of in Advanced Hurricane WRF Model Simulations Diamilet Pérez-Betancourt

Academic Affiliation, Fall 2011: Senior, University of at Mayagüez

SOARS® Summer 2011

Science Research Mentor: Christopher Davis Writing and Communication Mentor: Kimberly Kosmenko

ABSTRACT Rapid intensification (RI) is a form of tropical (TC) intensification that is particularly challenging to predict. Understanding RI is essential for improving hurricane intensity forecast skill and hurricane preparedness. However, the physical processes that lead to RI are not well understood. Hurricane Earl, the second-strongest TC of the 2010 North Atlantic basin season, underwent RI on August 29, in its first day as a hurricane. The Advanced Hurricane Weather Research and Forecasting (AHW) model produced a simulation of Earl that was unsuccessful in predicting intensification, followed closely by a simulation successfully forecasting RI. The purpose of this study is to investigate the environmental and structure characteristics that led to Earl’s simulated RI. The unsuccessful and successful simulations were first compared in terms of the environmental vertical , because this parameter is often negatively correlated with TC intensification. Area averages of the deep-layer and mid-layer vertical wind shear over the storm suggest that this parameter did not influence the intensification of the successful simulation. Initially, the successful simulation featured greater relative humidity throughout the troposphere within the storm’s circulation. This led to a greater upward mass flux throughout the troposphere and a rapid intensification of circulation in the middle troposphere before any significant change occurred at the surface. After this deep vortex had been established, the mass flux (and lower-tropospheric convergence) continued to increase, leading to RI. These results provide a basis for further research to better understand and predict the development of RI.

The Significant Opportunities in Atmospheric Research and Science (SOARS) Program is managed by the University Corporation for Atmospheric Research (UCAR) with support from participating universities. SOARS is funded by the National Science Foundation, the National Oceanic and Atmospheric Administration (NOAA) Climate Program Office, the NOAA Oceans and Human Health Initiative, the Center for Multi-Scale Modeling of Atmospheric Processes at Colorado State University, and the Cooperative Institute for Research in Environmental Sciences. 1. Introduction Tropical (TCs), named hurricanes in the North Atlantic and Eastern Pacific basins, are among Earth’s most devastating natural phenomena. This threat has motivated the efforts to better predict the TC’s track and intensity changes. Despite these efforts, predictions of TC intensification continue to have limited success (DeMaria et al. 2007). One particularly difficult to predict form of TC intensification is rapid intensification (RI), which the National Hurricane Center (NHC) defines as an increase in the maximum sustained surface wind speed of at least 15.4 ms-1 in a 24-hour period (NHC 2011). The physical mechanisms that lead to RI are still poorly understood.

Several studies have assessed the processes that influence TC development and intensification (e. g., Gray 1968, Merrill 1988, Frank and Ritchie 1999). However, only a few studies have explicitly examined RI (Kaplan and DeMaria 2003). We base our study on the factors that have been generally related to TC intensification. These factors, which are briefly reviewed below, may be categorized as: atmosphere-ocean interactions, environmental influences, and inner core dynamics (Bosart et al. 2000).

Studies focusing on atmosphere-ocean interactions have recognized (SST) as an indicator of the maximum potential intensity of hurricanes (e.g., Malkus and Riehl 1960; Emanuel 1986). High SSTs enhance the latent heat flux, or transfer of energy, between the sea surface and the atmosphere. This flux supports , a mechanism for cloud formation, as water evaporates from the ocean surface. Therefore, high SSTs are expected to be favorable for RI. Nevertheless, Merrill (1988) argued that most hurricanes never reach the SST-dependent maximum intensity. He suggested that unfavorable atmospheric conditions prevent these TCs from intensifying. Thus, Merrill focused on the TC environment-structure interactions. He compared the atmospheric conditions surrounding intensifying and non-intensifying hurricanes with current intensity far below their maximum potential. Merril’s findings indicate that low environmental vertical wind shear is also favorable for TC intensification. This conclusion has been confirmed by more recent studies (Frank and Ritchie 1999, 2001). Thus, the general influence of large-scale environmental flows over TC intensification has been well established. However, we still have much to learn about the specifics of the complex physical processes involved in this interaction, specifically during RI events.

