Validation of a microscale wind model using ultrasonic-anemometer data M. Wichmarin-Fiebig Northrhine- Westphalia State Environment Agency, Abstract Wind data recorded by an ultrasonic-anemometer over a time of eight months were used to validate a nonhydrostatic microscale model Observations took place in the vicinity of a pigsty at a height of 9.5m above ground. The height of the building was about 8m. Model runs were performed with various wind speeds and directions as well as different thermal stratifications. Comparison between observations and model results in general showed good agreement. However, downdrafts in the lee circulation were overestimated by the model Sensitivity to thermal stratification showed to be neglectable concerning observations as well as model results. 1 Introduction Dispersion modelling in the vicinity of buildings is of growing importance in recent time. However this requires determination of the wind field taking into account the influence of the building. Models used for this purpose are mostly validated by comparison with wind tunnel data. However, open air observations in the vicinity of real buildings are preferable for validation because they represent less idealised conditions. Moreover adiabatic cases would be included in the data
66 Computer Simulation Having at hand half a year of data recorded by an ultrasonic-anemometer situated near a pigsty a validation of a microscale model was performed. Various wind speeds and directions as well as different thermal stratifications were considered. In addition data for a smaller height and a second position were available at least for shorter time periods. In the following the observation site will be presented and data quality will be discussed. Thereafter a short description of the model to be validated will be given and results will be discussed. Finally model evaluation will be presented followed by some conclusions concerning dispersion modelling. 2 Data acquisition and validation Fig.l shows a horizontal sketch of the observation site. Positions Ml, M2 and N of the wind masts together with north orientation are indicated The longest data series were taken at M2 in 9.5m above ground from 19.08.92 to 12.03.93 with only some short interrupts of about four weeks altogether. In between data were taken for three weeks at a height of 3m above ground (Ml). For a shorter period of about four days data were also taken at point N in 8m above ground. The height of the two pigsties was about 8m. The narrow building in between was 5m high. Up to a distance of 500m no other buildings were found. Figure 1: Observation sites with north orientation and buildings. Data acquisition took place by ultrasonic-anemometers (USAT) [1] with a frequency of 20Hz. Based on these raw data 30min-means of the three-dimensional wind vector, temperature and turbulent fluxes of momentum and heat were calculated. For the purpose of data validation USAT data at site M2 were compared to those taken simultaneously at site N. Although data were taken on different
Computer Simulation 67 sides of the pigsties horizontal wind speed and direction at the two sites (Fig.2ab) show good agreement over the considered period of about three days. Time series of friction velocity u* (Fig.2c) are similar to each other, too, because the turbulent momentum flux is proportional to horizontal wind speed. Observed differences are due to upwind and downwind position in respect to the pigsties. As should be expected larger differences were found in vertical wind speed (Fig.2d) due to up- and downwind effects. Westward winds cause stronger upwinds at site N compared to M2 downwind at N. _c 12 Q> O X) a ^ c ++ o N J= Transactions on Ecology and the Environment vol 3, 1994 WIT Press, www.witpress.com, ISSN 1743-3541 4 M N 2 n c^ :> o o o o o uo r- co CM r*» LO %- co CM r^ while more southerly directions lead to 300 270 = I % I 5 240 ^c o 5 210 = W 180 = o o o o o o LO «- CM CO id r- CM co r^ k M Q8id CO CM CO CN ID CO «- CM Figure 2: Data series of (a) horizontal wind speed, (b), (c) u* and (d) vertical wind speed at observation sites Ml and N from 15:40h, 02.11 92, to 10:10h, 04.11.92.
