• Avila, L. A., 2009: Tropical cyclone report: Hurricane Cristobal. NOAA/National Hurricane Center Tech. Rep. AL032008, 8 pp. [Available online at http://www.nhc.noaa.gov/data/tcr/AL032008_Cristobal.pdf.]

  • Avila, L. A., , and Cangialosi J. , 2012: Tropical cyclone report: Hurricane Irene. NOAA/National Hurricane Center Tech. Rep. AL092011, 45 pp. [Available online at http://www.nhc.noaa.gov/data/tcr/AL092011_Irene.pdf.]

  • Bailey, C. M., , Hartfield G. , , Lackmann G. M. , , Keeter K. , , and Sharp S. , 2003: An objective climatology, classification scheme, and assessment of sensible weather impacts for Appalachian cold-air damming. Wea. Forecasting, 18, 641661, doi:10.1175/1520-0434(2003)018<0641:AOCCSA>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Benjamin, S. G., and et al. , 2004: An hourly assimilation forecast cycle: The RUC. Mon. Wea. Rev., 132, 495, doi:10.1175/1520-0493(2004)132<0495:AHACTR>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Brown, D., 2008: Tropical cyclone report: Hurricane Gabrielle. NOAA/National Hurricane Center Tech. Rep. AL072007, 13 pp. [Available online at http://www.nhc.noaa.gov/data/tcr/AL072007_Gabrielle.pdf.]

  • Childs, B., , and Lewis R. , 2001: Investigating the impact of changing the ASOS wind gust averaging period from five to three seconds. Preprints, 11th Symp. on Meteorological Observations and Instrumentation, Albuquerque, NM, Amer. Meteor. Soc., 2.5.[Available online at https://ams.confex.com/ams/annual2001/techprogram/paper_18249.htm.]

  • Davies, J., 2006: Hurricane and tropical cyclone tornado environments from RUC proximity soundings. Preprints, 23rd Conf. Severe Local Storms, St. Louis, MO, Amer. Meteor. Soc., 12.6A. [Available online at https://ams.confex.com/ams/pdfpapers/115483.pdf.]

  • Durst, C. S., 1960: Wind speeds over short periods of time. Meteor. Mag., 89, 181187.

  • Eastin, M. D., , Gardner T. L. , , Link M. C. , , and Smith K. C. , 2012: Surface cold pools in the outer rainbands of Tropical Storm Hanna (2008) near landfall. Mon. Wea. Rev., 140, 471491, doi:10.1175/MWR-D-11-00099.1.

    • Search Google Scholar
    • Export Citation
  • Glahn, B., 2005: Comments on implementation and refinement of digital forecasting databases. Bull. Amer. Meteor. Soc., 86, 13151318, doi:10.1175/BAMS-86-9-1315.

    • Search Google Scholar
    • Export Citation
  • Glahn, H. R., , and Ruth D. P. , 2003: The New Digital Forecast Database of the National Weather Service. Bull. Amer. Meteor. Soc., 84, 195201, doi:10.1175/BAMS-84-2-195.

    • Search Google Scholar
    • Export Citation
  • Hanson, T., , Matthewson M. , , and Romberg M. , 2001: Forecast methodology using the GFE Suite. Preprints, 17th Int. Conf. on Interactive Information and Processing Systems for Meteorology, Oceanography, and Hydrology, Albuquerque, NM, Amer. Meteor. Soc., 1.13. [Available online at https://ams.confex.com/ams/annual2001/techprogram/paper_18722.htm.]

  • Harper, B. A., , Kepert J. D. , , and Ginger J. D. , 2010: Guidelines for converting between various wind averaging periods in tropical cyclone conditions. World Meteorological Organization Tech. Rep. WMO/TD-1555, 54 pp.

  • Hsu, S. A., 2001: Spatial variations in gust factor across the coastal zone during Hurricane Opal in 1995. Natl. Wea. Dig., 25, 2123. [Available online at http://www.nwas.org/digest/papers/2001/Vol25No12/Pg21-Hsu.pdf.]

    • Search Google Scholar
    • Export Citation
  • Hsu, S. A., 2003: Nowcasting the variation of wind speed with height using gust factor measurement. Natl. Wea. Dig., 27, 6769. [Available online at http://www.nwas.org/digest/papers/2003/Vol27No1/Pg67-Hsu.pdf.]

    • Search Google Scholar
    • Export Citation
  • Kaplan, J., , and Demaria M. , 1995: A simple empirical model for predicting the decay of tropical cyclone winds after landfall. J. Appl. Meteor., 34, 24992512, doi:10.1175/1520-0450(1995)034<2499:ASEMFP>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Knabb, R. D., , and Mainelli M. , 2007: Tropical cyclone report: Hurricane Ernesto. NOAA/National Hurricane Center Tech. Rep. AL062006, 36 pp. [Available online at http://www.nhc.noaa.gov/data/tcr/AL062006_Ernesto.pdf.]

  • Krayer, W., , and Marshall R. D. , 1992: Gust factors applied to hurricane winds. Bull. Amer. Meteor. Soc., 73, 613618, doi:10.1175/1520-0477(1992)073<0613:GFATHW>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Malkin, W., 1959: Filling and intensity changes in hurricanes over land. National Hurricane Research Rep. 34, U.S. Weather Bureau, Washington, DC, 18 pp.

  • Masters, F. J., , Vickery P. J. , , Bacon P. , , and Rappaport E. N. , 2010: Toward objective, standardized intensity estimates from surface wind speed observations. Bull. Amer. Meteor. Soc., 91, 16651681, doi:10.1175/2010BAMS2942.1.

    • Search Google Scholar
    • Export Citation
  • Moore, B. J., , Bosart L. F. , , Keyser D. , , and Jurewicz M. L. , 2013: Synoptic-scale environments of predecessor rain events occurring east of the Rocky Mountains in association with Atlantic basin tropical cyclones. Mon. Wea. Rev., 141, 10221047, doi:10.1175/MWR-D-12-00178.1.

    • Search Google Scholar
    • Export Citation
  • Mueller, K. J., , Demaria M. , , Knaff J. , , Kossin J. P. , , and Vonder Haar T. H. , 2006: Objective estimation of tropical cyclone wind structure from infrared satellite data. Wea. Forecasting, 21, 990, doi:10.1175/WAF955.1.

    • Search Google Scholar
    • Export Citation
  • Paulsen, B., , and Schroeder J. , 2005: An examination of tropical and extratropical gust factors and the associated wind speed histograms. J. Appl. Meteor., 44, 270280, doi:10.1175/JAM2199.1.

    • Search Google Scholar
    • Export Citation
  • Powell, M. D., , and Houston S. H. , 1998: Surface wind fields of 1995 Hurricanes Erin, Opal, Luis, Marilyn, and Roxanne at landfall. Mon. Wea. Rev., 126, 1259, doi:10.1175/1520-0493(1998)126<1259:SWFOHE>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Powell, M. D., , Houston S. H. , , Amat L. R. , , and Morisseau-Leroy N. , 1998: The HRD Real-time Hurricane Wind Analysis System. J. Wind Eng. Ind. Aerodyn., 77–78, 5364, doi:10.1016/S0167-6105(98)00131-7.

    • Search Google Scholar
    • Export Citation
  • Schwerdt, R. W., , Ho F. P. , , and Watkins R. R. , 1979: Meteorological criteria for standard project hurricane and probable maximum hurricane wind fields, Gulf and East Coasts of the United States. NOAA Tech. Rep. NWS 23, 317 pp. [Available online at http://www.nws.noaa.gov/oh/hdsc/Technical_reports/TR23.pdf.]

  • Srock, A. F., , and Bosart L. F. , 2009: Heavy precipitation associated with southern Appalachian cold-air damming and Carolina coastal frontogenesis in advance of weak landfalling Tropical Storm Marco (1990). Mon. Wea. Rev., 137, 2448, doi:10.1175/2009MWR2819.1.

    • Search Google Scholar
    • Export Citation
  • Vickery, P., , and Skerlj P. , 2005: Hurricane gust factors revisited. J. Struct. Eng., 131, 825832, doi:10.1061/(ASCE)0733-9445(2005)131:5(825).

    • Search Google Scholar
    • Export Citation
  • Walsh, K. J. E., , Sandery P. , , Brassington G. B. , , Entel M. , , Siegenthaler-Ledrian C. , , Kepert J. D. , , and Darbyshire R. , 2010: Constraints on drag and exchange coefficients at extreme wind speeds. J. Geophys. Res., 115, C09007, doi:10.1029/2010JA015748.

    • Search Google Scholar
    • Export Citation
  • Yu, B., , and Chowdhury A. G. , 2009: Gust factors and turbulence intensities for the tropical cyclone environment. J. Appl. Meteor. Climatol., 48, 534552, doi:10.1175/2008JAMC1906.1.

    • Search Google Scholar
    • Export Citation
  • View in gallery

    The analysis region, along with key cities that are mentioned throughout the paper.

  • View in gallery

    Tracks for all storms in which NDFD forecasts were examined. Red, orange, green, and blue lines indicate max sustained winds of category 2 hurricanes, category 1 hurricanes, tropical storms, and tropical depressions, respectively.

  • View in gallery

    Ernesto (2006) (a) max H*Wind-analyzed wind speed (kt) over all available analysis times, (b) max CRONOS wind speed (kt) station data, (c) wind speed difference between NDFD and H*Wind (NDFD − H*Wind; kt), and (d) wind speed difference between NDFD and CRONOS (NDFD − CRONOS; kt).

