Proceedings PHYSMOD 2007

8/20/2019 Proceedings PHYSMOD 2007

1/260

PHYSMOD 2007

International Workshop

on Physical Modelling of Flow

and Dispersion Phenomena

August 29-31, 2007

Orléans, France


2/260

ISBN 2-913454-32-1

EAN 9782913454323

© Presses universitaires d’Orléans

Crédits photos : - Université d’Orléans / Communication / JSL - David Hall


3/260

Proceedings of

International Workshop on Physical Modelling ofFlow and Dispersion Phenomena

August 29-31, 2007. Orléans, France

Scientific committee:Dr. Sandrine Aubrun. L.M.E., University of Orléans, France.Dr. Daniele Contini. Institute of Atmospheric Sciences and Climate, CNR, italy.Dr. David Hall. Envirobods, Ltd. UK.Prof. Jerry Havens. University of Arkansas, USA.Dr. David Heist. U. S. Environmental Protection Agency, N Carolina, USA.Prof. Zbynek Janour. Academy of Sciences, Prag, Czech Republic.Prof. Tamas Lajos. Budapest University of Technology and Economics, Hungary.Dr. Bernd Leitl. University of Hamburg, Germany.Prof. Masaaki Ohba. Tokyo Polytechnic University, Japan.Prof. Hideharu Makita. Toyohashi University of Technology, Japan.Dr. Robert Meroney. Colorado State University, USA.Prof. Alan Robins. University of Surrey, UK.Dr. Eric Savory. University of Western Ontario, London, Canada.Prof. Michael Schatzmann. Universiy of Hamburg, Germany.Prof. Ted Stathopoulos. Concordia University, Montreal, Canada.Prof. Jeroen Van Beeck. Von Karman Institute, Rhode-St.-Genèse, Belgium.

Editor :Dr. Sandrine Aubrun. L.M.E., University of Orléans, France.

Workshop organised by:Université d’OrléansPolytech’OrléansLaboratoire de Mécanique et d’Energétique

Special thanks for support to:Université d’Orléans

Polytech’OrléansLaboratoire de Mécanique et d’Energétique

Région CentreVille d’Orléans

Presses Universitaires d’Orléans


4/260


5/260

Preface

On behalf of the Scientific Committee, it is my great pleasure to present to you theproceedings of the international workshop on Physical Modelling of Flow and DispersionPhenomena, PHYSMOD 2007, held at Orléans, France, on August 29-31, 2007.

The workshop had 46 submissions and each of these was reviewed by two members ofthe Scientific Committee. We have selected 32 as oral presentations and 14 as posterpresentations. Related full-length papers are gathered in these proceedings. I thank themembers of the Scientific Committee for their important contributions to the workshop,for providing major input to the formulation of the workshop goals and their carefulreviews of submissions.

The objective of PHYSMOD is to bring together the community active in physical

modelling of flow and dispersion in wind tunnels or water channels. It is intended todiscuss, assess and report on the state-of-the art of experimental work in this field,define directions of future research and encourage wider collaboration betweenresearch institutes. It is intended to provide a wide platform for information exchangeand knowledge transfer, and participating institutions and laboratories are encouragedto also bring their undergraduate and postgraduate students to present their work and toincorporate them into the active fluid modelling community.

PHYSMOD 2007 concentrates on the physical modelling of flow and dispersion in thenatural environment, referring mainly to the following topics:

Heat and mass transfer due to atmospheric dispersion in urban areas (alsoincluding heat island problems)

Unsteady properties of the dispersion process

Building effects on the flow characteristics in urban areas Validation of numerical and analytical modelling methods Improvement and validation of atmospheric flow and dispersion modelling

techniques Quality assurance in physical modelling Use of boundary layer modelling for wind technology

I wish you all a fruitful workshop and a nice stay in Orléans!

Dr. Sandrine Aubrun


6/260


7/260

CONTENTS

KEY-NOTE LECTURE

Compiling validation data sets from systematic wind tunnel measurementsrequirements and pitfalls ........................................................................................................................ 1B. Leitl, M. Schatzmann

DENSITY EFFECTS

Effect of Semi-circular Windbreak Array on the Heavy Gas Plume Dispersion in Urban Areas............ 7B.S. Shiau, Y.C. Wu

Wind tunnel studies of LNG vapor dispersion from impoundments ..................................................... 15J. Havens, T. Spicer, W. Sheppard

The Effect of Release Time on the Dispersion of a Fixed Inventory of Heavy Gas A Wind Tunnel Model Study ................................................................................................................. 17D.J. Hall, V, Kukadia, S. Walker, P. Tily, G.W. Marsland.

QUALITY ASSURANCE

Quality assurance of micro-scale meteorological models – Action COST 732.................................... 23M. Schatzmann, B. Leitl.

How comprehensive is comprehensive enough?

Model-specific reference data for the validation of micro-scale LES flow and dispersion models....... 27F. Harms, B. Leitl, M. Schatzmann

How dense is dense enough?Systematic evaluation of the spatial representativeness of flow measurements in urban areas......... 33D. Repschies, B. Leitl, M. Schatzmann.

How close is close enough?Sensitivity of wind tunnel results with respect to changing approach flow conditions.......................... 41I. Herbst, B. Leitl, M. Schatzmann

WIND TECHNOLOGY

Properties of the far wake of a wind turbine in an atmospheric boundary layer .................................. 47G. España, S. Aubrun, P. Devinant, L. Laporte, E. Dupont.

Aerodynamic Design of the Princess Elizabeth Belgian Antarctic Research Station........................... 53J. Sanz Rodrigo, C.Gorle, J. Van Beeck, P.Planquart.

Feasibility study of tests on scale models for the evaluation of the overpressures inducedby a passing train on adjacent structures............................................................................................. 59L. Procino, G. Bartoli, C. Borri, A. Borsani.

Wind Tunnel Simulations of Pollution from Roadways......................................................................... 63D. K. Heist, S. G. Perry, L. A. Brixey, G. E. Bowker.


8/260

URBAN FLOWS

How rough is rough?Characterization of turbulent fluxes within and above an idealized urban roughness ......................... 69M. Schultz, B. Leitl, M. Schatzmann.

Effect of roofshape on unsteady flow dynamics in street canyons....................................................... 75J. Barlow and B. Leitl.

Study of Flow Fields in Asymmetric Step-Down Street Canyons......................................................... 79B. Addepalli, E.R. Pardyjak

Spanwise variation of drag on roughness elementsin a nominally two-dimensional boundary layer.................................................................................... 87P. Hayden, T. Mapurisa and A.G. Robins

CFD analyses of flow in stratified atmosphere..................................................................................... 93G. Kristóf , N. Rácz,, M. Balogh

BUILDING EFFECTS ON PLUMES

Wind tunnel study of the concentration fields in a plume emission...................................................... 99 A.R. Wittwer, F. De Paoli, A.M. Loredo-Souza, E.B.Camano Schettini

Improved Building Dimension Inputs for AERMOD Modelingof the Mirant Potomac River Generating Station............................................ .................................... 105R. L. Petersen, J.J. Carter

Physical modeling of the downwash effect of rooftop structures ....................................................... 111 A. Gupta, P. Saathoff, T. Stathopoulos

DISPERSION IN URBAN AREAS

Dispersion from an area source in urban-like roughness................................................................... 113F. Pascheke, J.F. Barlow and A. Robins

Dispersion of traffic exhausts in urban street canyons with tree plantings.Experimental and numerical investigations ....................................................... ................................ 121C. Gromke,J. Denev, B. Ruck

Flow and dispersion study in the simplified urban area ..................................................................... 129H. Sedenkova, Z. Janour

Measurements on the Dispersion of Pollution in Urban Environment of Cubic Building Arrays

with Different Wind Directions ............................................................................................................ 135B.S. Shiau, Y.S. Lin

Specifying exhaust systems that avoid fume reentry and adverse Health Effects............................. 143R.L. Petersen, J.J. Carter


9/260

JETS AND PLUMES

Comparison of near-field behaviour of jets and plumes in a crosswind............................................. 149E. Savory, M. Restorick, Z. Duan.

Comparison of experimental and modelled plume rise in stable environments................................. 157D. Contini, A. Donateo, D. Cesari, A. G. Robins.

Validation of LES on plume dispersionin the convective boundary layer capped by a temperature inversion ............................................... 165T. Tamura, H. Nakayama, K. Ohta.