68 Computer Simulation Ultrasonic data were also compared to wind observations taken at a nearby conventionally instrumented wind mast on an airbase (Fig.3). These data were classified in steps of 0.5m/s. The data sets show good agreement concerning but systematic differences in wind speed. Higher values of the ultrasonic-anemometer especially at low wind speed reflect the higher sensitivity of this instrument. At wind speeds greater than about 3.5m/s, however, airbase data exceed USAT-values perhaps because of an overspeeding of the conventional anemometer at the airbase. 360 0 90 180 270 360 USAT Direction.1 0 1 2 3 4 5 6 7 USAT wind speed in m/s Figure 3: Comparison of USAT (a) and (b) wind speed with data of a nearby airbase. Altogether results of data validation assure us to have at hand a suitable data set for the intended model validation 3 Model description and results A nonhydrostatic microscale model [2] was used to calculate the flow field in the vicinity of the pigsties. The initial wind field is defined as horizontal homogeneous assuming, however, zero wind speed inside of buildings. The vertical wind profile is determined by the wind at the model top, which corresponds to the top of the surface layer, and a turbulence parameterisation following a mixing length approach. Mixing length is assumed to be proportional to the minimum of the distance to the nearest horizontal or vertical surface.
Computer Simulation 69 Concentrations resulting from the models wind- and turbulence fields downwind of a cube were compared with wind tunnel data [3] whereby in general good agreement was found. For comparison with data described above wind speed at model top was set to 2.5m/s, 5.5m/s and 8.5m/s, respectively Wind direction was varied in steps of 30. Sensitivity studies concerning thermal stratification showed no significant influence on the results So it was neutral. norm. hor. wind speed Transactions on Ecology and the Environment vol 3, 1994 WIT Press, www.witpress.com, ISSN 1743-3541 I.U - 2.5m/s 0.8 **'^ 0.8 0.6 0.4,^,A \.vv ^ *^ 0.2 0.0 0 I 0.6 5 0.4 JC E 0.2 (b) E 0.2 / - / 90* 180* ' ' 0.0 * ' < 270* 360* 0* 90* 180* 270* 360 wind drection ^ 5.5m/s obs.d.2 set to 90* 180* 270* 360* 90* 180* 270* 360* 1 (O I 0.8 * 8.5m/s obs.d.3 90* 180* 270* 360* 90* 180* 270* 360* Figure 4: Normalised calculated and observed horizontal wind speeds at site M2. Figure 5: As Fig.4 but at site Ml.
70 Computer Simulation To study sensitivity of horizontal wind speed to direction calculated values at points Ml, M2 and N were normalised with the respective wind speed at the model top. Results are shown in Fig.4 to 6. Normalised wind speeds show to be insensitive to top speed in general. However, comparison between parts a, b and c of the figures shows some systematic differences especially between 2.5m/s runs and the higher velocities. model 1 model 360 90 IOU Figure 6: As Fig.4 but at site N. Figure 7: Normalised calculated and observed vertical wind speeds at site (a) M2, (b) Ml and (c) N.
Computer Simulation 71 It can be seen that horizontal wind speed is reduced if the flow is directed from or to the building in respect to the observation site, i.e. 16 and 196 at Ml and M2 as well as sectors 136 to 196 and 286 to 16 at N. Reduction is strongest in the lee of the building. However, at site N maximum reduction is observed if the flow is directed perpendicular to the walls of the building ( i.e. 16 and 286 ). Diagonal flow results in a significant smaller reduction because then horizontal deviation of the flow dominates in relation to convergence and deceleration. At sites Ml and M2 top speeds of 5.5m/s and 8.5m/s show a reduction of wind speed also at 96, which corresponds to a direction parallel to the building This is supposed to be due to dissipation instead of convergence and therefore is not expected to increase vertical velocities Calculated vertical wind speeds at sites Ml, M2 and N were normalised with the horizontal velocities calculated at the respective sites. Results are shown in Fig.7. Dependence of normalised vertical velocities on top wind speed showed to be neglectable. Therefore only 5.5m/s-results are shown for each site. As expected values are positive on the upwind side of the building (106 to 286 at sites Ml and M2, 210 to 90 at N) and negative in the lee Nondimensionalised values of Ml and M2 differ hardly in the upwind sector while differences grow in the downwind region. Here Ml values are much larger than those at M2. Moreover the maximum downward wind speed at 16 is about two times larger than the maximum upward wind speed at 196. Observed reduction of horizontal wind speed at 106 is not reflected in nondimensionalised vertical wind speeds at M2 and Ml This confirms the assumption made above that the horizontal minimum is to due to dissipation instead of convergence. Due to its position near a corner of the pigsty site N shows upwind over a wind sector of 240. Direction sensitivity in this sector itself is very small compared to Ml and M2. Because site N is situated nearer to the building upwind values are larger than at sites Ml and M2 Summarising the structure of the model results seems to be reasonable. Analysis of observations will be used for a comprehensive validation.
