UNIVERSITY OF HAWAI'I LIBRARY
HELICAL CIRCULATIONS IN THE TYPHOON BOUNDARY LAYER
A THESIS SUBMITIED TO THE GRADUATE DMSION OF THE UNIVERSITY OF HAWAI'I IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF
MASTER OF SCIENCE
IN
METEOROLOGY
AUGUST 2008
By Ryan T. Ellis
Thesis Committee:
Steven Businger, Chairperson Gary Barnes Thomas A. Schroeder We certify that we have read this thesis and that, in our opinion, it is satisfactory in scope and quality as a thesis for the degree of Master of Science in Meteorology.
TIIESIS COMMITTEE < ~ ft---.
ii ACKNOWLEGDEMENTS
I would like to thank Dr. Steven Businger for his guidance throughout this project and his support over the past few years. Dr. Gary Barnes and Dr. Tom Schroeder were a vital part of the editing process. Their guidance along the way has also been greatly appreciated.
I am grateful to Peter Dodge for his support with radar analysis software.
Countless homs of data processing were done by Jennifer Meehan. Special thanks to Ian
Morrison for his advice throughout this study. Level IT WSR-88D data were provided through the National Climatic Data Center. Thanks to Mark Lander and Chip Guard for their hospitality and logistical help during a field visit to Guam. This work was supported under contract from the United States Army Corps of Engineers (#W912HZ-
05-C-0036).
ill ABSTRACI'
Low-level wind data from the WSR-88D in Guam obtained in Typhoon Dale
(1996) and Typhoon Keith (1997) are analyzed for coherent structures. Consistent with the results of previous studies of Atlantic hurricanes, velocity anomalies associated with coherent structures were found in the boundary layer of both storms.
Coherent structures that are consistent with roll vortices were documented in 99 total cases during a six-hour evaluation periods dtning each storm. Storm relative roll location, roll vorticity, asymmetries in roll updrafts and downdrafts as well as differences in the upward and downward momentum fluxes associated with roll circulations are addressed in this study. A velocity azimuth display (VAD) technique is used to obtain wind profiles using Doppler velocity data to determine the depth of the roll circulation.
My work shows the effects of terrain and convective elements, such as rainbands. on the formation and maintenance of rolls. Finally, signatures of transverse circulations normal to the mean flow are explored. Convergence and divergence patterns and transverse circulations of ~5 m s·\ were observed perpendicular to the mean flow are and presented here.
This research supports and extends prior findings of roll observations and theory.
The results presented here can be used to help validate theoretical and numerical models of coherent structures within tropical cyclones. Moreover, the wind variations documented in this study have application for wave run-up and the structural wind damage potential in tropical cyclones.
iv TABLE OF CONTENTS i\ckJlovvl~eIIlellts ...... •.••...... •...... iii i\bstract...•....•...•...... ••.•.•...... •...... •. .iv
List of Tables ...•...... vi
L1st· 0 fF·19ures ••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••• V11..
Chapter 1: Introduction ...... •...... •...... 1
1.1 Previous Studies...... •... 1
1.2 Goals•.....••...... •...... •••.••••.••••••••..••.••.•.••...... ••••••••••..•••.....•.•••••••••• 6
Chapter 2: Data and Methods ...... 8
2.1 WSR-88D...... •...... •...... •...... •.•...... •....• 8
2.2 Roll Analysis Methods ...... •.....•..•...... _..•..... 8
2.3 Overvievv of Typhoons t\nalyzed...... 16
Chapter 3: Results...... •...... 19
3.1 Roll Vortices in the Typhoon Boundary Layer...... 19
Chapter 4: Conclusions and Future Work...... •.•...... 27
4.1 Conclusions...... 27
4.2 Discussion...... 29
4.3 Future Work...... ••...... •...... •..•. 31
Tables...... 33
Figures...... •.•...... ••.•.•...... ••. 35
References...... 63
v LIST OF TABLES
Table ~
1. Storm data from Typhoons Dale, Keith and Paka...... •...... •.. 33
2. Comparison of results to those of Morrison et aI. (2005) ....•...... 34
vi LIST OF FIGURES
Figure ~
1. Elevation angles and beam width ofWSR-88D...... ••...... 35
2. Example of V AD method ...... •...... •...... 36
3. Example of tracking roll vortices in WSR-88D data ...... 37
4. Exp1anation of positive and negative residuals ...... •...... •...... •.. 38
5. V AD profiles for Typhoons Dale and Keith with methodology for determining roll
depth ...... 39
6. Reliefmap of Guam showing locations ofWSR-88D radar ..••••••••••••••.•..•••••• .40
7. Map of best tracks for Typhoons Keith, Dale and Paka...... 41
8. Best track minimum sea level pressure trace of Typhoon Dale...... •...... •. 42
9. Best track minimum sea level pressure trace of Typhoon Keith...... 43
10. Best track minimum sea level pressure trace of Typhoon Paka...... 44
vii 11. Storm relative roll locations ...... •...... 45
12. Height and range of observed rolls in Dale and Keith...... •...... •...... 46
13. Positive horizontal velocity anomalies ...... 47
14. Negative horizontal velocity anomalies...... • 48
15. Roll orientation relative to the mean wind...... •...... •....•...... •.. .49
16. Roll wavelength...... •...... •...... 50
17. Vorticity observed in roll vortices ...•...•.•..•...... 51
18. Roll depth...... 52
19. Rolls observed per hour in Dale and Keith with associated fetch ...... •...... 53
20. Roll aspect ratio ...... 54
21. Radial and tangential wind velocities from VAD Profiles in Typhoons Dale and
Keith...... 55
viii 22. Updraft and downdraft in roll vortices...... 56
23. Downward momentum flux in roll vortices ...... •...... •.... 57
24. Upward momentum flux in roll vortices ...... •...... _•...... 58
25. Areal-averaged momentum flux in roll vortices ...... 59
26. Reflectivity and residual velocities in rainband and non-rainband situations...... 60
27. Transverse circulations observed perpendicular to the mean wind...... 61
28. Height and range of along-flow and normal-flow cases of rolls...... 63
ix CHAPTER 1. INTRODUCTION
1.1 Previous Studies
Roll vortices in the planetm:y boundary layer (PBL) have been documented extensively. Early laboratory work (Faller 1965) introduced the idea that dynamical as well as thermal processes may playa role in convection. Faller investigated the possibility that shear flow in the atmosphere may cause convective motions not associated with thermal processes. He concluded that an adiabatic atmospheric boundary layer should contain large eddies with a wavelength of -1-2 km in the form of roll vortices, and this roll-vortex structure should occur for a large range of conditions due to an approximately constant turbulent Reynolds number. A numerical integration of the equations of motion was evaluated to obtain critical values of a Reynolds number for laminar Ekman flow (Faller and Kaylor 1966).
Early observational work used aircraft and tower data to determine the characteristic structure and dynamics of roll vortices in the PBL. Wmd and temperature fields accompanying the rolls, their phase speeds, orientation, wind, and static stability conditions favorable for formation were investigated (LeMone 1973).
Browning and Wexler (1968) developed a technique using a single Doppler radar that estimates the wind field divergence, translation and deformation through harmonic analysis of the conventional velocity azimuth display (VAD). Since estimation errors were more pronounced at greater ranges, an optimum scanning procedure was outlined that emphasized the boundary layer. This led to the study ofhorizonta1 roll convection in various phenomena using single Doppler and dual-Doppler analysis techniques (Carbone
1985). Kelly (1982) used Doppler radar to explore cloud streets and rolls during lake
I effect snowstorms over Lake Michigan. Christian and Wakimoto (1989) observed cloud streets along the western flank of a ridge in Northeastern Colorado. Convective rolls that appeared in the radar echoes were linked to cloud streets. These rolls were present in the
top of the boundary layer and aligned with the mean wind. Mass et al. (1995) noted pulses in wind speed occurring over a period of -1 0 min during a winter storm in northern Washington that could be attributed to boundary layer rolls. The Washington
case was caused by high pressure forcing the roll circulation through a mountain gap.
Rolls have also been known to exist in association with cold air outbreaks
occurring off of the ice pack in the Bering Sea (Mourad and Walter 1996). Cloud streets
observed using satellite-based synthetic aperture radar and advanced very high-resolution
radiometer imagery showed streets spaced -5 km apart and 5 times the depth of the boundary layer. The roll circulation evolved into closed cell, mesoscale cellular
convection -200 km offshore.
Many previous studies suggested that a moderate surface sensible heat flux and
some vertical wind shear are necessary for rolls to exist (Weckwerth et aI. 1997).
Weckwerth concluded that a minimum wind speed of 5.5 m S·1 throughout the convective
boundary layer (CBL) was required for the existence of rolls. Different from prior
3 studies, Weckwerth observed rolls in low CBL shear conditions (2 X 10. S·I). Low-level
shear was found to be more important than shear across the entire CBL. Roll
wavelengths were proportional to the depth of the CBL and the roll aspect ratio, defined
as the wavelength divided by the depth, was found to increase with CBL instability. Roll
orientation was highly correlated with wind direction because no directional shear was
required for the formation of rolls (Weckwerth et al. 1997).