NEW TECHNOLOGICAL SOLUTIONS

Simulation of long time averaged concentration under actual meteorological conditions ................. 167

T. Hara, R. Ohba, K. Okabayashi, J. Yoneda, H. Nagai, T. Hayashi

A proposal for a new atmospheric boundary layer wind tunnel in the Netherlands ........................... 175P. Builtjes, H. van Dop, B. Holtslag, H. Jonker

Active driving of a multi-fan wind tunnel ............................................................................................. 177K. Sassa and H. Miyagi

Determination of Spatial Structure of Internal Gravity Waveby Multi-channel Thermo-Anemometer Measurement....................................................................... 179H. Makita and K. Ohba

POSTERS

Use of detection of coherent flow structuresfor better understanding of 3D flow fields in urban environment........................................................ 185T. Régert, I. Goricsán, M. Balczó, K. Czáder, T. Lajos

Air-quality and spatial planning .......................................................................................................... 191FL.H. Vanweert,. J.I.J.H. van Rooij

Charateristics of the low–speed wind tunnel of the LMF, France ...................................................... 197L. Perret

Evaluation of pollution dispersion prediction using RANSwith turbulence models available in FLUENT 6.3............................................................................... 201R. Izarra-Garcia, J. Franke

Flow and dispersion around tall buildings .......................................................................................... 203T. Lawton, A. Robins

Quantifying the temporal representativeness of flowand dispersion measurement in a complex urban area ..................................................................... 207M. Rix, M. Schtzmann, B. Leitl.

How often is sufficient? A program for the statistical analysisof puff dispersion in urban environments ........................................................................................... 213R. Fischer, M. Schtzmann, B. Leitl


10/260

Issues concerning wind tunnel modelling of heavy gas dispersionwith focus on risk assessment............................................................................................................ 221K. Bezpalcová, M. Ohba

Visualisation of dispersion processes in the surrounding of livestock buildings ................................ 225

K. von Bobrutzki, H.J. Müller

Wake development and interactions within an array of large wind turbines ...................................... 229F. Pascheke and P.E. Hancock

Wind-Tunnel Modelling and Numerical Simulation Within Urban Intersection................................... 231R. Kellnerová, M. F. Yassin, Z. Jaour

Adaptation of the « Lucien Malavard »Wind Tunnel as an Atmospheric Boundary Layer Wind Tunnel......................................................... 237S. Aubrun, G. España, P. Devinant

Numerical Simulation of Room Air Motion, Physics and CFD Modeling....... ..................................... 241V. Esfahanian, E. Moallem

INDEX OF AUTHORS ............................................................................................. 243


11/260

LAVISION GMBH

ANNA-VANDENHOECK-RING 19

37081 GOETTINGEN

TEL.: +49 (0)551 9004 0

EMAIL: [email protected]

WWW.LAVISION.COM


12/260

T S I E u r o p e

TSI Incorporated designs and manufactures precision instruments

used to measure flow, particulate, and other key parameters in

environments the world over.

TSI provides worldwide solutions for industry, governments,

research institutions, and universities, with applications ranging

from pure research to primary manufacturing. Innovative thinking

keeps TSI Incorporated a leader in the markets we serve.

We remain committed to helping customers solve their

measurement problems, and we have global technical talent,

resources, and commitment to find unique solutions for

difficult applications.

Call us:

TSI France

Europarc Bât D, 26 rue John Meynard Keynes, Technopôle de Château-Gombert, 13013 MARSEILLE Tél : +33 (0)4.91.95.21.90 - [email protected] - Site internet : www.tsiinc.fr

TSI - Leaders in Instrumentation


13/260

For more information : www.quantel-laser.com


14/260


15/260

For improved

quality of life

Aerodynamics and

hydrodynamics

Optimisation of

materials and

components

Optimisation of product

design and combustion

•Imaging Systems- Global flow

measurements

- Combustion

diagnostics

- Particle size

measurements

•Fluid Dynamics- Air and gas flow

measurements

- Liquid flow measurements

- Particle size

measurements

•Indoor Climate- Thermal Comfort

- Air Quality

- Ventilation Efficiency

•Solid Mechanics- Strain & Stressmeasurements- Vibration analysis- Non-destructive testing


16/260


17/260

1

Compiling validation data sets from systematic wind tunnelmeasurements – requirements and pitfalls

B Leitl, M Schatzmann

Center for Marine and Atmospheric Research, Environmental Wind Tunnel LaboratoryUniversity of Hamburg,Hamburg, Germany

[email protected]

Abstract - Laboratory data are often used as areference for validating micro-scale flow and

dispersion models. In contrast to field data,

carefully generated test data sets compiled fromsystematic wind tunnel tests provide a number of

advantages regarding the consistency and accuracyof the test data, the completeness of a dataset or

concerning the documentation of the boundaryconditions of an experiment. Based on more than

10 years of experience in generating validation dataat the Environmental Wind Tunnel Laboratory

(EWTL), the paper is illustrating common pitfallsand problems related to the compilation of

reference data sets. From a wider, project-independent perspective it is intended to define

minimum modeling quality requirements in order tofurther promote boundary layer wind tunnels as a

source of reliable reference data for atmosphericflow and dispersion modeling.

Key words – physical modeling, atmospheric flow anddispersion, quality assurance, modeling standard.

Introduction

Physical modeling of environmental flow anddispersion phenomena is providing reliable and detailedinformation on wind driven pollutant transport in built-upareas for several decades. Whereas wind tunnelmodeling was used for almost any kind of environmentalflow and dispersion problem within the atmosphericboundary layer in the past, nowadays a large fraction ofenvironmental flow and dispersion phenomena can bemodeled numerically with accuracy sufficient for manypractical applications. However, physical modeling in aboundary layer is still the preferred choice if local-scale,transient flow and dispersion phenomena in the range ofa few hundred meters and at time scales in the order ofminutes are a matter of particular interest. Based oncareful and diligent physical modeling and with the helpof state-of-the-art instrumentation, information and datacan be provided with a temporal and spatial resolutionwhich cannot yet be achieved by numerical modeling.The high resolution of wind tunnel test data acquiredunder controlled boundary conditions also enablesvalidation data to be compiled from systematic windtunnel tests. Based on systematic reference data, thequality of different types of numerical models can beevaluated and the 'fitness for purpose' for specific modelapplications can be tested.

In general, the availability of test data and theirquality and accuracy significantly affect the outcome ofmodel validation procedures. In a strict sense, theresults of a model evaluation are as good or asuncertain as the reference data used (Britter andSchatzmann, 2006). Depending on the point of view

from which the validation problem is seen, different andonly partially overlapping data quality requirements canbe derived. Perhaps the most global set of data qualityrequirements and constraints is developing from a strictphysical perspective but those requirements are mostlikely not to be met entirely by field or laboratory data.Seeing the problem from a numerical modeling point ofview also leads to very demanding data qualityrequirements which are unlikely to be met entirely byregular sources of data. From a physical modelers pointof view, the data to be delivered will satisfy what is 'stateof the art' in instrumentation and practically possible in awind tunnel, but not necessarily what is needed bynumerical modeling.

Until now, there is no clear definition of what shouldbe called 'validation data' and what quality criteriashould be met particularly by systematic test dataderived from wind tunnel tests. Consequently, varioustest data sets have been used and sometimes evenmisused for 'validating' atmospheric flow and dispersionmodels. Since it is obvious, that reliable and quality-assured reference data are as important as a soundvalidation strategy for testing accuracy and reliability ofatmospheric flow and dispersion models it is intended topromote a more complete approach to the validation

data problem. Within the scope of the following paperthe basic requirements on validation data developedand the methodology applied at the EWTL are outlinedand a methodology for compiling model- andapplication-specific reference data is drafted.

1 Data Requirements and Constraints

In principal both, field data and laboratory data canbe used for testing the quality and accuracy of micro-scale flow and dispersion models. Nevertheless,laboratory data compiled from systematicmeasurements in a boundary layer wind tunnel providea number of benefits regarding the evaluation ofnumerical models and the most prominent general datarequirements are listed below.

Whereas field data in most cases represent verycomplex flow and dispersion situations, in a wind tunnelthe degree of complexity can be chosen as needed andreference data of different complexity can beprovided. Particularly the more simple configurationssuch as individual structures or regular arrays ofobstacles enable a numerical code to be tested carefullywith respect to the physical behavior under well-definedgeometrical conditions. A unique feature of laboratorytest data sets certainly is, that even block-structured'numerical' representations of a certain urban structurecan be tested in detail for fixed, well-documentedboundary conditions (Leitl et al, 2001).

