’ orientation was coupled with the translation vector to determine the crosswise motion of the features (Vc).

  3. The crosswise motion was then coupled with the wavelength distribution to determine the ground-relative frequency distribution using the following relationship: F = Vc/λ, where F is the feature’s frequency in hertz, Vc is in meters per second, and λ is the wavelength in meters.

1) Estimation of the translation vector

The translation vector of the features was determined using the tracking radar echoes by correlation (TREC) method (Tuttle and Foote 1990). This technique uses the cross correlation between two radar reflectivity (or radial velocity) images obtained at two different times to determine the local advection. To determine the translation vector of the HBL small-scale features, the residual velocity field of the TEMP scanning strategy was used. The evaluation included 68 and 110 PPIs from Hurricanes Isabel and Frances, respectively. For each PPI, three subareas of approximately 2.5 km × 1.5 km were chosen. The correlation was determined for each subarea between an initial image and the three subsequent scans. The time between the initial scan and the subsequent scans (Δt) was limited to less than 60 s; the timestep was restricted to such a short duration to limit large changes in depth between the initial and the tracked arrays (Δh). The analysis of the vertical structure has shown that there were no noticeable changes with height of the features over small depths. For the analysis to be as accurate as possible, only correlation coefficients greater that 0.6 were considered. Table 6 provides the resulting translation speed and direction for the two storms. For both cases the range of the translation speed and direction is relatively narrow. For Hurricane Isabel the translation speed ranges from 21 to 28 m s−1 and the translation direction is between 348° and 3°. For Hurricane Frances, the translation speed ranges between 22 and 29 m s−1 with directions between 37° and 47°. These results indicate that on average the small-scale features were moving at a slower speed than the mean surface wind speed, but on rare occasions higher translation speeds did occur. The translation direction was close to the azimuthal wind in both storms.

2) Estimation of the crosswise motion

An accurate assessment of the crosswise motion of the HBL small-scale features is essential to calculate their ground-relative frequency, as it will determine how fast the features move laterally across their wavelength. The crosswise motion was calculated by using the mean orientation for the considered time period and the translation vector. The crosswise motion (Vc) is calculated using the simple trigonometric formula:
i1520-0493-136-8-2871-e1
where Vt is the translation speed, a is the features’ orientation angle, and b is the translation direction angle.

The features were found to laterally move slowly inward toward the center of the storm. Figure 19 represents the crosswise motion distribution for Hurricane Frances. The distribution presents values below 8 m s−1 with a mean of 5.5 m s−1. Similar results were found for Hurricane Isabel, with a mean of 3.7 m s−1. The difference in amplitude of Vc between the two storms is likely to be related to the difference in feature translation speed (Table 6). These results agree with the inward phase velocities of 3–5 m s−1 found in Nolan (2005). However, they differ from Foster’s (2005) preliminary work, which found phase velocities of ∼4.5 m s−1 within a radius of 30 km from the center.

3) Ground-relative frequency

The ground-relative frequency distribution of the HBL small-scale features was calculated using the formula:
i1520-0493-136-8-2871-e2
where F(h) is the ground-relative frequency at a particular height h and λ(h) is the mean of the wavelength values between 200 and 650 m. Figure 20 is a histogram distribution of the ground-relative frequency with height for Hurricane Frances. The overall distribution ranges from 6 × 10−3 to 2 × 10−2 Hz, with a maximum around 1.3 × 10−2 Hz, and exhibits a vertical coherency over the entire analyzed depth. Similar results were found for Hurricane Isabel. In this case, while the overall frequency distribution ranged from 3 × 10−3 to 16 × 10−3 Hz with a maximum around 9 × 10−3 Hz, the maximum value was located slightly to the right of the maximum found in Hurricane Frances. This difference is likely due to the difference in Vc values. It is interesting to note that, given the rapid feature translation speed, high ground-relative frequencies could have been expected. However, the features exhibit little motion in the radial direction, and yield to relatively small ground-relative frequencies.

4) Frequency comparison

To compare the derived ground-relative frequency of the HBL features with the tower wind speed data, PSDs were generated using a nonparametric method. Two hours of tower data were considered, centered on the time of the radar analysis. The final PSD for each hurricane was generated by averaging the individual PSDs computed using 20-min sliding windows with 50% overlap. At each iteration, the mean wind speed value was removed and the fast Fourier transform (FFT) was computed.

Figures 21 and 22 are the PSDs generated using wind speed tower data from Hurricane Isabel and Frances, respectively. The ordinate represents the normalized power1 and the abscissa is the frequency. A logarithmic (Fig. 21) and a linear power axis (Fig. 22) representation of the PSDs are presented as both axes can provide different information. The shaded areas on both figures represent the ground-relative frequency range of the HBL features computed in the previous section. Some noise issue, most probably in the acquisition system, caused a leveling off of the power in the high frequencies in Fig. 22; however, this does not affect the interpretation of the data. In both cases the frequency range of the features matches almost exactly with the high-energy area in the low-frequency range of the near-surface wind PSD. This result strongly suggests that the features have an influence on the surface wind field. This assertion is strongly corroborated by the fact that the observed shift in the frequency range of the features in Hurricane Isabel and Hurricane Frances appears also in the PSDs, which reinforces the idea that these high-energy areas are linked to the HBL small-scale features.

