Gust factors during thunderstorm episodes versus non-thunderstorm episodes in the midwest.

Akyuz, F. Adnan

Abstract

The statistical relationship between observed peak gust velocities and simultaneously measured fastest-minute wind was examined in order to determine "gust factors" appropriate to thunderstorm episodes in the Midwestern United States. This paper focuses on a pilot study conducted on wind data from 1984 through 1991 for the four first-order National Weather Service (NWS) stations in Missouri: Columbia Regional (COU), Kansas City International (MCI), St. Louis International (STL), and Springfield Regional (SGF) airports.

The majority of the annual wind damage to structures in the US is produced by gusts associated with straight-line thunderstorm winds as opposed to tornadoes and hurricanes. Gust factor, which is defined as the ratio of the peak gusts to the sustained winds during thunderstorm episodes, is quite different from those published gust factors based on observation of gusts and sustained wind during episodes of mechanical turbulence. The study showed that the gust factors during thunderstorm episodes ranged from 1.39 to 1.5 while the gust factors during non-thunderstorm episodes ranged from 1.37 to 1.38. The purpose of this paper is to examine the observed relationship between recorded peak gusts and simultaneously observed sustained winds, and to estimate appropriate gust response factors for those localities where wind damage will most likely be due to thunderstorm winds.

1. Introduction

In the design of structures, the structural engineer must consider the likelihood of wind damage to the structure or portions thereof over the lifetime of the structure. Wind damage to structures is normally caused by or initiated by very short-term (instantaneous) peak gusts superimposed on a sustained background wind of longer duration but of lesser average speed. (Definitions of sustained wind speed, peak gust and gust factor are given in the Appendix section). Up to 1985, peak gust velocities were not routinely measured and archived at the National Weather Service (NWS) offices. The recorded and archived maximum wind data have consisted of either the maximum of wind averaged over a one-minute interval (fastest-minute wind) or the fastest mile of wind passing the anemometer (fastest-mile wind), during each 24-h day.

Wind effects on buildings and structures registered first conceptualized attention at international conferences in 1963 (Proceedings of Conference on Buildings and Structures, 1963; Cermak, 1985; Mehta, 1988). A more comprehensive consideration of wind effects was initiated in 1970 (Roskho, 1970) at a conference where the designation of "Wind Engineering" was adopted to identify the new discipline. A pioneering gust factor study was one by Durst (1960) using data that had been obtained with Dines recorders at Cardington, England. The analyses reported by Durst involved averaging time departures from 10-minute mean speeds as well as from 1-h mean speeds, with the reported gust factors being referenced to 1-h means. Thereafter, most studies relating average wind speeds to shorter-term gust speeds referred to results of Durst's study. Vellozzi et al. (1970), for example, relied on Durst's results in their study discussing methods of calculating the dynamic response of tall, flexible structures, such as towers, stacks and masts to wind loading. Davis et al. (1968) and Gill (1969) studied gust factor variation with height and with mean wind speed, which depended on the averaging time. Tattelman (1975) studied gustiness and wind speed range as a function of averaging time interval and mean wind speed. The results of that study showed that, in general, the gust factors decrease with increasing wind speed and as the averaging time decreases, the wind speed increases, results noted by Davis et al. (1968). A diagrammatic illustration of this effect is shown figure 1. It shows the relationship between average wind speed and the averaging time interval. As the averaging time increases, the average wind during that time interval centered on a peak wind will decrease. An arrow in the figure indicates the peak gust in the center of the averaging time interval.

[FIGURE 1 OMITTED]

Vellozzi et al. (1967) indicated a probable gust factor of 1.56 for peak gusts of 1 sec. duration when calculated from sustained winds averaged over one hour. The most straightforward gust factor calculation was made by Brekke (1959) by simply dividing the gust speeds by the fastest-mile wind. The gust factors varied between 1.3 and 1.8 over a range of averaging time dependent on the speed of the wind during the fastest-mile of wind passage. Cramer (1960) used a rule-of-thumb gust factor of 1.62 based on the ratio of instantaneous peak gust to 10-minute average wind for 99% of all cases and 1.38 for 99% of very high wind speeds. The decreasing gust factor values with higher wind speeds agree with the study of Tattelman (1975). The gust factor curves that the study presented showed that, in general, the gust factor decreases with increasing wind speed. Faber et al. (1963) had undertaken their study based on typhoons, and Krayer et al. (1992) studied gust factors applied to hurricanes. Faber et al. (1963) found gust factors as high as 2.05, while Krayer et al. (1992) found an average gust factor of 1.78. However, Krayer et al. (1992) compared gust factors derived from hurricane winds with those derived from open-scale records such as by Durst (1960). Shellard (1965) conducted a study of gust factors for an open exposure near a coastline. His gust factor results varied between 1.3 and 1.9 with the suggested average value of 1.48. Gust factors computed by Deese (1964) for heights greater than 60 feet and 5-minute average wind speeds greater than 30 mph were less than 1.4. Table 1, as developed by Davis and Newstein (1968) with the additional results of Akyuz (1994), Krayer and Marshall (1992), and Beebe (1977), shows historical computed gust factors.

