EACWE 5 Florence, Italy 19 th 23 rd July 29 Flying Sphere image Museo Ideale L. Da Vinci Chasing gust fronts - wind measurements at the airport Munich, Germany E. Agu, M. Kasperski Ruhr-University Bochum Department of Civil and Environmental Engineering Sciences Ecevit.Agu@rub.de Michael.Kasperski@rub.de Keywords: frontal depression, gust front, storm trap, wind field measurement ABSTRACT The wind climate of Germany is governed by three storm types: strong frontal depressions, gust fronts and thunderstorms. While there are sophisticated models to describe the turbulent wind fields in strong frontal depressions, knowledge on gust fronts and thunderstorms is limited. Since 8.8.8 a new facility at the Munich airport is in operation which observes the turbulent wind fields permanently. The paper describes the experimental setup and gives some first results based on two gust fronts which have already been caught in the storm trap. Since the duration of strong wind conditions in gust fronts and thunderstorms is considerably smaller than the usual reference period of one hour, one minute is proposed as new basic reference period. To identify differences in the wind fields of the three storm types, the natural scatter of the describing parameters has to be analysed for the storms induced by strong frontal depressions. This can be done based on simulations using as basic input the knowledge summarized in the respective ESDU-documents. For a consistent load model, which includes gust fronts and thunderstorms, the duration of individual events and the number of events per independent storm are required. An appropriate statistical model, however, requires more measured data which will be collected in the next years. Contact person: M. Kasperski, Ruhr-University Bochum Department of Civil and Environmental Engineering Sciences, Building IA 4/32, Tel: +49 234 32 24148, FAX: +49 234 32 14317 E-mail Michael.Kasperski@rub.de
INTRODUCTION The knowledge on wind fields in the different storm types governing the strong wind climate is considerably biased. While wind engineers can obtain an extremely sophisticated model for the wind fields in strong frontal depressions from the ESDU-data [1], there is very limited information on gust fronts. These strong wind conditions are connected to strong frontal depressions and are induced by convective processes, e.g. heavy rain [2]. The corresponding gust wind speeds are much larger than is predicted by the ESDU-model, and they are high enough to test structures in regard to their wind resistance. There is no doubt that for a consistent design concept these storm types have to be considered. However, the substantial lack of knowledge allows the consideration of these important storms based on only some simplifying assumptions. The most important question is if the wind fields and thus the basic building aerodynamics in the different storm types are same or at least similar. Further question deals with the duration of strong wind conditions in these storm systems and the number of events per individual storm. Unfortunately, the data basis provided by the German Weather Service (Deutscher Wetterdienst - DWD) does not allow answering these questions. The usual set of data per day consists of 24 values for the hourly mean wind speed and the corresponding wind direction. Additionally, the daily peak gust wind speed is monitored together with its time of occurrence in hours and minutes. These data may be used for sampling extreme wind speeds for different storm types, however, the resulting ensembles for thunderstorms and gust fronts will suffer from the basic shortcoming that per day only one event can be sampled. Strictly speaking, thunderstorms have two gust fronts, an upwind and a downwind one, and only the larger event occurs in the data set of the DWD. Furthermore, there is no reason to assume that there is only one gust front in a strong frontal depression which may have a duration of several hours. To contribute to these burning questions and to study the wind fields of gust fronts and thunderstorms in Germany in more detail, a storm trap has been erected at Munich airport with the financial support of the German Ministry of Research and Education. The facility started its operation on August, the 8 th 28. The paper gives a short overview on the basic experimental set-up. Since the duration of strong wind conditions in gust fronts and thunderstorms is considerably smaller than the usual reference period of one hour, one minute is proposed as new basic reference period. To identify differences in the wind fields of the three storm types, the natural scatter of the describing parameters has to be analysed for the storms induced by strong frontal depressions. This can be done based on simulations using as basic input the knowledge summarized in the respective ESDU-documents. The observations so far contain at least two synoptic events with considerable additional gust fronts. Some first results of the wind field analysis are presented. LAYOUT OF THE STORM TRAP Airports provide an ideal opportunity for wind measurements since there are large areas without distortions from buildings or trees. For the storm trap, a position beside the runway has been chosen. The basic idea is to provide 3-D information for the turbulent velocity fields. Therefore, an arrangement of four masts has been designed which has three 1m high masts at the corners of an equilateral triangle with side length of 1 m and a further 2m high mast in the balance point of the triangle. While the three corner masts have a single wind sensor at their top, the centre mast has additional sensors at 5m and 1m. The wind sensors are Young 3D Sonic Anemometer (model 81VRE). Additionally, the atmospheric pressure is measured at the centre mast at a height of 9.8m with a Young sensor 6122V. The orientation of the triangle corresponds to the dominant wind direction of strong frontal depressions (figure 1).