Because the comparison of PSDs could be biased by the magnitude of the mean wind speed and because the mean surface wind speed was greater in Hurricane Frances, PSDs with reduced frequency were constructed. The reduced frequency is a nondimensional parameter obtained by multiplying the original frequency by the height at which the measurements were acquired and then dividing by the mean wind speed of the record. Reduced frequency is a valuable tool to compare spectra obtained for data collected in different conditions, such as datasets with different mean wind speeds (Geurts 1997). Figure 23 shows that the high-energy areas under investigation are at the same reduced frequencies in both storms, confirming that the small-scale features contribute to these high-energy areas at the surface and also confirming that the shift in the frequency range of the HBL features was due to the difference in the wind speed magnitude.

The spectra were compared with data from different terrain and synoptic conditions to assess the generality of the results (Van der Hoven 1957). A comparison of the PSDs generated with universal spectra (Geurts 1997) was completed to determine if these high-energy areas at relatively low frequencies are typical of standard near-surface wind flows or if it is a characteristic of hurricane near-surface wind flow. Examination of Fig. 24 illustrates two major differences. In the high-frequency portion of the spectra there is a lack of energy in the observed PSDs for both storms, partly due to the response of the R.M. Young propeller anemometer (Schroeder and Smith 2003). Another reason for such a lack of energy could be due to the normalization of the graph. Indeed, in order to keep the area under the curve equal to one, the added low-frequency energy from the small-scale features would result in a decrease of the high-frequency energy. At low frequencies (∼10−3 Hz) there is an increase of energy in the observed PSDs, relative to the universal PSDs. This excess energy in this frequency range is mentioned by Powell et al. (1996) and Schroeder and Smith (2003) as being characteristic of the hurricane spectrum and matches very well with the ground-relative frequency of the HBL small-scale features. This result implies that the features have an influence on the near-surface wind field, and result suggests that the kinematic signature of the features is an integral part of the HBL near-surface wind flow. However, because of the existence of other scales of motion, it is not clear how much of this high energy is associated with the small-scale features.

6. Conclusions

Understanding the HBL and embedded small-scale features is crucial for an accurate parameterization in NWP. This study was conducted with the intention of providing valuable information concerning the structure of the HBL coherent linear features identified in previous research. Because HBL small-scale features have been potentially identified as being responsible for the damage irregularities another goal of this study was to investigate their reflection on the near-surface wind field.

Although the conclusions drawn here might not be applicable for every tropical cyclone at landfall, the precipitation structure in Hurricanes Isabel and Frances were quite different (Hurricane Isabel exhibited relatively stratiform precipitation while Hurricane Frances was quite convective) suggesting that the small-scale features can exist in various hurricanes with different precipitation structures. The features were observed in data collected from onshore flow near the coast and at a significant distance from the eyewall in Hurricane Frances, and in offshore flow much farther inland and closer to the eyewall in Hurricane Isabel. This result indicates that the small-scale features are present in various types of flow and hurricane regions. The result agrees with Wurman and Winslow (1998) who also observed these types of features in Hurricane Fran’s eyewall.

Examination of radial velocity data from the two hurricanes revealed that HBL small-scale features were present in all inspected data. This result suggests that the features are an integral part of HBLs and not occasional features. The fact that the features are omnipresent in the HBL flow is very important and emphasizes the need to better understand their behavior for a necessary parameterization of the HBL in NWP. It should be noted that the very outskirts of each storm was not sampled since wind speeds were minimal. Hence, the acquired data do not allow conclusions about the possibility that the features require a minimal wind speed to exist.

This study offers for the first time a well-defined representation of the vertical and horizontal coherency of the HBL small-scale wind perturbations from observational data. Hurricane Isabel’s residual radial velocity data indicate that the small-scale features embedded in the BL were vertically coherent, with a more cellular-like pattern, which could be due to an increased frictional effect. In Hurricane Frances the features appeared more defined and maintained a similar vertical coherency. Individual features were not always identified with alternating positive and negative residual velocities because of the superposition of different scales of motion.

The orientation of the features was retrieved. On average, the features were oriented approximately 7° to the left of the mean near-surface wind direction. Slight veering of the features with height was observed; however, it is questionable whether or not this change of orientation with height is real or if it is due to a possible superposition of different scales of motion that would be oriented differently at different heights.

The wavelength of the features was objectively assessed. The wavelengths were similar for both hurricanes and approximately ranged between 200 and 650 m. The wavelength distribution was vertically and horizontally coherent. These results are very important as they contrast in some respects with results presented in previous numerical and observational studies. The current findings are in agreement with the study of Wurman and Winslow (1998), who documented average wavelengths of 600 m and suggested that there could be features with wavelengths as small as 100 m in the lowest levels of the HBL.