2 Initial theoretical results by Lilly (1966), supported the laboratory work of Faller
(1965) and presented further evidence that boundary layer rolls are the natural state in a turbulent environment An application of cloud street theory to actual atmospheric conditions indicated that in strong flows heated from below, longitudinal rolls doubled their height in -10 min. It was also specUlated that wind streaks on the ocean surface could be related to similar mechanisms as those attributed to rolls (Kuettner 1971). The observational work done by LeMone (1973) led to a theory suggesting that rolls are maintained primarily through buoyancy and the production of energy from the cross wind component of the mean PBL wind spiral. Brown (1980) stated that both theory and observations agree that convection in the presence of shear and the dynamic inflection point instabilities of the Ekman layer lead to the formation of roll vortices. A review by
Etling and Brown (1993) outlined the effects of rolls on flux modeling in the planetary boundary layer.
Glendening (1996) used a large eddy model to simulate rolls in a marine boundary layer with strong wind shear and weak buoyancy flux. Fluxes of momentum. heat, and water vapor were separated into roll and non-roll contributions to examine the importance of roll circulations in the boundary layer. Rolls were found to be transient in nature with a lifespan of -8 h and roll aspect ratios increased with time. Rolls propagated with a speed slightly slower than the vertically averaged boundary layer velocity. The study found that roll influence is largest at mid boundary level for momentmn and moisture fluxes but largest near the top of the boundary layer for heat flux.
On several occasions, damage swaths extending several hundred meters in the direction of the mean wind have been revealed in the wake of tropical cyclones (Ft:!iita
3 1992; Wakimoto and Black 1994). The damage patterns can range from a few hundred meters to a kilometer wide. Damage swaths were attributed to downward momentum transport resulting from helical roll circulations. Doppler on Wheels (Wurman and
Winslow 1998) and synthetic aperture radar (Katsaros et al. 2000) observations supplied evidence that small-scale secondary circulations with wavelengths on the order of 1 km are prevalent within the hurricane boundary layer. For the purpose of this study, the hurricane boundary layer is defined as the region of the lower atmosphere that is most
directly affected by friction against the ocean surface and the transport of sensible and
latent heat across the ocean surface (Morrison et al. 2005).
Morrison et al. (2005); hereafter (M2005), used Level IT WSR-88D Doppler radar to observe rolls in data that were collected in Hurricanes Fran (1996), Bonnie (1998), and
Georges (1998). Georges made landfalls in Puerto Rico and Key West, FL. A residual
velocity field was created by subtracting the mean velocity from the VAD. In the three
storms, 149 cases of roll vortices were documented with 56,44,24, and 25 rolls
identified in Fran, Bonnie, Georges (Key West) and Georges (Puerto Rico), respectively.
Roll statistics compiled included roll wavelength, depth. aspect ratio, wind speed,
orientation, and momentum flux. Using Doppler radar, M2005 demonstrated that the
mean boundary layer flow in hurricanes frequently contains horizontal roll vortices,
which are approximately aligned along the mean wind, and which span the depth of the
boundary layer. These rolls induce lineally organized wind speed anomalies on the order
of 10m S·1 in the direction of the mean wind extending over kilometer-scale or sub
kilometer-scale distances. An estimate of the averaged momentum flux when rolls are
present in the hurricane boundary layer is two to three times the estimate from standard
4 boundary layer parameterizations applied to the observed mean flow profiles. Rolls were present 35-69"10 of the time in the sampled Doppler radar volumes during four separate hurricane landfalls.
Foster (2005) explored the linear and nonlinear dynamics of the hurricane boundary layer to investigate the mechanisms for the formation and maintenance of rolls.
Rolls are shown to be the result of three-dimensional nonlinear instabilities of the mean boundary layer flow. While both shear and buoyancy contribute to the roll instability, the shear instability is dominant and determines the basic characteristics of the rolls. The results presented are consistent with the observations of wavelength, depth, aspect ratio, anomaly wind speed, orientation and areal-averaged momentum flux provided by M2005.
The study predicts features not yet confirmed in roll observations, such as the asymmetry in perturbation winds between roll updrafts and downdrafts, warmer near-surface air below updraft regions, and the dependence of the roll drift velocity on the radius from storm center.
Three pieces of evidence from Foster (2005) and M2005 suggest that roll vortices in a modified mean flow are the expected basic state in the hurricane boundary layer.
They are the prevalence of rolls in the observations, the excellent agreement between the nonlinear roll theory and the observed characteristics of rolls, and the conclusion from the theory that roll instabilities reach finite amplitude.
Brown and Zheng (2001) used a two-layer similarity hurricane boundary layer model to simulate roll circulations in tropical cyclones. Model results were compared to global positioning system dropsonde observations. Results show nonlinear roll effects have a large influence on the vertical transport ofmomentum. moisture, and heat.
5 In recent years, tropical cyclone track forecasting has improved dramatically where as tropical cyclone intensity forecasts have not. Numerical weather prediction models are unable to predict the influences of boundary layer rolls on tropical cyclone intensity. In these models, boundary layer rolls are not resolved or are inadequately parameterized. Inclusion of roll effects would improve upon NWP model boundary layer schemes, which could lead to more accurate tropical cyclone intensity forecasts.
1.2 Goals
Roll vortices have been suggested to be prevalent within the hurricane boundary layer and have an effect on the vertical transport of momentum, moisture and heat within a tropical cyclone (M2OOS). While rolls in tropical cyclones have been previously documented in the Atlantic, this study represents the first time roll vortices have been observed in typhoons. The island of Guam is an advantageous position for studying roll vortices in tropical cyclones becanse of its isolated position in the Pacific Ocean. This reduces the effects of landfall and allows for relatively unobstructed observation of the structure of rolls in all quadrants of the storm.
This work investigates the general hypothesis that roll vortices are prevalent within the typhoon boundary layer and can be a significant factor in storm energetics.
This study compares and contrasts the results of M200S regarding the along roll residual velocity, depth, wavelength, aspect ratio, orientation and momentwn flux. It expands the
understanding of physical characteristics of rolls beyond M200S by addressing the
vertical component of roll vorticity, asymmetries between the updraft and downdraft portions of the roll circulation, vertical momentwn transport and documenting storm
6 relative roll locations. In addition, this work illustrates how large-scale convective structures such as rainbands can affect roll circulation and the genesis of new rolls. The effects of terrain on roll circulation are also explored here. An increased knowledge of rolls in tropical cyclones should help improve the quality of boundary layer physics in tropical cyclone models and provide more insight of the low-level wind structure in tropical cyclones.
This study supports the objectives of the Pacific Island Land Ocean Typhoon
(PILOT) experiment, under contract with the U.S. Army Corps of Engineers, through analysis of Doppler radar derived low-level wind observations during iaDdfalIing typhoons. The ultimate goal of PILOT is to create a coupled air-sea wave uprush model.
Low-level wind data will be utilized by the PILOT experiment to better define the physics of coastal waves, wind-driven uprush, water-level, and terrain influenced winds to obtain an improved ability to predict the effects of these processes during a typhoon landfall. This stody has supported the PILOT experiment by providing Doppler radar derived low-level wind observations during land fa11ing typhoons, specifically VAD profiles and low level Doppler velocity fields in two typhoons.
7 CHAPTER 2. DATA AND MEmODS
2.1 WSR-88D
Tropical cyclone data were obtained from the Guam WSR-88D radar during
Typhoon Dale in 1996 and Typhoon Keith in 1997 to analyze coherent structures known as roll vortices in the typhoon boundary layer. Level II NEXRAD data were used in this study.
Data for this experiment were collected with the WSR-88D operating in precipitation mode, as is customary during tropical-cyclone passage. In precipitation mode, the WSR-88D completes a volume scan in 6 minutes. The volume scan contains data from elevation angles ranging from 0.5° to 19S obtained in 14 conical sweeps. The instrument has a range of 125 km and a radial resolution of250 m.. The angular beam width of the radar is ....().95°. For this study only the lowest six elevation angles were used (-25 s apart) to ensure that the data remain within the boundary layer (Fig. 1). Data were surveyed out to a range of 10 km. At this distance the beam width is -160 m, which represents the limit of the vertical and horizontal resolution at maximum range in this study. The Guam radar is 30 m above ground level. Crum and Alberty (1993) provide a more detailed account of the specifications of the WSR-88D.
2.2 Roll Analysis Methods
To track potential roll vortices using the WSR-88D a residual velocity field was created for each elevation angle using a VAD analysis. The residual velocity field was obtained by subtracting the Doppler radial velocity from the first harmonic of the VAD derived velocity in each radial band (Fig. 2). Once the residual velocity field was
8 acquired, a threshold of ±3 m S·1 was applied to the field to facilitate the identification of larger coherent structures and eliminate ground clutter. The residual velocity fields were then examined for pairs of adjacent positive and negative velocity anomalies. The magnitudes of residuals are usually largest within 15° of the mean typhoon boundary layer (lBL) wind direction because no directional shear is required for the existence of rolls. To ensure that an observed anomaly pair is indeed a roll circulation, it was tracked in successive scans that occur -25 s apart (Fig. 3). If the structure was preserved for at least two scans, occurring at multiple elevation angles, it was classified as a roll and its physical characteristics were recorded. From this point on, this will be the definition of ao individual roll case that is found para1lel to the mean flow. Roll signatures found perpendicular to the mean flow will be discussed later. Rolls were tracked until they ceased to exist within the residual velocity field or until they exited the 10 km range of study.