runway position of the masts basic arrangement of the masts Figure 1: Basic layout of the storm trap at Munich airport The data acquisition system is based on three removable hard drives. In case of power failure, an independent power supply guarantees that there is no loss of data. The data acquisition works permanently, the data are stored for 5 minute intervals and compressed to one-day packages. Per wind sensor and day, the amount of data is about 25.7 Mb. The output-files give values for the three wind speed components u, v and w, the speed of sound and the sonic temperature. Measuring errors are indicated in the corresponding protocol in a separate column. For sensor 1, a further column specifies the atmospheric pressure. RE-ANALYSIS OF TURBULENT WIND FIELDS IN FRONTAL DEPRESSIONS The basic parameters describing the turbulent wind field are the profile of the mean wind speed, the turbulence intensity, the spectral density of the velocity fluctuations, the integral length scales and the correlation of the velocity fluctuations. To identify differences in these parameters for the different storm types, first, the natural scatter of these parameters has to be studied for frontal depressions based on the new reference period of one minute. This is done in the following by simulations assuming, applying as basic input an hourly mean wind speed of 2 m/s at 1 m height above ground, a corresponding turbulence intensity of.19, an integral length scale of 1 m and a mean wind speed profile exponent of.16. Time histories corresponding to 1 individual hours in full scale are generated with a sampling frequency of 1 Hz for simultaneous wind speeds at 5m, 1m and 2m above ground. As a matter of fact, none of the describing parameters is statistically stable if estimated on only a single set of data corresponding to the reference period of one hour. Consequently, the natural scatter of the basic parameters with reference to one minute is fairly large. In figure 2, the results are shown in form of histograms for the 1-minute mean wind speed at 1m height, the corresponding turbulence intensity and the velocity ratios between 2m and 5m, respectively, to the 1m value. Obviously, the mean wind speed profile in frontal depressions is rather stable even for short averaging periods of only one minute, i.e. compared to the 1-minute mean value at 1m height there always will be a clear wind speed decrease for 5m and a clear wind speed increase for 2m. On average, the velocity ratios for 1-minute means are virtually the same as for the hourly means. The turbulence intensity, however, decreases to a new mean value of.17. The natural scatter of the
turbulence intensity ranges from 1% to 2%. The natural scatter in the wind speed ratios for different heights can be used to identify special features for the two storm types gust front and thunderstorm. If for instance during the passage of a gust front wind speeds at the three observed heights turn out to be almost the same, a statistical significant difference is obtained. The integral length scale is obtained from applying the model of frozen turbulence, i.e. from multiplying the integral time scale with the mean wind speed. The integral time scale is estimated based on an integration of the autocorrelation function until the first zero-crossing. Both, the integral time scale and the integral length scale are extremely instable even for one hour. For shorter periods, the natural scatter will increase further. Results for the new sub-period of one minute are shown in figure 3. The mean value of the integral time scale drops from the initial value of 5s to 3.33s; the mean value of the integral length scale drops from initially 1m to 66m. For the integral length scale, a minimum value of about 1m is obtained in this example. 1 m 5 m 14 16 18 2 22 24 26 1-min mean wind speed [m/s].75.8.85.9.95 1. velocity factor 1 m 2 m.3.35.4 turbulence intensity 1. 1.5 1.1 1.15 1.2 1.25 velocity factor Figure 2: Natural scatter for the 1-min mean values at 1 m height and the corresponding turbelence intensity, scatter of the relative wind speeds at 5 and 2 m height applying the ESDU-model 5 1 15 integral time scale [s] 1 2 3 integral length scale [m] Figure 3: Natural scatter of the integral time and length scale for averaging period of one minute and typical time-history of the integral length scale