Studies of the inner core dynamics of TCs link intensity changes with eyewall replacement cycles (e. g., Shapiro and Willoughby 1982, Willoughby et al. 1982). Other studies link TC intensification with potential vorticity anomalies (e.g., Montgomery and Kallenbach 1997, Molinari et al. 1998, Möller and Montgomery 2000). Numerical Weather Prediction (NWP) models, such as the Advanced Hurricane Weather Research and Forecasting (AHW) model, are utilized to predict the evolution of the TC inner core and TC intensity changes. These models are limited by initial condition errors related to the representation of the TC inner structure (Torn 2010). Understanding the TC inner core dynamics, as well as their relationship with the environment, is necessary to predict RI occurrence. Davis et al. (2008) evaluated the performance of the AHW in forecasting such rapid intensity changes during the 2005 North Atlantic season. The authors found

SOARS® 2011, Diamilet Pérez-Betancourt, 2 that the AHW had difficulty capturing RI, particularly when RI occurred soon after initialization.

We perform a case study of Hurricane Earl’s RI using AHW model simulations described by Davis et. al. (2010). According to the NHC TC report (Cangialosi 2011), Earl was one of the five major hurricanes of the 2010 North Atlantic season that experienced RI. Data from reconnaissance flights and satellite imagery indicate that Earl’s maximum sustained surface wind speed went from 38.6 ms-1 (August 29, 18:00 UTC) to 59.2 ms-1 (August 30, 18:00 UTC), increasing 20.6 ms-1 over a day. Interestingly, the AHW produced a simulation of Hurricane Earl with almost no intensification (0.7 ms-1 increase in maximum wind in 24-hours) followed closely by a simulation that captured the RI event more accurately (14.0 ms-1 increase in maximum wind in 24-hours) (Figure 1).

Figure 1. Evolution of Hurricane Earl’s maximum wind speed. The black curve shows the official estimate from the NHC. The red curve shows the results of the AHW simulation starting in August 27, which produced almost no intensification. The blue curve shows the results of the AHW simulation starting in August 28, which followed closely the observed RI.

The purpose of this study is to assess the environmental and structural characteristics that led to Hurricane Earl’s simulated RI. Our approach is to first compare the aforementioned unsuccessful and successful AHW simulations on a large-scale by examining the environmental vertical wind shear. We chose this parameter because it is often considered an unfavorable factor for TC intensification. Vertical wind shear can advect the heat and moisture away from the inner core of the storm necessary for convection (Gray 1968), or can tilt the vortex, producing an anomaly that inhibits convection (DeMaria 1996).

SOARS® 2011, Diamilet Pérez-Betancourt, 3

Subsequently, we compare the simulations in terms of Hurricane Earl’s vortex characteristics. We examine the local rate of change of the relative vorticity (ζ) in isobaric coordinates , which is given by: ! !! ! !! ! ! %!V ( = "V •#(! + f ) "" " (! + f )#•V + k •' $ #"* (1) t p p ! ! & ! ) where p is the pressure, is the horizontal wind velocity vector, is the absolute vorticity or Coriolis parameter, and ω is the vertical wind velocity in pressure coordinates (!p / dt ). The area-average of the relative vorticity is defined as the circulation of the storm. Thus, we examine the vorticity equation (1) to identify the processes by which the successful simulation developed a stronger circulation (Figure 2).

Initialized at 2010-08-27_00:00:00 Initialized at 2010-08-28_00:00:00 (a) (b)

Figure 2. Horizontal wind field of Hurricane Earl at August 30 00:00 UTC from (a) the unsuccesful simulation, and (b) the successful simulation. The successful simulation developed the typical hurricane structure, including the and the highest winds in the eyewall.

Our expectation is that the mechanism by which the AHW rapidly intensified Earl in the successful simulation is the mechanism by which the observed Earl rapidly intensified. To verify this, our analysis needs to be validated with observations of the storm’s environmental and structural characteristics. This analysis contributes to the understanding of how rapid intensification occurs, therefore improving the accuracy of intensity forecasting.

The rest of the paper is organized as follows. Section 2 provides a synopsis of the history of Hurricane Earl. Section 3 describes the data set and the methods applied to

SOARS® 2011, Diamilet Pérez-Betancourt, 4 analyze Earl’s simulated RI, while section 4 discusses the results of this analysis. We provide some concluding remarks in section 5.