Once a roll was identified, a variety of characteristics were documented in order to better understand the physical properties associated with roll vortices aod their environment. The time aod location within the plan position indicator (PPI) that a roll occurs was recorded. The roll location was mapped relative to storm center using best track data to estimate the cyclone position. The elevation aogle at which the roll was first observed was recorded along with range from the radar. The range of the roll observation was measured from the center of the residual velocity couplet to the center of the radar.
The height of the observation was then derived from this information.
For each roll documented in the residual velocity field there is a positive and a negative radial velocity aoomaly (relative to the radar) associated with the roll
9 circulation. For the purpose of this study, positive and negative velocity anomalies
(relative to the roll) refer to the along wind variation. Positive velocity anomalies move faster than the average wind speed and negative velocity anomalies move slower than the average wind speed. To calculate positive and negative residuals of the rolls, the maximum positive and negative radial anomaly wind speeds associated with the roll were recorded. It was then determined whether the roll was moving towards or away from the radar. This affected the sign of the radial velocity. When the mean wind direction at the location the roll is observed is toward the radar, the positive radial velocity being recorded is actually the negative residual relative to the roll. The negative radial velocity recorded is then the positive residual relative to the roll When the mean wind direction is away from the radar, the positive radial velocity is the positive residual relative to the roll and the negative radial velocity is the negative residual relative to the roll (Fig. 4).
The positive and negative residuals relative to the rolls were used for all statistical calculations in this study.
Due to the fact that measurements of positive and negative residual velocities taken by the radar were at various elevation angles, it was important to note that the measurements are not exactly horizontal. This effect increases with ascending elevation angle. In this study, a measurement taken at an elevation angle of 1.4° is -98% horizontal and -2% vertical. Measurements taken at an elevation angle of 6° are -90% horizontal and -10% vertical.
The orientation of the roll with respect to the mean typhoon boundary layer wind was computed by measuring the distance in degrees between the VAD wind direction and
10 the roll orientation angle. The roll orientation angle was obtained by averaging the direction of propagation of the positive and negative residual velocities.
The distance between the positive and negative anomalies in the residual velocity field represents half of the wavelength of the roll circulation. This distance was measured from the center of the maximum positive residual velocity observation to the center of the maximum negative residual velocity observation in each roll case. The vertical component of vorticity was calculated for each roll by subtracting the negative residual velocity from the positive residual velocity and dividing by half the wavelength.
The results are both cyclonic and anti-cyclonic in nature but were assumed to be cYclonic
(positive) for this study as we were more interested in obtaining an accurate estimate of the magnitude of the vorticity.
A Fourier analysis of the VAD was done to construct wind profiles from the
Doppler radar data (Browning and Wexler 1968). Wmd speed and direction estimates were taken from 20 radial bands at 250 m resolution between 5.00 km and 9.75 km for each elevation angle (Marks et al. 1999). This allows for maximum vertical resolution and less ground clutter. This methodology was used in M2005 and allows for estimates starting at -100m above ground level. Estimates occur every -25 m up to a height of
-3.4 Ian. The VAD wind profiles were evaluated throughout the passage of the tropical cyclone and helped to determine roll depth, aspect ratio and momentom flux. VAD wind profiles were divided into radial and tangeutial component wind profiles for each storm.
The radial and tangential components in this case are relative to the radar, and not the tropical cyclone. Storm motion was removed from the profiles. The results were color coded by observation time to show the evolution of the radial and tangential wind profiles
11 as the storms passed Guam. The V AD assumes uniformity and eliminates any gradients that may occur in the space of the PPI scan. Wind speed values that make up V AD profiles are thus averages over the area of each scan rather than values of wind speed at an individual point This needs to be taken into account when comparing V AD generated results with conventional observatious.
The depth of a typhoon boundary-layer roll was estimated through the use of the
residual velocity field and the tangential component of the V AD wind profile. The
assumption that was made was that the roll circulation transports higher momentum air
downward and lower momentum air upward from the level that the maximum residual
velocity is observed in the radar data. Therefore, we can estimate the depth of the roll by matching positive and negative velocities to the mean tangential V AD wind profile to
estimate the top and bottom of the roll circulation (Fig. 5). After estimating the depth of the roll, an aspect ratio was computed by dividing the wavelength of the roll by its depth.
Based on the definition of the aspect ratio, a roll of equal width and height will have an
aspect ratio of2:1.
Mixing length theory, based on principals developed by Prandtl in 1925, was used
to estimate vertical velocity in roll circulations (M2005). The mixing length theory
assumes that turbulence is generated mechanically and so it is only valid in statically
neutral situations. The theory also assumes linear gradients of wind and moisture. In the real atmosphere these gradients are likely to remain linear only over small distances
(Stul11988). Mixing length theory states that
U ,"'- (au),iJz z
12 and au w'=k az z'
In the above equations, k is the von Karman constant and its value is 0.4. au I iJz is the magnitude of the shear over the depth of the roll and z' is the depth of the boundary layer roll circulation. u' and w' represent the horizontal and vertical velocity of the parcel, respectively. In cases where ali/ iJz > 0, w' = - /al' and when ali/ iJz < 0, mixing length theory states that w' = k u'. The mixing length equations illustrate that the horizontal wind shear over the depth of the layer is important in estimating horizontal velocity anomalies. Applying the von Karman constant to the magnitude ofthe wind shear yields an estimate of the vertical velocity anomaly associated with the roll circulation. The mixing length approach works best in surface layer cases but is weaker over the entire boundary layer. For the purpose of this study, u' is the positive or negative residual and w' represents the corresponding updraft or downdraft portion of the roll circulation. In each roll case, the updraft portion of the roll circulation is associated with the negative horizontal velocity anomaly and the downdraft is associated with the positive residual velocity anomaly.
The vertical velocity fluctuations from the mixing length approach were then used in conjunction with the horizontal velocity fluctuations obtained from the residual velocity field to estimate momentum flux associated with the updraft and downdraft portions of the roll circulation using the equation Fm = u'w'. Calculating momentum flux
13 for updraft and downdrafts within the roll circulation is useful in gauging the direct vertical transport of momentum by updrafts and downdrafts in rolls.
The momentum flux associated with rolls was also calculated using an areaI- average assuming that the flux varied across the roll and is roughly sinusoidal.
The equation for the areaI-averaged momentum flux integrates the flux with respect to a length equal to one half of the wavelength ofthe roll to quantifY the vertical transport of momentum by the entire roll circulation. The flux has a maximum in the updraft and downdraft portions of the roll and tends to zero in between (M2005). Observations and estimates of all prior mentioned statistics and their corresponding standard deviations were compared and contrasted with the results ofM2005.
Rainbands and other convective elements have been known to disrupt the roll circulation (M2005). To evaluate this, a rainband case was examined from Typhoon
Keith. Reflectivity during and after the passage of the rainband was compared with the corresponding residual velocity fields obtained from the WSR-88D. This information was used to approximate how much of the evaluated area is actually covered by roll vortices in the two situations. The geometry associated with the radar makes it very difficult to obtain precise values of how much area is covered by roll circulations. Since the radar is only looking at radial velocities, only those areas that are within --±JO° of the mean wind direction can be evaluated. The elevation angle of the scan also makes it
14 difficult to quantify roll coverage because the radar beam is sampling through different sections of the roll circulation at various ranges.
The fetch to the radar was evaluated according to terrain to see how topography and changes in roughness effected the roll circulation. The fetch was classified as either over sea. over flat land, or over mountainous terrain. The number of rolls per hour was then compared with the average wind direction and to fetch type in both Typhoon Dale and Typhoon Keith.
The flow perpendicular to the mean wind direction was analyzed in Typhoon
Keith for evidence of roll structures. Typhoon Keith was ultimately chosen for this part of the study becanse it passed closer to Guam and yielded twice as many along-flow cases of roll vortices than Typhoon Dale. Since transverse roll circulation is harder to obtain than the positive and negative residuals using the WSR-88D due to a smaller difference in the magnitude of the wind speeds, Typhoon Keith was the most logical choice. From this point on, evidence of roll circulations found perpendicular to the mean flow will be referred to as "normal-flow" cases or transverse circulations. Evidence of rolls found parallel to the mean wind will be referred to as "along-flow" cases when being compared to normal-flow cases. The goal of this analysis was to be able to detect evidence of the convergence and divergence that is present at the base or the top of the roll circulation. Linear patterns of alternating positive and negative radial velocities were documented normal to the mean flow. For a normal-flow case to be documented, at least one radar observation of converging and one observation of diverging winds had to be present perpendicular to the mean flow. Since the residual velocities in the v-direction tended to be smaller than those in the u-direction, the threshold of ±3 m S·I was reduced
15 to ±2 m S·1 to better identify coherent structures while still reducing the amount of noise in the data. The times that normal flow cases were observed were documented and compared to those observed along the mean flow. The elevation angle and range at which normal-flow cases were observed were documented so that the levels of normal flow observations could be compared with along-flow cases.
2.3 Overview o/Typhoons Analyzed
Data were available for two typhoons that passed close enouglt to Guam to observe roll characteristics. These were Typhoon Dale in 1996 and Typhoon Keith in
1997. Althouglt Typhoon Paka was a near direct hit on the island of Guam in 1997, level
II radar data were not recorded due to a power failure of the recording mechanism as the storm approached the island. It remains important to discuss the characteristics of
Typhoon Palm as it is often used as a benchmark for Typhoons affecting the island of
Guam.