A second major advantage of wind tunnel data to beused for model validation purposes is the potential to


18/260

2

generate systematic test data sets. The ability of anumerical model to replicate one particular test caserepresenting one particular configuration is notnecessarily documenting the models performance. Adifference between model results and reference datamight be acceptable, as long as the model is behaving'physically correct'. Testing the physically correct

behavior of a numerical model requires reference dataseries with singular test parameters such as the meanwind direction, the size of the model domain, thelocation of emission sources or the type of sources to begradually varied in order to test the sensitivity of a modelto certain changes or uncertainties in model input. Theavailability of systematic test data is mandatory fortesting the physical performance and the required typeof data can be provided with justifiable effort basedlaboratory experiments only.

The completeness of reference data sets isanother point where wind tunnel test data can surpassthe quality of most of the field data sets. As it is nearlyimpossible to completely measure and document thephysical and geometrical boundary conditions of a flowand dispersion experiment in the field, laboratory

experiments provide at least a chance to measure alldriving boundary conditions required for a 'completedocumentation' of a data set. As a foundation for arealistic evaluation reliable information on theaccuracy/uncertainty of the reference data is anothernecessary requirement for a complete documentationwhich can be provided not always for field experimentsbut which is readily available from carefully preparedwind tunnel tests.

As a further requirement, the representativeness of

validation data is listed here. For model evaluation, thereference data must be of known statistical, temporaland spatial representativeness. While it is again not aserious problem to define the representativeness ofmeasured data within the scope of a diligently carried

out wind tunnel experiment it is more or less impossibleto document the representativeness of field data withoutfurther assumptions or without the need of further inputfrom numerical or physical modeling. However, it mustbe stated clearly, that defining and documenting the'representativeness' of reference data requiressubstantial efforts in addition to 'regular' experiments,but without this information, data is of no value for modelvalidation.

Finally, the 'fitness for purpose' of validation datasets must be mentioned. As long as the 'fitness forpurpose' is understood as providing test data for aparticular environmental flow and/or dispersion situation,there is no real preference for data from fieldexperiments or laboratory data from a wind tunnelexperiment. However, if the fitness for purpose is

understood as the need for providing validation datawhich fit the needs of a certain type of numerical modelto be validated, there are again several advantages inusing specifically generated laboratory data. Forexample, a RANS type code would require a statisticallyrepresentative mean flow and dispersion field to be usedas reference, which cannot be derived from field data atall because the quasi-stationary conditions assumed ina RANS-based model approach do not exist at fullscale. Similarly, the validation of LES-based numericalcodes is most likely to be based on statistic andprobabilistic measures of representative ensembles oftransient flow and dispersion data which require (long-term) data collection for quasi-stationary boundaryconditions as well.

Despite the advantages the use of carefully compiledlaboratory data has, the biggest constraint is of coursethat wind tunnel data are model results only. As it is thecase for numerical models, the results from wind tunnelmodeling are significantly affected by the choice and thequality of boundary conditions applied in the model andthe simplifications introduced for example regarding the

effects of thermal stratification or the spectral range ofturbulent eddies replicated in the wind tunnel.Consequently, wind tunnel results do not automaticallyprovide reliable results for model validation or for up-scaling to full-scale conditions when a model looks'geometrically similar'. In fact, the scatter or uncertaintyin wind tunnel data can distort the physical resultsentirely, when no state-of-the-art modeling rules areapplied and physical similarity is not sufficiently fulfilled.On the other hand, results from independent wind tunneltest campaigns or even from completely different windtunnel facilities are expected to match within the boundof the measurement accuracy/repeatability of modeltests, if the same geometrical and physical boundaryconditions are realized.

2 Methodical Requirements

2.1 General RemarksUp to now, there is no clear definition of what the

'state-of-the-art' in wind tunnel modeling is. Most of thewind tunnel laboratories follow their own modelingstandard, which is driven mainly by the purpose of aparticular wind tunnel test. A comprehensive summaryof the similarity concepts of boundary layer wind tunnelmodeling is given in the 'Guideline for Fluid Modeling of Atmospheric Diffusion' (Snyder, 1981). Although thephysical background as well as most of themeteorological references is still appropriate, the mainfocus of wind tunnel modeling has changed noticeably inthe past. Long-range pollutant transport from elevated

point sources, flow and dispersion in the upper part ofthe atmospheric boundary layer (Ekman-layer) or flowand dispersion phenomena under stratified atmosphericconditions are nowadays modeled with sufficientaccuracy using numerical tools instead of a physicalmodel. A wind tunnel model remains the preferential toolfor modeling flow and dispersion phenomena in thelower atmospheric boundary layer, where flow field isdominated by the effects from the underlying roughness. Accordingly, the modeling requirements must beadjusted.

A more recent attempt to establish a standard forenvironmental flow and dispersion modeling has beenmade by the German Engineering Society VDI. The'standard' defined in VDI 3783/12 (2000) wasestablished based on a guideline developed by theWTG (Windtechnologische Gesellschaft) specifically forbuilding aerodynamics. The modeling qualityrequirements in general and the specification of targetvalues given in the guideline were based on ainterlaboratory comparison of flow and dispersionmeasurements around a single cubic obstacle. As aresult, the uncertainty documented in evaluation/targetdata cannot be counted as quality criteria. In contrast, itis documenting the modeling uncertainty when windtunnel modeling is based on the VDI standard only.Whereas an uncertainty of model results of more than50% might be acceptable for some practicalapplications, this order of magnitude expected as scatterin model results is clearly unacceptable for referencedata to be used for model validation.


19/260


20/260

4

2.4 Model Quality and Suitability ofExperimental Setup

In a wind tunnel experiment, not necessarily themodel with the highest geometrical detail or the 'mostrealistic appearance' is giving the best or most realisticanswers. Designing and constructing 'aerodynamicmodels' of urban structures also differs significantly from

common practice in aviation research, architecturaldesign or in the models used for investigations ofvehicle aerodynamics for instance. Surface roughnessand surface structure have to be considered as a majorfactor influencing the results of near-ground or near-wallflow and dispersion measurements. Unfortunately, thereare no common standards on how to built wind tunnelmodels suitable for aerodynamic testing of urbanstructures, except perhaps the local roughnessReynolds number criteria to be considered, and it isdifficult to judge the 'fitness for purpose' of a model if thetest documentation is missing the information requestedin paragraph 2.3. A prominent example of how differentaerodynamic models might look like is the modelrepresentation of trees in a wind tunnel. Whereas at theEWTL a aerodynamic model of groups of trees wasdeveloped for flow and dispersion measurements in andabove forest areas (Aubrun et al, 2004), other windtunnel laboratories still prefer the use of model treeswith a high degree of geometrical similarity.

Dispersion modeling also requires the emissionsources to be modeled properly in order to simulate aparticular dispersion process. To replicate for exampleemissions from road traffic in a wind tunnel, an artificial'line source' must be built into the model but the way linesources are actually built normally differs. It is known,that a well-defined release of emissions from any kind ofwind tunnel model source requires a sufficient pressureloss across the entire release system to ensure thesource is working independently from the pressurepatterns introduced by obstacles/model buildings

surrounding the source area (Meroney et al, 1996;Pascheke et al, 2003). If the release is not independentfrom the pressure field above the source for examplechanging the wind direction in the model will alsochange the emission pattern and from concentrationmeasurements it is most likely not clear, whether achange in the concentration pattern is caused bychanges in the flow field or whether it is resulting from adifferent emission source behavior. A second problemregarding emission source modeling is a possiblyunrealistic distortion of the mass balance or the localflow field due to the release of large amounts of tracer. Again, the street canyon case can be seen as aprominent example, where the use of too high sourcestrength or the introduction of a significant momentumdue to jet-like releases can downgrade the accuracy and

reliability of the results significantly. The suitability of apassive emission source must be documented at leastby giving experimental prove that the results fromnearby concentration measurements scaled with thesource strength give constant values over a wide rangeof release rates.

There are several other issues regarding thesuitability of an experimental setup which cannot bediscussed in detail here. For instance, the need formodeling traffic induced turbulence and how to modelcar turbulence is still an open question. However, at thispoint it was intended to make the clear statement, thatnot just the (geometrical) wind tunnel model but theentire experimental setup is defining the accuracy ofwind tunnel test results.

