ARTICLE IN PRESS

Journal of Wind Engineering and Industrial Aerodynamics 92 (2004) 275290

Field measurement and experimental study of wind speed prole during thunderstorms
Edmund C. C. Choi*
School of Civil & Environmental Engineering, Nanyang Technological University, Nanyang Avenue, Singapore 639798, Singapore Received 18 September 2003; received in revised form 2 December 2003; accepted 8 December 2003

Abstract This paper investigates the variation of wind velocity with height during thunderstorms. Wind velocities during thunderstorms are measured at ve levels on a 150 m tall tower. Velocity proles for more than 50 thunderstorms are studied. In the analysis, these proles are classied into four types according to the prole shape and the height of the highest wind speed. It is observed that wind proles during thunderstorms are affected by several factors with the major ones being distance from the thunderstorm cell centre, intensity of the storm and ground roughness. Experimental studies on the simulation of thunderstorm wind using an impinging jet are also carried out to further investigate the effect of these parameters. r 2003 Elsevier Ltd. All rights reserved.
Keywords: Thunderstorm; Downburst; Velocity prole; Wind speed; Wind direction; Terrain effect

1. Introduction Thunderstorm wind has been a topic of interest to engineers in the past decade. Recently, there are more indications that thunderstorm wind plays a signicant part in dening the strong wind characteristics of a place [1,2]. Many investigations are carried out to study the wind characteristics during a thunderstorm; of particular interest is the variation of wind speed with height of a downburst which as dened by Fujita [3] as a strong downdraft which induces an outburst of damaging winds on or near the ground. Such investigations included eld observations and monitoring [46] as well as physical, empirical and numerical modelling [710].
*Tel.: 67905273; fax: 67910676. E-mail address: [email protected] (E.C.C. Choi). 0167-6105/$ - see front matter r 2003 Elsevier Ltd. All rights reserved. doi:10.1016/j.jweia.2003.12.001

ARTICLE IN PRESS
276 E.C.C. Choi / J. Wind Eng. Ind. Aerodyn. 92 (2004) 275290

1.1. Thunderstorm Thunderstorms are common occurrence around the world. There are nearly 2000 thunderstorms in progress at any time over the earths surface. Every thunderstorm is produced by a cumulonimbus cloud and is always accompanied by lightning and thunder. Although its occurrence is worldwide, the frequency of thunderstorm occurrence varies greatly in different areas. Most thunderstorms are found in tropical and sub-tropical regions. A thunderstorm usually lasts no more than 1 h and it begins when a parcel of air, either warmer than the surrounding or being pushed up by colder encroaching air such as a cold front, begins to rise. This unstable updraught forms the embryo of the thunderstorm. If the rising air is sufciently moist, condensation will occur to form cloud droplets. A large amount of moisture condenses and precipitation particles begin to grow as the cell continues to build up. The latent heat released during condensation adds buoyancy to the vapour. The rising air stream forms an updraft that is strong enough to keep the water droplets and ice crystals suspended in the clouds. The continuous release of latent heat supplies energy to the system and further accelerates the updraught. As precipitation gathers momentum, the frictional drag induces downdraught. During the mature stage of the thunderstorm, downdraught intensies. This is the most active period of a thunderstorm with gusty winds and heavy precipitation. The leading edge of the cold air spreading out over the ground resulting from the downdraft is called the gust front. In severe thunderstorms, the strong downdraft develops into a downburst, which diverges horizontally as an outburst of destructive winds upon reaching the ground. The gust front usually moves ahead of the thunderstorm. Unlike the monsoon wind or even the typhoon wind where the wind moves in the horizontal direction over a substantial stretch of ground, winds from thunderstorm rush vertically down and spread outwards. Thus, the ground roughness has little effect on the wind. The wind characteristics of thunderstorm wind, e.g. the mean wind prole, gust prole, gust factor and turbulence intensity, are expected to be quite different from those of the large-scale winds.