2. Overview of Hurricane Earl

Earl had a long life as a TC, reaching TC status on August 25 and moving through phases as a tropical storm, hurricane and major hurricane until September 4 (Figure 2). According to the NHC TC report (Cangialosi 2011), Earl developed from a that moved off the African coast on August 23. The wave strengthened to a tropical depression on August 25. Later that day, the depression was declared a tropical storm south of the Islands. It was the third storm of the 2010 season to form in the Main Development Region (10-20 N, 20-60 W). The storm moved westward to northwestward through the Atlantic until it became a Category 1 (Saffir-Simpson scale; Simpson 1974) hurricane on August 29 (12:00 UTC) east to the northern Caribbean . Then, it rapidly intensified and was declared a Category 4 hurricane on August 30 (18:00 UTC).

Figure 3. Best track positions for Hurricane Earl based on the analyses from the NHC, the National Oceanic and Atmospheric Administration’s Ocean Prediction Center and the Canadian Hurricane Centre. The gray shadow marks the period when Earl experienced RI (August 29 18:00 UTC- August 30 18:00 UTC).

SOARS® 2011, Diamilet Pérez-Betancourt, 5

After the RI event, the hurricane underwent a concentric eyewall replacement cycle, as observed by the Puerto Rico Doppler radar and reconnaissance aircrafts. Earl briefly weakened to Category 3 on September 1 as it continued moving northwestward over the Atlantic waters. It re-intensified to Category 4 by 18:00 UTC southeast of . As it turned northward, Earl rapidly weakened to Category 2 intensity (September 3). On September 4, it made as a Category 1 hurricane on , , and as a tropical storm on . Finally, it became an extratropical system by September 5 and merged with an upper-level low on September 6.

The meteorological and damage statistics indicate that Earl produced from 3 to 5 inches of rainfall over the northern Caribbean Leeward Islands, as well as over portions of North Carolina, , and . It also produced up to 3 ft of surge across the coast from North Carolina to Maine. Five casualties were reported related to Earl’s surge in different parts of East Coast of the . Damage from this hurricane was estimated at 45 million dollars. Unlike other case studies that focus on higher-impact TCs, we chose to analyze Earl because it provides an opportunity to better understand the RI process. RI could contribute to greater damage by future land-threatening hurricanes.

3. Methods

The simulations of Hurricane Earl analyzed in this study were produced by the AHW model, which is derived from the Advanced Research Weather Research and Forecasting (ARW) model (Davis et al. 2008). These simulations, based on ARW version 3.2, were generated as part of the National Oceanic and Atmospheric Administration’s Hurricane Forecast Improvement Project (HFIP). One goal of HFIP is to improve the RI forecast skill (Davis et al. 2010). The HFIP does this by performing retrospective model runs for testing. The retrospective forecasts are simulations produced after the hurricane season with the same data available for initialization that existed in real time, but with an improved version of the model. In other words, an upgraded version of the AHW was used to reproduce forecasts for the past season. In this study, two retrospective simulations of Hurricane Earl were examined.

For these runs, the AHW was initialized via an ensemble Kalman filter data assimilation system consisting of 96 members (Torn 2010). An outer domain of coarse 36 km horizontal resolution was used to simulate the Atlantic basin. Inside this parent domain, a double-nested domain was placed to generate high resolution forecasts. These nests had horizontal resolutions of 12 km and 4 km. They were centered on each storm as described by Davis et al. (2008). We examined the model output from the coarse domain because it provides better spacial coverage to study large-scale processes, and from the 4 km nested domain because it has the best spacial resolution to examine the storm’s inner structure.

SOARS® 2011, Diamilet Pérez-Betancourt, 6 The initial time of the AHW unsuccessful simulation is August 27, 2010 at 00:00 UTC. This simulation produced only a 0.7 ms-1 maximum wind increase during the 24 hours when the observed RI of Earl occurred. The initial time of the sucessful simulation, is August 28, 00:00 UTC. This simulation was closer to the observations, producing a 14.0 ms-1 maximum wind increase. Although these forecasts go out to at least 5 days, we hypothesize that the differences between the simulations can be identified within the first 72-hours. Therefore, we examined the parameters near the time the forecasts were initialized.