2.5 Wind Tunnel Instrumentation andExperimental Facility

A few remarks on the instrumentation used for flowand dispersion measurements shall be made here. Evenif it is assumed to be a matter of course, theinstrumentation utilized in a wind tunnel experimentmust be suitable for the type of measurements as well.

Again, there is no common sense, what exactly isneeded for a particular type of flow and dispersionmeasurements and suitability is more often definedbased on the availability of instruments than on thecapabilities of measurement devices. Just as anexample, the required reference wind speedmeasurement – at a location with sufficiently lowturbulence levels – can still be carried out reliably bymeans of a pneumatic probe and a calibrated pressuretransducer. In contrast, flow measurements within anurban structure or next to obstacles will require LDA orcomponent-resolving PIV for reliable measurements.Even a multi-component hotwire will not give properresults because of the complexity and threedimensionality of the flow field.

For dispersion measurements, most often FIDsystems are used, giving sufficiently reliable andaccurate measurement results, provided thatinstrumentation is used properly. Dispersionmeasurements also require a sufficient monitoringand/or control of the tracer release or emission sourcestrength as well.

In general, a properly compiled data report forvalidation data must necessarily contain sufficientinformation on the type of instrumentation used for aspecific measurement task. In addition, the accuracy ofthe measurement devices as well as their calibrationstatus must be documented clearly in order to provideconfidence in the data to be used as a reference formodel validation.

Regarding the wind tunnel itself, there is no

fundamental restriction on what kind of wind tunnel canbe used. Even if specially built boundary layer windtunnels might be counted as the most efficient type offacility, the same or sometimes even better quality ofphysical modeling of environmental flows can bereached in adapted conventional wind tunnel facilities.The method of boundary layer modeling might, however,differ. The common approach of using a boundary layerspecific set of turbulence generators and a specific floorroughness is successfully applied particularly in windtunnels with a long test section or fetch upwind of themodel area. For smaller model scales, conventionalwind tunnels have been equipped with grid structures orhorizontal bars, enabling homogenous model boundarylayer flows to be generated within shorter test sections.

Of course, the size of the test section is clearly

defining the model scale to be realized in a specific windtunnel facility. From the experiences made in differentboundary layer wind tunnels at EWTL one can concludethat in an approximately 1.5 m wide test section thebiggest model scale providing sufficient agreement withfull-scale conditions in the lower atmospheric boundarylayer is in the order of 1:500 to 1:400. A larger modelscale in the order of 1:200 or 1:300, as it might berequired for reaching sufficient spatial resolution duringlocal scale flow and dispersion measurements withinurban structures, already requires a wind tunnel facilitywith an approximately 3 to 4 m wide test section whenurban type boundary layer flows have to be modeled.When using a large scale geometrical model of an urbanstructure in a narrow test section, physical similarity is


21/260

5

significantly limited because large scale low frequencyturbulent eddies are certainly not properly modeled atthe right spatial and temporal scale.

2.6 Post-Processing of Laboratory Data Another set of problems and uncertainties of

laboratory based validation data sets is related to the

post processing and 'proper' scaling of wind tunnelresults. Usually, wind tunnel data are provided to the'end user' in non-dimensional form, for example as C*-concentration values. This enables the end user of thedata – theoretically – to scale the wind tunnel dataaccording to different/particular field conditions to beused during a model validation procedure. In order tore-scale data properly, it must be documented clearly,how the raw measurement data was post-processedand prepared and how the dimensionless values havebeen calculated in order to ensure 'compatibility' ofdifferent information sources. Quite often, different andsometimes even wrong scaling is applied to wind tunneldata because the data provider does not distinguishbetween the type of the source (line/point/area) or'universal scaling' does not apply in a particular case.

For example dispersion measurements in the near fieldof a momentum-depended release such as an exhaust jet should not be scaled using a C*-concept because itimplies that the results could be transferred to anycombination of wind speed and source strength, which isnot at all the case.

When data is handed out to the end user in non-dimensional form, another basic requirement regardingthe documentation of the experiments is also to providethe measurement uncertainty and repeatability in similar,non-dimensional form. From this it becomes clearinstantly, that for example not only a very preciseconcentration measurement is necessary in the case ofdispersion measurements. The control andmeasurement of the emission source strength and thereference wind speed can contribute significantly to the

overall uncertainty of the wind tunnel results as well.Just as another example, if the accuracy of the releasedgas volume flow is measured with an accuracy of 20%only, the minimum accuracy of the concentrationmeasurements cannot be better than 20%. In fact,experience shows that most of the 'standard dispersionmeasurements' suffer from a severe lack of accuracy inthe control and measurement of the 'initial and boundaryconditions' more than suffering from a lack of accuracyin the concentration measurements itself. In some,hopefully rare, cases an in depth analysis of 'standarddispersion measurements' reveals that they cannot bemore precise than ±50%. On the other hand, carefullycarried out dispersion measurements can reach anaccuracy of better than 5%.

3 Summary and Outlook

From the comments given above it is clear, that acommon standard for wind tunnel modeling ofatmospheric flow and dispersion phenomena is much-needed. Otherwise, the scatter found in similarexperimental results compiled in different wind tunnellaboratories might be counted as a general limitation ofphysical modeling and not as what it really is, anuncertainty due to different boundary conditions anddifferent efforts spent for wind tunnel testing. Qualityassurance becomes even more important if results fromphysical modeling are used as a reference fordevelopment and testing of numerical flow anddispersion models.

In addition to the already established technicalstandards and common practices in the wind loads /wind engineering community, a common practiceguideline for physical modeling of micro-scaleatmospheric flow and dispersion phenomena should bedeveloped which is sufficiently considering the particularmodel quality requirements of dispersion modeling. In

order to further promote wind tunnel testing as a reliablesource of reference data, a quality standard must beaccepted among wind tunnel modelers, which is notbased on the 'least common denominator' only butbased on what is needed to assure reference quality ofdata and based on what is technically possible with'state-of-the-art' facilities and instrumentation.

References

Britter, R. and Schatzmann, M. (Ed.) (2006)."Background and justification document to support themodel evaluation guidance and protocol", COST Action 732, Quality Assurance and Improvement ofMicroscale Meteorological Models, ISBN 3-00-018312-4.

Leitl, B.; Chauvet, C. and Schatzmann, M. (2001)."Effects of Geometrical Simplification and Idealizationon the Accuracy of Microscale Dispersion Modeling"Proc. 3

rd Int. Conf. Urban Air Quality , March 19-23

2001, Loutraki, GreeceSnyder, W.H. (1981) "Guideline for Fluid Modeling of Atmopsheric Diffusion" US EPA, EnvironmentalSciences Research Laboratory , Research TrianglePark, NC 27711

VDI 3783/12 (2000) "Environmental Meteorology –Physical modeling of flow and dispersion processes inthe atmospheric boundary layer, Application of windtunnels" Verein Deutsche Ingenieure VDI, HandbuchReinhaltung der Luft, Band 1B (in German & English)

Liedtke, J; Leitl, B.; Schatzmann, M. (1998) "Carexhaust dispersion in a street canyon – Wind tunneldata for validating numerical dispersion models" Proc.2

nd East European Conference on Wind Engineering,

Prague, 7-11 September 1998, vol. 1, pp. 291-297Schatzmann, M.; Bächlin, W.; Emeis, S.; Kühlwein, J.;

Leitl, B.; Müller, W.J., Schäfer, K.; Schlünzen, H.(2006) "Development and validation of tools for theimplementation of european air quality policy inGermany (Project VALIUM)" Atmos. Chem. Phys., vol.6; pp. 3077-3083

Herbst, I.; Leitl, B.; Schatzmann, M. (2007) "How closeis close enough? – Sensitivity of wind tunnel resultswith respect to changing approach flow conditions"Physmod2007, ibid.

Aubrun, S.; Leitl, B. (2004) "Development of animproved physical modelling of a forest area in a wind

tunnel" Atmospheric Environment, vol 38, 2004, pp.2797-2801

Meroney,R.N.; Pavageau, M.; Rafailidis, S.;Schatzmann, M. (1996) "Study of line sourcecharacteristics for 2-D physical modelling of pollutantdispersion in street canyons". J. Wind Eng. Ind. Aerodyn. 62 (1996), pp. 37–56.