2. Field measurement Singapore is an island situated close to the equator. In general, the climate of Singapore is made up of two monsoon seasons, northeast monsoon from December to March and southwest monsoon from June to September, and frequent thunderstorms during the inter-monsoon periods. There are more than 100 thunderstorm days in a typical year. Thus, Singapore is an ideal place for thunderstorm research. 2.1. Tuas wind measuring station A wind measuring station is set up at the western end of the Singapore island. It is located at the Tuas industrial area. The terrain is reasonably at and surrounded by

ARTICLE IN PRESS
E.C.C. Choi / J. Wind Eng. Ind. Aerodyn. 92 (2004) 275290 277

Fig. 1. Site of Tuas wind measuring station.

almost homogeneous industrial buildings with the open sea about 24 km in the S and W directions (Fig. 1). Towards north and east, there are isolated taller buildings further away. Three-component ultra sonic high-frequency anemometers are used to measure the wind speed. These anemometers are mounted at ve levels on a tower, at heights of 35.7, 62, 80, 125, and 152 m: Temperature and humidity sensors are also installed at these levels. Normally, statistics (mean, r.m.s., maximum, minimum) of wind speed and direction are recorded continuously every 10 min: When triggered into high-speed mode (1 min mean speed exceeding 8 m=s) wind speed and direction data are recorded at 50 Hz: 2.2. Wind systems In the present study, weather events are classied into two categories, the thunderstorm (TS) and non-thunderstorm (NTS) events. TS events are identied by the thunder, inspection of the 10-min mean wind speed, r.m.s. speed, wind direction, temperature, humidity, and pressure data. A typical TS event is characterized by a sudden increase in wind speed, a sudden change in wind direction, high r.m.s. value, a rapid drop of temperature and an increase in humidity as shown in Fig. 2. Radar plots are also used to obtain more information and further conrmation of the event. NTS events are selected from occasions of ne days with a high wind speed. These are the monsoon events. It was found that the NE monsoon winds are slightly stronger than that of the SW monsoon winds. Studies on the variation of wind speed with height are carried out for over 90 NTS events. The averaged power exponents of the power-law prole for winds coming over land (N to E) is found to be 0.22 and over sea (W to S) is 0.17. These values reect the terrain roughness and are used as

ARTICLE IN PRESS
278 E.C.C. Choi / J. Wind Eng. Ind. Aerodyn. 92 (2004) 275290

Fig. 2. Typical plots of a thunderstorm event.

datum for TS prole investigation. Furthermore as reported by Choi [2], studies of extreme wind in Singapore showed that the 10-min mean and the 3-s gust extreme wind are governed by small scale storms. The 50-year return wind speeds are respectively 17 and 29 m=s for the 10-min mean and the 3-s gust. These give an overall picture of the wind climate in Singapore. 2.3. Thunderstorm wind While NTS events are considered to be large scale, with winds move more or less horizontally over the ground, and wind characteristics largely governed by terrain

ARTICLE IN PRESS
E.C.C. Choi / J. Wind Eng. Ind. Aerodyn. 92 (2004) 275290 279

roughness; thunderstorm winds rush vertically down and when the downdraft reaches the ground, it spreads out forming the out-ow and the gust front. It is generally accepted that at the gust front and out-ow regions wind speed increases with height from the ground upwards, reaching a peak at a height of about 100 200 m and the wind speed starts to decrease. While this is a general statement, it is understood that the shape of the wind prole is affected by many factors, e.g. the intensity of the storm, the size of the storm cell, distance of the prole from the cell centre and the ground roughness. Out of the 50 over TS events 36 have been selected for analysis in the present investigation. For the thunderstorm velocity prole study, 5-min mean wind speeds are calculated. A 5-min mean is used as TS events are very localized in space and time and a 10-min mean is observed to be too long to reect the high wind characteristics. The variations of the 5-min mean wind speed with height (at the time where the 5-min running mean gives the highest value for the storm) during TS events indicate different shapes of the wind prole. The highest wind speed is observed sometimes at the lower levels, but sometimes also at the highest level. Using the exponent obtained from the NTS events and tting a power law prole to the data point at the bottom level and also the data point at the top level, leads to the classication of the prole shape into four types. For Type I, all data points t reasonably well with the power prole of the NTS exponent, and with the highest wind speed at the top level. For Type II, the higher wind speeds occur at the lower levels. Wind speeds of all levels fall within the NTS power exponent prole tted to the lowest level data point (Fig. 3). Type III is similar to Type II except that wind speeds at other levels exceed the proles tted to lowest level as well as the top level data points. For Type IV, the highest wind speed is recorded at the top level. All wind speeds are less than the prole tted to the top level data point but exceed the prole tted to the bottom point. 2.4. Case studies To gain further insight to the different proles, the movement and location of the thunderstorm cell(s) is studied. In general, there is a diversity of variations of the thunderstorm cell movement and location within each type of prole; however, certain trend of variations can be observed. The following are discussions of some TS events. On 23 March 2001, storm clouds were observed over the southern part of Singapore. In Fig. 4(a), the radar picture shows the more intense rain cloud is more than 10 km to the south east of Tuas Station. The velocity prole is shown in Fig. 4(b) and it belongs to Type I. The envelope of the maximum gust (dotted line) is also shown in the gure. As shown in Fig. 5, the radar picture showed that on 25 December 2000 the storm clouds were located almost directly over Tuas. The radar picture was recorded at 15:46. The velocity prole for the higher wind speed at 16:02 showed a Type II velocity prole.