An earlier study by Huss (1974) showed that the average distribution of ratios of maximum wind speed to average wind speeds for one location could be used for another location of a similar exposure to estimate the expected maxima when there is no existing wind history for that location. Average wind speed could also be used to estimate the peak gust at different wind speed intervals for a location where the peak gust data is absent.

Brook et al. (1970) also studied wind gusts due to mechanical turbulence that also permitted structural engineers and others to interpret presently available, extreme-wind gust data, in terms of height and averaging times appropriate to their specific applications.

The fastest-mile wind has been the most conventionally reported wind statistic for almost all stations in the United States for many years prior to 1989. Starting in September 1989 continuing through 1994, the Weather Service adopted the fastest-minute wind. On the other hand, peak gust observations have only recently been made routinely and archived after the advent of the Automated Surface Observing Stations (ASOS). Beebe (1977) studied peak gust-fastest-mile wind relationship and came up with a linear relationship. The study showed that a reasonable estimate of the peak gust could be found from fastest-mile wind by just multiplying the fastest-mile wind by the gust factor of 1.2. This is the same gust factor value that Hollister (1970) found for the average wind speeds of 60 mph. The relationships found in either case may not be valid in the areas where thunderstorms are present.

Galway (1975) analyzed data based on a comprehensive listing of reported maximum thunderstorm gusts greater than 50 knots (57.5 mph) in the central Great Plains of the United States to respond to the request from architects, design engineers and the others who are interested in peak wind loading on buildings and the other construction.

One of the latest studies based on peak gusts for different weather types by Peterson et al. (1993) determined a linear relationship between peak gusts and fastest-minute winds reported at five stations for various periods prior to 1991. The slope of the regression for total weather events was found to be 1.16 with the constant of 5.82 and r-squared value of 0.70 when all data for five stations were composited.

The longest averaging time used for wind speed for the operational period of a measuring station (at least 30-yr long) is one hour. This long-term average is referred to as annual mean. Although information on this speed is important for wind energy utilization, it is useless for wind load on structures because only high winds of short durations (peak gust in most cases) are of interest in this case (Liu, 1991). In the absence of an extensive history of peak gust observations, the probability of peak gust values can therefore be estimated by applying a peak gust factor to sustained wind speeds.

In much of the United States and in particular in the Midwest, between the Rocky Mountains and the Appalachian Mountains, the strongest episodes of straight-line wind damage are associated with thunderstorm outflow. Straight-line wind damage due to thunderstorm activity in the Midwest is far more frequent, widespread and costly than wind damage due to other causes, such as tornadoes. In fact, the probability that a particular structure in the Midwest will be hit by straight-line thunderstorm wind gusts of say, 100 mph is ten times as great as that of a tornado with the same wind speed (Darkow, 1986). The majority of damage due to straight-line thunderstorm wind gusts, much of which could be prevented by proper understanding of basic wind-gust relationship and proper construction design, occurs on farms and in small communities of rural America.

There is no evidence showing that the relationship of peak gusts to sustained winds, or gust factors, during thunderstorm outflow episodes, should be the same as the published gust factors based on non-thunderstorm episodes, measurement of winds during episodes of mechanical turbulence.

It is the purpose of this paper to examine the observed relationship between recorded peak gusts and simultaneously observed background-sustained winds, and to estimate appropriate gust response factors for those localities where wind damage will most likely be due to thunderstorm winds.

The turbulence in these studies was generated as mechanical turbulence due to strong flow over roughened underlying terrain. The studies did not include episodes of turbulence associated with intense convectively-induced turbulence such as that produced by strong outflow or downbursts.

2. Data Selection

Data for this study was extracted from Local Climatological Data. Monthly and Annual Summaries for four stations in Missouri: Columbia Regional Airport (COU), Kansas City International Airport (MCI), Springfield Regional Airport (SGF), and St. Louis International Airport (STL). The data covered the period from 1984 to 1991 for this study. Daily peak gust and fastest one-minute wind information are reported for each station as well as the monthly maximum of both in miles per hour for every month of the study period. The Local Climatological Data also provided "Weather Types" information. The thunderstorm occurrences were selected based on the information given under the "Weather Types" column. It was assumed that the maximum peak gust and the associated fastest one-minute wind were measured during the thunderstorm event reported. The data was stratified to separate thunderstorm-related events from non-thunderstorm cases. The peak gust, fastest-minute wind and the date on which they occurred were recorded for both thunderstorm related and non thunderstorm related cases.