The environmental and inner core characteristics of both successful and unsuccessful simulations of Earl were examined by computing area-averages of the quantities over four different radius (r) from the center of the storm (200, 300, 400 and 500 km). This was done to vertify that the differences between the simulations were consistent at multiple scales. The storm center was obtained from the AHW track forecasts (Figure 4). There is no significant difference in SST along the two paths during the first 72 hours, therefore it is not likely that the track difference caused the intensity difference. Rather, the tracks diverged after the difference in intensity became clear. Therefore, the differences in track are not analyzed in detail in this study.

Figure 4. Hurricane Earl’s track from the AHW unsuccesful simulation (red line) and the successful simulation (blue line).

The influence of the environmental vertical wind shear was examined by calculating the area average of the wind vector difference between two pressure levels. The calculation of the standard deep-layer wind shear (850-200 hPa) was followed by the calculation of the mid-layer wind shear (850-500 hPa). The mid-layer wind shear could

SOARS® 2011, Diamilet Pérez-Betancourt, 7 be more relevant at early stages of TC development where the circulation may not extent to 200 hPa.

The storm’s inner structure was studied by examinining the divergence ! (stretching) term of the vorticity equation (1), !(! + f )"•V . The divergence term can change the relative vorticity in the model runs in two different ways: (i) At similar divergence, the simulation that starts with a stronger cyclonic circulation (higher relative vorticity) near the surface will continue to have a stronger circulation. (ii) At similar initial vorticity, the local rate of change of the relative vorticity will depend on the divergence or convergence of air.

The divergence of a fluid can be expressed through the mass continuity equation as:

! #! !•V = " (2) #p Under hydrostatic balance,

! #! #w !•V = " $ " #p #z where w is the vertical wind velocity in height coordinates ( dz / dt ). Upward motion is associated with convergence near the surface, while downward motion is associated with divergence near the surface.

The mass flux (σ), or area-average vertical transport of mass, indicates whether upward or downward motion is occurring in the atmosphere. This parameter is given by:

! = "#A (3) where ρ is the air density and A is the averaging area. Vertical profiles of the mass flux were compared between the two simulations to identify convergence or divergence of air. The mass flux can be affected by the water vapor content near the clouds. Entrainment of dry air into the inner core in the mid-troposphere can produce downdrafts and divergence of air near the surface. Thus, vertical profiles of the average relative humidity were also produced for both simulations. We examine the environmental influences of the relative humidity by using an averaging radius of 400 km and at the initial time. For an averaging radius of 200 km, the storm evolution strongly influences the humidity profile.

The influence of the advection terms of the vorticity equation described by the first and second terms in (1) was left for future investigation. The horizontal advection of absolute vorticity may be related to the environmental vertical wind shear. Finally, the influence of the tilting term (fourth term in equation (1)) was not considered because it is typically less important on horizontal scales of 100 km or more.

SOARS® 2011, Diamilet Pérez-Betancourt, 8 4. Results and Discussion a. Environmental characteristics

Time series of the average deep-layer vertical wind shear are shown in Figure 5a. For all four averaging radii, and for both simulations, the initial vertical wind shear was low (1-3 ms-1) compared to what is commonly considered unfavorable for TC intensification (about 10 ms-1). While the unsuccessful simulation experienced stronger deep-layer wind shear, it did not do so until after the intensities in the two simulations started diverging from each other. These results suggest that the wind shear did not influence significantly the intensification of the successful simulation. Time series of the average shallow-layer vertical wind shear (Figure 5b) confirm this result. For all four averaging radii, the shallow-layer wind shear was similar for both simulations at the beginning. After the first 12 hours into the successful simulation (24 hours into the unsuccessful simulation), the shallow-layer wind shear became stronger in the successful simulation than in the unsuccessful simulation.

(a)

(b)

Figure 5. Average (a) 850-200 hPa vertical wind shear and (b) 850-500 hPa vertical wind shear for the unsuccessful simulation (red lines) and the successful simulation (blue lines). The environmental vertical wind shear was not significantly different between the simulations at initial times.