  • View in gallery

    As in Fig. 3, but for Gabrielle (2007).

  • View in gallery

    CWAs for the various NWS WFOs in the study region.

  • View in gallery

    As in Fig. 3, but for Cristobal (2008).

  • View in gallery

    As in Fig. 3, but for Hanna (2008).

  • View in gallery

    As in Fig. 3, but for Earl (2010).

  • View in gallery

    As in Fig. 3, but for Irene (2011).

  • View in gallery

    The number of missing hourly observations for select ASOS stations impacted by Irene (2011). The station abbreviations are as listed in Table 4 (here the initial “K” is omitted), and RWI: Rocky Mount, NC; MRH: Beaufort, NC; AKQ: Wakefield, VA.

  • View in gallery

    Storm tracks for Isabel (2003) and Irene (2011). Red, orange, and green lines indicate max sustained winds of category 2 hurricanes, category 1 hurricanes, and tropical depressions, respectively.

  • View in gallery

    Median RUC-analyzed four-quadrant wind speed (kt) within the 34-kt max wind radius at various times after landfall for Isabel (2003) (solid) along with a 12-h linear interpolation (dashed). The four quadrants represented are (a) northwest, (b) northeast, (c) southeast, and (d) southwest.

  • View in gallery

    As in Fig. 12, but for the 50-kt max wind.

  • View in gallery

    As in Fig. 12, but for Irene (2011). The vertical lines indicate the approximate time of second landfall for Irene (2011).

  • View in gallery

    As in Fig. 14, but for the 50-kt max wind.

  • View in gallery

    Histograms of gust factors for sustained wind speeds: (a) 10–30, (b) 30–40, (c) 40–50, and (d) greater than 50 kt.

  • View in gallery

    Histogram of gust factors for (a) northwesterly, (b) northeasterly, (c) southwesterly, and (d) southeasterly flow.

  • View in gallery

    As in Fig. 17, but for gust factors at Ocean City.

  • View in gallery

    As in Fig. 17, but for gust factors at Salisbury.

  • View in gallery

    Time series of sustained wind speed, wind gust, and gust factor for (a) Raleigh and (b) New Bern during Irene (2011). Data are plotted at 10-min intervals.

All Time Past Year Past 30 Days
Abstract Views 0 0 0
Full Text Views 162 162 12
PDF Downloads 125 125 11

An Examination of Wind Decay, Sustained Wind Speed Forecasts, and Gust Factors for Recent Tropical Cyclones in the Mid-Atlantic Region of the United States

View More View Less
  • 1 North Carolina State University, Raleigh, North Carolina
  • | 2 NOAA/NWS, Raleigh, North Carolina
  • | 3 NOAA/NWS, Wilmington, North Carolina
© Get Permissions
Full access

Abstract

In this study, several analyses were conducted that were aimed at improving sustained wind speed and gust forecasts for tropical cyclones (TCs) affecting coastal regions. An objective wind speed forecast analysis of recent TCs affecting the mid-Atlantic region was first conducted to set a benchmark for improvement. Forecasts from the National Digital Forecast Database were compared to observations and surface wind analyses in the region. The analysis suggests a general overprediction of sustained wind speeds, especially for areas affected by the strongest winds. Currently, National Weather Service Weather Forecast Offices use a software tool known as the Tropical Cyclone Forecast/Advisory (TCM) wind tool (TCMWindTool) to develop their wind forecast grids. The tool assumes linear decay in the sustained wind speeds when interpolating the National Hurricane Center 12–24-hourly TCM product to hourly grids. An analysis of postlandfall wind decay for recent TCs was conducted to evaluate this assumption. Results indicate that large errors in the forecasted wind speeds can emerge, especially for stronger storms. Finally, an analysis of gust factors for recent TCs affecting the region was conducted. Gust factors associated with weak sustained wind speeds are shown to be highly variable but average around 1.5. The gust factors decrease to values around 1.2 for wind speeds above 40 knots (kt; 1 kt = 0.51 m s−1) and are in general insensitive to the wind direction, suggesting local rather than upstream surface roughness largely dictates the gust factor at a given location. Forecasters are encouraged to increase land reduction factors used in the TCMWindTool and to modify gust factors to account for factors including the sustained wind speed and local surface roughness.

Corresponding author address: Bryce Tyner, Dept. of Marine, Earth, and Atmospheric Sciences, 2800 Faucette Drive, Rm. 1125 Jordan Hall, North Carolina State University, Raleigh, NC 27695-8208. E-mail: bptyner@ncsu.edu

Abstract

In this study, several analyses were conducted that were aimed at improving sustained wind speed and gust forecasts for tropical cyclones (TCs) affecting coastal regions. An objective wind speed forecast analysis of recent TCs affecting the mid-Atlantic region was first conducted to set a benchmark for improvement. Forecasts from the National Digital Forecast Database were compared to observations and surface wind analyses in the region. The analysis suggests a general overprediction of sustained wind speeds, especially for areas affected by the strongest winds. Currently, National Weather Service Weather Forecast Offices use a software tool known as the Tropical Cyclone Forecast/Advisory (TCM) wind tool (TCMWindTool) to develop their wind forecast grids. The tool assumes linear decay in the sustained wind speeds when interpolating the National Hurricane Center 12–24-hourly TCM product to hourly grids. An analysis of postlandfall wind decay for recent TCs was conducted to evaluate this assumption. Results indicate that large errors in the forecasted wind speeds can emerge, especially for stronger storms. Finally, an analysis of gust factors for recent TCs affecting the region was conducted. Gust factors associated with weak sustained wind speeds are shown to be highly variable but average around 1.5. The gust factors decrease to values around 1.2 for wind speeds above 40 knots (kt; 1 kt = 0.51 m s−1) and are in general insensitive to the wind direction, suggesting local rather than upstream surface roughness largely dictates the gust factor at a given location. Forecasters are encouraged to increase land reduction factors used in the TCMWindTool and to modify gust factors to account for factors including the sustained wind speed and local surface roughness.

Corresponding author address: Bryce Tyner, Dept. of Marine, Earth, and Atmospheric Sciences, 2800 Faucette Drive, Rm. 1125 Jordan Hall, North Carolina State University, Raleigh, NC 27695-8208. E-mail: bptyner@ncsu.edu

1. Introduction

Developing gridded forecasts of sustained wind speeds and gusts associated with landfalling tropical cyclones (TCs) remains a significant challenge posed to National Weather Service (NWS) forecasters in the mid-Atlantic region of the United States. A software tool known as the Tropical Cyclone Forecast/Advisory (TCM) wind tool (TCMWindTool) included in the Gridded Forecast Editor (GFE; Hanson et al. 2001) is currently used by NWS Weather Forecast Offices (WFOs) to develop wind forecast grids when a given region is impacted by a TC. The TCMWindTool automatically interpolates the 34-, 50-, and 64-knot (kt; 1 kt = 0.51 m s−1) 12–24-hourly four-quadrant maximum wind radii forecasts from the National Hurricane Center’s (NHC) Tropical Cyclone Forecast/Advisory (TCM) product to an hourly 2.5 km × 2.5 km grid. The algorithm assumes linear temporal decay in wind speeds when interpolating the 12–24-hourly TCM product to hourly grids. The bilinear interpolation in space is conducted assuming a modified Rankine vortex wind field [see Mueller et al. (2006) for details on the modified Rankine vortex]. Forecasters at the WFOs select various options from the TCMWindTool with the aim of improving the raw output from the TCM product. These include a background wind field used to smooth the winds outside of the TCM product wind field and a universal multiplicative factor by which to decrease the surface wind speeds over land to account for surface friction. Furthermore, the forecaster chooses the number of pie slices to which the winds are interpolated, with more pie slices creating a smoother transition between the four quadrants with different radii. Based on these parameters, the TCMWindTool generates an output grid that represents a base sustained wind speed forecast. Forecasters at the WFOs are then tasked with altering these grids by “applying local knowledge and mesoscale expertise to produce the final set of explicit/deterministic wind forecasts for the WFO’s Area of Responsibility” (information available online at http://www.srh.noaa.gov/rtimages/crp/tig/2011_TCMWindTool.pdf).

The decay of TCs over land has been studied in a number of empirical studies. Malkin (1959) examined Atlantic TCs making landfall and determined that there was a tendency for the most intense hurricanes to weaken most rapidly once over land. Furthermore, as the fraction of the storm that remained over water after landfall decreased, the rate of weakening increased. In Schwerdt et al. (1979), the authors provide evidence showing the rate of decay in wind speeds once over land also varies based on the geographical region. The rate of decay was shown to be slowest over the Gulf of Mexico coastline, quickest over Florida, and medium along the rest of the East Coast. It was concluded in both studies that any linear interpolation scheme employed to estimate wind speeds at intermediate time steps can be inaccurate. Furthermore, decay rate was shown to be proportional to intensity and largest just after the time of landfall. Motivated by some of these early studies, Kaplan and Demaria (1995) developed an empirical model for predicting sustained TC wind speeds after landfall that avoids a linear interpolation assumption, based on data from all TCs making landfall in the United States during 1967–93.