Pascheke, F.; Leitl, B.; Schatzmann, M. (2003)"EVALUATION OF FIELD TRACER EXPERIMENTSWITH RESPECT TO COMPLEX URBAN TYPEBOUNDARY CONDITIONS". In Manfrida, Giampaoloand Contini, Daniele, Eds. ProceedingsPHYSMOD2003: International Workshop on PhysicalModelling of Flow and Dispersion Phenomena, Prato-Italia.


22/260


23/260


24/260

8

which is with a molecular weight of 44 (1.52 timesmolecular weight of air). Dispersion experiments wereexecuted by continuously spill the carbon dioxide (CO2) ofheavy gas at a controlled flow rate from an elevated pointsource. Sampling tracer of the heavy gas was carried outfor various spilling flow rates and windbreak opening ratiosfor downstream distances away from source at differentdownstream stations. For concentration measurements,tracer gas sampling system was developed. It is

composed of 15 sampling tubes that arranged in a rake.The 15 tubes attach to 15 air bags and suck the tracer gasby pumps. In order to obtain sufficient tracer concentrationanalysis, 5 minutes of sampling time was executed inevery downstream station. Mean concentrations of thesampled tracer gas in air bags were obtained by using theanalyzer of Cole-Parmer carbon dioxide detector. Thedetector has the concentration range of 0~100000 ppmwith a resolution of 10 ppm.

sh

Fig.1 Schematic diagram of the arrangement of theexperiments

3 Similarity Analysis of Concentrationand Dispersion Scale Parameters

From fluid flow modelling laws, some similarityrequirements have to be performed for transferring smallscales of wind tunnel model results to prototype scales.The similarity laws are obtained by dimensional analysis.The sketch of the continuously release of the heavy gasand windbreak location is shown as in Fig.1. The localheavy gas concentration downstream of the spill sourcecan be expressed as the following function:

},,

,,),(,,,,,,,,,,{1

P

Z nhU U H Dh g z y x F C

aa

ref sa s s s s s

(1)

where coordinates : x is the downstream direction, y is thelateral direction, and z is the vertical direction. hs: elevatedheight of source; Ds: diameter of source, H: windbreakheight, P : opening ratio of windbreak array, Us: spillvelocity of heavy gas, s :density of spilled heavy gas;

)( sa hU :cross mean wind velocity at the height hs; n:exponent of power law for mean velocity profile ofapproaching flow; ref Z :boundary layer thickness; a :ambient air density; a :viscosity of the air; C is heavy gasconcentration at ( x,y,z ).

After the dimensional analysis, some dimensionlessgroups were obtained, and equation (1) can betransformed into the following dimensionless form asshown in equation (2).

Re},Re,,,,,,,{2 sa

s Fr P n H

z

H

y

H

x F K

(2)

where ( H h s ): dimensionless source height; a s :density ratio; where Fr sa s s s gDU )( :densimetric Froude number; s s s s s U D /Re : Reynoldsnumber in source; aaU L /Re 00 : Reynolds number of

ambient flow. The dimensionless concentration K is scaledas:

H DU

CQ K

s s

(3)

As shown in equation 2, it indicates that thedimensionless local concentration K in both of the modeland prototype is the same at the locations X/H , Y/H , Z/H ,provided that all the remaining dimensionless parameterson the right-hand side of equation can be matched in thewind tunnel experiments. In the present study, all theparameters are set the same for model and prototypeexcept two parameters, such as: ambient Reynoldsnumber, Re and spill source Reynolds number, Res. Theambient Reynolds numbers kept exceeding the criticalnumber (10

4) for ensuring flow becomes turbulent in the

experiments. A cross wire was placed at the exit of theelevated source to trip the outflow. This is to ensure itbecome turbulent flow.

The experiments carried out in the present study wererun for some important dimensionless parameters whichmake sure of both safe experimental operating conditionsand a variation of the similarity parameters dominating theheavy gas dispersion.

For non-Gaussian distributions associated with theheavy gas dispersion behind the semi-circular windbreakarray, the parameters y and z can be viewed as thestandard deviation of the concentration distributions inlateral and vertical directions, respectively. Theparameters are defined by,

dy z y xC

dy z y xC y y

y),,(

),,()( 22

(4)

where

dy z y xC

dy z y x yC y

),,(

),,(

(5)

dz z y xC

dz z y xC z z

z ),,(

),,()( 22

(6)


25/260

9

where

dy z y xC

dy z y x zC z

),,(

),,(

(7)

4 Results and Discussion

4.1 Approaching FlowThe urban terrain type of neutral turbulent boundary

layer was generated as the approaching flow. Meanvelocity profile of the simulated turbulent boundary layerflow is approximated by the power law shown as equation(8).

n

ref ref Z

Z

U

Z U )(

)( (8)

where U(Z) is the mean velocity at height of Z , ref U isthe free stream velocity, and ref Z is the boundary layerthickness. In the present study, a rural terrain type of

neutral atmospheric boundary layer was simulated with amodel scale of 1/500. The free stream velocity is

ref U =3.5 m/s; and the boundary layer thickness, ref Z isabout 100 cm. The measured mean velocity profile isshown in Fig. 2. Results indicate that the power exponentn is 0.222. This value lies in the range of 0.21 to 0.40 asproposed by Counihan [10] for the urban terrain type ofneural atmospheric boundary layer flow.

0.4 0.6 0.8 1

U(z)/Uref

0

0.2

0.4

0.6

0.8

1

Z/Zref

U(Z)/Uref =(Z/Zref )n

X=10m,Y=1m

n=0.222

Fig.2 Mean velocity profile of approaching flow

The simulated longitudinal turbulence intensity profileis shown in Fig. 3. The longitudinal turbulence intensity isdefined as:

U

u I rmsu (9)

Here urms is the root mean square of turbulent velocityfluctuation, ,and U is the local mean velocity. It is seen thatthe simulated longitudinal turbulence intensity close to thewall is about 23%. Counihan [10] had indicated that thelongitudinal turbulence intensity close to the ground-levelin the urban terrain areas fell in the range of 20% to 35 %.

The Reynolds stress profile of the simulated

approaching flow is shown in Fig.4. Here '' wu is theReynolds stress. As seen in the figure, a constant stresslayer exists near the ground of the boundary layer. Thisagrees with the results of field observations reviewed byCounihan [10].

Fig.3 Turbulence intensity profile of approaching flow

0 0.002 0.004 0.006 0.008

u'w'/Uref 2

0

0.2

0.4

0.6

0.8

1

Z/Zref

n=0.222

Zref =100cm

Fig.4 Reynolds stress profile of approaching flow

4.2 Vertical ConcentrationFig.5 shows the vertical concentration profiles for

different opening ratios of windbreak array at variousdownstream distances with Fr=22. Results indicate thatheavy gas concentration is higher for opening ratio 50 %than that of 70 % and 85 % at the downstream stationx/H=11.667. It is also seen that the vertical concentrationprofiles have no significant change as decreasing theopening ratio of windbreak array at downstream distancex/H>25. Others cases of Fr=16 and 30 have presentedsimilar results. In summary, the results show thatsemi-circular windbreak array inhibits the dispersion of theheavy gas plume and reduces the plume concentrationsignificantly. When the opening ratio of the windbreak is50%, the concentration reduction downwind of thewindbreak is more effective than that of opening ratio 70%and 85%.

The vertical concentration profiles for different initial


26/260


27/260

11

that the ground level concentration firstly increases andthen decreases along the downstream distance for P=50%. The other two cases of P=70 % and 85 % only exhibitthe decrease of ground level concentration along thedownstream distance.

Fig.8 is the ground level concentration variations alongthe downstream distance for different initial densimetric

Froude numbers with P= 50 %.