The procedure allowed entering all thunderstorm cases which yielded more than one entry into the thunderstorm cases for the months that had more than one thunderstorm day. On the other hand, there was one entry per month in the non-thunderstorm cases, the maximum wind for the given month. The entry was omitted when the simultaneously measured peak gust data was not readily available since the data set required both fastest-minute and peak gust values.

3. Data Analysis and Results

A regression analysis between peak gusts and fastest-minute average winds was made for each station and for the combined data for the four stations for both thunderstorm and non-thunderstorm cases. Another regression analysis was run between the gust factor and the associated fastest-minute wind speeds to determine how the gust factor changes with fastest-minute wind speed changes. A gust factor histogram was generated to display the frequency distribution of the gust factor.

Figures 2 and 3 show the comparison between relationship of peak gust and average wind (averaged over one-minute period) during non-thunderstorm and thunderstorm activities, respectively. None of the cases with wind speeds less than 10 mph were taken into consideration. The following generalizations emerge from the regression analysis between peak gust and average wind for non-thunderstorm and thunderstorm cases:

* Peak gusts show more variability around the line of best fit during thunderstorm cases than during the non-thunderstorm cases. Note the correlation of determination, r2, of 0.766 and 0.745 for non-thunderstorm and thunderstorm cases, respectively.

* Peak gust variability decreases as wind speed increases during thunderstorm events.

* Peak gust variability is rather constant with the wind speed during non-thunderstorm events.

[FIGURES 2-3 OMITTED]

Figures 4 and 5 are the comparison between relationship of gust factor and average wind (averaged over one-minute period) during non-thunderstorm and thunderstorm activities, respectively. They show how the gust factor varies with wind speed averaged over a one-minute period. The following generalizations may be made from the regression analysis between the gust factor and average wind for non-thunderstorm and thunderstorm cases:

* In general, the gust factor decreases with increasing wind speed. This agrees with Tattelman's results (Tattlelman, 1975) and those found by Davis et al. (1968).

* Neither Figure 4 nor Figure 5 suggested a strong correlation. However, gust factors and average winds are more correlated during thunderstorm events (r = 0.2) than non-thunderstorm events (r = 0.07). Even though the figures may suggest that there is more variability of gust factor to average wind during thunderstorm events, one must keep in mind that there are many data points that coincides near the estimation line for the thunderstorm events.

[FIGURES 4-6 OMITTED]

Figure 6 is the frequency histogram of the gust factor during thunderstorm activities. It reveals that 67% of all thunderstorm cases had a gust factor between 1.3 and 1.8 with the mode of 1.50.

[FIGURE 6 OMITTED]

Table 2 is an overall population of the gust factors (GF) during thunderstorm and non-thunderstorm occurrences. Each station is weighted individually in two speed intervals and two cases as well as in combination as if they were one composite station. The wind speed range greater than 10 mph included all cases. The range greater than 30 mph excluded 30-mph and less wind speeds. All stations are also analyzed as if they were one composite station. The number of occurrences, N, is given for each case. The results emerging from the Table 2 are as follows:

* In general, the gust factors are lower at the higher wind speed interval for both thunderstorm and non-thunderstorm cases.

* The gust factors showed negligible variation between the individual stations.

* In general, the frequency distribution of GF is narrower at the higher speed interval for thunderstorm cases and is independent of wind speed for non-thunderstorm cases.

* In general, gust factors are higher during thunderstorm events than non-thunderstorm events at all wind speed ranges, especially at lower speeds.

4. Conclusion

A structural engineer must allow for the likelihood of wind damage initiated by very short-term peak gusts in addition to that caused by the sustained winds of longer duration but of lesser average wind speeds. In the absence of an extensive history of peak gust observations, a more appropriate gust factor than the existing gust factors must be applied in order to determine the probable peak gust wind speed values. This study shows that the gust factors appropriate to thunderstorm episodes of the Midwest are much greater that those associated with non-thunderstorm episodes. A structure located anywhere in the Midwest is more likely to be damaged by thunderstorm outflow straight-line winds than by a tornado. In fact, for wind speeds less than 125 mph, the probability of a structure being affected by straight-line thunderstorm winds is greater than the probability of being struck by a tornado. Also, thunderstorm winds cause more cumulative damage than tornadoes. Therefore, the structures located in the Midwest must be designed or reinforced to withstand the peak gusts of thunderstorm straight winds. The Author suggests the use of Figure 6, frequency distribution of the gust factor during thunderstorm activity. Thus, the mode of the gust factor distribution of 1.5 must be multiplied by sustained wind speed averaged over a one-minute period in order to estimate the design peak gust for a location in the Midwestern United States.