After the sustained wind forecast has been developed, forecasters are then tasked with creating a wind gust forecast grid, developed using gust factor values. The gust factor is typically defined as the ratio of the peak wind speed to the sustained wind speed. A number of studies have examined gust factors when applied to TC conditions; however, few of the studies have examined these gust factors in terms of the 2-min averaging period that NWS forecasters use for their forecasts. The 2-min average is calculated by averaging 24 discrete 5-s wind samples, which are based on 1-s wind observations. Most gust factor studies are based on a 10-min averaging period, consistent with practices of wind engineers (e.g., Durst 1960; Krayer and Marshall 1992; Yu and Chowdhury 2009). In Hsu (2003), 148 samples of Automated Surface Observing System (ASOS) data at various airports while impacted by 11 Atlantic TCs were used to calculate the gust factor based on the 2-min averaging period. The calculated mean gust factor from the data was 1.42, with a standard deviation of 0.18. In Hsu (2001), 2-min gust factor data for Hurricane Opal (1995) for select offshore, coastal, and inland locations were compared to an empirical formula for calculating the gust factor. The results suggested an increase in gust factor associated with onshore conditions compared to offshore conditions. Further analysis needs to be conducted in order to support the mean gust factor on these shorter wind averaging periods both spatially and temporally. Furthermore, the variations in the gust factor must also be examined in order to aid forecasters developing wind gust grids.

We conducted an informal survey of 13 forecasters from NWS WFOs in the mid-Atlantic region to motivate the current research. The results of the survey suggested a lack of consistency in scientific reasoning for developing the final sustained wind speed and gust grids in TC regimes. Following are some of the key survey results alluding to this subjectivity in the forecast process:

  • Most of the surveyed forecasters stated that they were not aware of a formal climatological analysis of land reduction factors over the study region. A few responders referred to an informal, unpublished study conducted by former hurricane specialist Dr. J. Pelissier at NHC. The study was for Atlantic TCs during 1999–2005 and suggested a wind reduction of 10% within 5 mi of the coast, with 20% to be used at areas farther inland to account for surface frictional effects.
  • Forecasters diverged on how to develop and even define a gust factor. While most forecasters indicated a percentage above sustained wind speeds is appropriate, some suggested a percentage of the maximum low-level mixed layer winds may be more effective.
  • Forecasters reported using a wide range of percentages above the sustained wind speeds to use as a gust factor. Suggestions ranged from 15% to 40% above the sustained wind speeds, depending on such factors as the degree of mixing and downward momentum transport, low-level lapse rates, presence/absence of convective precipitation, wind direction, and distance from coastline and storm center. In addition, there were large discrepancies in suggested values among forecasters within several of the same WFOs.

The survey results suggest a need for a comprehensive examination of the wind fields for recent TCs affecting the study region. Furthermore, the lack of consistency in standardizing wind averaging periods as well as instrumentation heights in past gust factor studies motivates this analysis. The focuses for this study are on the steps of the current forecast process at the various WFOs and providing insight into potential areas of improvement. The primary goals of the study are to

  • conduct an objective analysis of NWS sustained wind speed forecasts during times of TC influence,
  • examine the rates of decay in the sustained wind speeds for a TC after landfall, and
  • examine gust factors in the study region, specifically examining variability based on wind direction, sustained wind speed, and proximity to coastline.

The structure of this paper is as follows. The data and analysis tools used for the study are enumerated in section 2 and results are presented in section 3. The final section of the paper provides some discussion of the results as well as areas of future work that will be investigated by the authors.

2. Data and methods overview

For this study, the analysis region was defined as the area encompassing Maryland, Virginia, North Carolina, and South Carolina. A map of the analysis region along with key cities mentioned throughout the paper is shown in Fig. 1. It is important to examine recent wind speed forecasts made by the various NWS forecast offices within the study region in order to set a benchmark for improvement. All available National Digital Forecast Database (NDFD) sustained wind speed forecasts were obtained from the National Climatic Data Center (NCDC) for the study region and were used to conduct the forecast analysis. The NDFD was available starting in 2006, when gridded wind speed forecasts became operational at the NWS offices. Remnant storms that made landfall along the Gulf Coast and later propagated into the study region were not examined, since these are not often representative of true TC environments. TCs that formed off the mid-Atlantic coast and impacted the study region were included in the selection of storms. Since the number of storms was limited in the study period, the NDFD analysis was extended beyond landfalling TCs to include storms grazing the coastline. Table 1 lists the six TCs and the analysis dates for which NDFD forecasts were examined based on these criteria.

Fig. 1.
Fig. 1.

The analysis region, along with key cities that are mentioned throughout the paper.

Citation: Weather and Forecasting 30, 1; 10.1175/WAF-D-13-00125.1

Table 1.

Storms used to examine NDFD forecasts. The number of available CRONOS stations is also listed for each storm.

Table 1.

The analysis was completed using a combination of the Hurricane Research Division’s (HRD) Real-time Hurricane Wind Analysis System (H*Wind) and hourly surface observations from the State Climate Office of North Carolina Climate Retrieval and Observations Network of the Southeast Database (CRONOS; available online at http://www.nc-climate.ncsu.edu/cronos). The H*Wind surface analyses are advantageous in that they blend model data with observations from U.S. Air Force and NOAA aircraft, ships, buoys, and land-based surface platforms (Powell and Houston 1998). All data are quality controlled and then standardized to a 10-m height and 1-min wind speed averaging period. After landfall, when dropsondes can no longer be deployed, the analyses are largely driven by available surface data as well as surface-adjusted reconnaissance observations converted to open terrain (Powell et al. 1998). In times of significant data outages and limited reconnaissance data, the H*Wind analyses still provide an estimate of the analyzed wind speeds, but are largely weighted toward short-term model forecasts. Unfortunately, H*Wind analyses are often only created for short periods after landfall, as shown in Table 1. Furthermore, the analyses are only available at approximately 3-h intervals, leading to some discontinuity in the maximum winds at a given location throughout the duration of a TC.

Because of the inherent limitations in the H*Wind analyses, hourly surface observations from CRONOS were also examined. CRONOS included available ASOS, Automated Weather Observing System (AWOS), and North Carolina Environment and Climate Observing Network (ECONet) stations. A standard 10-m observation height was used for all of the stations, allowing for direct comparison to the NDFD forecasts. For quality control, probability distribution functions of the sustained wind speeds were calculated for all stations during times of storm impact. Stations with greater than 40% of the sustained wind observations equaling less than 2 kt were removed from the analysis. Routine inspection of the sustained wind speed data for the remaining stations suggests hourly observations are not available for all locations and times of storm impact. The missing data can be attributed to mechanical problems as well as communication failures and power outages at the various stations, especially during times of significant storm impact. Due to the inherent limitations in each dataset, both data sources are examined and incorporated into the analysis. It should be noted that the NDFD sustained wind forecasts are based on a 2-min averaging period, consistent with CRONOS. The H*Wind analyses are based on a 1-min averaging time, leading to potential error when comparing to NDFD. However, the maximum error accounting for these different averaging times is expected to be less than 13% (Harper et al. 2010).

To reduce bias in the wind forecasts due to TC track and intensity uncertainty prior to affecting the region, only the NDFD forecasts issued immediately prior to the analysis times were examined in this study. Following this method, the study uses 1-h NDFD wind forecasts valid at each hourly analysis time (Glahn and Ruth 2003; Glahn 2005). The NDFD forecasts and H*Wind analyses were bilinearly interpolated onto a common grid to allow for direct comparison of forecasted and analyzed wind speeds at various locations. The interpolated common grid contained 220 × 220 grid points within the latitude range of 30°–40°N and the longitude range of 75°–85°W, with a grid spacing of approximately 0.045° × 0.045°. Similarly, the NDFD forecasts were bilinearly interpolated to the CRONOS station locations. For both analyses, bias was calculated as the difference between the maximum NDFD-forecasted wind speed at each interpolated grid point over the period of analysis and the maximum analyzed value at the grid point over the same time period. For example, Table 1 indicates H*Wind analyses are available for Ernesto (2006) during the period from 0130 UTC 31 August to 0430 UTC 1 September. To allow for direction comparison to forecasts, only NDFD forecasts valid from 0100 UTC 31 August to 0400 UTC 1 September were examined for the H*Wind analysis. To account for the spatial discontinuity in the H*Wind analyses and because H*Wind analyses were often not available beyond several hours after TC impact, only grid points that had analyses available for at least half of the forecast valid times were used in the study. Also following Table 1, maximum data available from CRONOS during times of TC impact were examined and compared to NDFD forecasts valid for the same period of TC impact.

As previously noted, the TCMWindTool interpolates the 12–24-hourly wind field forecast from NHC to hourly forecast grids. The tool assumes linear changes in the wind speeds within each 12–24-h interval. Hourly Rapid Update Cycle (RUC) analyses with a horizontal grid spacing of 20 km (Benjamin et al. 2004) are available from NCDC for 2002–11. The RUC analyses were obtained for all landfalling TCs in the study region during this period of availability in order to evaluate the assumption of linear change. While there are inherent limitations of the RUC analysis resolving TC intensity, a comparison to H*Wind analyses showed the hourly analyses were still able to resolve the general storm structure as well as changes in the wind and pressure fields with time (not shown). Furthermore, RUC analyses have been used to study TC structure in several past studies. For example, Davies (2006) examined hourly RUC soundings at select locations to study the characteristics of TC tornadic environments, including storm-relative helicity at the locations.

After a comparison to NHC TC best-track data (available online at http://www.nhc.noaa.gov/data/#hurdat), it was determined that eight TCs made landfall in the study region during the selected analysis years. A list of the eight analyzed storms and their approximate times of landfall is presented in Table 2. The land decay analysis was conducted based on storm quadrant in order to account for the azimuthal differences in storm structure and respective wind fields. To allow for comparison of storms of various sizes, analyzed wind speeds were examined within the sectors consistent with the NHC best-track 34- and 50-kt maximum wind radii. The storm center was calculated based on RUC-analyzed hourly minimum sea level pressure associated with the TCs.

