Wind Tunnel Modelling in Support of Model Validation

A contribution to EUROTRAC 2-subproject SATURN

M. Schatzmann, B. Leitl, J. Liedtke

Meteorologisches Institut, ZMK, Universität Hamburg, Bundesstr. 55, D-20146 Hamburg, Germany


 


1. Summary

Validating conventional grid models means to intercompare their results with field or laboratory data. However, field data, laboratory data and numerical model results contain some fundamental differences. To simply relate experimental data and model results to each other comes, therefore, often close to the proverbial comparison of apples with oranges. A systematic wind tunnel study was carried out within which these differences were investigated and quantified. This was done at the example of a monitoring station located in a busy four-lane street canyon in Hanover, Germany.
 

2. Aim of the research

Model validation deals with the comparison of results from a numerical model with data sets. It is not always understood that this is a rather complex task. It was the objective of last years work to propagate awareness of the fact that such comparisons have to be made with care. This is subsequently demonstrated at the example of an obstacle resolving conventional grid model with full turbulence parameterisation and data from a street canyon monitoring station.
 

3. Activities during the year

The concept for our work is demonstrated in Fig. 1 at the example of a small area source which continuously discharges a passive tracer into a street canyon. Shown are the traces of concentration versus time (in excess above background) at the same receptor point and under identical steady-state ambient conditions as they might be found in (a) a field experiment, (b) in a wind-tunnel experiment, or (c) in a numerical simulation with full turbulence parameterisation.
 

Field experiments:

High resolution field measurements provide usually highly intermittent signals, i.e. periods of near zero concentration are interspersed with non-zero fluctuating concentrations. It is to be expected that the intermittency of the signal depends largely on the turbulence structure within the canyon and the wind direction fluctuations. In general, the source dimensions relative to the cross-sectional area of the canyon and the distance between the source and the receptor point should also influence the intermittency.

If the concentration versus time trace varies as shown in Fig. 1 (top), long averaging times are required in order to produce a meaningful time-mean-value. It is to be expected that the commonly used 10 min or 30 min measurement cycles are not long enough. Longer averaging times, however, are usually not feasible since the atmospheric boundary conditions change during the diurnal cycle. The conclusion is that the repeatability of results is poor and that large error bars should be attached to time-averaged concentrations determined in field situations as described.
 

Laboratory experiments:

When the same dispersion problem is modelled in a wind tunnel or water channel, the concentration signal presented in Fig. 1 (center) is obtained. If all main similarity parameters were matched properly in the small scale simulation, the time series should resemble that of the field test, but it would be somewhat less intermittent since the low frequency wind direction variations are weaker in a ducted flow. Therefore, time mean concentration maxima determined in laboratory experiments are usually larger than those obtained in the field. The degree of overestimation depends again on the source dimensions, the source/receptor distance and the turbulence structure of the ambient flow.

An important advantage of wind tunnel measurements in comparison to field tests, however, is that the boundary conditions can be chosen to be appropriate to the problem being solved, and that numerous repetitions of the same case can be made in order to determine the inherent variability of the dispersing cloud characteristics.
 

 Numerical model results: Finally, at the bottom of Fig. 1 the concentration versus time trace, as obtained from a common grid model, is displayed. Provided the model considers turbulent fluctuations only in parameterised form, than with constant boundary conditions it delivers a stationary concentration value. In contrast to the point-by-point experimental data, this value represents not only a time-mean but also a space-mean concentration representative of the characteristics of the whole volume of the grid cell.

Fig. 1: Comparison of concentration versus time traces for field measurements (top), wind tunnel measurements (center) and numerical results (bottom) (concentrations inexcess above ambient only).

 
  This example clearly demonstrates that we should expect remarkable differences in the results from the 3 different sources. It seemed to be a worthwile task to try to quantify those differences.

In order to do so, a scale model (1:200) of a street canyon monitoring station operated by the State Environmental Agency of Lower-Saxony in Hanover, Germany (NLO, 1995), was built (Fig. 2). The monitoring station is located in a busy four-lane street canyon. Based on automatic traffic counts and information on the composition of the German vehicle fleet, good estimates of pollutant emission rates were available. The above-roof wind and background concentrations also measured.
 

 

Fig. 2: View on the wind tunnel model of the Hanover site. The rod indicates the position of the monitoring station.


 4. Principal results

The measurements in the field were replicated in a boundary layer wind tunnel under carefully controlled conditions (Liedtke and Schatzmann, 1998). In order to compare the results from the field and the wind tunnel with each other, NOx concentrations observed in the field over a period of one year (1994) were grouped according to the wind direction (10° steps) and brought into the non-dimensional form c* = C · uref · H/(Q/L) where C is the time mean value of the measured concentration (30 min average), uref is a reference velocity, taken at a height of 100 m, H is a characteristic length (the average height of the surrounding buildings) and (Q/L) is the source strength of the line source.
 
 

Fig. 3 shows the results. The curve marked with triangles represents the field measurements and that marked with circles the wind tunnel data. The agreement is generally fair with the exception of wind directions around 280° (wind about from the right to the left in Fig. 2). Small shifts of the probe show that for this wind direction sector the monitoring station is located in a zone with large concentration gradients (i.e. very small probe positioning errors have large effects). The concentrations in the wind tunnel were measured utilizing a fast Flame Ionisation Detector with a frequency response of approximately 400 Hz. This high resolution in time enabled us to collect time series which subsequently were averaged over different time intervals. Since in the wind tunnel experiments all initial and boundary conditions were carefully controlled and kept constant, the circles represent steady-state results. The error bars attached to the wind tunnel data points indicate the variation of mean concentrations when averaging intervals of only 9s (which corresponds to 30 min in the field) were chosen. They should be attached to the field data points. They respresent the large inherent variability of the field data caused by the unsteadyness of the wind vector within the urban canopy layer.
 

Fig. 3: Comparison of results from field and wind tunnel measurements. For explanations see text.
 

5. Main conclusions

The results confirm that in cases as described here, model validation is a complex task, and that the common believe, field data would represent the truth, is obviously not always justified.
 

6. Aim for the coming year

To back up these findings, similar work will be carried out at the example of the Jagtvej monitoring station in Copenhagen, Denmark.
 

7. Acknowledgements

The authors are grateful for financial support from PEF (Projekt Europaisches Forschungszentrum fur Massnahmen zur Luftreinhaltung, Forschungszentrum Karlsruhe), and UBA (German Federal Environmental Agency, Berlin).
 

8. References

Liedtke, J. and M. Schatzmann; Autoabgasausbreitung in Strassenschluchten - Vergleich von Windkanal- und Naturmessungen an einem konkreten Fallbeispiel, Internet publication http://bwplus.fzk.de/pef/diskpef98/schatzmann/schatzm.htm (1998).

NLO; Lufthygienisches Uberwachungssystem Niedersachsen - Standortbeschreibung der NLO Stationen, Bericht Niedersachsisches Landesamt fur Okologie, Gottinger Str. 14, 30449 Hannover (1995).