While the correlation of wind speeds with reference to one hour is a fairly stable statistical parameter, the correlations will show larger scatters for a reduced reference period of one minute. The corresponding results are shown in figure 4 for the two correlations between wind speeds at 5 m and 2 m height, respectively, to the wind speeds at 1 m height. The mean value of the correlations are slightly reduced, with a value of.74 for the correlation between 5 and 1 m and.66 for the correlation between 1 and 2 m...2.4.6.8 1. correlation..2.4.6.8 1. correlation 5 m 1 m reference value for one hour:.8 1 m 2 m reference value for one hour:.74 Figure 4: Natural scatter of the correlation of wind speeds in one-minute sub-periods ANALYSIS OF GUST FRONTS The data sets so far contain two synoptic events with large overshooting gusts. The first event occurs on February 1th, 29 leading to a gust wind speed of 25.8 m/s with a corresponding hourly mean wind speed of 13.2 m/s. The gust factor is 1.95. The second event on March 23 rd, 29 delivers a mean value of 11.2 m/s and a gust wind speed of 22.4 m/s, i.e. a gust factor of 2. (figure 4). A 1-minute close-up identifies for the first event a single gust front starting at about 2 s; for the second event, two gusts fronts which are separated by about 4 minutes (figure 5) are found. 3 3 25 25 wind speed [m/s] 2 15 1 wind speed [m/s] 2 15 1 5 5 6 12 18 24 3 36 42 48 54 6 66 72 time [s] 6 12 18 24 3 36 42 48 54 6 66 72 time [s] February 1 th, 17: 19: March 23 rd, 22:-24: Figure 4: Two synoptic events with overshooting gusts in 29
For the further analysis, the observed time histories have to be decomposed into a time-varying mean wind speed and a remaining fluctuating wind speed component. In a first approach, the time-varying mean wind speed is modelled by a moving average. The appropriate depth of the moving average is identified by trial and error. For the two observed events, a moving average with 2 s seems to be appropriate. All three gust fronts have fairly short durations, i.e. they have passed in periods smaller than two minutes. The next step in the analysis deals with the velocity ratios at different heights. In figure 6, the velocity ratios for one-minute sub-periods are shown for the range of two observation hours. For both events, the mean wind speed profile averaged over the complete two hours is virtually the same, with an average value of 1.1 for the wind speeds at 2m at 1m height and an average value of.94 for the wind speeds at 5m and 1m height. The respective time histories for 1-minute sub-periods give no indication that the mean wind speed profile is influenced by the gust fronts. 3 original time history time depending mean wind speed fluctuating wind speed 3 original time history time depending mean wind speed fluctuating wind speed 25 25 2 2 wind speed [m/s] 15 1 5 wind speed [m/s] 15 1 5-5 -5-1 15 156 162 168 174 18 186 192 198 24 21 time [s] -1 456 462 468 474 48 486 492 498 54 51 516 time [s] February 1 th, 17: 19: March 23 rd, 22:-24: Figure 5: Decomposition based on 2s moving average for the time dependent mean value 1.25 v 1min (z=2m) / v 1min (z=1m) v 1min (z=5m) / v 1min (z=1m) 1.25 v 1min (z=2m) / v 1min (z=1m) v 1min (z=5m) / v 1min (z=1m) 1.2 1.2 1.15 1.15 1.1 1.1 velocity ratio 1.5 1. velocity ratio 1.5 1..95.95.9.9.85 2 4 6 8 1 12 time [min].85 2 4 6 8 1 12 time [min] February 1 th, 17: 19: March 23 rd, 22:-24: Figure 6: Development of the wind speed profile over the observation period of two hours