Table 2.

Landfalling TCs in the study region, 2002–11, for which the land decay analysis was conducted.

Table 2.

ASOS 1-min sustained wind speed and gust data are available to the public from NCDC since 2000. For the gust factor analysis, this ASOS data were obtained for all available locations in the study region during TC impact from 2000 to 2011, as listed in Table 3. It is important to note that the averaging period for calculating wind gusts has not been consistent at these ASOS stations. Prior to 2005, wind gusts were calculated as the maximum 5-s wind speed within the past minute of the 2-min averaging time used to calculate the sustained wind speed. Between 2005 and 2009, the ASOS stations were upgraded to Ice Free Wind Sensors, and the averaging period for wind gusts was reduced to 3 s. Childs and Lewis (2001) investigated the impact of changing the ASOS wind gust averaging period from 5 to 3 s using data from Sterling, Virginia. The results indicated a positive bias for the shorter averaging period, but the bias was found to be only 0–2 kt. Based on the results of this study, the ASOS sensor change is not expected to affect the overall results of the gust factor analysis presented in this paper. For each storm, gust factors were calculated at each ASOS station during times of TC influence. The gust factor was calculated at each time and location as the ratio of the wind gust to the sustained wind speed value.

Table 3.

All TCs impacting the study region for which the gust factors were analyzed, 2000–11.

Table 3.

3. Results

a. NDFD analysis

1) Ernesto (2006)

The wind forecasts for six TCs affecting the study area were examined, and the tracks of these storms are shown in Fig. 2. According to the NHC storm report, Ernesto (2006) made landfall near Oak Island, North Carolina, at approximately 0340 UTC 1 September, with maximum sustained winds near 60 kt and a minimum central pressure of 985 hPa (Knabb and Mainelli 2007). Before the passage of the TC, a predecessor rain event occurred over much of Virginia and North Carolina. Furthermore, northerly flow from a surface high pressure located over southern Canada led to a weak cold-air damming event for the region beginning on 31 August (Moore et al. 2013). After landfall, the tropical storm gradually weakened as it moved northward through northeastern North Carolina and over eastern Virginia before becoming extratropical at around 1800 UTC 1 September.

Fig. 2.
Fig. 2.

Tracks for all storms in which NDFD forecasts were examined. Red, orange, green, and blue lines indicate max sustained winds of category 2 hurricanes, category 1 hurricanes, tropical storms, and tropical depressions, respectively.

Citation: Weather and Forecasting 30, 1; 10.1175/WAF-D-13-00125.1

Plots of maximum H*Wind-analyzed 10-m wind speed as well as the difference between the maximum NDFD- and H*Wind-analyzed wind speed are shown in Figs. 3a and 3c. For the available H*Wind analysis times, the core of maximum winds over land was located from central North Carolina eastward to the coastline (Fig. 3a). Maximum sustained surface wind speeds were analyzed to be around 32–36 kt over much of this region, with a local maximum of 40–44 kt near the location of landfall in southeastern North Carolina. There is a widespread underprediction of sustained wind speeds over much of east-central North Carolina and southeastern Virginia of 4–8 kt during the analysis time (Fig. 3c). The maximum underprediction of these wind speeds was located in southeastern North Carolina, with an underprediction of 16–20 kt in a localized area. The H*Wind analyses were only available up to 0430 UTC 1 September, which was near the time of landfall (Table 1). As a result, much of the region had not yet been affected by the stronger TC winds located farther to the south near the storm core.

Fig. 3.
Fig. 3.

Ernesto (2006) (a) max H*Wind-analyzed wind speed (kt) over all available analysis times, (b) max CRONOS wind speed (kt) station data, (c) wind speed difference between NDFD and H*Wind (NDFD − H*Wind; kt), and (d) wind speed difference between NDFD and CRONOS (NDFD − CRONOS; kt).

Citation: Weather and Forecasting 30, 1; 10.1175/WAF-D-13-00125.1

Because the H*Wind analyses are not available for Ernesto (2006) beyond the time of landfall, the H*Wind analysis presented is not representative of the maximum wind speeds throughout the storm duration for much of the study region. Figure 3d shows the difference between the maximum NDFD-forecasted sustained wind speed and the maximum hourly observed wind speed at the available CRONOS stations. With these data available, the maximum sustained wind speeds over much of eastern South Carolina, North Carolina, and Virginia appeared to have been overpredicted. This overprediction was highest in northeastern South Carolina and southeastern North Carolina, where overprediction of up to 20 kt is seen at some CRONOS stations. Weak overprediction is also apparent over central North Carolina and Virginia, with overprediction values near 4–12 kt.

In summary, prior to the landfall of Ernesto (2006), there was a general underprediction of wind speeds over much of east-central North Carolina and southeastern Virginia. After landfall, when the strongest wind speeds affected much of the study region, a general overprediction of wind speeds was observed, following the CRONOS analysis. It is hypothesized that the competing influence of boundary layer stabilization from the predecessor rain event along with increased gradient wind flow from the cold-air damming led to a sustained wind forecasting challenge. A literature survey suggests most cold-air damming observational and modeling studies have focused on TC precipitation distribution impacts (e.g., Srock and Bosart 2009). To the knowledge of the authors, a comprehensive investigation into the influence of cold-air damming on the surface winds as it interacts with a TC in the region is lacking. Cold-air damming can be thought of as having competing impacts on the surface winds for an impending TC. The cold air from the associated parent surface high pressure over the northeastern region of the United States stabilizes the environment, reducing the near-surface lapse rates and preventing vertical mixing of winds aloft to the surface. However, the parent high also increases the pressure gradient as the TC approaches. A future comprehensive study of this nature would help improve wind forecasts for these complex scenarios.

2) Gabrielle (2007)

Tropical Storm Gabrielle (2007) made landfall near Cape Lookout, North Carolina, at 1400 UTC 9 September with maximum sustained winds near 50 kt, based on the NHC best-track data. After making landfall, the storm moved quickly toward the northeast and off the coastline near Kill Devil Hills, North Carolina. As the NHC storm report indicates, strong northerly upper-level winds sheared the convection, keeping the strongest winds offshore (Brown 2008). Figures 4a and 4c show the maximum H*Wind-analyzed wind speed and the difference between the maximum NDFD-forecasted and available H*Wind-analyzed wind speed. The H*Wind analyses are only available for approximately 2 h after landfall (Table 1). However, as a result of the strongest winds to the east of the storm center and the rapid northeasterly track of the storm, the period in which the region was most directly impacted is captured from the H*Wind analysis. The maximum H*Wind analysis shows the strongest winds were indeed confined to the coastline and areas just offshore, consistent with reduced surface roughness in those locations as well as the increased proximity to the storm center. There is a large region of 28–36-kt maximum sustained winds in eastern North Carolina. The peak winds over land were confined to the Outer Banks, with maximum analyzed surface winds over 44 kt in a narrow region over Cape Hatteras. Farther to the west, maximum sustained wind speeds gradually decreased, with maximum analyzed wind speeds over central North Carolina near 8–16 kt.

Fig. 4.
Fig. 4.

As in Fig. 3, but for Gabrielle (2007).

Citation: Weather and Forecasting 30, 1; 10.1175/WAF-D-13-00125.1

Figure 4c indicates a widespread region of overprediction of maximum wind speeds in eastern North Carolina by approximately 4–8 kt. The overprediction was largest over much of the Outer Banks, where the strongest observed wind speeds occurred. In this region, there was a widespread overprediction of wind speeds by approximately 8–12 kt. Farther west, there was large region of underprediction of wind speeds of 4–8 kt in the region of weaker analyzed wind speeds. Figure 5 shows the County Warning Areas (CWA) of the various NWS WFOs. It is important to note that the underprediction–overprediction dipole closely aligns with the various WFO boundaries. The overprediction of wind speeds in North Carolina was restricted to east-central North Carolina, with underprediction of wind speeds in the bordering WFOs.

Fig. 5.
Fig. 5.

CWAs for the various NWS WFOs in the study region.

Citation: Weather and Forecasting 30, 1; 10.1175/WAF-D-13-00125.1

The analysis using CRONOS stations in Fig. 4d is quite similar to the analysis using H*Wind data. This includes a large region of overprediction of wind speeds of 4–8 kt in eastern North Carolina, with even higher overprediction in select locations. In the areas farther inland that were affected by weaker wind speeds, a widespread underprediction of wind speeds of around 4–8 kt is visible. In between these regions, there is an area where the observed maximum sustained wind speeds closely matched the NDFD-forecasted wind speeds.

In summary, the H*Wind and CRONOS wind speed analyses yield similar overall results, with strong overprediction in wind speeds for areas affected by the strongest sustained wind speeds. Sharp gradients in the forecasted wind speeds across WFO boundaries suggest a need for greater collaboration on land reduction factors used in the TCMWindTool during the forecast process. A GFE tool has been developed that will promote this collaboration of land reduction factors. Details of this tool are discussed more thoroughly in section 4 of this paper.

3) Cristobal (2008)

According to the NHC storm report, Cristobal (2008) developed from a decaying frontal boundary off the mid-Atlantic coastline (Avila 2009). As the system drifted westward toward the Outer Banks on 20 July, it strengthened into a weak tropical storm. The storm then gradually accelerated northeastward in the Atlantic away from land. As a result of this northeasterly track, the H*Wind analyses were available for sufficient time to capture the strongest wind speeds impacting the study region. The storm never made landfall and the strongest wind speeds remained off the coastline (Fig. 6a). The H*Wind surface analysis indicates that maximum sustained wind speeds over 25 kt were confined to parts of the Outer Banks. Furthermore, the analysis indicates there was a sharp decrease in wind speeds away from the coastline, with locations just slightly inland experiencing maximum sustained winds of less than 16 kt.