ARTICLE IN PRESS
280 E.C.C. Choi / J. Wind Eng. Ind. Aerodyn. 92 (2004) 275290

Fig. 3. Wind proles during TS events.

For event 29 January 2001, the general weather system was observed to move slowly from east to west. Fig. 6(a) shows the radar picture at 11.21 am with the more intense rain cloud almost directly over the Tuas station. Higher wind speeds occurred at 11.29 and the velocity prole is shown in Fig. 6(b). It can be seen to be a Type III prole. On 21 May 2002, storm clouds were approaching Singapore from the west as shown in Fig. 7. Velocity prole for the higher wind speed which occurred at 4.43 is presented in the gure. The prole is classied to be Type IV. Sometimes the situation is more complicated. Fig. 8 shows the radar picture for 8 April 2001 at 16.31 and 17.01. From the wind records, the general wind direction before the storm was from the south with a wind speed of about 4 m=s recorded at Tuas. Wind speed started to decrease at 16.00 and decreased to almost zero at 16.30

ARTICLE IN PRESS
E.C.C. Choi / J. Wind Eng. Ind. Aerodyn. 92 (2004) 275290 281

Fig. 4. (a) TS 23Mar01 Radar picture and (b) 5-min mean and gust proles.

(at which time the wind direction started to change to the north). After this the wind speed started to increase and reached values of more than 8 m=s at 16.50 before it started to decrease again. Fig. 9(a) shows the variation of the 10-min mean wind speed and mean direction with time for the event. For this event, the storm cell was more than 15 km away from the Tuas station. The velocity prole, also shown in Fig. 9(b), belongs to Type IV. It seems for this event, a storm cell occurring at the north of Tuas during a boundary layer wind blowing from the south. The outow from the storm, reaching Tuas from the north, counter acted the opposing boundary layer wind and subsequently with the northerly out ow velocity bigger than the southerly wind, the wind direction changed to north. This event has a Type IV prole with the higher

ARTICLE IN PRESS
282 E.C.C. Choi / J. Wind Eng. Ind. Aerodyn. 92 (2004) 275290

Fig. 5. (a) TS 25Dec00 Radar picture and (b) 5-min mean and gust proles.

recorded wind speed at the top level. This might be due to that the highest elevation of measurement is not high enough or some other unknown factor.

3. Experimental studies From the eld measurement reported in the previous section, it seems the wind prole observed for a thunderstorm with cell centre far away from the monitoring

ARTICLE IN PRESS
E.C.C. Choi / J. Wind Eng. Ind. Aerodyn. 92 (2004) 275290 283

Fig. 6. (a) TS 29Jan01 Radar picture and (b) 5-min mean and gust proles.

location approached that of the large-scale wind prole. However, the wind prole observed for a thunderstorm with cell centre close to the monitoring location exhibited the feature where higher wind speeds were observed at the lower elevations. In order to investigate how the distance of the wind prole from the cell centre affects its shape, velocity measurements with experimental thunderstorm simulation were carried out. Various techniques were used in the experimental simulation of thunderstorm downdraft, e.g. Holmes [11] using an impinging jet and Mason et al. [12] using a pulsed jet. Other works using similar method was carried out by Wood et al. [8],

ARTICLE IN PRESS
284 E.C.C. Choi / J. Wind Eng. Ind. Aerodyn. 92 (2004) 275290

Fig. 7. (a) TS 21May02 Radar picture and (b) 5-min mean and gust proles.