The present study also revealed that different scales of motion were present in the HBL and were superimposed on the features of interest, which could make a subjective wavelength analysis challenging. Using thresholding methods to assess the wavelength of the features significantly alters the resulting wavelength distributions. The incorporated technique allowed the analysis of small scales of motion that were overlooked by other studies and at the same time was able to document the presence of larger scales of motion. Consideration of results from this study and previous studies suggests that there is a large spectrum of coherent features with various scales of motion within the HBL.

The relationship between the near-surface wind field and the HBL small-scale features was assessed using observational data. First, a direct comparison between radar and tower time histories showed a definite correlation between the residual velocities of the features and the tower-derived radial wind speeds. An indirect method provided the ground-relative frequency of the HBL features. For both hurricanes the ground-relative frequency distribution was vertically coherent with a narrow and well-defined range frequency (from 3 × 10−3 to 16 × 10−3 Hz for Hurricane Isabel and from 6 × 10−3 to 2 × 10−2 Hz for Hurricane Frances). These frequency ranges were then compared with PSDs generated from tower-observed wind speed time histories.

The comparisons revealed that the ground-relative frequency ranges of the features determined using radar data were collocated with high-energy areas in PSDs derived from the tower data, illustrating that the features contribute to these high-energy areas. At this point it is not clear how much they contribute to these high-energy peaks in comparison with other scales of motion identified in other studies. The tower PSDs were compared with universal spectra which showed that the HBL contained excess energy at relative low-frequency ranges (2 × 10−3–2 × 10−2 Hz, 0.8–8 min) confirming results from previous studies (Powell et al. 1996; Schroeder and Smith 2003) and suggesting that the kinematic signature of the features could be a characteristic of the HBL near-surface flow.

The magnitude of the perturbations appeared to be rather modest. The coherent features were also found to be omnipresent in the HBL flow. The facts suggest that it is unlikely that the features (by themselves) could be responsible for damage irregularities. Thus, although these small-scale features have a reflection in the near-surface wind field, other phenomena with larger scales of motion are more likely to contribute in the energy transport within the HBL and cause the noted damage irregularities. Results from the present study and other previous studies (Nolan 2005; Foster 2005, Morrison et al. 2005; and Wurman and Winslow 1998) led to the hypothesis that these features could interact with other larger scales of motion, which in turn could be responsible for a more substantial amount of the HBL energy transport.

Although important contributions were made in this study, it is still not possible to completely define the HBL small-scale features. The principal reason for this is that the cross-feature flow could not be determined with the available dataset. The limited vertical extent of the analysis made it even more difficult to formally characterize the features. In some instances the features had characteristics similar to streaks. For example, they seemed to lose definition with height as seen in Khanna and Brasseur (1998) and exhibit small wavelengths (∼400 m), as reported by Drobinski and Foster (2003). The fact that the characteristics of the features are affected by the underlying roughness also suggests that the features could be streaks. However, their vertical extent, although not fully resolved, was more characteristic of the rolls. The orientation agrees with other HBL roll studies (Morrison et al. 2005; Foster 2005), but does not completely disagree with the documented orientation of streaks (Drobinski and Foster 2003). Further research is still needed on these features to allow further characterization and identify possible interaction with larger scales of motion. Moreover, the determination of the crosswise motion revealed that the features were moving in the inward direction (toward the storm center), which implies that the features might be involved in inward transport of momentum. Thus, lateral transport of energy by the features is an important subject and should be addressed in the future.

Acknowledgments

This research was supported by National Institute of Standards and Technology (Department of Commerce NIST/TTU Cooperative Agreement Award 70NANB8H0059) and the National Scientific Foundation (ATM-0134188). The lead author was also funded by the Conseil Regional de la Martinique. The authors thank Dr. Mike Biggerstaff and Gordon Carrie for their help in deploying the SMART radars and the TTU graduate students and faculty for their help during the field effort.

REFERENCES

  • Bendat, J. S., and A. G. Piersol, 1986: Random Data, Analysis and Measurement Procedures. John Wiley and Sons, 566 pp.

  • Biggerstaff, M. I., and Coauthors, 2005: The Shared Mobile Atmospheric Research and Teaching (SMART) Radar: A collaboration to enhance research and teaching. Bull. Amer. Meteor. Soc., 86 , 12631274.

    • Search Google Scholar
    • Export Citation
  • Browning, K. A., and R. Wexler, 1968: The determination of kinematic properties of a wind field using Doppler radar. J. Appl. Meteor., 7 , 105113.

    • Search Google Scholar
    • Export Citation
  • Cressman, G. P., 1959: An operational objective analysis system. Mon. Wea. Rev., 87 , 367374.