Fig. 6.
Fig. 6.

As in Fig. 3, but for Cristobal (2008).

Citation: Weather and Forecasting 30, 1; 10.1175/WAF-D-13-00125.1

There was a general overprediction in sustained wind speeds throughout the eastern portions of the study region (Fig. 6c). As was the case in Gabrielle (2007), the areas with the strongest sustained wind speeds were consistent with the largest values of overprediction. This includes much of extreme eastern North Carolina, with a general overprediction in wind speeds near 8–12 kt. Farther to the north over eastern Virginia and farther southwest over eastern South Carolina, the analysis indicates widespread weak overprediction of wind speeds, with values near 4–8 kt. Over central North Carolina and Virginia, where the wind speeds were the weakest, Fig. 6 shows widespread close agreement between the forecasted and analyzed sustained wind speeds, with just a few locations observing slight underprediction of wind speeds.

A comparison to the CRONOS analysis for Cristobal (2008) reveals many of these same spatial features (Fig. 6d). Over the coastal regions, where the strongest wind speeds occurred, an overprediction of approximately 8–12 kt for many locations is apparent, consistent with the H*Wind analysis. Over central and western North Carolina and Virginia, in the areas affected by the periphery of the storm, the forecasted maximum wind speeds closely matched the CRONOS observations.

4) Hanna (2008)

Hanna (2008) made landfall near the border between North and South Carolina at 0700 UTC 6 September with maximum sustained winds near 60 kt. After making landfall, the storm continued to track northward over northeastern North Carolina and southeastern Virginia and through the mid-Atlantic states. As described in Eastin et al. (2012), outer rainbands associated with Hanna led to the development of a strong surface cold pool over much of central North Carolina and Virginia. This surface cold pool developed as the precipitation fell into dry low-level air, eventually resulting in an ageostrophic wind adjustment of flow from the northeast. The diabatically forced northeasterly flow was consistent with an in situ cold-air damming event for the region (Bailey et al. 2003).

The H*Wind analyses for Hanna (2008) were only available shortly after landfall (Table 1). As a result, the verification using H*Wind analysis is for the period prior to the strongest wind speeds impacting much of northeastern North Carolina and eastern Virginia. Maximum wind speeds greater than 20 kt occurred over much of eastern South Carolina and North Carolina during the analysis period (Fig. 7a). Similar to Gabrielle (2007) and Cristobal (2008), the strongest wind speeds were confined to locations east of the storm center, where maximum sustained wind speeds were analyzed to be greater than 28 kt. The maximum difference plot shows an extensive region of overprediction in maximum wind speeds throughout much of east-central North Carolina and Virginia (Fig. 7c). The overprediction was approximately 8–12 kt ahead of the main TC circulation. Farther to the west and east of this region, the forecasts matched closely with the H*Wind-analyzed maximum sustained wind speeds.

Fig. 7.
Fig. 7.

As in Fig. 3, but for Hanna (2008).

Citation: Weather and Forecasting 30, 1; 10.1175/WAF-D-13-00125.1

When the entire period of storm impact is considered, the CRONOS analysis in Fig. 7d suggests more extensive overprediction of wind speeds associated with Hanna (2008). Over much of the eastern portion of the study region, the overprediction was between 12 and 20 kt in many areas. The overprediction is much lower toward western North Carolina, in the region affected by weaker wind speeds.

In summary, the region affected by the strongest wind speeds displayed an overprediction for Hanna (2008). The overprediction was also observed ahead of the storm prior to the strongest wind speeds impacting the region, as indicated by the H*Wind analysis. In the regions affected by weaker winds to the west, the forecasted winds were much closer to the observed and analyzed wind field. Similar to Ernesto (2006), it is hypothesized that the cold-air damming ahead of the storm complicated the forecast process for the wind speeds. As previously mentioned, future work should thoroughly examine the impact of cold-air damming on near-surface wind speeds during times of TC impact.

5) Earl (2010)

Earl (2010) approached the southeastern United States as a category 2 hurricane (Fig. 2). During its recurvature to the northeast, it weakened quickly from a category 2 hurricane at 1200 UTC 2 September to a category 1 storm at 1200 UTC 3 September. At its closest approach to the coastline, the storm passed approximately 75 miles to the east of Cape Hatteras on 3 September. After impacting the North Carolina coastline, the storm accelerated to the northeast. Similar to Gabrielle (2007), the maximum wind speeds for Earl (2010) were confined to portions of the Outer Banks (Fig. 8a). In this area, maximum wind speeds were around 30–36 kt, with a peak over Cape Hatteras of near 48 kt. There was a sharp spatial gradient in the maximum wind speeds, with areas over east-central North Carolina and Virginia experiencing weaker sustained winds of around 12–20 kt. For locations west of this region, maximum analyzed wind speed values were below 12 kt.

Fig. 8.
Fig. 8.

As in Fig. 3, but for Earl (2010).

Citation: Weather and Forecasting 30, 1; 10.1175/WAF-D-13-00125.1

Most of the eastern portions of the study region saw a strong overprediction in maximum surface wind speeds. As in the case of Gabrielle (2007) and Cristobal (2008), the areas experiencing some of the strongest wind speeds were in line with areas of strongest overprediction. In the Outer Banks, the overprediction peaked near 8–16 kt (Fig. 8c). The large overprediction in the coastal regions was also present when evaluating with the CRONOS stations (Fig. 8d). The maximum overprediction was near 12–16 kt over much of this region, which is quantitatively consistent with the H*Wind analysis. The overprediction is not as large over portions of east-central North Carolina and extreme southeastern Virginia, with values near 4–8 kt over much of the area. In the region affected by the weaker wind speeds farther to the west, the forecasted wind speeds closely matched the H*Wind analyzed as well the wind speeds observed from the CRONOS stations.

6) Irene (2011)

Irene (2011) made landfall near Cape Lookout, North Carolina, at 1200 UTC 27 August with maximum sustained winds of 75 kt, according to the NHC storm report (Avila and Cangialosi 2012). After landfall, the storm continued to move northeast over eastern North Carolina and Virginia and along the coastline up to New England. The strongest winds were mainly confined to along and east of the storm track (Fig. 9a). This included areas along the Outer Banks, where widespread maximum sustained wind speeds over 62 kt were analyzed. Over much of east-central North Carolina and Virginia, maximum wind speeds were near 32–40 kt, diminishing to 16–24 kt over the central portions of these states.

Fig. 9.
Fig. 9.

As in Fig. 3, but for Irene (2011).

Citation: Weather and Forecasting 30, 1; 10.1175/WAF-D-13-00125.1

H*Wind surface analyses were created by HRD long after landfall for Irene (2011), providing a more complete picture for the NDFD analysis than for several of the other storms examined. The maximum difference plot reveals a strong overprediction in wind speeds over much of the study region. The H*Wind analysis indicates the maximum overprediction was across portions of eastern Virginia and east-central North Carolina, where maximum overprediction was around 16–20 kt (Fig. 9c). There was a sharp reduction in the maximum wind speed error over central North Carolina following within the Raleigh WFO area of responsibility, shown previously in Fig. 5. Based on some of the preliminary results of some of the work presented here, forecasters at the Raleigh WFO noted using exceptionally high land reduction factors in the TCMWindTool of 30%–35% compared to neighboring offices (G. Hartfield, NOAA/NWS Raleigh, 2011, personal communication). This suggests the increased land reduction factors may be necessary for operational forecasters in future landfalling TCs, especially for locations near the coastline experiencing some of the strongest maximum sustained winds.

The CRONOS analysis indicates an overprediction of wind speeds along the coastline of South Carolina, North Carolina, and Virginia (Fig. 9d). The neutral colors over central North Carolina suggest the forecasted maximum winds were comparable to observations for areas over the Raleigh WFO, consistent with the H*Wind analysis. The numbers of hourly observations that were unavailable for select stations impacted by the storm are plotted in Fig. 10. Several locations in eastern North Carolina and Virginia reported missing observations as a result of power outages. Because of these power outages, many of these stations were unable to record observations when the strongest wind speeds impacted the location. Thus, the extreme overprediction for much of eastern North Carolina and Virginia is not likely representative of the true overprediction in the region (Fig. 9d). The overprediction would likely have been reduced if the stronger wind speeds during major TC impact had been recorded. Hence, for this storm, it is suggested that the H*Wind analyses better quantitatively capture the overprediction in the areas impacted by the stronger wind speeds than analyses using CRONOS.

Fig. 10.
Fig. 10.

The number of missing hourly observations for select ASOS stations impacted by Irene (2011). The station abbreviations are as listed in Table 4 (here the initial “K” is omitted), and RWI: Rocky Mount, NC; MRH: Beaufort, NC; AKQ: Wakefield, VA.

Citation: Weather and Forecasting 30, 1; 10.1175/WAF-D-13-00125.1

b. Land decay analysis

As previously mentioned, eight TCs made landfall in the study region during the selected analysis years. The hourly four-quadrant 10-m winds within the NHC best-track 34- and 50-kt maximum wind radii were examined for these eight TCs. Consistent with the results presented in Kaplan and Demaria (1995), it was determined that the weaker storms had much slower rates of intensity change after making landfall. Because of these slower rates of decay, the assumption of linear decay was not shown to result in significant error for these TCs. However, two of the TCs examined were much more intense and hence the results of the land decay for these storms are presented in detail.