Definitions

Sustained Wind Speed (V): Daily maximum wind speed averaged over a 1-minute interval (fastest-minute wind) or averaged over time during a passage of 1 mile of wind (fastest-mile wind) at 33 ft (10 m) above ground in open terrain as reported in column 18, Local Climatological Data (LCD) (1), National Oceanic and Atmospheric Administration (NOAA).

Peak Gust (PG): Daily peak wind speed at 33 ft (10 m) above ground in open terrain as reported in column 16, LCD.

Gust Factor (GF): Ratio of the peak gust to sustained wind speed.

References

Akyuz, F. A., 1994: Thunderstorm Peak Gust Estimation for Structural Engineering Design. Dissertation. Atmospheric Science Department. University of Missouri-Columbia. pp 83.

Beebe, R. G., 1977: Wind, Fastest Mile vs. Peak Gust. Technical Attachment 77-6, March 1977.

Brekke, G. N., 1959: Wind Pressures In Various Areas of the United States. Building Materials and Structures Rept. National Bureau of Standards. 152, 8 pp.

Brook, R. R., and K. T. Spillane, 1970: On the Variation of Maximum Wind Gust With Height. Journal of Applied Meteorology. 9, 7278.

Cermak, I. E., 1985: Wind Engineering Applications to Buildings and Structures. Wind Effects. UMC's Tenth Annual Short Course on Wind Load. Department of Civil Engineering, College of Engineering and Engineering Extension, University of Missouri-Columbia.

Cramer, H. E., 1960: Use of Power Spectra and Scale of Turbulence in Estimating wind Loads. Meteor. Monogr., 4, No.22, 12-18.

Darkow, a. L., 1986: Tornado Wind Probabilities for Engineers, Course Notes, 11th Annual Continuing Education Short Course on Wind Effects on Buildings and Structures, Engineering Extension, University of Missouri-Columbia.

Davis, F. K., and Newstein, 1968: The Variation of Gust Factors With Mean Wind Speed and With Height. Journal of Applied Meteorology, 7, 372-378.

Deese, J. H., 1964: Problem of Low Level Wind Distribution. J. F. Kennedy Space Center, NASA, Cape Kennedy, FL., 88 pp.

Durst, C. S. 1960: Wind Speeds Over Short Periods of Time. Meteor. Mag. Vol. 89, 181-186.

Faber, S. E., and G. J. Bell, 1963: Typhoons in Hong Kong and Building Design. The Engineering Society of Hong Kong, 28 pp.

Galway, J. G., 1975: Maximum Thunderstom Gusts in the NWS Central Region. Technical Attachment 75-8. CRH-SSD.

Gill, G. C., 1969: Comments on "The Variation of Gust Factors with Mean Wind Speed and with Height". Notes and Correspondence, Journal of Applied Meteorology, 8, 167-167.

Gumbel, E. 1J, 1958: Statistics of Extremes. Columbia University Press. New York. New York. 375 pp.

Hollister, S. C., 1970: "The Engineering Interpretation of Weather Bureau Records for Wind Loading on Structures. Proceedings of Technical Meeting Concerning Wind Loads on Buildings and Structures. Washington D.C. 1970.

Huss. P. O. 1974: Estimation of Distributions and Maximum Values of Horizontal Wind Speeds. Journal of Applied Meteorology. 13, 647-653.

Krayer, W. R., and R. D. Marshall, 1992: Gust Factors Applied to Hurricane Winds. Bulletin American Meteorological Society. Vol. 73, No.5, May 1992, 613-617.

Liu, H., 1991: Wind Engineering: A Handbook for Structural Engineers. Prentice-Hall, NJ. 209 PP.

Metha. K. C., 1988: Guide to the Use of the Wind Load Provisions of ANSI A58.1. NSF Granr ECE-8512044. Institute for Disaster Research Texas Tech University. Lubbock, Texas.

Peterson, R. E., S. D. Gollestain and K. C. Metha, 1993: Analysis of Peak Gusts vs. Fastest-Mile Wind Statistics. Proceedings, Third Asia-Pacific Symposium on Wind Engineering, 13-15 December 1993, Hong Kong.