0 50 100 150 200 250 300

K

0

1

2

3

4

Z/H

X/H=8.333Fr=30

Fr=22

Fr=16

0 100 200 300

K

0

1

2

3

4

Z/H

X/H=11.667Fr=30

Fr=22

Fr=16

0 100 200 300

K

0

1

2

3

4

Z/H

X/H=16.667Fr=30

Fr=22

Fr=16

0 100 200 300

K

0

1

2

3

4

Z/H

X/H=25Fr=30

Fr=22

Fr=16

0 100 200 300

K

0

1

2

3

4

Z/H

X/H=41.667Fr=30

Fr=22

Fr=16

0 100 200 300

K

0

1

2

3

4

Z/H

X/H=58.333Fr=30

Fr=22

Fr=16

Fig.6 Vertical concentration profiles for different initialdensimetric Froude number at various downstream

stations; P=85 %

4.4 Dispersion Length Scales in Vertical andHorizontal Directions

The dispersion length scales are commonly used asindication of the spread extent of the heavy gas as it isspilled. The vertical dispersion length scale can becalculated by applying the equations (6) and (7) for thenon-Gaussian concentration distribution of heavy gas


28/260

12

dispersion over the semi-circular windbreak. The verticaldispersion length scales for different opening ratios ofsemi-circular windbreak array with Fr=16 are plotted inFig.9. Results show that the vertical dispersion lengthscales increase as the downstream distance increases forcases with opening ratio of windbreak P=50 %, 70 %, and85 %. The length scales approach to a constant value at a

far downstream distance for all cases. The shelter effecton the dispersion of heavy gas plume is better for P=50 %than the other cases of 70 % and 85 %. So the verticallength scale is smaller for the case P=50 % at thelocations not far from the windbreak (i.e. x/H


29/260

13

reduction downwind of the windbreak is more effectivethan that of opening ratio 70% and 85%. The horizontaldispersion scales for different downstream distances of thewindbreak approach to the constant values at a fardownstream distance. The approaching to the constantvalues of horizontal dispersion scales is reached quicklyas the windbreak opening ratio decreases. The vertical

dispersion scales appear to close to the constant values ata far downstream distance behind the windbreak fordifferent opening ratios of windbreak array

6 References

[1] Schatzmann, M., (1993), ”Experiments with Heavy GasJets in Laminar and Turbulent Cross-Flows”, Atmospheric Environment , Vol. 27A, pp. 1105-1116.

[2] Schatzmann, M., (1995), ”Accidental Release of HeavyGases in Urban Areas”, Wind Climate in Cities,Cermak, J.E., et al. (Eds.), pp. 555-574.

[3] Britter, R.E., (1989), ”Atmospheric Dispersion of DenseGases,” Annual Review of Fluid Mechancis, Vol.21,pp.317-344.

[4] Robins, A., Castro, I., Hayden, P., Steggel, N., Contini,D., and Heist, D., (2001), ”A Wind Tunnel Study ofDense Gas Dispersionin a Neutral Boundary Layerover a Rough Surface,” Atmospheric Environment ,Vol.35, pp. 2243-2252.

[5] Zhu, G., Arya, S.P., and Snyder, W.H., (1998), ”AnExperimental Study of the Flow Structure within aDense Gas Plume,” Journal of Hazardous Materials,Vol.62, pp. 161-186.

[6] Donat, J. And Schatzmann, M., (1999), ”Wind TunnelExperiments of Single-phase Heavy Gas JetsReleased under Various Angles into Turbulent

Cross Flows”, Journal of Wind Engineering andIndustrial Aerodynamics, Vol. 83, pp. 361-370.[7] Nielsen, M., Ott, S., Jorgensen, H.E., Bengtsson, R.,

Nyren, K., Winter, S., Ride, D., and Jones, C.(1997), ”Field Experiments with Dispersion of PressureLiquefied Ammonia”, Journal of Hazardous Materials,Vol. 56, pp. 59-105.

[8] Khan, F. and Abbasi, S.A., (2000), ”Modeling andSimulation of Heavy Gas Dispersion on the Basis ofModification in Plume Path Theory”, Journal ofHazardous Materials, Vol. 80, pp. 15-30.

[9] Duijm, N. J, Carssimo, B., Mercer, A., Bartholome, C.,and Giesbrecht, H., (1997), “Development and Test ofan Evaluation Protocol for Heavy Gas DispersionModels”, Journal of Hazardous Materials, Vol. 56, pp.273-285.

[10] Counihan, J., (1975), ”Adiabatic Atmospheric

Boundary Layers: A Review and Analysis of the Datafrom the Period 1880-1972”, AtmosphericEnvironment , Vol. 9, pp. 871-905.


30/260


31/260

15

WIND TUNNEL STUDIES OF LNG VAPOR DISPERSION FROM

IMPOUNDMENTS

Jerry Havens, Tom Spicer, and Wendy SheppardChemical Hazards Research Center

Ralph E. Martin Department of Chemical EngineeringUniversity of Arkansas

Fayetteville, AREmail : [email protected]

Abstract - After briefly outlining the development ofthe present regulatory requirements for LNG vapor

dispersion modeling in support of LNG terminal

siting applications in the United States, this paperpresents data from recent wind tunnel experiments

designed to determine experimentally, using carbondioxide gas with density similar to LNG vapor, the

effects upon dispersion of an impounding dike, withor without an LNG tank centered inside the dike,

and compares these dispersion scenarios with thatwhich would result in the absence of either a tank or

dike. The results, while confirming an expectedreduction in downwind exclusion zone

requirements due to the presence of a dike andtank, indicate that the reduction in these tests is

primarily due to the tank rather than the dike.

However, the reduction in exclusion zone distance

in this data set, due to the presence of a tank anddike, appears to be masked by the atypically high

value of surface roughness used in the wind tunnel,suggesting that the relative effect of the tank would

be even greater if more typical, smaller, siteroughness were present. Surprisingly, the

measured distance to the endpoint concentrationwas shown to be greater when the dike (alone) is

present than when it is absent. This finding isexplained as resulting from the dike’s restriction of

gravity spreading, which decreases the overallplanar area for vertical mixing of the vapor cloud,

thus increasing the downwind distance to theendpoint concentration.


32/260


33/260

17

The Effect of Release Time on the Dispersion of a Fixed Inventory of

Heavy Gas – A Wind Tunnel Model Study.

D.J. Hall*, V. Kukadia S.Walker**, P. Tily, G.W. MarslandBuilding Research Establishment, * Now at Envirobods Ltd

Garston, Watford, Herts. ** Now at Dstl Porton Down WD25 9XX, UK Email: [email protected]

[email protected] ; [email protected]

Abstract – The paper describes wind tunnelexperiments of releases of a fixed inventory pf

dense gas, with release times varying from nearlyinstantaneous to times long enough to produce

continuous plume behaviour. This type ofexperiment more closely resembles many

accidental release scenarios and should lead tothe most extensive gas clouds occurring at

intermediate release rates. The experiments, forvarying gas release Richardson numbers, show

this to be generally true, with greatest downwind

distances to specific gas cloud concentrationsdistances occurring at intermediate bulk

Richardson numbers and release times betweenthe instantaneous and near continuous release

rates.

Key words – Dispersion, dense gases, fixedinventory, effect of release rates.

Introduction

Risk assessment studies for the dispersion ofaccidental releases of toxic and flammable gas cloudsare based either on direct experimental data or, morefrequently, on numerical models derived from this sortof data. Though there is a body of large scale

experimental field data of this sort, most experimentaldata is obtained using small scale models in windtunnels and water channels, providing both genericstudies and specific site investigations. The majorityof these measurements cover two types of release,‘instantaneous’ (see, for example, Spicer andHavens(1984) and Meroney and Lohmeyer(1983))and ‘continuous’ (see, for example, Koning–Langloand Schatzmann(1991)). ‘Instantaneous’ releasemodels mostly use the sudden exposure of a fixedvolume of gas, initially contained in a tent of somesort, to the prevailing wind conditions. Though therelease is instantaneous, it is also passive in thesense that it has no initial energy, following the‘Thorney Island’ type of field trial (Mcquaid and

Roebuck(1983)). Continuous releases are those forwhich the release time is long enough for the gasplume to have reached a steady state condition withno further development. In the case of heavier-than-air gas clouds this time can be considerable, theequivalent of hours in large scale releases.

The reality of most release scenarios issomewhat different. Usually there is a fixed inventory,a tank- or pipe-full of material under pressure forexample, and the variables in the release scenario arethe form of release, its direction, discharge velocityand the time over which the release occurs. Therelease time is the most critical

parameter since the release rate of the discharge is ininverse proportion to the time and in conventionalplumes the concentration is proportional to therelease rate. One might expect, therefore, that withreducing release time the extent of the plume to agiven boundary would increase. However, as therelease time reduces the amount of energy in therelease increases and for relatively short releasetimes starts to produce some initial mixing of the gascloud, which in turn reduces concentrations in thecloud. The balance between these effects impliesthat the most extensive gas clouds (in terms ofdistance to a given concentration) would occur at

some intermediate release time. Predicting this typeof behaviour is difficult as it falls between the twocommon types of experiment previously described, sothat there is a lack of a satisfactory experimental database on this type of release. Interpolating betweenthe two extreme conditions, for which data does exist,is unsatisfactory and at present unreliable.