Isabel (2003) and Irene (2011) were strong TCs that impacted the region during the period of study. The tracks of Isabel (2003) and Irene (2011) are shown in Fig. 11. Time series of median four-quadrant 10-m winds within the NHC best-track 34- and 50-kt maximum wind radii are shown for Isabel (2003) in Figs. 12 and 13. The plots indicate that in all four storm quadrants, there was a 2–4-h period of rapid decrease in wind speeds. The period of decay of 10-m wind speeds for the west side of the storm preceded the decay to the east, consistent with the left quadrants of the TC moving over land prior to the right. For example, in the southwestern quadrant, the median wind speed decreased within the 50-kt maximum wind radius from around 42 kt 2 h after landfall to approximately 24 kt 1 h later (Fig. 13). In the southeastern quadrant, the period of decay began around 5 h after landfall, with the median wind speed within this storm radius decaying by approximately 22 kt within a 2-h period.

Fig. 11.
Fig. 11.

Storm tracks for Isabel (2003) and Irene (2011). Red, orange, and green lines indicate max sustained winds of category 2 hurricanes, category 1 hurricanes, and tropical depressions, respectively.

Citation: Weather and Forecasting 30, 1; 10.1175/WAF-D-13-00125.1

Fig. 12.
Fig. 12.

Median RUC-analyzed four-quadrant wind speed (kt) within the 34-kt max wind radius at various times after landfall for Isabel (2003) (solid) along with a 12-h linear interpolation (dashed). The four quadrants represented are (a) northwest, (b) northeast, (c) southeast, and (d) southwest.

Citation: Weather and Forecasting 30, 1; 10.1175/WAF-D-13-00125.1

Fig. 13.
Fig. 13.

As in Fig. 12, but for the 50-kt max wind.

Citation: Weather and Forecasting 30, 1; 10.1175/WAF-D-13-00125.1

Isabel (2003) took a nearly perpendicular angle of approach to the coastline, with the western quadrants of the storm reaching land first (Fig. 11). The rapid decay once over land is a reflection of this direct angle of approach at landfall, consistent with the results presented in Malkin (1959). This direct angle of approach resulted in rapid weakening of the storm, as well as increased surface roughness, leading to frictional reduction in near-surface wind speeds. The rapid weakening occurred on time scales much shorter than the available 12–24-hourly NHC forecast data. The linear decay assumption used to create the hourly wind grids in the TCMWindTool is thus unable to capture the decay in wind speeds after landfall that was observed in the RUC analyses. Applying a linear decay function would lead to strong error in wind speeds after the storm makes landfall, given the rapid decay in wind speeds over a short time interval. To better depict this, dashed lines in Figs. 12 and 13 are overlaid, showing linear changes in wind speed over the 12-h period analyzed for all four quadrants. Because of the rapid decay, the interpolated wind speeds are too low prior to the period of rapid decay and too high afterward. The amount of error in the wind speeds is as high as 50%.

Irene (2011) made landfall in southeastern North Carolina before moving northeastward and exiting the coastline near the North Carolina–Virginia border (Fig. 11). The storm then continued to move northward, moving roughly parallel to the coastline before a second landfall over southern New Jersey. Temporal analyses of four-quadrant decay in median wind speeds for the 34- and 50-kt maximum wind radii are plotted in Figs. 14 and 15, respectively. For the western quadrants of the storms, the wind speeds underwent a brief period of reduction within the first few hours of landfall. The wind speeds increased as the storm moved off the coastline approximately 12 h after landfall, particularly for the median wind within the 50-kt maximum wind radius. The eastern quadrants did not undergo this period of reduction in wind speeds, consistent with reduced surface roughness for these quadrants as well as continued strong convection on the eastern hemisphere of the storm. As the storm made its second landfall over southern New Jersey, the wind speeds underwent a period of rapid decay. The plots indicate a short period of 2–4 h of rapid decay approximately 24 h after landfall, especially for the northeastern and southeastern quadrants within the 50-kt maximum wind radius. As in the case of Isabel (2003), the application of linear decay would lead to strong bias in wind speeds during the period of rapid decay, resulting in large errors for the hourly grids. The decay of wind speeds within the weaker 34-kt maximum wind radii was much more gradual, resulting in minimal error when assuming linear decay over the periods of 12–24 h.

Fig. 14.
Fig. 14.

As in Fig. 12, but for Irene (2011). The vertical lines indicate the approximate time of second landfall for Irene (2011).

Citation: Weather and Forecasting 30, 1; 10.1175/WAF-D-13-00125.1

Fig. 15.
Fig. 15.

As in Fig. 14, but for the 50-kt max wind.

Citation: Weather and Forecasting 30, 1; 10.1175/WAF-D-13-00125.1

As previously mentioned, the current tool only allows the use of one universal land decay value. An improved version of the TCMWindTool would help take into account periods of rapid decay, both as a result of hourly changes in storm strength as well as differences in surface roughness at the various locations affected by the wind speeds. This could be achieved by creating a grid-to-model reduction due to surface roughness as well as adding a slider bar within the TCMWindTool to alter the linear change within the 24-h four-quadrant maximum wind radii. These options are currently being tested by the collaborators for the project. As previously discussed, though the 20-km RUC grid spacing is relatively coarse to represent the strength of the TC, a comparison to various analysis datasets indicates that the general wind structure is still well represented. Furthermore, Malkin (1959) showed that the rate of decay over land increases with increasing wind speed. Based on this, it is suggested that the violation of the linear rate of decay in wind speeds presented in this analysis may even be stronger in actuality.

c. Gust factor analysis

ASOS wind data were examined for all TCs impacting the region from 2000 to 2011, listed in Table 3. Gust factors were calculated for each available observation. As previously mentioned, the gust factors were calculated as a ratio of wind gust to sustained wind speed value. The gust factors were not examined when the sustained wind speed reported was less than 10 kt. Gust factors were analyzed based on sustained wind speed as well as on wind direction over the study region. Furthermore, a temporal analysis was conducted for each storm at each station to examine how the gust factors evolved as the TC passed.

1) Gust factor versus sustained wind speed

Figure 16 shows histograms of available 1-min gust factors as a function of sustained wind speed. The results indicate a large variability in the gust factor for weak sustained wind speeds. For wind speed values less than 30 kt, gust factors have a large degree of spread, with values from 1.0 to 2.2. In fact, there are near-equal frequencies of gust factors of 1.15–1.2, 1.2–1.25, and 1.25–1.30. For sustained wind speed values between 30 and 40 kt, the gust factors shift slightly to lower values, with most gust factors falling between 1.15 and 1.25. However, there is a tail in the distribution of gust factors, with several observations greater than 1.5. The shift to lower gust factor values is much more evident at sustained wind speeds between 40 and 50 kt, with very few observed gust factors greater than 1.5. Although there are few data for sustained wind speeds greater than 50 kt, the available data indicate very little spread in gust factors for wind speeds over 50 kt, with gust factors remaining relatively constant near 1.20. To confirm these results, the standard deviation of the gust factors for the various wind speed bins was calculated. The value for sustained wind speeds less than 30 kt was 0.21, compared to a much smaller value of 0.14 for wind speeds greater than 50 kt. The reduced variability in gust factors observed at high wind speeds is consistent with the recent work of Walsh et al. (2010), where drag coefficients were shown to level off or even decay at high TC sustained wind speeds. Overall, the results suggest the methodology used by many forecasters in which a single gust factor is used across the entire forecast area may be a poor approach. It is suggested that forecasters consider using higher gust factors for locations impacted by weaker sustained wind speeds than locations impacted by stronger TC sustained wind speeds.

Fig. 16.
Fig. 16.

Histograms of gust factors for sustained wind speeds: (a) 10–30, (b) 30–40, (c) 40–50, and (d) greater than 50 kt.

Citation: Weather and Forecasting 30, 1; 10.1175/WAF-D-13-00125.1

2) Gust factor versus wind direction

In Paulsen and Schroeder (2005), the authors compare gust factors at 10-m averaging periods for TC environments and non-TC environments. The results showed that in both environments, gust factors were higher when the wind came from a direction associated with higher upstream surface roughness. Based on this result, we hypothesized that the gust factors at the various ASOS stations with 2-min averaging periods would show bias based on wind direction. In particular, values at locations near the coastline were expected to be higher for continental flow with higher associated upstream surface roughness compared to maritime flow. Gust factors from 14 coastal sites in the study region were calculated for the storms and analyzed according to wind direction. A list of these coastal stations is presented in Table 4. Figure 17 shows a histogram of gust factors as a function of wind direction for these coastal sites. The histograms do not indicate a conclusive difference in magnitudes of gust factors based on wind direction. There is a clustering of gust factors near 1.25–1.4, with a gradual spread in other values for all four wind directions. The mean gust factor from all four directions is approximately the same, consistent with the histogram results (Table 5). The results suggest that other factors besides upstream surface roughness are important in determining the gust factors. This is consistent with the results presented in Vickery and Skerlj (2005), where local topography and associated surface roughness was found to have a heavy influence on the gust factors associated with TCs.

Table 4.

List of coastal ASOS sites for which gust factors were analyzed according to wind direction.

Table 4.
Fig. 17.
Fig. 17.

Histogram of gust factors for (a) northwesterly, (b) northeasterly, (c) southwesterly, and (d) southeasterly flow.

Citation: Weather and Forecasting 30, 1; 10.1175/WAF-D-13-00125.1

Table 5.