The terrain of Guam is characterized by gently rolling mesa-like terrain in the northern portion of the island that slopes dramatically at the coast. Andersen Air Force
Base in extreme northern Guam is -600 ft (183 m) above sea level and is less than a haIf mile from the coast. Mataguak: Hill represents the highest point in northern Guam at
-800 ft (245 m). The terrain of southern Guam is dramatically different from the north.
Larger, more rugged mountains make up most of southern Guam and highlight the southwestern coastline. Most of the mountains in southern Guam measure between
1000-1300 ft above sea level. Mount Humuyong Manglo is the tallest peak on the island, measuring 1334 ft (407 m) above sea level. A relief map of Guam showing the locations
16 of the WSR-88D radar and the terrain of Guam is provided (Fig. 6). Best tracks (Fig. 7) and variable statistics regarding data availability, winds and pressure observations on
Guam (fable 1) are available for all three storms.
Typhoon Dale formed at the eastern end of the near-equatorial trough as the result of a westerly wind burst in October of 1996. The storm passed 203 km to the south of the
Guam radar and resulted in an estimated $3.5 million in damages to the island. Peak gusts on Guam were measured at 38 m s·' at Apra Harbor and the University of Guam in
Mangilao. The Guam WSR-88D detected winds of 5 1 m s·' at heights of 6000 - 12000 feet above the surface. The minimmn sea level pressure recorded on Guam during the passage of Dale was 987 mb. A best track minimmn sea level pressure trace is provided
(Fig.8). High seas and surf on the western coast of the island closed roads and overtopped sea cliffs measuring -30 m. Higher shear values in Typhoon Dale made it the most ideal case of the observed storms for studying the existence of rolls in the typhoon boundary layer.
Typhoon Keith formed between twin near-equatorial troughs during October of
1997. Keith formed in the Marshall Islands and recurved once it passed through the
Northern Marianas Islands. The typhoon achieved and remained at super typhoon intensity levels for 3.5 days. During that time, the storm passed between the islands of
Tinian and Rota, 129 km to the north of the Guam radar, hitting neither island directly.
As the storm passed to the north, the Guam WSR-88D observed wind gusts of72 m s" in the eyewa11. Peak wind gusts observed on Guam were 26-31 m s·'. The Joint Typhoon
Warning Center on Nimitz Hill (altitude 183 m) recorded a minimmn pressure of969 mb, which translates to a minimmn sea level pressure of 987 mb, identical to that of Typhoon
17 Dale. A best track minimwn sea level pressure trace is also provided for Typhoon Keith
(Fig. 9).
In December of 1997. Typhoon Paka formed in association with a westerly wind burst. Paka was a central Pacific storm that formed to the southwest of Hawaii and tracked due west through the Marshall Islands. The eyewa11 ofTyphoon Paka passed over the northern portions of Guam. The storm caused hundreds of millions of dollars in damage to the island. On Guam, maximum sustained winds observed on the ground were at 51 m S·1 with gusts up to 77 m S·I. The minimwn sea level pressure observed was 948 mb at Andersen Air Force Base. The best track data shows that Paka reached an estimated low pressure of -876 mb on 17 December 1997 (Fig. 10). The Guam WSR-
88D entered standby mode during the onset of the storm and could not be reset
18 CHAPTER 3. RESULTS
3.1 Roll Vortices in the Typhoon Boundary Layer
During Typhoons Dale and Keith, 99 roll vortices were documented. Thirty nine roll vortices were observed by the Guam WSR-88D as Typhoon Dale passed -200 km to the south of the island during a six-hour window from 1100 - 1700 UTC on 7 November
1996. The remaining 60 rolls were detected during the period from 0800 -1400 UTC on
2 November 1997 as Typhoon Keith passed -130 km to the north of Guam. The six-hour periods were selected for observation because of the storms' proximity to Guam.
Rolls in both storms were found at distances of 100 - 300 km from the storm center (Fig. 11). Rolls were found in all quadrants of the storm (including those rolls documented by M2005) outside of areas coDtaining larger scale features such as organized rainbands or the eyewall. All rolls were mapped relative to storm center along with those documented in M2005. The combined data from both studies does not suggest that anyone quadrant is more favorable for roll development than another. The subset from M2oo5 was closer to storm center and further supports the conclusion that rolls are prevalent within the tropical cyclone boundary layer and can form in any sector of the storm.
The height of the roll base, in this case defined as the first observation of a roll circulation in the WSR-88D. was plotted against the range from the radar (Fig. 12). The greatest amount of roll activity was found at heights of -1 00 - 200 In. This activity was dominant along the 1.4° elevation angle, at a range of3000 - 6000 m from the radar.
Roll activity decreased with increasing elevation angle and as a result, increasing height
19 When rolIs were found in the typhoon boundary layer, an analysis was performed to evaluate several physical properties associated with the roll vortices as outlined in chapter 2. Positive residuals velocities averaged -6 m S·1 for both storms combined, with a standard deviation of2 m S·1 (Fig.13). Individual results were slightly higher in
Typhoon Dale than in Typhoon Keith. The positive anomalies in Typhoon Dale averaged
-7 m S·1 with a standard deviation of -2 m S·I. In Typhoon Keith the average positive anomaly was -6 m S·1 with a standard deviation of -I m S·I.
Negative residual velocities were a similar magnitude to the positive velocity
1 anomalies, averaging - -7 m 8 with a standard deviation of2 m S·1 for both storms (Fig.
14). Negative anomaly wind speeds were higher in Typhoon Dale, averaging -8 m S·1 with a standard deviation of2 m S·1 compared with an average of -6 m S·1 and standard deviation of 2 m S·1 in Keith. Sustained wind speeds observed on Guam during the passage of Typhoon Dale were greater than during the passage of Typhoon Keith. This likely accoWlts for greater residual wind speed observations in Typhoon Dale.
Roll orientation is defined with respect to the direction of the mean wind with negative values oriented to the left of the mean wind and positive values oriented to the right of the mean wind. On average rolIs in both storms combined are oriented - _2° with a standard deviation of ....go (Fig. 15). Rolls in Typhoon Dale were oriented - -40 with a standard deviation of ....go. Rolls in Typhoon Keith were more aligned with the mean wind averaging - -1° with a standard deviation of -8°. Orientation was the only statistic surveyed that showed results arranged in a Gaussian type distribution instead of a lognormal distribution. Orientation was to the left of the mean wind in 64 of the 99 rolls observed.
20 The average wavelength observed for both storms combined was ~1350 m and varied widely with a standard deviation of ~550 m (Fig. 16). Wavelength averaged
-1250 m in Typhoon Dale and ~ 1400 m in Typhoon Keith. The standard deviation for wavelength in Typhoon Dale was -500 m compared with --600 m in Typhoon Keith.
Roll vorticity calculations in both storms combined revealed an average value of
2 1 2 2.4 x 10- 8 with a standard deviation of 1.3 x 10- S-I (Fig. 17). Vorticity averaged 29 x
10-2 S-I with a standard deviation of20 x 10-2 S-I in Typhoon Dale and 2.0 x 10-2 S-I with
3 a standard deviation of 9.0 x 10- S-1 in Typhoon Keith.
Velocity Azimuth Display (VAD) wind profiles were calculated for each roll event. The average VAD profile is used to estimate the depth of the average roll circulation as explained in the previous chapter (Fig 5). Using this method, the average roll depth was estimated to be --680 m in Typhoon Dale and - 730 m in Typhoon Keith.
To arrive at this value, height of the base of the roll circulation was subtracted from the height of the top of the roll circulation. The top ofthe roll circulation was estimated to be
-740 m in Typhoon Dale with the base estimated to be at --60 m. In Typhoon Keith the top of the roll circulation was estimated to be at -820 m with a roll base at -90 m.
Instead of first averaging the V AD profiles and using average values of the positive and negative residuals to determine the average roll depth. each individual roll depth can be evaluated using the same method as above with the specific VAD profile corresponding to the time of the roll observation using the positive and negative residual velocity from the observed roll case (Fig. 18). Using this method, the average roll depth for rolls in both storms combined was -730 m with a standard deviation of -300 m. The average depth for rolls in Typhoon Dale was ~740 m with a standard deviation of ~300
21 m. The average roll depth was -730 m in Typhoon Keith also with a standard deviation of -300 m. Similar results for roll depth in both typhoons are unexpected because flow from Typhoon Dale approaches the radar from the sea, while the flow from Typhoon
Keith approaches over more rugged terrain. The difference in roughness between the two fetches should disturb the roll cirCulation enough to yield different results for depth.
A comparison of the number of rolls formed versus the average wind direction was done to find out if the difference in fetch affected roll formation (Fig. 19). As
Typhoon Keith passed to the north of Guam, the flow approaching the radar first comes from the west, passing over a small amount of land with minimal terrain (Fig. 6). As the average wind direction shifted to the southwest, the distance from the radar to the coastline increased. The terrain in this direction is more rugged, reaching elevations above 1000 ft. During this period the number of rolls documented decreased -75% from
17 cases observed down to 4 cases observed. When the average wind direction changed to the south, on the opposite side of the ridge, roll activity increased from 4 cases observed to 8. Flow approaching the radar in Typhoon Dale was consistently from the northeast throughout the observation period Terrain in this direction is lower in elevation and less rugged. Roll observations in Typhoon Dale were less variable. A maximum at 1300 UTe coincides with the closest approach of Typhoon Dale to the
Guam radar.