The experiments described here address thisproblem directly and cover the release of a fixedinventory of gas over a range of release times. Theywere some of the first to do this and used a smallscale wind tunnel model, at about 1/100 of the scaleof the Thorney Island releases. Only a summary ofthe work is given here. Full details of the experiment

and it’s results can be found in Hall et al(1996).

1 Experimental Details and Scaling.

The experiments were carried out in the BREdispersion modelling wind tunnel, then at WarrenSpring Laboratory, which has a working section 22mlong, 4.3m wide and 1.5m high. A simulatedboundary layer for a surface roughness, z0, of 0.4mmwas used in the experiments. The gas cloud, of afixed volume, V, of 2000 ml, of gas was released intothe wind tunnel over a variable time, T., dischargedfrom a flush opening in the tunnel floor of 100mmdiameter (some exper iments using smallerdiameters). The gas discharge system (set below thetunnel floor) is shown in Figure 1. It used a rolling

diaphragm cylinder in which the gases each side ofthe piston are contained in leak proof rolling rubberdiaphragms, with the discharging gas contained in theupper cylinder by a commercial pinch valve, whichpermits filling the cylinder but allows a very free flowof the discharging gas. The gas cloud wasdischarged by pressurising the lower side of thecylinder and the discharge rate controlled by varyingthis pressure and the use of constricting orifices.Discharge times varied between 0.1 and 30s.Concentration/time measurements were made alongthe release centreline in the wind direction and bothlaterally and vertically at two downstream distances,of 150mm and 600mm. Ten repeat measurementswere made at each sampling point both to show the


34/260

18

repeat variability in this type of release and to obtainreliable mean concentration data.

The characteristic length scale, L, of the experimentwas defined as,

1/ 3L V .

The fixed discharge volume, of 2000 ml, gives a value

of L of 0.126m. The height for the reference windspeed, U, was also taken as L. The reference windspeed varied between 0.7 and 1.2 m s

-1. The release

volume and source size matched several previousexperiments using instantaneous releases of theThorney Island type, so that a direct data comparisonwith the earlier work was possible.

Time scales, T, were non-dimensionalised withrespect to the characteristic length scale, L, andreference wind speed U to provide the dimensionlesstime,

UT

L,

and the dimensionless release time,

r UTL

where Tr was the release time of the gas cloud.The release times, Tr , used, between 0.1 (the shortestthat could be readily managed) and30s, produced dimensionless release times betweenabout 0.5 and 250. The shortest values ofUTr /L corresponded to the external airflow travellingdistances less than the source diameter during thedischarge, close to an instantaneous release, thelongest times produced a near continuous plume inthe downwind direction in most cases.

The relative stability of the gas release was defined bythe bulk Richardson number Ri,

i 2LR g U

where, = g-a, where g is the density of thereleased gas and a is the ambient density. A value ofRi of zero represents a neutrally buoyant release anda value of 10 a very stable gas cloud exhibiting strongstability effects approaching those of still airconditions. The values of Ri used here were, 0, 0.5, 1,2, 5 and 10, obtained by varying both wind speed andsource gas density.

The source discharge momentum, m, is of interest asthe shorter release times produced significantmomentum. The dimensionless dischargemomentum, M, is defined as,

g 2

2 2

a

mM , where m w AU L .

w is the discharge velocity of the gas from the sourceof area A. Note that this momentum is relative to theambient density, a, not absolute, though thedifference is small since the ambient density of air isapproximately 1.2kg m

-3.

Table 1, shows the basic data for the experimentaltest conditions as an array of Richardson numbersand release times over which measurements weremade. Each square in these columns shows thedimensionless release time in bold and thedimensionless discharge momentum in italics. The

shaded squares are for conditions with negligibledischarge momentum.

2 Results of Experiments andAnalysis

Figure 2 shows examples of ground level

concentration/time traces for low (Ri=0.5) and high(Ri=5) Richardson numbers at a distance of 600mm(X/L = 4.8) from the source. The left hand pair ofplots are for a short release time (T r = 0.3s) and theright hand pair are for a long release time (Tr = 10s)Each plot shows the mean, 90%ile and 10%ileconcentration distribution derived from the ten repeatmeasurements. All subsequent data in this paper areof the mean values of the repeated measurementsonly. The traces are all consistent with other heavygas release data. The short time high Richardsonnumber release data shows the characteristic sharpinitial peak in concentration due to the gravity currentfront of the gas cloud from short term releases, whilethe other traces show a more Gaussian type ofdistribution, due either to longer release times or tolower Richardson numbers.

Figure 3 shows a complete set of ground level meanconcentration time traces for all the releaseconditions, measured 600mm (X/L = 4.8) from thesource centre, so the relative effects of theRichardson number and release time can be clearlyseen. Because of the initial release energy theshortest release time gave the lowest concentration atthe ground, the highest concentrations occurring atintermediate release times between 0.3s and 3s. Itcan be seen from the traces that the release timeapparently had a greater effect than the Richardsonnumber on the gas cloud. The traces pass from thosecharacteristic of instantaneous releases at short

release times to those characteristic of plumesegments at the longest release time of 30s. Therelease times for the highest concentrations in the gasclouds varied with the Richardson number. For theneutrally buoyant releases the highest concentrationin the gas cloud occurred for the 3s release time,while at higher Richardson numbers it occurred attime between 0.3s and 3s..

In practice the gas cloud concentrations shown inFigure 3 are additionally dependent on whether therelease became elevated due to the dischargemomentum. For the neutrally buoyant releases thiswas significant and the low concentrations were partlydue to elevation of the release. However, at thehighest Richardson number the gas cloud remained

largely on the ground and the low concentrations atthe shortest release times were mostly due toenhanced initial mixing of the gas cloud.

One of the main interests in dense gas clouddispersion is the downwind distance to a specific gascloud concentration. Figure 4 shows contour maps ofthe dimensionless distances (X/L) to the maximumand the mean gas cloud concentration at 1% and 10%of the source concentration. Both maximum andmean concentrations show similar behaviour. It canbe seen that the greatest distances for bothconcentrations occurred in dense gas clouds ofrelatively low Richardson number and intermediaterelease times.


35/260

19

Some facets of the gas cloud behaviour can be seenin Figure 5, which shows plots of ground level gascloud concentration with distance from the source forall the release times at four Richardson numbers. Ingeneral the release time has the greatest effect onground level concentration in neutrally buoyant

releases, and the least effect at the highest cloudRichardson number. For the shortest release timesall the dense gas clouds show an inflection in thereduction of concentration with distance, implyingsome elevated part of the gas cloud contacting theground further downwind. No confirmatory flowvisualisation was possible in these experiments, butthe fact that all the concentration distributionsotherwise show a falling concentration with increasingdistance suggest that a major part of the gas cloudalways remained in contact with the ground,irrespective of the discharge momentum. There issome evidence from the vertical concentrationdistributions of the cloud that it partly bifurcated intoairborne and ground based components.

3 Discussion and Conclusions

1. The principle outlined at the start of this paper, thatthe most extensive gas clouds from adischarge of a fixed inventory occur atrelease times intermediate betweeninstantaneous and continuous, is borne outby the experiments. Nearly all the measuredparameters showed this type ofcharacteristic.

2. The most extensive gas clouds occurred atrelatively short release times, correspondingto values of the dimensionless release timearound 2-10, the time reducing withincreasing Richardson number.

3. The main cause of this effect was the initial mixingof gas clouds generated by their releaseenergy when the release time was short,which reduces the source concentration.

4. The overall gas cloud behaviour was the result of acomplex interaction between release rate asa generator of initial mixing, release rate as aregulator of concentration and the size of thegas cloud, gas cloud stability (Richardsonnumber) source size and the details of therelease.

5. The basic release showed a mixture of omni-directional and vertical discharge whichresulted at times in two-component plumes.These could generate sudden increases inconcentration at the ground as the two partsof the plume came together.

6. For longer release times, the effect of the dischargemomentum diminished and over about halfthe of the measurements, the lower right halfof the conditions laid out in Table 1, it isdoubtful whether the details of the release,other than the release time, were of anyparticular importance to the resultant plumedispersion.

7.There is a complete archive of this data setavailable for further investigations.

Acknowledgments

The work described here was jointly funded by the UKHealth and Safety Executive and the European

Commission as part of the FLADIS (CT90-0125)project in the Major Technical Hazards Programme.