72 Computer Simulation Model validation As no information about wind speed at the top of the surface layer during observations was available the question was how to normalise measured horizontal wind speeds We decided to build three classes corresponding to the three top wind speeds used in the model runs Boundary velocities vy between classes i and i+1 were determined by averaging the corresponding model results v at site S vy (class i, class i+1, 6, S) = (v(class i, 9, S) + v(class i+1, 9, S))/2, where dependence on 9 was also considered. Following this classification scheme measured wind speed data in class 1 were normalised by 2.5m/s, data in class 2 by 5.5m/s and data in class 3 by 8.5m/s. Concerning site M2 - where eight months of data were available - it can be seen that direction dependencies in class 1 and 2 correspond quite well to one another and to the model results (Fig.4a-b). Class 3, however, obviously contains some larger values than those represented by 8.5m/s top wind speed (Fig.4c). Nevertheless in all classes minima at 196 and 16 as well as maxima in between are clearly reflected in the observations at M2. The local minimum at 196 discussed above seems to be confirmed by the data for classes 2 and 3 but not for class 1 which agrees with the model results. This leads to the conclusion that scaling of the flow should be done very careful when wind speed is low. Due to the short data acquisition period observations at points Ml and N are very sparse. Yet minima at 106 and 196 are also indicated at Ml (Fig.5). At site N the minimum below 186 and at 286 can be identified, too (Fig.6). Observed vertical velocities are normalised by the corresponding simultaneously observed horizontal wind speeds. Upwind cases at sites Ml and M2 (Fig.7a-b) agree quite good with model results, although calculated values are systematically smaller. Concerning downwind cases, however, observations at site M2 do not reflect the large values of the model. Especially between 346 and 46 very poor agreement is found. Unfortunately data at Ml are so sparse in this wind sector that no comparison with calculated values is possible. Also observations at site N are very sparse they indicate the same pattern as
Computer Simulation 73 described above: Upwind values are too low while downwind values tend to be too large (Fig.Tc). Conclusions Results of the non hydrostatic microscale model agree in general with data of an ultrasonic-anemometer taken at three sites in the vicinity of a pigsty However downward wind velocities are significantly overestimated by the model. From the observations it could not be decided whether the calculated lee circulation is too intensive or the region of the circulation extends too far downward. Results of dispersion modelling are expected to be sensitive to these differences because concentrations downwind of a building are largely affected by the structure of the lee circulation. Further studies also including the turbulence data might give further insight in the cause of the differences between observations and model results This should enable us to further improve the model. Acknowledgement I wish to thank the Niedersachsisches Landesamt fiir Okologie for performing observations at site N and placing the data at out disposal. I am also grateful to the German Weather Service who delivered data from Norvenich airbase. Last but not least thanks to all colleagues who helped to record and analyse the data References 1. Thielen, H. Zur Struktur der Turbulenz in der atmospharischen Grenzschicht. Institut fur Geophysik und Meteorologie, University of Cologne, 1994. 2. Eichhorn, J Entwicklung und Anwendung eines dreidimensionalen mikroskaligen Stadtklima-Modells. Ph.D. Thesis, University of Mainz, 1989. 3. Wichmann, M. Numerische Simulation der Ausbreitung luftverunreinigender Stoffe im Nahbereich von Gebauden. Aus der Tatigkeit der LIS, Essen, 1992.