Roll aspect ratios were similar for both storms with Dale averaging -2.1 with a standard deviation of 1.5 and Keith -2.2 with a standard deviation of -1.2 (Fig. 20). Roll aspect ratios for both storms averaged -2.2 with a standard deviation of -1.3. The aspect ratios show that the observed rolls are slightly wider as they are deep.
22 Radial and tangential VAD wind speed profiles, relative to the radar, were evaluated in the two storms (Fig. 21). After removing the storm motion, the radial wind profiles in Typhoon Dale exhibited low-level inflow up to an altitude of -3000 m, with outflow above. Tangential wind speeds in Dale increased from near zero at the surface to
-35 m S·l aloft. Typhoon Keith also showed inflow up to 3000 m with outflow aloft and tangential winds reached a maximum speed of -35 m S·l. Typhoon Keith reached maximum tangential wind speed at a height of -2500 m and Typhoon Dale reached maximum tangential wind speeds at an elevation of -7000 m. This result is consistent with the Joint Typhoon Warning Center's annual tropical cyclone report for Typhoon
Dale in 1996. Color-coding the profiles by observation time showed that in both storms radial inflow increased with time. Both storms also show tangential wind speeds increased with time and then decreased once the storms passed the closest approach to
Guam.
The horizontal velocity anomalies found in the residual velocity fields were used to estimate the updraft and downdraft wind speeds within the roll vortices through application of the mixing length theory (Fig. 22). Positive residuals are related to the downdraft portion of the rolls and responsible for downward momentwn transport, and the negative residuals correspond to the updraft portion of the roll circulation and thus upward transport of momentum. On average, the updrafts of the roll vortices were slightly stronger than the downdrafts. For both storms combined, updrafts averaged -3 m S·l with a standard deviation of -1 m S·l and downdrafts averaged - -3 m S·l with a standard deviation of -1 m S·l. Updraft estimates in Typhoon Dale averaged -3 m S·l with a standard deviation of -1 m s·l and downdrafts averaged - -3 m S·l with standard
23 deviation of -1. Updrafts in Typhoon Keith averaged -2 m S·1 with a standard deviation of -1 m S·1 and downdrafts averaged - -2m S·1 with a standard deviation of -1 m S·I.
The vertical velocity results allow for the estimation of the momentum flux. The average value for downward momentum flux specifically associated with the downdraft portion of the roll circulation in Typhoons Dale and Keith was -17 m2S·2 with a standard deviation of -10 m2 S·2 (Fig. 23). Downward momentum flux in Typhoon Dale averaged
- -21 m2S·2 with a standard deviation of 13 m2 S·2. Downward momentum flux in
Typhoon Keith averaged - -14 m2 S·2 with a standard deviation of -8 m2 s·2. Upward momentum flux associated specifically with the updraft portion of the roll circulation for both storms combined was 20 m2S·2 with a standard deviation of -12 m2 S·2 (Fig. 24).
Upward momentum flux for rolls in Typhoon Dale averaged -27 m2 s·2 with a standard deviation of -15 m2 s'2and averaged -16 m2s'2 with a standard deviation of.....l) m2 s·2 in
Typhoon Keith.
Estimations of areal-averaged momentum flux over the entire roll circulation averaged .....I) m2S·2 for both storms combined with a standard deviation of -5 m2 s·2 (Fig.
25). In Typhoon Dale areal-averaged momentum flux was -12 m2 S·2 with a standard deviation of -6 m2 S·2. and was -8 m2S·2 in Typhoon Keith with a standard deviation of
-4 m2 s·2. Results were consistent with the findings ofM2005 (fable 2).
One topic that M2005 addresses is that large convective features within the hurricane, such as rainbands, tend to disrupt the roll circulation and can deter the genesis of roll vortices. During Typhoon Keith a rainband was clearly captured by the Guam
WSR-88D (Fig. 26). One halfhour after the passage of the feature, the radar recorded
24 low reflectivity values. The reflectivity fields were compared with the corresponding residual velocity field in order to assess roll activity in each situation.
It is clear from the results stated regarding convective elements and terrain effects on roll circulation that the area of the tropical cyclone that is directly affected by rolls varies in different circumstances. In the residual velocity snapshots presented in Fig. 26, the areas that fall within ±30· of the mean wind direction are the areas of the WSR-88D radar scan that were analyzed for roll circulation. The geometry of the radar scan does not allow for the roll sigual to appear elsewhere in the scan with the exception of ~ :1:15· perpendicular to the mean wind. Scanning at various elevation angles will not allow the
WSR-88D to observe the entirety of the roll circulation. Of the areas that fall within 30· of the mean wind, - 60% of the area is covered with possible roll circulations in the non rainband case. This nwnber includes the areas in between and directly adjacent to the residuals, as these areas would contain portions of the roll circulation undetectable by the
WSR-88D. The area of the radar scan covered by rolls falls to -10% in the rainband case, where convective activity is clearly deterring roll genesis. There was significantly more roll activity in the low reflectivity situation than during the passage of the rainband.
It is also clear that despite the presence of the rainband, there was still some roll activity in the residual velocity field. As stated earlier, terrain has an effect on roll genesis as well and in Typhoon Keith, cut roll formation by -15% when the mean flow was over rugged terrain. A subset of residual velocity fields was surveyed for roll coverage including 10 scans at an elevation angle of 1.4· in convective situations and another 10 scans in non-convective situations in Typhoon Keith. The results support the findings of the examples presented in Fig. 26 with ~SO% of the scan covered in non-convective
25 situations with a standard deviation of :1:10%. In convective situations the results showed -100/0 coverage with a standard deviation of :1:10%.
Evidence of rolls perpendicular to the mean flow was examined in Typhoon
Keith. The results yielded 25 additional cases of apparent roll circulations. Two examples of normal flow cases of rolls observed by the WSR-88D are shown (Fig. 27).
Of the 25 normal-flow cases, 16 corresponded with observations of along-flow roll cases.
Radial wind couplets highlighting areas of convergence were compared with those highlighting areas of divergence. Normal-flow cases showed that areas of convergence were similar, but slightly stronger (1.3 x 10 -2 S-l) than areas of divergence (1.2 x 10-2 S-l).
These results highlight the near symmetry of the roll circulation and suggest that the updraft portion of the roll circulation may be slightly stronger than the downdraft portion.
Normal-flow cases of rolls were most commonly found at an elevation angle of2.4° and a range of -5 kIn (Fig. 28). This yielded a maximum residual observation height of -210 m which is in agreement with -200 m for along-flow cases. Since it is unlikely that the transverse flow observed by the radar represents the top of the roll circulation. the 200 m level could be the favored base height for rolls in this particular tropical cyclone. The average wavelength of normal-flow cases of roll vortices was 1800 In. Although this is larger than the 1400 m average wavelength from the along roll cases, the number is consistent with the theory that only the larger roll circulations will be observed by the radar using this method because the cross roll residual velocities tend to be smaller than the along roll residuals, cansing the velocity threshold to mask the smaller circulations.
26 CHAPTER 4. CONCLUSIONS AND FUTURE WORK
4.1 Conclusions
Doppler radar derived low-level winds on Guam during Typhoons Dale (1996) and Keith (1997) were analyzed for coherent structures. The results yielded the first observational analysis of roll vortices in tropical cyclones over the Pacific Ocean. The
WSR-88D radar maintained by Andersen Air Force Base, Guam. These data support the conclusion that rolls are significant in the boundary layer of tropical cyclones over the
Pacific Ocean. In total, 99 cases of roll vortices (as defined by this study) were observed.
A total of 678 Plan Position Indicator (PPI) scans were examined with roll activity present in 312 of these scans, or -46% of the time. Rolls were found to cover -SO"lo of the area scanned by the radar in non-convective situations and -10% of the area scanned in convective situations. Enhanced terrain was found to reduce the genesis of rolls by
-7S% in Typhoon Keith.
The results of this work support and extend the findings ofM200S (Table 2). The locations of rolls within the tropical cyclone were explored, including rolls documented in M200S. The location of roll observations was mapped relative to storm center.
Combining the results of this study and M2OOS, rolls were observed in all four quadrants of the tropical cyclone. Rolls in this study were most likely to be found at an elevation angle of 1.4° and a distance of 3-6 km from the radar. This translates to a height of -100
- 200 m above ground level.
A typical boundary layer roll in a hurricane environment has a wavelength of
-1400 m and a depth of-700 In, and an aspect ratio of -2, or 2:1. Positive and negative horizontal velocity anomalies, associated with roll circulations, average -7 m S·l and the
27 2 estimated areal-averaged momentum flux generated by a typical roll is 8-9 m S·2. This
2 value increases to -17 m S·2 when considering momentum flux associated with
downdraft portion of the roll only and -20 m?- {2 associated with the roll updraft.
Orientation of the rolls is consistently _2° to the left of the mean wind and the vertical
2 component of vorticity produced by rolls in on the order of 10. S·I.