  • Drobinski, P., and R. C. Foster, 2003: On the origin of near-surface streaks in the neutrally-stratified planetary boundary layer. Bound.-Layer Meteor., 108 , 247256.

    • Search Google Scholar
    • Export Citation
  • Faller, A. J., 1961: An experimental analogy to and proposed explanation of hurricane spiral rainbands. Preprints, Second Tech. Conf. on Hurricanes, Miami, FL, Amer. Meteor. Soc., 307–313.

  • Foster, R. C., 2005: Why rolls are prevalent in the hurricane boundary layer. J. Atmos. Sci., 62 , 26472661.

  • Franklin, J. L., R. J. Pasch, L. A. Avila, J. L. Beven II, M. B. Lawrence, S. R. Stewart, and E. S. Blake, 2006: Atlantic hurricane season of 2004. Mon. Wea. Rev., 134 , 9811025.

    • Search Google Scholar
    • Export Citation
  • Fujita, T. T., 1992: Damage survey of Hurricane Andrew in south Florida. Storm Data, Vol. 34, NCDC, 176 pp.

  • Gall, R., J. Tuttle, and P. Hildebrand, 1998: Small-scale spiral bands observed in Hurricanes Andrew, Hugo, and Erin. Mon. Wea. Rev., 126 , 17491766.

    • Search Google Scholar
    • Export Citation
  • Geurts, C. P. W., 1997: Wind-Induced Pressure Fluctuations on Buildings Facades. Technische Universiteit Eindhoven, 255 pp.

  • Kaimal, J. C., and J. J. Finnigan, 1994: Atmospheric Boundary Layer Flows: Their Structure and Measurement. Oxford University Press, 289 pp.

    • Search Google Scholar
    • Export Citation
  • Khanna, S., and J. G. Brasseur, 1998: Three-dimensional buoyancy- and shear-induced local structure of the atmospheric boundary layer. J. Atmos. Sci., 55 , 710743.

    • Search Google Scholar
    • Export Citation
  • Lawrence, M. B., L. A. Avila, J. L. Beven, J. L. Franklin, R. J. Pasch, and S. R. Stewart, 2005: Atlantic hurricane season of 2003. Mon. Wea. Rev., 133 , 17441773.

    • Search Google Scholar
    • Export Citation
  • Lee, W-C., C. A. Walther, and R. Oye, 1994: Proposed Doppler radar data exchange format: DORADE. NCAR Tech. Note NCAR/TN 4031IA, 36 pp.

  • Moeng, C-H., and P. P. Sullivan, 1994: A comparison of shear- and buoyancy driven planetary boundary layer flows. J. Atmos. Sci., 51 , 9991022.

    • Search Google Scholar
    • Export Citation
  • Morrison, I., S. Businger, F. Marks, P. Dodge, and J. A. Businger, 2005: An observational case for the prevalence of roll vortices in the hurricane boundary layer. J. Atmos. Sci., 62 , 26622673.

    • Search Google Scholar
    • Export Citation
  • Nettleton, L., S. Daud, R. Neitzel, C. Burghart, W. C. Lee, and P. Hildebrand, 1993: SOLO: A program to peruse and edit radar data. Preprints, 26th Conf. on Radar Meteorology, Norman, OK, Amer. Meteor. Soc., 338–339.

  • Nolan, D. S., 2005: Instabilities in hurricane-like boundary layers. Dyn. Atmos. Oceans, 40 , 209236.

  • Powell, M. D., S. Houston, and T. Reinhold, 1992: Standardizing wind measurements for the documentation of surface wind fields in Hurricane Andrew. Proc. ASC Specialty Conf: Hurricanes of 1992, Miami, FL, ASCE, 52–68.

  • Powell, M. D., S. Houston, and T. Reinhold, 1996: Hurricane Andrew’s landfall in south Florida. Part I: Standardizing measurements for documentation of surface wind fields. Wea. Forecasting, 11 , 304328.

    • Search Google Scholar
    • Export Citation
  • Schroeder, J. L., and D. A. Smith, 2003: Hurricane Bonnie wind flow characteristics. J. Wind Eng. Indust. Aerodyn., 91 , 767789.

  • Tuttle, J. D., and G. B. Foote, 1990: Determination of the boundary layer airflow from a single Doppler radar. J. Atmos. Oceanic Technol., 7 , 218232.

    • Search Google Scholar
    • Export Citation
  • Van der Hoven, I., 1957: Power spectrum of horizontal wind speed in the frequency range from 0.0007 to 9000 cycles per hour. J. Meteor., 14 , 160164.

    • Search Google Scholar
    • Export Citation
  • Wakimoto, R. M., and P. G. Black, 1994: Damage survey of Hurricane Andrew and its relationship to the eyewall. Bull. Amer. Meteor. Soc., 75 , 189200.

    • Search Google Scholar
    • Export Citation
  • Weckwerth, T. M., J. W. Wilson, R. M. Wakimoto, and N. A. Crook, 1997: Horizontal convective rolls: Determining the environmental conditions supporting their existence and characteristics. Mon. Wea. Rev., 125 , 505526.