Mean gust factor values for the various locations and wind directions presented in the paper.

Table 5.

3) Gust factor analysis: Maritime versus continental flow

Based on the results presented in Hsu (2001), it was further hypothesized that locations immediately along the coastline would observe lower gust factors than when the wind was from a direction with a large fetch over land. Hence, for these locations right along the coastline, gust factors were analyzed based on wind direction, and our results support this hypothesis. Figure 18 shows a histogram of gust factors as a function of wind direction for Ocean City, Maryland. The histogram suggests a preference in gust factors for this location based on wind direction. There is a tight clustering of gust factors below 1.3 when the wind direction is from the northeast or southeast (Figs. 18b,d). Ocean City is located right along the Atlantic coastline (Fig. 1). Hence, these wind directions are associated with maritime flow. This clustering of gust factors toward lower values for maritime flow is consistent with the results presented in Hsu (2001), where gust factors observed by buoys were found to be clustered near 1.3. The gust factors display more spread when the wind direction is from the southwest and northwest (Figs. 18a,c). Table 5 indicates the mean gust factors are lower when the wind direction is from a maritime direction, particularly from the northeast. After conducting a Student’s t test, it was shown that the differences in the means for the maritime flow compared to the continental flow were significant to the 99% confidence level. In contrast, Fig. 19 shows the gust factor as a function of wind direction for Salisbury, Maryland, a location approximately 50 km inland of Ocean City. For this station, the gust factors do not show a large preference based on wind direction. The mean gust factor is approximately the same from all four wind quadrants (Table 5). The results, combined with the results of the previous section, suggest gust factors are largely determined by local surface roughness conditions. For most locations, the local upstream surface roughness is not vastly different based on wind direction. However, for locations immediately adjacent to the coastline, the local upstream surface conditions are inherently different, resulting in a preference for lower gust factors for this maritime flow.

Fig. 18.
Fig. 18.

As in Fig. 17, but for gust factors at Ocean City.

Citation: Weather and Forecasting 30, 1; 10.1175/WAF-D-13-00125.1

Fig. 19.
Fig. 19.

As in Fig. 17, but for gust factors at Salisbury.

Citation: Weather and Forecasting 30, 1; 10.1175/WAF-D-13-00125.1

4) Gust factor temporal analysis: Irene (2011)

Gust factors were also examined at select stations throughout the various stages of TC impact. A time series of gust factors is presented in Fig. 20 for Raleigh and New Bern during Irene (2011). Wind speed observations are plotted at 10-min intervals in the figure. As the TC approached on 27 August, sustained wind speeds peaked at around 17 kt at 1200 UTC 27 August at Raleigh. With these low wind speed values, gust factors exhibited a large degree of spread, with values ranging from 1.2 to 2.0 throughout the time of TC impact. The high degree in spread of gust factors for lower sustained wind speeds is consistent with the previous results presented for all times of TC impact for the study region.

Fig. 20.
Fig. 20.

Time series of sustained wind speed, wind gust, and gust factor for (a) Raleigh and (b) New Bern during Irene (2011). Data are plotted at 10-min intervals.

Citation: Weather and Forecasting 30, 1; 10.1175/WAF-D-13-00125.1

The gust factor temporal analysis for New Bern differs from that for Raleigh. As the TC approached, the observed sustained wind speeds increased to over 30 kt early on 27 August. The gust factor values were noticeably lower for New Bern and the variability was also largely reduced. This was especially the case when the wind speed values were greater than 20 kt after 0300 UTC 27 August, where the gust factors were closely clustered around 1.25. Unfortunately, because of a sensor outage, sustained wind speeds and gusts are not available from 0724 UTC 27 August to 2211 UTC 27 August, as the wind speeds continued to increase. The data outage occurred for many ASOS and AWOS stations along the North Carolina coastline (Fig. 10). As the wind speeds gradually decreased after the storm passage and the sustained wind speeds decreased on 28 August, the gust factors increased in variability, with a mean value increasing slightly to around 1.3. Consistent with the previously presented results, as the wind speed increased in New Bern, the gust factor variability was reduced and converged to a lower mean value.

4. Discussion and conclusions

A climatology of sustained wind speeds, gusts, and forecasts of recent TCs impacting the mid-Atlantic was presented. To set a benchmark for improvement, a systematic analysis of NDFD sustained wind speed forecasts was conducted for recent TCs impacting the study region. The NDFD analysis using both the observations from CRONOS and the H*Wind surface analyses suggested a general overprediction in sustained wind speeds for much of the study region. This overprediction was seen in areas impacted by the strongest wind speeds. Ernesto (2006) and Hanna (2008) were storms in which a cold-air damming event preceded the TC making landfall. Future studies should examine this influence of cold-air damming on observed TC wind speeds. Based on the analysis, forecasters are encouraged to consider using larger land reduction factors when running the TCMWindTool in order to reduce the overprediction in wind speeds, especially for areas affected by the strongest wind speeds. Forecasters are also encouraged to collaborate among WFOs on the land reduction factors they select when running the TCMWindTool to improve the sustained wind speed forecasts and consistency for future TCs. A land reduction factor of 5% is suggested along the immediate coastline, gradually increasing to approximately 15% slightly farther inland and, finally, to 35% for locations well inland. These suggested land reduction factor values are somewhat heuristically based. Future studies should strive to objectively determine these land reduction factor values to be used in the TCMWindTool.

Currently, the TCMWindTool used to develop the wind speed forecast grids assumes a linear decay within the 12–24-h wind speed forecast periods. Using RUC analyses, it was shown that recent strong TCs making landfall in the region underwent a period of rapid decay in terms of sustained wind speeds. The period of rapid decay was dependent on the angle of storm approach with respect to the coastline. The decay occurred over a period much shorter than the 12–24-hourly temporal resolution of the NHC wind forecast data. The linear assumption used in the TCMWindTool to create hourly grids resulted in large error in wind speeds that should be accounted for in future times of TC landfall. It is suggested that an alternative automatic interpolation option be added to the TCMWindTool to account for this lack of linear decay after landfall, where forecasters can select the period of rapid decay within the 12-h raw wind speed forecast guidance provided by NHC.

ASOS data for recent TCs impacting the region were used to examine elements that ultimately determine the gust factors for a given location. The gust factors were shown to have a high degree of spread at lower sustained wind speeds between 10 and 30 kt. Forecasters are encouraged to use the mean gust factor for a given sustained wind speed as a starting point. For the lower end of the 10–30-kt range, a gust factor of 1.5 is appropriate, while for a 30-kt wind a gust factor of around 1.35 is a suggested starting point. The variation in the gust factor decreased for winds of 30–40 kt and decreased significantly for winds greater than 40 kt. The gust factors appeared to nearly asymptotically decay from 1.3 to 1.2 for sustained wind speeds above 40 kt, although the sample size was much lower for these higher sustained wind speeds. No conclusive bias in gust factors was observed based on wind direction for most locations, indicating that large-scale upstream surface roughness does not largely determine the gust factor for a given location. Instead, local upstream surface roughness was found to be important, supported by the preference of lower gust factors for Ocean City when the wind direction was from the east. This location was located immediately along the coastline, where a difference in local upstream surface roughness would be significantly different based on wind direction. The results are consistent with the results of several past studies (e.g., Yu and Chowdhury 2009; Hsu 2001). Forecasters should consider the sustained wind speed, wind direction, and local upstream surface roughness when choosing a gust factor to create wind gust forecasts.

Currently, forecasters typically use the default land reduction factor in the TCMWindTool of 10% and apply a single gust factor for all locations in the WFO. The gust factor and NDFD analyses suggest that this is not an adequate practice. The dependence of both gust factors and land reduction factors on local surface roughness characteristics is consistent with a recent analysis of ASOS data presented in Masters et al. (2010). Participants in the project have developed a set of GFE tools and procedures to take into account the results of this study. The experimental WindReductionFactor and WindGustFactor tools provide forecasters the flexibility to vary land reduction and gust factors spatially and temporally across the forecast area. The tools allow forecasters to integrate the collective impact of nonlinear storm decay, friction, and fetch into the forecast process. These tools also allow forecasters to see the land reduction and gust factors from other WFOs, resulting in improved collaboration and more consistent NDFD forecast products. By using the same methodology, the tools can provide forecasters with a common starting point prior to making local edits and by retaining the previous shifts forecast edits, the methodology ensures more shift-to-shift continuity. Furthermore, with the addition of the tools, forecasters can create reduction and gust factors prior to the NHC TCM product’s arrival, resulting in more timely products. Future storms impacting the region will serve as a way to test improvements based on the WindReductionFactor and WindGustFactor tools. An NDFD verification will be conducted for sustained wind speeds and gusts for these future storms.

The results of the current study are based on various observational platforms and analyses that have their own limitations. Furthermore, the results are based on a relatively small sample size because of limitations on data availability for the various products. The results presented in this study will be compared to select high-resolution simulations of TCs impacting the study region. With the increase in computational power over recent years, the Weather Research and Forecasting (WRF) Model now allows for explicit simulation of the boundary layer via large eddy simulation capability. The simulations will examine the characteristics of large eddy circulations associated with TCs and how they impact surface sustained wind speeds, wind gusts, and gust factors. The results of the simulations will be combined with the results presented here to provide more complete guidance for NWS forecasters developing their wind grids during times of TC impact.