Foster (2005) noted asymmetries in the analytical analysis between the updraft
and downdraft portions ofroll circulations. In the analytical solution, vertical velocities
were -1.4 ms·1 in the updrafts and --1.1 ms·1 in the downdrafts (Table 2). Thevertical
velocities estimated in Dale and Keith also showed that the estimated updrafts were
larger than the downdrafts. Rolls updrafts were estimated to be -2.1 m S·I and
downdrafts - -1.9 m S"I. Estimated values in this study tended to be larger than those
predicted by the analytical solution, as was the case in M2005 (Table 2). Theoretical results from Foster (2005) showed an average roll to have a horizontal velocity anomaly
of2.2 m S·I, wavelength oft 004 In. depth of 50Om. aspect ratio of 2.4. and orientation of
3.5" to the left of the mean wind direction.
VAD profiles revealed typical roll circulation of -700 m deep. The results from
Typhoons Dale and Keith also showed that the roll circulation as a whole can occur at
different levels in the atmosphere. Color-coded VAD profiles showed the evolution of
the radial and tangential winds in both tropical cyclones.
Evidence of roll vortices was also documented by the WSR-88D at a direction perpendicular to the mean flow. Although smaller than the positive and negative residual velocities documented parallel to the mean flow, the radar depicted 25 cases of transverse roll circulation in Typhoon Keith at average wind speeds of -5 m S·I. Convergence and
28 divergence associated with the transverse flow gives some evidence supporting the idea that updrafts in rolls may be slightly larger than the downdrafts. Transverse circulations were more likely to be found one elevation angle higher than the along flow cases of rolls.
4.2 Discussion
The results presented here are largely in agreement and supportive of M200S.
Both studies combined revealed that rolls were observed in all four quadrants of tropical cyclones. Values of horizontal wind anomalies were -1 m S·l slower than those recorded in M200S. The difference could be attributed to the rolls being further away from the storm center with slower mean flow than those in M2OOS. Roll orientation was variable but on average tended to align slightly to the left of the mean wind.
Greater depths and shorter wavelengths reveal slightly lower aspect ratios than of those in M2ooS. This could be attributed to the fact that rolls in this study were found at larger radii from the storm center where the boundary layer tends to be deeper, extending the upper bound within which rolls are capable of furming.
Presenting the momentum flux associated with the updraft and downdraft portions of the rolls is an important addition to presenting the momentum flux as an areal average over the entire circulation of rolls in order to see the momentum flux associated with roll updrafts and downdrafts. While rare, individual cases in rolls showed upward and
2 downward momentum fluxes upwards of SO m S·2. This value is valid at the level of the estimated vertical velocity maximum which varies depending on what level the roll was observed. If the vertical velocity is zero at the top and bottom of the roll circulation it
29 would tend to a maximum at half the depth of the roll circulation which would be -350 m for the rolls presented here.
This study shows that in order to assess the percentage of the area of the tropical cyclone affected by rolls. the structure of each storm and its surrounding environment must be taken into account This study concluded that rugged terrain had an adverse effect on the genesis of rolls in Typhoon Keith. This effect can increase or decrease depending on the height and the alignment of the underlying terrain. Rainbands and other convective areas in Typhoon Keith also proved to deter roll genesis. To better assess the affect of rolls on the tropical cyclone boundary layer it is first necessary to evaluate the storm structure to determine how much of the storm is covered by convective areas. Once this is determined, a better estimate of roll coverage for an individual storm can be obtained.
Transverse circuIations observed perpendicular to the mean flow by the WSR-
88D give more evidence to support the significance of rolls in the tropical cyclone boundary layer. While it would be desirable to compare the number of observed normal flow cases directly to the number of observed along flow cases, it is important to highlight a notable difference in the sampling methods. This study did not require transverse circulations to appear in successive radar scans because the propagation speed of the rolls would not allow the circuIation sampled to remain perpendicular to the mean wind direction in successive radar scans. Regardless of sampling method, the fact that there is evidence of rolls perpendicular to the mean wind qualitatively gives more support to the theory that rolls are significant to the tropical cyclone boundary layer.
30 The standard deviations of various roll characteristics showed significant variability in many of the quantities analyzed. While the goal of this study, as well as
M200S, is to form a generalized concept about what a roll in the hurricane boundary layer looks like, it is important to recognize that no two storms are exactly the same. While it is difficult to quantify exactly to what extent roll circulations affect the energy budget of the tropical cyclone, it is reasonable to argue that they represent a significant portion and should be at least parameterized, ifnot resolved in the boundary layer schemes of tropical cyclone forecast models.
4.3 Future Work
The results presented here provide a validation target for :futW'e theoretical and nwnerical modeling work on roll vortices and hurricane boundary layer structure. The results suggest that rolls are as prevalent in tropical cyclones over the Pacific basin as they are in tropical cyclones over the Atlantic and exhibit similar characteristics. A larger sample size of rolls from many storms would be ideal to more completely analyze storm-relative extent of the rolls, however the data suggest that the occurrence of rolls is equally favorable in the four quadrants of tropical cyclones. Further observational studies using Dual-Doppler radar observations of rolls in the hurricane boundary layer would be useful to perform an analysis of how the physical characteristics of rolls change in relation to the intensity of the tropical cyclone. Additional studies would be needed to evaluate how the characteristics of rolls change in the hurricane boundary layer following tropical cyclone landfall.
31 The observations documented here create a fOlUlClation for including the effects of roll vortices in tropical cyclone model forecasts. One approach for including the effects
of rolls in tropical cyclone intensity forecasts is to parameterize the effects of rolls using theoretical results. Another approach would be to use explicit results from a large eddy
simulation and include them into the boundary layer schemes of tropical cyclone models.
One example of this would be to use the National Center for Atmospheric Research large eddy simulation software for including the effects of roll vortices on the tropical cyclone boundary layer (Moeng 1984). These results could be used in hurricane simulations using the Weather Research and Forecasting (WRF) model. The data presented in this
study suggest that terrain and large convective elements, such as the rainbands of a hurricane, disrupt the development of roll vortices. A modeling study using various types of terrain and land use could be done to better understand how terrain effects roll production. As the resolution of numerical models and observing instruments continues to improve, the understanding of small-scale boundary layer features like roll vortices will ultimately lead to better tropical cyclone intensity forecasts.
32 Tables
Dale Keith Paka
RADAR Data Yes Yes No
Closest Distance to 203km 129km 56km Guam
Winds (peak Gusts on 38 ms-I 31 ms-I 77ms-1 Guam)
MSLP(Guam) 987bPa 987bPa 948bPa
Table 1. Comparisons of statistics between Typhoons Dale (1996), Keith (1997), and
Paka (1997)
33 Ellis M2005 Foster 2005
Count 99 149 N/A
Wavelength 1350m 1450m 1004m
Depth 730m 660m 500m
Aspeet Ratio 2.1 2.4 2.4
Positive Residual 6.5 m s'\ 7.0 m s'\ 2.2 m s·\ Velocity (U)
Orientation _2.0· -3.9 • _3.5·
Momentnm Flu 9.3 m2s'2 8.0~S'2 N/A (Areal Average)
Vorticity 0.024 s·\ N/A N/A
Updraft 2.75 ms·\ N/A 1.4 m 8\
Downdraft 2.50 ms'\ N/A 1.1 m s·\
Table 2. Comparisons ofphysical characteristics of roll vortices found in this study, in
M2005, and in Foster (2005). Aspect ratio is the result of the wavelength divided by the depth. Orientation is with respect to the direction of the average velocity.
34 Figures
(a) z 6.00 1051 m
4.31;1 752 m
3.3' 577 m
2.4° 419 m
1.4(1 244 m
0.40 70m
30m NOT TO SCALE
10 km Height X Above Radar at 10 km Range
(b) y
NOT TO SCALE
X
Fig. 1. WSR-88D elevation angles used in this study (a) and their corresponding heights above the radar at the maximum 10 km range used in this study. Radar beam width at maximum range (b) is also shown here.
35
13 15 11 13 9 11 7 9 5 7
~ , 3 5 -3 3 -3 -5 .- -5 -7 , ., -7 -9 -9 -11 ~ -+- -11 -13 -13 -15 RIA (ms")
Fig. 2. (a) Radial velocity (red line) at 5 krn from the radar and sinusoidal component (1st harmonic) of the radial velocity (black line) from M2005. (b) Example VAD ofWSR-
88D Doppler velocity, (c) sinusoidal component of the radial velocity, and (d) residual velocity field (d) with 2 km range rings from the Guam WSR-88D at 0955 UTe on 2
November 1997 during Typhoon Keith. Velocity scale for (c) is same as shown for (b).
In all cases, positive velocity is directed away from the radar and negative velocity is towards the radar.
36 (m S l) lJ 15 11 lJ t- 9 11 7 9 !, ,~ \ A,~ \ 7 3 \ -3 3 ~ - ~ -\ -3 -7 -\ -9 -7 - 11 -9 -lJ-ll • - 1\ - 13 RIA
Fig. 3. A roll circulation is tracked from the Guam WSR-88D during Typhoon Keith.
Each frame is - 25 s apart at subsequent higher elevation angles. Range rings are 2 km apart with hatches representing 30° swaths. Ring without hatches represents the beginning range of observations.
37 Positive Res idual
Negative Res idua l
= Positive Radial Vel. o = Negat ive Radial Vel.