Letchford and Illidge [13] and the study of pressure distribution due to downburst on obstacles by Chay and Letchford [14,15]. In the present study, a continuous stream of air generated by a blower unit was directed to a at plane through a long circular pipe of 100 mm diameter. The velocity at the jet exit was more or less uniform but curved slightly close to the pipe wall. The turbulence intensity was about 5%. The experiment was carried out with the jet exit placed at distances ve times to eight times the jet diameter from the testing surface. Velocity at various heights from the plane of the surface and at various distance from the jet axis were measured using two-dimensional hot wire anemometer. The jet velocity at the exit of the pipe was also monitored. The study was carried out with three different types of surface roughness, ply-wood board (smooth), carpet (rough) and coarse grain particles (very rough) tted to the plane surface (Fig. 10). Horizontal velocity proles for different testing conditions are shown in Figs. 11(a),(b) and (c), respectively, for the smooth, carpet and coarse grain surfaces. The horizontal velocities normalized by the jet exit speed are plotted for locations with distances 1D (jet diameter) to 5.5D from the jet axis. It can be seen that the peak wind speed decreases with increasing distance and the height at which the peak

ARTICLE IN PRESS
E.C.C. Choi / J. Wind Eng. Ind. Aerodyn. 92 (2004) 275290 285

Fig. 8. (a,b) TS 08Apr01 radar pictures.

ARTICLE IN PRESS
286 E.C.C. Choi / J. Wind Eng. Ind. Aerodyn. 92 (2004) 275290

Fig. 9. (a) TS 08Apr01 wind speed, wind direction and (b) velocity prole.

occurs increase with distance from the jet axis. From the shape of the prole, it can be seen that close to the jet axis, the ground roughness effect is small. At larger distances from the jet axis, the effect of ground roughness is stronger and velocity gradually develops into the boundary layer prole. It can also be seen that the rougher the surface the higher is the elevation of the peak wind speed. Fig. 12 shows

ARTICLE IN PRESS
E.C.C. Choi / J. Wind Eng. Ind. Aerodyn. 92 (2004) 275290 287

Fig. 10. Experimental set-up for thunderstorm wind simulation.

a plot of the variation of the height of peak wind speed with distance to the jet axis. It clearly indicates that height of the peak wind speed increases with the distance from the jet axis and also increases with ground roughness. Further study was carried out on the proles. A power-law prole was tted to the portion of the prole below the peak wind speed. The power exponents obtained for proles at various distances from the jet axis are listed in Table 1. Although the values of the exponent are scattered, an increasing trend of value with distance can be observed. The experimental ndings conrm the on-site observations that as the measuring point gets further away from the thunderstorm cell, the observed wind prole is more inuenced by the terrain effect and the peak wind speed (of the prole) is pushed to higher elevations.

4. Discussion Field measurement of wind speed at different heights during thunderstorms is reported in this paper. Wind proles obtained for TS events are compared with wind proles for large-scale wind over the site. For discussion, the TS proles are classied into four types based on the comparison. Investigations are carried out in attempt to nd out the reasons for the different types of prole. Study on the wind speed, direction in conjunction with the radar picture for each TS event reviews certain trend of variations. In general, Type I prole is observed when the storm cell centre is far away (10 or more kilometres) from the measuring station. The out ow has to move over a considerable fetch before reaching the measuring station and thus is strongly affected by terrain effect. When the storm cell is very close to the station, a Type II or Type III prole is observed. The prole is less affected by terrain roughness and more dominated by the out spread of the downdraft. The occurrences of Type IV proles are more random, sometimes they are further away from the measuring station, but other times they are close. It seems that Type IV proles are from the more intense storms. This may result in a thicker out ow layer with the highest wind speed occurring at or above the top observation level.

ARTICLE IN PRESS
288 E.C.C. Choi / J. Wind Eng. Ind. Aerodyn. 92 (2004) 275290

Fig. 11. Velocity prole for (a) smooth surface; (b) carpet surface; (c) coarse grain surface.