    • Search Google Scholar
    • Export Citation
  • Wurman, J., and J. Winslow, 1998: Intense sub-kilometer boundary layer rolls in Hurricane Fran. Science, 280 , 555557.

  • Yau, M. K., Y. Liu, D-L. Zhang, and Y. Chen, 2004: A multiscale numerical study of Hurricane Andrew (1992). Part VI: Small-scale inner-core structures and wind streaks. Mon. Wea. Rev., 132 , 14101433.

    • Search Google Scholar
    • Export Citation
  • Zhang, J. A., K. B. Katsaros, P. G. Black, S. Lehner, J. R. French, and W. M. Drennan, 2008: Effects of roll vortices on turbulent fluxes in the hurricane boundary layer. Bound-Layer Meteor., in press.

    • Search Google Scholar
    • Export Citation

Fig. 1.
Fig. 1.

Position of the experimental site with respect to the track of Hurricane Isabel. The dotted line represents approximately the path of Hurricane Isabel, and the solid circle is the maximum range set for SR2.

Citation: Monthly Weather Review 136, 8; 10.1175/2008MWR2273.1

Fig. 2.
Fig. 2.

Tangential and radial wind profiles for Hurricane Isabel at 1700 UTC obtained using collected data by SR2 located at the New Bern, NC, site. The dashed line represents the radial wind speed, and the solid line is the tangential wind speed.

Citation: Monthly Weather Review 136, 8; 10.1175/2008MWR2273.1

Fig. 3.
Fig. 3.

Position of the experimental site with respect to the track of Hurricane Frances. The dotted straight line represents approximately the path of Hurricane Frances and the solid circle is the maximum range set for SR1 and SR2, while the dashed circles represent the range of the dual-Doppler experiment that was also conducted during the field project.

Citation: Monthly Weather Review 136, 8; 10.1175/2008MWR2273.1

Fig. 4.
Fig. 4.

Tangential and radial wind profiles for Hurricane Frances at 0400 UTC obtained using collected data by SR1 located at the Merrit Island, FL, site. The dashed line represents the radial wind speed, and the solid line is the tangential wind speed.

Citation: Monthly Weather Review 136, 8; 10.1175/2008MWR2273.1

Fig. 5.
Fig. 5.

Results of the VAD technique. (a) An example of a radial velocity scan from Hurricane Frances. (b) The resulting residual radial velocity after performing the VAD technique. The black arrow indicates the near-surface wind direction.

Citation: Monthly Weather Review 136, 8; 10.1175/2008MWR2273.1

Fig. 6.
Fig. 6.

Residual radial velocities obtained from a 3.2° sector scan in Hurricane Isabel. The black arrow indicates the near-surface wind direction.

Citation: Monthly Weather Review 136, 8; 10.1175/2008MWR2273.1

Fig. 7.
Fig. 7.

Horizontal cross section of residual radial velocities at 300 m AGL for Hurricane Frances.

Citation: Monthly Weather Review 136, 8; 10.1175/2008MWR2273.1

Fig. 8.
Fig. 8.

Three horizontal cross sections (200, 500, and 1000 m AGL) of residual radial velocities from a single radar volume acquired from Hurricane Frances.

Citation: Monthly Weather Review 136, 8; 10.1175/2008MWR2273.1

Fig. 9.
Fig. 9.

(a) Vertical cross section of residual radial velocities at 2-km range from the radar in Hurricane Frances. (b) The 1:1 ratio vertical cross section, close-up of (a).

Citation: Monthly Weather Review 136, 8; 10.1175/2008MWR2273.1

Fig. 10.
Fig. 10.

Example of one of the steps of the method of minimum variance. (a) The rotation angle applied on the residual radial velocity field. The vertical solid lines indicate how the cross sections were taken. (b) The variance as a function of rotation angle. In this case the minimum in variance is located at 48°.

Citation: Monthly Weather Review 136, 8; 10.1175/2008MWR2273.1

Fig. 11.
Fig. 11.

Histogram of the difference between orientation and mean surface wind using a dataset from both hurricanes.

Citation: Monthly Weather Review 136, 8; 10.1175/2008MWR2273.1

Fig. 12.
Fig. 12.

Residual radial velocity data extracted from a cross section perpendicular to the features at 250 m AGL in Hurricane Frances.

Citation: Monthly Weather Review 136, 8; 10.1175/2008MWR2273.1

Fig. 13.
Fig. 13.

Wavelength distribution with height for Hurricane Frances. The lightest shades represent the most common wavelengths, while the darkest represent the rarest wavelengths.

Citation: Monthly Weather Review 136, 8; 10.1175/2008MWR2273.1

Fig. 14.
Fig. 14.

As in Fig. 13, but only for data located at 1 km from the radar.