INFLUENCE OF GUST FRONTS ON THE EXCEEDANCE PROBABILITY OF THE DESIGN WIND LOAD The question arises how the observed additional gusts may influence the exceedance probability of a specific wind load level. Assuming in a first step that the basic building aerodynamics in gust fronts and normal storms induced by strong frontal depressions are the same allows using the same aerodynamic coefficients for both flows. Then, the following parameters have to be considered: the equivalent mean wind speed for the gust v m, eq, the duration of the gust T gust and the ratio of the equivalent mean wind speed to the hourly mean wind speed ε v. The equivalent wind speed is obtained by dividing the observed gust wind speed by the average or expected gust factor of a normal storm. The wind load is obtained as: 1 ρ (1) 2 2 w = v c with ρ - air density, v - wind speed and c - aerodynamic coefficient In a first step, the air density is assumed to be a constant or deterministic value. The exceedance probability of a specific wind load w ref in a single hour of a normal storm is given as p ref : p(w > w ref v m, 1 h) = pref (2) Since v m and ρ are constant, p ref specifies the exceedance probability of the aerodynamic coefficient as well, i.e.: p(c > c ) = p (3) ref ref Changing the wind speed level from v m to v m, eq can be considered in a change of the non-exceedance probability of the equivalent aerodynamic coefficient as follows: p(c c / ε ) = p (4) 2 ref v eq The duration of the gust front finally is considered in the following relation: T gust /Tref p(w > w ref v m, eq, T gust ) = 1 - peq (5) with T ref = 1 hour In figure 7, the exceedance probability is shown for a variation of T gust from 3 s to 6 s and for a variation of the wind speed ratio ε v from 1 to 2. As reference value, the 78%-fractile value for the aerodynamic coefficient is used, i.e. the exceedance probability is.22. This reference value corresponds to design wind speed level and design wind load level. For the extreme aerodynamic coefficients, an extreme value distribution type I is assumed with a variation coefficient of 15%. Basically, the exceedance probability increases if the duration of the gust front increases and if the equivalent mean wind speed increases. In the above example, the target value of.22 is exceeded if for the duration of 3 s the wind speed ratio exceeds 1.43. For the duration of 2 minutes, the target value is exceeded if the wind speed ratio becomes larger than 1.25. With an average gust factor of 1.6 in strong frontal depressions and a gust factor limit of 1.8 for the identification of overshooting gusts, a relevant contribution to the exceedance probability is obtained for any event exceeding the duration of 44 seconds.
2..1 1.8.2.3 1.6.4.5 ε v 1.4.6.7 1.2.8.9 1. 1. 6 12 18 24 3 36 42 48 54 6 T gust [s] Figure 7: Influence of the gust front on the exceedance probability of the wind load target value.22 for a single event SUMMARY AND CONCLUSIONS Today s concepts for specifying the design wind loads do not appropriately consider the strong wind fields induced by thunderstorms and gust fronts. This may lead to shortcomings in the design. For a consistent model, further knowledge on these two storm types is required, especially on the duration of strong wind conditions during these storms and on their occurrence rate per individual storm. Additionally, the question has to be answered if there are relevant differences in structure of the turbulent wind fields in different storm types. A new facility at Munich airport, consisting of four masts with a total number of six ultra-sonic anemometers, provides high quality data to study the wind fields in gust fronts and thunderstorms. As basic reference period for the description of the wind fields a period of one minute is proposed. First results suggest that the initial phase of gust fronts may have different durations and that the number of gust fronts per individual storm is a random parameter. ACKNOWLEDGEMENT Part of this work has been sponsored by the Federal Ministry of Research and Education under the scope of the joint-research project RegioExAKT. This support is gratefully acknowledged. The author also expresses his special thanks to the members of the management of the Flughafen München GmbH who after a long struggle finally allowed the erection of four masts in the immediate vicinity of the runway. REFERENCES [1] ESDU - Engineering Science Data Unit Characteristic of atmospheric turbulence near the ground Single point data for strong winds, ITEM 852, 1985 Variations in space and time for strong winds, ITEM 861, 1986 [2] M. Kasperski A new wind zone map for Germany Journal of Wind Engineering and Industrial Aerodynamics (9) 22 pp 1271-1287