Proceedings of Conference on Wind Effects on Buildings and Structures (Redisignated as First International Conference on Wind Engineering), National Physical Laboratory, Teddington, England, 26-28 June 1963, Vol. I and II, H.M.S.O. 1065, 851.

Roshko, A., (1970): Conference reports on Wind Loads on Structures--First U. S. National Conference on Wind Engineering Research, California Institute of Technology, Pasedena, CA, December 1970, 137.

Shellard, H. C., 1965: The Estimation of Design Wind Speeds. Wind Effects on Building on Buildings and Structures, National Physical Labratory Symp. No. 16, 30-51.

Tattelman, P., 1975: Surface Gustiness and Wind Speed Range as a Function of Time Interval and Mean Wind Speed. Journal of Applied Meteorology, 14, 1271-1276.

Vellozzi, J., and E. Cohen, 1967: Gust Response Factors for Buildings and other Structures. Conference Preprint No.434, American Society of Civil Engineers, Environmental Engineering Conf., Dallas, Texas, 31 pp.

Vellozzi, J., and E. Cohen, 1970: Dynamic response of Tall Flexible Structures to Wind Loading Proceedings of the Technical Meeting Concerning Wind Loads on Buildings and Structures. Building Science Series 30. National Bureau of Standards. Washington D.C. 115-128

(1) http://www.ncdc.noaa.gov./oa/pdfs/lcd.html

F. Adnan Akyuz (1) NOAA's National Weather Service Climate Services 7220 NW 101st Terrace, Kansas City, MO 64153

(1) Corresponding author address: Fikri Adnan Akyuz, NOAA's National Weather Service, Climate Services 7220 NW 101st Terrace, Kansas City, MO 64153, mailto:Adnan.Akyuz@noaa.gov Phone: 816- 891-7734 ex:706, Fax: 816- 891-7810.

Table 1. Historical Gust Factors (Davis et al., 1968
with additional results of Krayer et al., 1992; Beebe,
1977; and Akyuz, 1994).

 Time Average
 Range of Gust of Mean Duration of Max.
Investigator Factors Wind Speed Wind Speed

Akyuz (1994) 1.53-1.62 1 min Instantaneous
Beebe (1977) 1.20 Varies Instantaneous
Brekke (1959) 1.08-1.30 Varies --
Cramer (1960) 1.38-1.62 10 min Instantaneous
Deese (1964) 1.20-2.00 5 min Instantaneous
Durst (1960) 1.00-1.59 1 hr 0.5 sec to 1 hr
Faber and Bell 1.28-2.05 1 hr Instant to 1 min
 (1963)
Krayer and Marshall 1.78 10 min 2 sec
 (1992)
Shellard (1965) 1.30-1.90 10 min 3 to 5 sec
Vellozzi and Cohen 1.56 1 hr 1 sec
 (1967)

Table 2. Gust Factor Comparison Between Thunderstorm and Non-
Thunderstorm Events with Varying Wind Speed Range Observed at
Columbia Regional (COU), Kansas City International (MCI),
Springfield Regional (SGF), and St. Louis International (STL)
Airports.

 Wind
 Speed Standard
 Range Average Deviation
Cases Station mph GF of GF

 >30 1.41 0.19
 COU >10 1.51 0.32

 >30 1.43 0.16
 MCI >10 1.51 0.29

 >30 1.37 0.13
THUNDERSTORM SGF >10 1.52 0.28

 >30 1.35 0.14
 STL >10 1.46 0.23

 >30 1.39 0.16
 COMPOSITE >10 1.50 0.28

 >30 1.33 0.16
 COU >10 1.39 0.17

 >30 1.38 0.11
 MCI >10 1.38 0.13

 >30 1.35 0.14
NON- SGF >10 1.35 0.13
THUNDERSTORM
 >30 1.40 0.17
 STL >10 1.41 0.15

 >30 1.37 0.15
 COMPOSITE >10 1.38 0.15

 Maximum Minimum
Cases Station GF GF N

 2.00 1.17 27
 COU 3.41 1.05 188

 1.97 1.16 26
 MCI 2.82 1.00 187

 1.71 1.23 17
THUNDERSTORM SGF 3.12 1.00 201

 1.77 1.10 32
 STL 2.40 1.00 186

 2.00 1.10 102
 COMPOSITE 3.41 1.00 762

 1.77 1.08 26
 COU 1.91 1.08 78

 1.63 1.13 27
 MCI 1.68 1.04 59

 1.55 1.18 12
NON- SGF 1.60 1.04 65
THUNDERSTORM
 1.91 1.16 28
 STL 1.91 1.16 61

 1.91 1.08 93
 COMPOSITE 1.91 1.04 263