Citation: Monthly Weather Review 136, 8; 10.1175/2008MWR2273.1

Fig. 15.
Fig. 15.

As in Fig. 13, but only for data located at 3 km from the radar.

Citation: Monthly Weather Review 136, 8; 10.1175/2008MWR2273.1

Fig. 16.
Fig. 16.

Example of superposition of scales of motion. (a) Vertical cross section of residual radial velocity from Hurricane Frances. The solid line indicates the height at which the data presented in (b) were extracted. (b) The thin line represents the individual residual velocity data at 350 m AGL, and the dashed line outlines the larger scales of motion superimposed on the signal.

Citation: Monthly Weather Review 136, 8; 10.1175/2008MWR2273.1

Fig. 17.
Fig. 17.

Illustration of the effect of thresholding on wavelengths: (a) The wavelength distribution for one volume scan of WSR-88 D data collected in Hurricane Fran with no thresholding applied. (b) The wavelength distribution after thresholding between −3 and 3 m s−1.

Citation: Monthly Weather Review 136, 8; 10.1175/2008MWR2273.1

Fig. 18.
Fig. 18.

Radar and tower radial wind speed comparison for Hurricane Isabel. The dashed line represents the radar data, and the solid line represents the tower data.

Citation: Monthly Weather Review 136, 8; 10.1175/2008MWR2273.1

Fig. 19.
Fig. 19.

Histogram distribution of the crosswise motion in Hurricane Frances.

Citation: Monthly Weather Review 136, 8; 10.1175/2008MWR2273.1

Fig. 20.
Fig. 20.

Histogram distribution of ground-relative frequency with height for Hurricane Frances.

Citation: Monthly Weather Review 136, 8; 10.1175/2008MWR2273.1

Fig. 21.
Fig. 21.

PSD generated from 2 h of tower wind speed data overlaid with the ground-relative frequency range of the HBL features (shaded) for Hurricane Isabel.

Citation: Monthly Weather Review 136, 8; 10.1175/2008MWR2273.1

Fig. 22.
Fig. 22.

PSD generated from 2 h of tower wind speed data overlaid with the ground-relative frequency range of the HBL features (shaded) for Hurricane Frances.

Citation: Monthly Weather Review 136, 8; 10.1175/2008MWR2273.1

Fig. 23.
Fig. 23.

Comparison of the PSD from Figs. 21 and 22 when displayed on a reduced frequency axis. The solid line represents the PSD for Hurricane Isabel, and the dashed line represents Hurricane Frances.

Citation: Monthly Weather Review 136, 8; 10.1175/2008MWR2273.1

Fig. 24.
Fig. 24.

Comparison of observed PSD from Hurricane Isabel and universal PSDs. The solid line represents Hurricane Isabel PSD while the crossed line, the dotted line, and the line with circles represent the universal spectra.

Citation: Monthly Weather Review 136, 8; 10.1175/2008MWR2273.1

Table 1.

Scanning strategies used for the radar data collection.

Table 1.
Table 2.

The 10 m-wind characteristics at 1700 UTC for Hurricane Isabel at the New Bern deployment site.

Table 2.
Table 3.

The 10 m-wind characteristics at 0000 UTC for Hurricane Frances at the Titusville deployment site.

Table 3.
Table 4.

Orientation relative to true north.

Table 4.
Table 5.

Near-surface wind characteristics from radar data

Table 5.
Table 6.

Feature translation speed and direction for Hurricanes Isabel and Frances.

Table 6.

1

It should be noted that the normalized spectra in Figs. 21 –24 are normalized by σ2, which is not the usual micrometeorological normalization.

Save
  • Bendat, J. S., and A. G. Piersol, 1986: Random Data, Analysis and Measurement Procedures. John Wiley and Sons, 566 pp.

  • Biggerstaff, M. I., and Coauthors, 2005: The Shared Mobile Atmospheric Research and Teaching (SMART) Radar: A collaboration to enhance research and teaching. Bull. Amer. Meteor. Soc., 86 , 12631274.

    • Search Google Scholar
    • Export Citation
  • Browning, K. A., and R. Wexler, 1968: The determination of kinematic properties of a wind field using Doppler radar. J. Appl. Meteor., 7 , 105113.

    • Search Google Scholar
    • Export Citation
  • Cressman, G. P., 1959: An operational objective analysis system. Mon. Wea. Rev., 87 , 367374.

  • Drobinski, P., and R. C. Foster, 2003: On the origin of near-surface streaks in the neutrally-stratified planetary boundary layer. Bound.-Layer Meteor., 108 , 247256.

    • Search Google Scholar
    • Export Citation
  • Faller, A. J., 1961: An experimental analogy to and proposed explanation of hurricane spiral rainbands. Preprints, Second Tech. Conf. on Hurricanes, Miami, FL, Amer. Meteor. Soc., 307–313.

  • Foster, R. C., 2005: Why rolls are prevalent in the hurricane boundary layer. J. Atmos. Sci., 62 , 26472661.