SOARS® 2011, Diamilet Pérez-Betancourt, 9 b. Inner structure characteristics

Vertical profiles of the circulation for both simulations are shown in Figure 6. The successful simulation initiall had a stronger circulation than the unsuccessful simulation throughout the troposphere. By examining the stretching term of the vorticity equation (1), we could anticipate that the successful simulation would develop an even stronger circulation since the circulation was initially stronger.

(a) (b)

Figure 6. Circulation (area-average of the relative vorticity) for the unsuccessful simulation (red lines) and the successful simulation (blue lines) with averaging radius of (a) 200 km and (b) 400 km. The successful simulation started with a stronger circulation throughout the troposphere.

Vertical profiles of the relative humidity are shown in Figure 7. The unsuccessful simulation started with a drier profile throughout the troposphere. Dry air in the mid- troposphere makes the atmosphere succeptible to downdrafts (downward air motion). These downdrafts produce divergence in the lower troposphere, a negative factor for increasing cyclone intensity.

SOARS® 2011, Diamilet Pérez-Betancourt, 10

(a) (b)

Figure 7. Average relative humidity for the unsuccessful simulation (red lines) and the successful simulation (blue lines) with averaging radius of (a) 200 km and (b) 400 km. The successful simulation started with a higher relative humidity throughout the troposphere.

The vertical profiles of the mass flux, which were computed to examine the divergence in the stretching term of the vorticity equation (1), are shown in Figure 8. The mass flux increased rapidly in the mid-troposphere (strong mid-tropospheric convergence developed) within the first 24 hours in the successful forecast. This occurred before any significant change in the circulation occurred near the surface (Figure 6). We hypothesize that the rapid increase in the mass flux at the mid-troposphere is related to the initial higher moisture content in the successful simulation. At the same time, a rapid increase in the relative humidity at mid-levels was related to this rapid increase in mass flux in the successful simulation. Convergence of air at the surface is necessary for convection to occur. The convection itself moistens the air at mid-upper levels. Although the unsuccessful simulation experienced some moistening, the air was still dry compared to the successful simulation.

SOARS® 2011, Diamilet Pérez-Betancourt, 11 (a) (b)

Figure 8. Mass flux (area-average of the mass vertical transport) for the unsuccessful simulation (red lines) and the successful simulation (blue lines) with averaging radius of (a) 200 km and (b) 400 km. The mass flux rapidly increased in the successful simulation at the mid-troposphere before the surface RI occurred.

A rapid increase of the circulation at the mid-troposphere also occurred during the first 24 hours in the successful forecast. After this deep vortex had been established, mass flux (and lower tropospheric convergence) continued to increase. After 48 hours, the successful simulation developed the typical mass flux profile of a mature hurricane, with convergence in the lower and middle troposphere (positive slope) and divergence aloft (negative slope). This lead to Hurricane Earl’s RI at the surface.

5. Conclusions

The purpose of this study has been to identify the environmental and storm structure characteristics that led to Hurricane Earl’s simulated RI in the AHW model. The AHW model produced this successful simulation following a simulation that was unsuccessful in predicting Earl’s RI. We compared the two simulations by studying the environmental vertical wind shear and the vorticity equation. Our expectation is that the mechanism by which the AHW model produced Earl’s RI is the mechanism by which the observed RI of Hurricane Earl occurred. Therefore, we expect to provide insight into RI occurrence. On the large scale, our results suggest that the environmental vertical wind shear did not influence significantly the RI of the successful simulation. On the storm scale, we

SOARS® 2011, Diamilet Pérez-Betancourt, 12 found that the relative humidity and circulation were initially stronger for the successful simulation than for the unsuccessful simulation. We concluded that the high moisture content in the mid-troposphere in the successful simulation eased the rapid increase of the mass flux at this level. As a result, the humidity and circulation also increased rapidly at the mid-troposphere. RI at the lower troposphere followed. Although there is still much to examine, these results provide a basis for further research to better understand the development of RI. Future work should include examination of other structural aspects of the storm’s thermodynamics. In addition, it would be important to verify the results with other AHW simulations of Earl and compare the results to available observations to determine if the successful simulation was successful for the right reasons. Finally, the results should be verified by examining other that experienced RI.