Acknowledgments

The authors wish to thank the members of the CSTAR TC Inland Winds group for helpful comments and assistance throughout the course of this project. This includes Gail Hartfield (NOAA/NWS Raleigh, North Carolina), David Glenn and Carin Goodall (NOAA/NWS Newport, North Carolina), Robert Bright and Frank Alsheimer (NOAA/NWS Charleston, South Carolina), Carl Morgan (NOAA/NWS Wilmington, North Carolina), John Billet (NOAA/NWS Wakefield, Virginia), and Dr. Michael Brennan (NOAA/National Hurricane Center, Miami, Florida). The authors also appreciate the prompt and useful feedback from Brian Miretzky (NOAA/NWS). The paper greatly benefited from comments provided by three anonymous reviewers. This research was supported by CSTAR Grant NA10NWS4680007.

REFERENCES

  • Avila, L. A., 2009: Tropical cyclone report: Hurricane Cristobal. NOAA/National Hurricane Center Tech. Rep. AL032008, 8 pp. [Available online at http://www.nhc.noaa.gov/data/tcr/AL032008_Cristobal.pdf.]

  • Avila, L. A., , and Cangialosi J. , 2012: Tropical cyclone report: Hurricane Irene. NOAA/National Hurricane Center Tech. Rep. AL092011, 45 pp. [Available online at http://www.nhc.noaa.gov/data/tcr/AL092011_Irene.pdf.]

  • Bailey, C. M., , Hartfield G. , , Lackmann G. M. , , Keeter K. , , and Sharp S. , 2003: An objective climatology, classification scheme, and assessment of sensible weather impacts for Appalachian cold-air damming. Wea. Forecasting, 18, 641661, doi:10.1175/1520-0434(2003)018<0641:AOCCSA>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Benjamin, S. G., and et al. , 2004: An hourly assimilation forecast cycle: The RUC. Mon. Wea. Rev., 132, 495, doi:10.1175/1520-0493(2004)132<0495:AHACTR>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Brown, D., 2008: Tropical cyclone report: Hurricane Gabrielle. NOAA/National Hurricane Center Tech. Rep. AL072007, 13 pp. [Available online at http://www.nhc.noaa.gov/data/tcr/AL072007_Gabrielle.pdf.]

  • Childs, B., , and Lewis R. , 2001: Investigating the impact of changing the ASOS wind gust averaging period from five to three seconds. Preprints, 11th Symp. on Meteorological Observations and Instrumentation, Albuquerque, NM, Amer. Meteor. Soc., 2.5.[Available online at https://ams.confex.com/ams/annual2001/techprogram/paper_18249.htm.]

  • Davies, J., 2006: Hurricane and tropical cyclone tornado environments from RUC proximity soundings. Preprints, 23rd Conf. Severe Local Storms, St. Louis, MO, Amer. Meteor. Soc., 12.6A. [Available online at https://ams.confex.com/ams/pdfpapers/115483.pdf.]

  • Durst, C. S., 1960: Wind speeds over short periods of time. Meteor. Mag., 89, 181187.

  • Eastin, M. D., , Gardner T. L. , , Link M. C. , , and Smith K. C. , 2012: Surface cold pools in the outer rainbands of Tropical Storm Hanna (2008) near landfall. Mon. Wea. Rev., 140, 471491, doi:10.1175/MWR-D-11-00099.1.

    • Search Google Scholar
    • Export Citation
  • Glahn, B., 2005: Comments on implementation and refinement of digital forecasting databases. Bull. Amer. Meteor. Soc., 86, 13151318, doi:10.1175/BAMS-86-9-1315.

    • Search Google Scholar
    • Export Citation
  • Glahn, H. R., , and Ruth D. P. , 2003: The New Digital Forecast Database of the National Weather Service. Bull. Amer. Meteor. Soc., 84, 195201, doi:10.1175/BAMS-84-2-195.

    • Search Google Scholar
    • Export Citation
  • Hanson, T., , Matthewson M. , , and Romberg M. , 2001: Forecast methodology using the GFE Suite. Preprints, 17th Int. Conf. on Interactive Information and Processing Systems for Meteorology, Oceanography, and Hydrology, Albuquerque, NM, Amer. Meteor. Soc., 1.13. [Available online at https://ams.confex.com/ams/annual2001/techprogram/paper_18722.htm.]

  • Harper, B. A., , Kepert J. D. , , and Ginger J. D. , 2010: Guidelines for converting between various wind averaging periods in tropical cyclone conditions. World Meteorological Organization Tech. Rep. WMO/TD-1555, 54 pp.

  • Hsu, S. A., 2001: Spatial variations in gust factor across the coastal zone during Hurricane Opal in 1995. Natl. Wea. Dig., 25, 2123. [Available online at http://www.nwas.org/digest/papers/2001/Vol25No12/Pg21-Hsu.pdf.]

    • Search Google Scholar
    • Export Citation
  • Hsu, S. A., 2003: Nowcasting the variation of wind speed with height using gust factor measurement. Natl. Wea. Dig., 27, 6769. [Available online at http://www.nwas.org/digest/papers/2003/Vol27No1/Pg67-Hsu.pdf.]

    • Search Google Scholar
    • Export Citation
  • Kaplan, J., , and Demaria M. , 1995: A simple empirical model for predicting the decay of tropical cyclone winds after landfall. J. Appl. Meteor., 34, 24992512, doi:10.1175/1520-0450(1995)034<2499:ASEMFP>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Knabb, R. D., , and Mainelli M. , 2007: Tropical cyclone report: Hurricane Ernesto. NOAA/National Hurricane Center Tech. Rep. AL062006, 36 pp. [Available online at http://www.nhc.noaa.gov/data/tcr/AL062006_Ernesto.pdf.]

  • Krayer, W., , and Marshall R. D. , 1992: Gust factors applied to hurricane winds. Bull. Amer. Meteor. Soc., 73, 613618, doi:10.1175/1520-0477(1992)073<0613:GFATHW>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Malkin, W., 1959: Filling and intensity changes in hurricanes over land. National Hurricane Research Rep. 34, U.S. Weather Bureau, Washington, DC, 18 pp.

  • Masters, F. J., , Vickery P. J. , , Bacon P. , , and Rappaport E. N. , 2010: Toward objective, standardized intensity estimates from surface wind speed observations. Bull. Amer. Meteor. Soc., 91, 16651681, doi:10.1175/2010BAMS2942.1.

    • Search Google Scholar
    • Export Citation
  • Moore, B. J., , Bosart L. F. , , Keyser D. , , and Jurewicz M. L. , 2013: Synoptic-scale environments of predecessor rain events occurring east of the Rocky Mountains in association with Atlantic basin tropical cyclones. Mon. Wea. Rev., 141, 10221047, doi:10.1175/MWR-D-12-00178.1.

    • Search Google Scholar
    • Export Citation
  • Mueller, K. J., , Demaria M. , , Knaff J. , , Kossin J. P. , , and Vonder Haar T. H. , 2006: Objective estimation of tropical cyclone wind structure from infrared satellite data. Wea. Forecasting, 21, 990, doi:10.1175/WAF955.1.

    • Search Google Scholar
    • Export Citation
  • Paulsen, B., , and Schroeder J. , 2005: An examination of tropical and extratropical gust factors and the associated wind speed histograms. J. Appl. Meteor., 44, 270280, doi:10.1175/JAM2199.1.

    • Search Google Scholar
    • Export Citation
  • Powell, M. D., , and Houston S. H. , 1998: Surface wind fields of 1995 Hurricanes Erin, Opal, Luis, Marilyn, and Roxanne at landfall. Mon. Wea. Rev., 126, 1259, doi:10.1175/1520-0493(1998)126<1259:SWFOHE>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Powell, M. D., , Houston S. H. , , Amat L. R. , , and Morisseau-Leroy N. , 1998: The HRD Real-time Hurricane Wind Analysis System. J. Wind Eng. Ind. Aerodyn., 77–78, 5364, doi:10.1016/S0167-6105(98)00131-7.

    • Search Google Scholar
    • Export Citation
  • Schwerdt, R. W., , Ho F. P. , , and Watkins R. R. , 1979: Meteorological criteria for standard project hurricane and probable maximum hurricane wind fields, Gulf and East Coasts of the United States. NOAA Tech. Rep. NWS 23, 317 pp. [Available online at http://www.nws.noaa.gov/oh/hdsc/Technical_reports/TR23.pdf.]

  • Srock, A. F., , and Bosart L. F. , 2009: Heavy precipitation associated with southern Appalachian cold-air damming and Carolina coastal frontogenesis in advance of weak landfalling Tropical Storm Marco (1990). Mon. Wea. Rev., 137, 2448, doi:10.1175/2009MWR2819.1.

    • Search Google Scholar
    • Export Citation
  • Vickery, P., , and Skerlj P. , 2005: Hurricane gust factors revisited. J. Struct. Eng., 131, 825832, doi:10.1061/(ASCE)0733-9445(2005)131:5(825).

    • Search Google Scholar
    • Export Citation
  • Walsh, K. J. E., , Sandery P. , , Brassington G. B. , , Entel M. , , Siegenthaler-Ledrian C. , , Kepert J. D. , , and Darbyshire R. , 2010: Constraints on drag and exchange coefficients at extreme wind speeds. J. Geophys. Res., 115, C09007, doi:10.1029/2010JA015748.

    • Search Google Scholar
    • Export Citation
  • Yu, B., , and Chowdhury A. G. , 2009: Gust factors and turbulence intensities for the tropical cyclone environment. J. Appl. Meteor. Climatol., 48, 534552, doi:10.1175/2008JAMC1906.1.

    • Search Google Scholar
    • Export Citation
Save