  • Franklin, J. L., R. J. Pasch, L. A. Avila, J. L. Beven II, M. B. Lawrence, S. R. Stewart, and E. S. Blake, 2006: Atlantic hurricane season of 2004. Mon. Wea. Rev., 134 , 9811025.

    • Search Google Scholar
    • Export Citation
  • Fujita, T. T., 1992: Damage survey of Hurricane Andrew in south Florida. Storm Data, Vol. 34, NCDC, 176 pp.

  • Gall, R., J. Tuttle, and P. Hildebrand, 1998: Small-scale spiral bands observed in Hurricanes Andrew, Hugo, and Erin. Mon. Wea. Rev., 126 , 17491766.

    • Search Google Scholar
    • Export Citation
  • Geurts, C. P. W., 1997: Wind-Induced Pressure Fluctuations on Buildings Facades. Technische Universiteit Eindhoven, 255 pp.

  • Kaimal, J. C., and J. J. Finnigan, 1994: Atmospheric Boundary Layer Flows: Their Structure and Measurement. Oxford University Press, 289 pp.

    • Search Google Scholar
    • Export Citation
  • Khanna, S., and J. G. Brasseur, 1998: Three-dimensional buoyancy- and shear-induced local structure of the atmospheric boundary layer. J. Atmos. Sci., 55 , 710743.

    • Search Google Scholar
    • Export Citation
  • Lawrence, M. B., L. A. Avila, J. L. Beven, J. L. Franklin, R. J. Pasch, and S. R. Stewart, 2005: Atlantic hurricane season of 2003. Mon. Wea. Rev., 133 , 17441773.

    • Search Google Scholar
    • Export Citation
  • Lee, W-C., C. A. Walther, and R. Oye, 1994: Proposed Doppler radar data exchange format: DORADE. NCAR Tech. Note NCAR/TN 4031IA, 36 pp.

  • Moeng, C-H., and P. P. Sullivan, 1994: A comparison of shear- and buoyancy driven planetary boundary layer flows. J. Atmos. Sci., 51 , 9991022.

    • Search Google Scholar
    • Export Citation
  • Morrison, I., S. Businger, F. Marks, P. Dodge, and J. A. Businger, 2005: An observational case for the prevalence of roll vortices in the hurricane boundary layer. J. Atmos. Sci., 62 , 26622673.

    • Search Google Scholar
    • Export Citation
  • Nettleton, L., S. Daud, R. Neitzel, C. Burghart, W. C. Lee, and P. Hildebrand, 1993: SOLO: A program to peruse and edit radar data. Preprints, 26th Conf. on Radar Meteorology, Norman, OK, Amer. Meteor. Soc., 338–339.

  • Nolan, D. S., 2005: Instabilities in hurricane-like boundary layers. Dyn. Atmos. Oceans, 40 , 209236.

  • Powell, M. D., S. Houston, and T. Reinhold, 1992: Standardizing wind measurements for the documentation of surface wind fields in Hurricane Andrew. Proc. ASC Specialty Conf: Hurricanes of 1992, Miami, FL, ASCE, 52–68.

  • Powell, M. D., S. Houston, and T. Reinhold, 1996: Hurricane Andrew’s landfall in south Florida. Part I: Standardizing measurements for documentation of surface wind fields. Wea. Forecasting, 11 , 304328.

    • Search Google Scholar
    • Export Citation
  • Schroeder, J. L., and D. A. Smith, 2003: Hurricane Bonnie wind flow characteristics. J. Wind Eng. Indust. Aerodyn., 91 , 767789.

  • Tuttle, J. D., and G. B. Foote, 1990: Determination of the boundary layer airflow from a single Doppler radar. J. Atmos. Oceanic Technol., 7 , 218232.

    • Search Google Scholar
    • Export Citation
  • Van der Hoven, I., 1957: Power spectrum of horizontal wind speed in the frequency range from 0.0007 to 9000 cycles per hour. J. Meteor., 14 , 160164.

    • Search Google Scholar
    • Export Citation
  • Wakimoto, R. M., and P. G. Black, 1994: Damage survey of Hurricane Andrew and its relationship to the eyewall. Bull. Amer. Meteor. Soc., 75 , 189200.

    • Search Google Scholar
    • Export Citation
  • Weckwerth, T. M., J. W. Wilson, R. M. Wakimoto, and N. A. Crook, 1997: Horizontal convective rolls: Determining the environmental conditions supporting their existence and characteristics. Mon. Wea. Rev., 125 , 505526.

    • Search Google Scholar
    • Export Citation
  • Wurman, J., and J. Winslow, 1998: Intense sub-kilometer boundary layer rolls in Hurricane Fran. Science, 280 , 555557.

  • Yau, M. K., Y. Liu, D-L. Zhang, and Y. Chen, 2004: A multiscale numerical study of Hurricane Andrew (1992). Part VI: Small-scale inner-core structures and wind streaks. Mon. Wea. Rev., 132 , 14101433.