Fig. 4. Schematic depicting positive and negative residuals associated with rolls moving toward and away from the radar.
38 1000r------r------,r---,------, Typhoon Dale (a) 900
800
700
_ 600 -S 1: 500 + C> iD Roll depthI -680 m I 400
300
200
100 " 50 ~~~~~~~~~~~~~~~~~~~~~~~~. a 5 10 15 20 25 30 35 40 Negati ve residuat; 7.9 m s·1 Wi nd speed (m s") Positive residual ; 7. 1 m s·1
1000 I Typhoon Ke ith (b) 900
800 ~
700
_600 - +- - -S VAD profile E 500 - .>1' Roll deplh -730 ~ I'" 400 -- +
300 Height of max residual = 203 m 200 j - 100 / \ 50 /"" J- ~ a 5 10 15 20 25 30 35 40 Windspeed(m s") \ . . . Negative residual; 6.2 m s·1 Pos itive residual ; 5.8 m s·1
Fig. 5. Average VAD profiles over the six-hour observation period for Typhoon Dal e (a) and Typhoon Keith (b). Shading indicates average depth of the boundary layer rolls.
39 Fig. 6. Relief map of Guam showing terrain of the island and highlighting the location of the WSR-88D radar.
40 BestTrack Data for Typhoons Dale (1996), Keith (1997), and Paka (1997)
15 ..
----,,..0"'600 UTe~ 2.. Nov 19~ 7
14 Paka
Q) "0 .'":J "l!uam -ctl ...J 13
+
12 1200 UTe - 7 Nov 1996 Dale
11 145 146 147 Longitude
Fig. 7. Map of Guam showing best track data for Typhoons Dale ( 1996), Keith (1997) and Paka (1 997) plotted every six hours (black dots).
41 Minimum Sea Level Pressure for Typhoon Daie (November 1996) 10~r---r---.---,----r---r--~~~---'----r---~--.---"
1000
980
920
900
~OL-__L- __~ __~ __-k __ -L __ ~ __~ __~ ____L- __L- __~ __~ 2 3 4 5 6 7 8 9 10 11 12 13 14 Time (dale)
Fig. 8. Best track minimum sea level pressure trace for Typhoon Dale with observation period highlighted in red.
42 Minimum Sea Level Pressure for Typhoon KeHh (Oct. - Nov. 1997) 10ro~--'--r-'--'-~--'--r-'--~~--'--r-'r-'--.--'--r-'
1000
980
920
900
980
m M ~ ~ V ~ ~ ~ 31 1 2 3 4 5 6 7 8 9 10 Time (da1e)
Fig. 9. Best track minimmn sea level pressure trace for Typhoon Keith with observation period highlighted in red.
43 Minimum Sea Level Pressure for Typhoon Paka (Nov. - Dec. 1997) 1020
1000
990
960 :c.s ~ 940
Q.~
920
900
880
29 3 5 7 9 11 13 15 17 19 21 Time (dale)
Fig. 10. Best track minimum sea level pressure trace for Typhoon Paka.
44 0 300 330 30
200
300 60 , 100 J , I 270 \ 90 Ke;th ( " ~ Morrison Dale ,• 240 • 120
210 150 180
Fig. II . Storm-relative roll locations for Dale (red) and Keith (blue) compared with rolls observed in M2005 (black). The zero degree mark refers to the direction of storm motion.
45 Roll Height and Range Relative to PGUA Radar 500 ~ Keith ~ Dale x 450 o x 400 cj< o 350 0 ® E 300 c§) E >® X C> ,p ~ 250 <9 00 X X <9 0 X 200 0 X ® 0 ~O >¢< 150 xO X ~~ 100 ® OO
50 ---'----- X-'------L -'- 2000 3000 4000 5000 6000 7000 8000 9000 Range 1m)
Fig. 12. Height and range of initial roll observations by the WSR-88D radar in Typhoon
Dale (red) and Typhoon Keith (blue).
46 25 25 Dale la) Keith (b) 20 Avg.= 7 20 Avg.= 6 5; 2 5= 1 C 15 C 15 ~ ~ 0 0 u 10 u 10
5 5
0 0 0 2 4 6 8 10 12 14 0 2 4 6 8 10 12 14 Wind Speed Ims- ') Wind Speed (ms- ')
30 Talal 25 Avg.= 6 20 5= 2 C il 15 u 10
5
0 0 2 4 6 8 10 12 14 Wind Speed (ms- ')
Fig. 13. Histogram of positive horizontal wind anomalies (faster than the mean wind speed) for all rolls found in Typhoon Dale (a), Keith (b), and total count for both storms
(c); s: Standard deviation.
47 25 25 Dale (a) Keith (b) 20 20 Avg.= -8 Avg.= - 6 s; 2 5= 2 15 C 15 C ~ ~ 0 0 u 10 u 10
5 5
0 0 - 14 - 12 - 10 -8 -6 -4 - 2 0 - 14 - 12 - 10 -8 -6 -4 -2 0 Wind Speed (ms- ') Wind Speed (ms- ')
30 Total (c) 25
20 5= 2 C ~ 0 15 u 10
5
0 - 14 - 12 -10 -8 -6 -4 -2 0 Wind Speed (ms- ')
Fig. 14. Histogram of negative horizontal wind anomalies (slower than the mean wind speed) for all rolls found in Typhoon Dale (a), Keith (b), and total count for both storms
(C) ; s: Standard deviation.
48 20 20 ,--- Dale Keith
15 15
C c :> 10 6 10 8 U
5 5
o ~~· o -30 -20 -10 o 10 20 30 - 30 - 20 - 10 o 10 20 30 Orientation (0)
30 ,------, Total 25
20 c 6 15 U 10
5 o -30 - 20 - 10 o 10 20 30 Orientation (0)
Fig. 15. Histogram of roll orientation relative to the mean wind direction in Typhoons
Dale (a) Keith (b), and total count for both storms (c); s: Standard deviation.
49 10 10 Oale (al l Keith (bl 8 Avg.= 1250 s= 500 E 6 E ~ ~ 0 0 u 4 u
2
0 200 1000 1800 2600 200 1000 1800 2600 Wavelength (m) Wavelength (m)
20 Total (el
15 Avg.= 1350 s= 550 C ~ 10 u0
5
0 200 1000 1800 2600 Wavelength (m)
Fig. 16. Histogram of wavelength of rolls in Typhoon Dale (a), Keith (b), and total count for both storms (c); s: Standard deviation.
50 Fig. 17. Histogram of the magnitude of roll vorticity for all rolls found in Typhoon Dale
(a), Keith (b), and total count for both storms (c); s: Standard deviation.
51 25 25 Dale (a) Keith (b) 20 Avy .= 740 20 Avg .= 730 5= 300 5=300 15 C 15 C ~ ~ 0 0 II 10 II 10
5 5
0 0 0 500 1000 1500 2000 0 500 1000 1500 2000 Depth (m) Depth (m)
35 Tolal (e) 30 Avg.= 730 25 s= 300 C 20 ~ 0 II 15 10 5 0 0 500 1000 1500 2000 Depth (m)
Fig. 18. Histogram of depth for all rolls found in Typhoon Dale (a), Keith (b), and combined results for both storms ( c); s: Standard deviation.