Another factor that needs to be considered is the wind condition just before and at the arrival of the gust front. The measured velocity reects the vector sum of the wind speed due to the thunderstorm downburst and that of the overall weather

ARTICLE IN PRESS
E.C.C. Choi / J. Wind Eng. Ind. Aerodyn. 92 (2004) 275290 289

Fig. 12. Variation of the height of peak wind speed with distance to jet axis. Table 1 Power-law exponent for proles at different distances from the jet axis Distance to jet axis (jet diameter) Smooth Carpet Coarse grain 1D NA NA NA 2D NA 0.110 0.134 3D NA 0.085 0.169 4D 0.046 0.111 0.162 5D 0.059 0.085 0.209 5.5D 0.074 0.199 0.198

system. Preliminary observations also indicate that with the storm cell moving with the weather system, the measured wind prole suggests that wind speeds are the vector sum of the thunderstorm outow wind and the boundary layer prole wind of the weather system, and not the translational speed of the storm cell. To gain further insight to the eld measurements, physical simulation of thunderstorm wind using an impinging jet is also carried out. The experimental result conrms that close to the jet axis (i.e., cell centre of thunderstorm storm) the height of the peak wind speed of the prole is lower and the shape is less affected by terrain roughness. For wind proles monitored at a larger distance from the jet axis, the elevation of the peak wind speed becomes higher and the prole is more affected by terrain roughness effect. The prole gradually develops into the boundary layer prole appropriate for that terrain roughness. These are in line with the eld observations of thunderstorm wind proles.

Acknowledgements The author wishes to express his gratitude to Meteorological Services Singapore for providing radar image data for this study.

References
[1] C.W. Letchford, C. Mans, M.T. Chay, Thunderstorms, their importance in wind engineeringa case for the next generation wind tunnel, J. Wind Eng. Ind. Aerodyn. 90 (2002) 14151433.

ARTICLE IN PRESS
290 E.C.C. Choi / J. Wind Eng. Ind. Aerodyn. 92 (2004) 275290 [2] E.C.C. Choi, A. Tanurdjaja, Extreme wind studies in Singapore. An area with mixed weather system, J. Wind Eng. Ind. Aerodyn. Netherlands 90 (2002) 16111630. [3] T.T. Fujita, The Downburst, Report of Projects NIMROD and JAW, University of Chicago, 1985. [4] T.T. Fujita, Tornadoes and downbursts in the context of generalized planetary scales, J. Atmos. Sci. 38 (8) (1981) 15111534. [5] E.C.C. Choi, Some observation on thunderstorms, Proceeding of Inaugural meeting of the Wind Engineering Research Centre, Tokyo Institute of Polytechnics, Japan, October 2001, pp. 4562. [6] K.D. Gast, J.L. Schroeder, Supercell rear-ank downdraft as sampled in the 2002 thunderstorm outow experiment, Preprints 11th International Conference on Wind Engineering, Vol. 2, June 25, 2003, Lubbock TX, USA, WISE, pp. 22332240. [7] J.D. Holmes, S.E. Oliver, An empirical model of a downburst, J. Ind. Aerodyn. 22 (2000) 11671172. [8] G.S. Wood, K.C.S. Kwok, N.A. Motteram, D.F. Flecher, Physical and numerical modeling of thunderstorm downbursts, J. Ind. Aerodyn. 89 (2001) 535552. [9] R.M. Oseguera, R.L. Bowles, A simple analytic 3-dimensional downburst model based on boundary layer stagnation ow, NASA Technical Memorandum 100632, July 1988. [10] H.D. Hangan, Roberts, Z. Xu, J.-D. Kim, Downburst simulations. Experimental and numerical challenges, Preprints 11th International Conference on Wind Engineering, Vol. 2, June 25, 2003, Lubbock TX, USA, WISE, pp. 22412248. [11] J.D. Holmes, Physical modeling of thunderstorm downdrafts by wind-tunnel jet, Proceedings of the 2nd AWES Workshop Melbourne, February 1992, pp. 2932. [12] M.S. Mason, C.W. Letchford, D.L. James, Pulsed jet simulation of a thunderstorm downburst, Preprints 11th International Conference on Wind Engineering, Vol. 2, June 25, 2003, Lubbock TX, USA, WISE, pp. 22492256. [13] C.W. Letchford, G. Illidge, Turbulence and topographic effects in simulated thunderstorm downdrafts by wind tunnel jet, Proceedings 10th International Conference on Wind Engineering, 1999, pp. 19071912. [14] M.T. Chay, C.W. Letchford, Pressure distributions on a cube in a simulated thunderstorm downburst, Part A: stationary downburst observations, J. Wind Eng. Ind. Aerodyn. 90 (2002) 711732. [15] C.W. Letchford, M.T. Chay, Pressure distributions on a cube in a simulated thunderstorm downburst, Part B: moving downburst observations, J. Wind Eng. Ind. Aerodyn. 90 (2002) 733753.