    • Search Google Scholar
    • Export Citation
  • Zhang, J. A., K. B. Katsaros, P. G. Black, S. Lehner, J. R. French, and W. M. Drennan, 2008: Effects of roll vortices on turbulent fluxes in the hurricane boundary layer. Bound-Layer Meteor., in press.

    • Search Google Scholar
    • Export Citation
  • Fig. 1.

    Position of the experimental site with respect to the track of Hurricane Isabel. The dotted line represents approximately the path of Hurricane Isabel, and the solid circle is the maximum range set for SR2.

  • Fig. 2.

    Tangential and radial wind profiles for Hurricane Isabel at 1700 UTC obtained using collected data by SR2 located at the New Bern, NC, site. The dashed line represents the radial wind speed, and the solid line is the tangential wind speed.

  • Fig. 3.

    Position of the experimental site with respect to the track of Hurricane Frances. The dotted straight line represents approximately the path of Hurricane Frances and the solid circle is the maximum range set for SR1 and SR2, while the dashed circles represent the range of the dual-Doppler experiment that was also conducted during the field project.

  • Fig. 4.

    Tangential and radial wind profiles for Hurricane Frances at 0400 UTC obtained using collected data by SR1 located at the Merrit Island, FL, site. The dashed line represents the radial wind speed, and the solid line is the tangential wind speed.

  • Fig. 5.

    Results of the VAD technique. (a) An example of a radial velocity scan from Hurricane Frances. (b) The resulting residual radial velocity after performing the VAD technique. The black arrow indicates the near-surface wind direction.

  • Fig. 6.

    Residual radial velocities obtained from a 3.2° sector scan in Hurricane Isabel. The black arrow indicates the near-surface wind direction.

  • Fig. 7.

    Horizontal cross section of residual radial velocities at 300 m AGL for Hurricane Frances.

  • Fig. 8.

    Three horizontal cross sections (200, 500, and 1000 m AGL) of residual radial velocities from a single radar volume acquired from Hurricane Frances.

  • Fig. 9.

    (a) Vertical cross section of residual radial velocities at 2-km range from the radar in Hurricane Frances. (b) The 1:1 ratio vertical cross section, close-up of (a).

  • Fig. 10.

    Example of one of the steps of the method of minimum variance. (a) The rotation angle applied on the residual radial velocity field. The vertical solid lines indicate how the cross sections were taken. (b) The variance as a function of rotation angle. In this case the minimum in variance is located at 48°.

  • Fig. 11.

    Histogram of the difference between orientation and mean surface wind using a dataset from both hurricanes.

  • Fig. 12.

    Residual radial velocity data extracted from a cross section perpendicular to the features at 250 m AGL in Hurricane Frances.

  • Fig. 13.

    Wavelength distribution with height for Hurricane Frances. The lightest shades represent the most common wavelengths, while the darkest represent the rarest wavelengths.

  • Fig. 14.

    As in Fig. 13, but only for data located at 1 km from the radar.

  • Fig. 15.

    As in Fig. 13, but only for data located at 3 km from the radar.

  • Fig. 16.

    Example of superposition of scales of motion. (a) Vertical cross section of residual radial velocity from Hurricane Frances. The solid line indicates the height at which the data presented in (b) were extracted. (b) The thin line represents the individual residual velocity data at 350 m AGL, and the dashed line outlines the larger scales of motion superimposed on the signal.

  • Fig. 17.

    Illustration of the effect of thresholding on wavelengths: (a) The wavelength distribution for one volume scan of WSR-88 D data collected in Hurricane Fran with no thresholding applied. (b) The wavelength distribution after thresholding between −3 and 3 m s−1.

  • Fig. 18.

    Radar and tower radial wind speed comparison for Hurricane Isabel. The dashed line represents the radar data, and the solid line represents the tower data.

  • Fig. 19.

    Histogram distribution of the crosswise motion in Hurricane Frances.

  • Fig. 20.

    Histogram distribution of ground-relative frequency with height for Hurricane Frances.

  • Fig. 21.

    PSD generated from 2 h of tower wind speed data overlaid with the ground-relative frequency range of the HBL features (shaded) for Hurricane Isabel.

  • Fig. 22.

    PSD generated from 2 h of tower wind speed data overlaid with the ground-relative frequency range of the HBL features (shaded) for Hurricane Frances.

  • Fig. 23.

    Comparison of the PSD from Figs. 21 and 22 when displayed on a reduced frequency axis. The solid line represents the PSD for Hurricane Isabel, and the dashed line represents Hurricane Frances.

  • Fig. 24.

    Comparison of observed PSD from Hurricane Isabel and universal PSDs. The solid line represents Hurricane Isabel PSD while the crossed line, the dotted line, and the line with circles represent the universal spectra.

All Time Past Year Past 30 Days
Abstract Views 0 0 0
Full Text Views 483 143 16
PDF Downloads 370 110 15