52 Number of Rolls per Hour in Typhoon Dale 20 Sea
§ " 0 " '0 0 0 11 00 .200 ' 300 1400 '500 '600 Hour (UTC) 360 (a) Number of Roll s per Hour in Typhoon Keith 20 Sea Flat land Mountam Flat land " 50 " .0 0 0 oeoo 0900 .000 1100 .200 ' 300 Hour (UTC) 360 (b) Fig. 19. Roll cases observed per hour with associated fetch as compared to the average V AD wind direction in Typhoon Dale (a), and Typhoon Keith (b). 53 15 15 Dale (a) Keith (b) Avg .= 2.1 Avg.= 2.2 $= 1.5 5;; 1.2 10 10 C C ~ ~ 0 0 U u 5 5 0 0 0 2 3 4 5 6 7 0 2 3 4 5 6 7 Aspect Ratio Aspect Ralio Total (e) Avg .= 2.1 s= 1.3 C 5 u 6 7 Aspect RallO Fig. 20. Histogram of aspect ratio from ro lls observed in Typhoon Dale (a), Keith (b), and total count for both storms (c); s: Standard deviation. 54 8000 ~ Oale Radial Tangential 7000 6000 5000 I S4000 ..I R.d 1 1 UTC 3000 Blue 13-15UTC Green 15 17 UTC 2000 woo - 2 (a) ,- KeIth Radial Red 8 1 UTC Blue 10 12 UTe Green 12 - 14 UTe __L - 2 (b) Fig. 2 1. Radial (left) and tangential (right) wind profiles from Typhoon Dale (a) and Keith (b). 55 50 -- . • ~-- . Dale la) Keith Ib) 40 -Avg.=-3 Avg .; 3 40 -Avg.;-2 Avg .; 2 s= 1 5= 1 5= 1 -c 30 C 30 o~ o~ u 20 u 20 10 10 oL---.. oL----' -7-6-5-4-3-2-1 0 1 2 3 4 5 6 7 -7-6-5-4-3-2-1 0 1 2 3 4 5 6 7 I Vertical Velocity (ms- ) Vertical Velocity (ms-') 70 ,------, Total Ie) 60 -Avg.;-3 Avg .; 3 50 s= 1 $;1 E 40 5 u 30 20 10 oL-.-. -7-6-5-4-3-2-1 0 1 2 3 4 5 6 7 I Vertical Velo ci ty (ms- ) Fig. 22. Histogram of updrafts (green) and downdrafts (red) estimated in rolls documented in Typhoon Dale (a), Keith (b), and total count for both storms (c); s: Standard deviation. 56 50 50 Oale (a) Keith Avg .= -21 40 Avg.= - 14 s= 13 s= 8 c C 30 o" 0" u u 20 10 o L-_~ _____ ~ -70 - 60 -50 - 40 - 30 - 20 -10 0 -80 - 70 -60 -50 -40 -30 -20 - 10 0 2 2 2 Downward Momentum Flux (m s - 2) Downward Momentum Flux (m s- ) Total (e) Avg .= - 17 s= 10 c 15 u 0 '-"--...... --- - 80 - 70 -60 -50 -40 - 30 -20 -10 0 Downward Momentum Flux (m2s-2) Fig. 23. Histogram of downward momentum flux estimated in Typhoon Dale (a), Keith (b), and combine results for both storms (c); s: Standard deviation. 57 Dale Keith (a) (b) c c a~ a~ () () 01020304050607080 o 10 20 30 40 50 60 70 80 2 2 2 Upward Momentum Flux (m s-2) Upward Momentum Flu x (m s- ) 50 ,------, Total (e) 40 C 30 5 () 20 10 o 10 20 30 40 50 60 70 80 2 2 Upward Momentum Flux (m s- ) Fig. 24. Histogram of total upward momentum flux associated with rolls in Typhoons Dale (a), Keith (b), and combined results for both storms (c); s: Standard deviation. 58 25 Dale (a) Keith (b) 20 Avg.= 12 Avg.= 8 $= 6 s= 4 c c o~ o~ U U 25 30 5 10 15 20 25 30 2 2 Areal Averaged Momentum Flux (m- 2/s-2) Areal Averaged Momentum Flux (m- /s- ) Total (e) Avg.: 9 5= 5 c 5 u o 5 10 15 20 25 30 2 2 Areal Averaged Momentum Flux (m- Js- ) Fig. 25. Histogram of areal-averaged momentum flux observed in rolls vortices in Typhoon Dale (a), Keith (b), and combined results for both storms (c); s: Standard deviation. 59 11 /2/970834 UTe Reflectivity 11 /2/97 0857 UTe (d 6z) 52 47 43 39 • 35 .~ . I \ 31 I t t + j i 27 t 22 '--+ ' . 18 14 10 +- - 6 t~ 2 Res i dual ~~~1~1~/2~/97 0857 UTe (m 5" ) 15 13 11 7 9 5 7 3 5 -3 3 -3 -5 -5 -7 -7 -9 -9 -11 -11 -13 -13 -15 RIA Fig. 26. Examples of refl ectivity (top) and residual velocity (bottom) for rainband (left) and non-rainband (right) cases in Typhoon Keith. Range rings are 2 km apart. 60 1042 UTe 2 Nov. 1997 EI 1.4° (m s ') 12 14 10 12 8 10 6 8 4 6 2 4 - 2 2 - 2 - 4 - 4 - 6 - 6 - 8 - 8 - 10 - 10 - 12 - 12 - 14 RIA Fig. 27. Examples of transverse circulations (inside red circles) detected normal to the mean wind direction (red arrow) using a 2 m S- I velocity threshold. 61 Roll Height and Range Relative to PGUA Radar 700 r-:-:--___ ----, X Keith I O Dale + Keith Normal + 600 500 + + X -P X _ 400 ()< .s 0 + E + C> '0; 0 + I 300 - c¢ + * JO >8 -+>< c§)c§) $ 0 X X+ $ X 200 00 ~$ ~ )¢(X ® X$ X ~ 100 ® OO X 0 --' J 2000 3000 4000 5000 6000 7000 8000 9000 Range (m) Fig. 28. Roll height and range from the Guam WSR-88D radar for along-flow cases of rolls observed in Typhoon Dale (red), Typhoon Keith (blue), with normal-flow cases of rolls observed in Typhoon Keith (green). 62 References Brown, R. A.,: 1980: Longitudinal instabilities and secondary flows in the planetary boundary layer: A review. Rev. Geophys. Space Phys., 18,683-697. --, and L. Zeng, 2001: Comparison of planetary boundary layer model winds with dropsonde observations in tropical cyclones. J. Appl. Meteor., 40, 1718-1723. Browning, K. A., and R. Wexler, 1968: The determination of kinematic properties of a wind field using Doppler radar. J. AppL Meteor., 7,105-113. Carbone, R. E., M. J. Carpenter, and C. D. Burghart, 1985: Doppler radar sampling limitations in convective storms. J. Atmos. Oceanic Technol., 2,357-361. Christian, T. W., and R. M. Wakimoto, 1989: The relationship between radar reflectivities and clouds associated with horizontal roll convection on 8 August 1982. Man. Wea. Rev., 117,1530-1544. Crum, T. D., and R. L. Alberty, 1993: The WSR-88D and the WSR-88D operational support facility. Bull. Amer. Meteor. Soc., 74. 1669-1687. Emanuel, K. A., 1999: Thermodynamic control of hurricane intensity. Natwe, 401,665- 669. Etling, D., and R. A. Brown, 1993: Roll vortices in the planetary boundary layer: A review. Bound-Layer Meteor., 65, 215-248. Faller, A. J., 1965: Large eddies in the atmospheric boundary layer and their possible role in the formation of cloud rows. J. Atmos. Sci, 22, 176-184. ----' and R. E. Kaylor, 1966: A numerical study of the instability of the laminar Ekman boundary layer. J. Atmos. Sci., 23,466-480. 63 Foster, R. C., 2005: Why rolls are prevalent in the hmricane booodary layer. J. Atmos. Sci., 62,2647-266l. Fujita, T. T., 1992: Damage survey of Hmricane Andrew in south Florida Storm Data, Vol. 34, No.8, 176 pp. Glendening, J. W., 1996: Lineal eddy features ooder strong shear conditions. J. Atmos. Sci., 53,3430--3449. Kaimal, J.C., 1979: Sonic Anemometer Measurement of Atmospheric Turbulence. Proc. Dynamic Flow Con/., 1978, pp. 551-65. Katsaros, K. B., P. W. Vachon, P. G. Black, P. P. Dodge, and E. W. Uhlhorn, 2000: Wind fields from SAR: Could they improve our ooderstanding of storm dynamics? John Hopkins APL Teck Dig., 211,86-93. Kelly, R. D., 1982: A single Doppler radar study of horizontal-roll convection in a lake effect snow storm. J. Atmos. Sci., 39, 1521-153l. Kuettner, J., 1971: Cloud bands in the earth's aunosphere: Observations and theory. Tellus, 23,404--425. LeMone, M. A, 1973: The structure and dynamics of horizontal roll vortices in the planetary boundary layer. J. Atmos. Sci., 30, 1077-109l. Lilly, D. K., 1966: On the instability of Ekman bowtdary flow. J. Atmos. Sci., 23,481- 490. Marks, F., P. Dodge, and C. Sandin, 1999: WSR-88D observations of hurricane aunospheric bowtdary layer structure at landfall. Preprints, 23d Con/. on Hurricanes and Tropical Meteorology, Dallas, TX, Amer. Meteor. Soc., 1051- 1054. 64 Mass, C. F., S. Businger, M. D. Albright, and Z. A. Tucker, 1995: A windstorm in the lee ofagap in a coastal mountain barrier. Mon. Wea. Rev., 123,315-331. Moeng, C-H, 1984: A Large-Eddy Simulation Model for the Study of Planetary Boundary-Layer turbulence. J. Atmos. &i., 41, 2052-2062. Morrison, I., S. Businger, F. Marks, P. Dodge, J. A. Businger, 2005: An observational case for the prevalence of roll vortices in the hmricane boundary layer. J. Atmos. &i.. 62, 2662-2673. Mourad, P. D., and B. A. Walter, 1996: Viewing a cold air outbreak using satellite-based synthetic aperture radar and Advanced Very High Resolution Radiometer imagery. J. Geophys. Res., 101, 16391-16400. Stull, R. B., 1988: An Introduction to Boundary Layer Meteorology. Kluwer Academic, 206-208. Wakimoto, R. M., and P. G. Black, 1994: Damage survey of Hmricane Andrew and its relationship to the eyewall. Bull. Amer. Meteor. Soc .• 75, 189-200. Weckwerth, T. M, J. W. Wtlson, R. M Wakimoto, and N. A. Crook, 1997: Horizontal convective rolls: Determining the environmental conditions supporting their existence and characteristics. Mon. Wea. Rev.• 125,505-526. Wilks D., 2005: Statistical Methods in the Atmospheric &iences. Academic Press, 648 pp. Wroe, D. R., and G. M. Barnes, 2003: Inflow layer energetics ofHmricane Bonnie (1998) near landfall. Mon. Wea. Rev., 131, 1600-1612. Wurman, J., and J. Wmslow, 1998: Intense sub-kilometer boundary layer rolls in Hmricane Fran. &ience, 280, 555-557. 65 66