6. Acknowledgements

This work was performed under the auspices of the Significant Opportunities in Atmospheric Research and Science Program. SOARS is managed by the University Corporation for Atmospheric Research and is funded by the National Science Foundation, the National Oceanic and Atmospheric Administration, the Cooperative Institute for Research in Environmental Sicence, and by the Center for Multiscale Modeling of Atmospheric Processes. The first author would like to thank Christopher Davis for his extensive collaboration with this project. Kimberly Kosmenko’s comments on the manuscript throughout its evolution are also much appreciated. Special thanks are extended to SOARS staff Rajul Pandya, Rebecca Haacker-Santos, and Moira Kennedy, and to the Mesoscale and Microscale Division. Ryan Torn (University at Albany-State University of ) produced the model output that was utilized in this project. Sherrie Fredrick and David Ahijevych kindly provided help with the NCAR Command Language scripts. Finally, the 2011 SOARS protégés and Andrea Feldman’s suggestions were quite helpful in improving the manuscript.

7. References

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

Cangialosi, J. P., 2011: report, Hurricane Earl, 25 August- 4 September 2010. NOAA/National Hurricane Center, 29 pp.

Davis, C., W. Wang, S. S. Chen, Y. Chen, K. Corbosiero, M. DeMaria, J. Dudhia, G. Holland, J. Klemp, J. Michalakes, H. Reeves, R. Rotunno, and Q. Xiao, 2008: Prediction of landfalling hurricanes with the Advanced Hurricane WRF Model. Mon. Wea. Rev., 136, 1990-2005.

_____ , _____ , and J. Dudhia, 2010: Does increased horizontal resolution improve hurricane wind forecasts?. Wea. Forecasting, 25, 1826-1841.

SOARS® 2011, Diamilet Pérez-Betancourt, 13

DeMaria, M., 1996: The effect of vertical shear on tropical cyclone intensity change. J. Atmos. Sci., 53, 2076-2087.

_____ , J.A. Knaff, and C. Sampson, 2007: Evaluation of long-term trends in tropical cyclone intensity forecasts. Meteor. Atmos. Phys., 97, 19-28.

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

Frank, W. M., and E. A. Ritchie, 1999: Effects of environmental flow upon tropical cyclone structure. Mon. Wea. Rev., 127, 2044-2061.

_____ , and _____ , 2001: Effects of vertical wind shear on the intensity and structure of numerically simulated hurricanes. Mon. Wea. Rev., 129, 2249-2269.

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

Kaplan, J., and M. DeMaria, 2003: Large-scale characteristics of rapidly intensifying tropical cyclones in the North Atlantic basin, Wea. Forecasting, 18,1093-1108.

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

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

Molinari, J., S. Skubis, D. Vollaro, F. Alsheimer, and H. E. Willoughby, 1998: Potential vorticity analysis on tropical cyclone intensification. J. Atmos. Sci., 55, 2632-2644.

Moller, D. J., and M. T. Montgomery, 2000: Tropical cyclone evoluton via potential vorticity anomalies in a three-dimensional balance model. J. Atmos. Sci., 57, 3366-3387.

Montgomery, M. T., and R. J. Kallenbach, 1997: A theory for vortex Rossby waves and its application to spiral bands and intensity changes in hurricanes. Quart. J. Roy. Meteor. Soc., 123, 435-465.

National Hurricane Center, cited 2011: Glossary of NHC terms. [Available online at http://www.nhc.noaa.gov/aboutgloss.shtml.]

Shapiro, L. J., and H. E. Willoughby, 1982: The response of balanced hurricanes to local sources of heat and momentum. J. Atmos. Sci., 39, 378-394.

Torn, R. D., 2010: Performance of a mesoscale ensemble Kalman filter (EnKF) during the NOAA High-Resolution Hurricane Test. Mon. Wea. Rev., 138, 4375-4392.

SOARS® 2011, Diamilet Pérez-Betancourt, 14

Willoughby, H.E., J.A. Clos, and M.G. Shoreibah, 1982: Concentric eyewalls, secondary wind maxima, and the evolution of the hurricane vortex. J. Atmos. Sci., 39, 395-411.

SOARS® 2011, Diamilet Pérez-Betancourt, 15