The experiments were carried out a the WarrenSpring Laboratory prior to its closure and the dataanalysis and report completed at the BuildingResearch Establishment, to which two of the authorsand the dispersion wind tunnel transferred.

References

Hall D.J. Kukadia V., Walker S., Tily P, MarslandG.W.(2006) “The Effects of release time on thedispersion of a fixed inventory of heavier-than-airgas – A wind tunnel model study. BuildingResearch Establishment, UK. Report No.

CR149/96.Hall D.J. and Waters R.A.(1989). Investigation of Two

Features of Continuously Released Heavy GasPlumes. Warren Spring Laboratory, UK. ReportNo.LR707(PA).

Havens J.A., Spicer T.O.(1983). “Gravity Spreadingand air Entrainment by Heavy GasesInstantaneously Released Into a Calm Atmosphere”. IUTAM Symposium on‘Atmospheric Dispersion of Heavy Gases andSmall Particles’, Delft, The Netherlands, August29-September 2cnd. Springer-Verlag. ISBN 3-540-13491-3.

Konig-Langlo G., and Schatzmann M.(1991). “WindTunnel Modelling of Heavy Gas Dispersion”. Atmospheric Environment , Vol. 25A, No. 7, pp.1189-1198.

McQuaid J.D., Roebuck B.(1984). “ Large Scale FieldTrials on Dense Vapour Dispersion. Final Reportto Sponsors on Heavy gas Dispersion Trials atThorney Island, 1982-1984”. Health and SafetyExecutive, Sheffield, UK.

Meroney R.N., Lohmeyer A.(1983). “StatisticalCharacteristics of Instantaneous Dense GasClouds Released in an Atmospheric BoundaryLayer Wind Tunnel”. Boundary LayerMeteorology . Vol. 28. pp. 1-22.


36/260

20

Table 1. Operating conditions for experiments.DImensionless release time, UT r /L, in boldand dimensionless discharge momentum, M,in italics.

- Additional measurements made usingReduced Diameter Source# Additional Measurements made using

Source Fitted with Raised Circular Cover

Shaded area represents approximate region wheresource discharge momentum effects are small.

Release Time, Tr (s)Ri Gas g /a /a

U

m s-1

0.1 0.3 1 3 10 30

0 Air 1 0 0.970.773.4#

2.310.38

7.70.034

#

23.10.0037

770.0003

2310.00004

0.5 Argon 1.38 0.38 0.970.774.7

2.310.52

7.70.047

23.10.0052

770.0005

2310.00005

1 Argon 1.38 0.38 0.690.559.12 #

1.64

1.03

5.470.091

16.40.010

54.70.0009

1640.00001

2

BCF/Air

50/50 3.37 2.37 1.207

0.96

7.3

2.87

0.82

9.6

0.073

28.7

0.0082

96

0.0007

287

0.00008

5 BCF 5.74 4.74 1.080.8615.5

2.570.1.74

8.570.15

25.70.017

85.70.0016

2570.00017

10 BCF 5.74 4.74 0.770.6131#

1.833.42

-

6.10.31

18.30.034

610.0031

1830.0003


37/260

21

Figure 1. Gas cloud discharge apparatus

Figure 3. Mean concentration/time traces for allgas cloud release conditions, measured atthe ground 600mm (X/L=4.8) from thesource.

Figure 2. Examples of centreline ground level gascloud concentration/time tracesmeasurements. 600mm (X/L=0.=4.8)downwind of the source.Left side T r = 0.3s, right side T r = 10s.


38/260

22

Figure 4. Contours of dimensionless distance (X/L) to 1% and 10% maximum (left) and mean (right) concentrationas a function of release Richardson number and release time.

Figure 5. Plots of maximum ground level gas cloud concentration against distance for four Richardson numbersand all gas cloud release times.


39/260

23

Quality assurance of micro-scale meteorological models –

Action COST 732

M. Schatzmann, B. Leitl

University of Hamburg, Meteorological Institute (ZMAW)Hamburg, Germany

[email protected]

Abstract – A COST action has been launched which

is tasked to assure the quality of micro-scale

meteorological models that are applied for

predicting flow and transport processes in urban or

industrial environments. Main objective of the action

is to develop a coherent and structured model

evaluation procedure. Part of the procedure is the

provision of appropriate validation data. The talk

will focus on validation data requirements and the

role of boundary layer wind tunnels within this

context.

Key words – quality assurance, obstacle resolvingnumerical models, field and wind tunnel data, urbancanopy layer .

Introduction

The emergence of increasingly powerful computersenabled the development of tools that have the potentialto predict flow and transport processes within the urbancanopy layer. These new tools are micro-scalemeteorological models of prognostic or diagnostic type.Prognostic models are based on the Reynolds-averagedNavier-Stokes (RANS) equations, whereas diagnosticmodels are less sophisticated and ensure only theconservation of mass. These two model types are

presently supplemented by even simpler engineeringtools. It is to be expected, however, that the latter willsooner or later be replaced by Computational FluidDynamics (CFD) tools such as RANS codes or the evenmore complex Large Eddy Simulation (LES) models.

Micro-scale meteorological models are special in sofar as they are tailored to the needs of meteorologists.They are adjusted to domain sizes of the order ofseveral decametres to a few kilometres (street canyons,city quarters). They usually use boundary conditionsbased on surface characteristics like land use,roughness and displacement thickness and they maycontain modules that have the potential to simulatechemical transformations, aerosol formation or otherimportant atmospheric physico-chemical processes.Typically these models contain a substantial amount ofempirical knowledge, not only in the turbulent closure

schemes but also in the use of wall functions and inother parameterisations.

Models play an important and often dominant role inenvironmental assessment and urban climate studiesthat are undertaken to investigate and quantify theeffects of human activity on air quality and the localclimate. Their increasing use is paralleled by a growingawareness that the most of these models have neverbeen subject to rigorous evaluation. Consequently thereis often a lack of confidence in the modelled results.

Objective and Methodology

The main objective of action COST 732 is to improveand assure the quality of micro-scale meteorologicalmodels that are applied for predicting flow and transportprocesses in urban or industrial environments. Inparticular it is intended

to develop a coherent and structured qualityassurance procedure for this type of model thatgives clear guidance to developers and users asto how to properly assure their quality and theirproper application,

to provide a systematically compiled set ofappropriate and sufficiently detailed data formodel validation work in a documented,convenient and generally accessible form (wwwdata bank),

to invite from all participating states modeldevelopers and users to apply the procedure andto prove its serviceability,

to build a consensus within the community ofmicro-scale model developers and usersregarding the usefulness of the procedure,

to stimulate a widespread application of theprocedure and the preparation of qualityassurance protocols which prove the ‘fitness forpurpose’ of all micro-scale meteorologicalmodels participating in this activity,

to contribute to the proper use of models bydisseminating information on the range ofapplicability, the potential and the limitations ofsuch models,

to establish a consensus on ‘best practises’ incurrent model use and

to give recommendations for focussedexperimental programmes in order to improvethe data base.

It is to be expected that the very existence of awidely accepted European standard for qualityassurance in the field of micro-scale meteorologicalmodels in combination with the provision of suitablevalidation data will significantly improve “the culture”within which such models are developed and applied.European model developers shall find step-by-stepguidance on how to demonstrate that their models are fit

for a particular purpose. Data sets (both flow andconcentration data) obtained from extensiveexperiments will be made accessible and more widelyexploited. Relevant expertise available within themember states will be brought together and combined todevelop a consensus for appropriate model use andmodel improvement.


40/260

24

Results and Discussion

The action started in July 2005 with a jointESF/COST 732 Exploratory Workshop on ‘Quality Assurance of Micro-Scale Meteorological Models’ inHamburg. About 45 scientists from Europe and the US(the number of participants was limited in order to allowample discussions) attended the workshop. The

workshop proceedings (Schatzmann and Britter 2005)contain a state of the art report on former qualityassurance initiatives in the field of micro-scalemeteorological models. These initiatives comprise the'General Requirements for a Quality Assurance ProjectPlan' by Borrego and Tchepel (1999), the 'Guidelines forModel Developers' and the 'Model Evaluation Protocol'which were worked out by the Model Evaluation Group(MEG, 1994) under the CECs Major Industrial HazardsProgramme, the US-Environmental Protection Agency’srequirements for quality assurance of atmosphericdispersion models (Irwin, 1998 and 1999) and theexperience gathered within the initiative forharmonization of atmospheric dispersion modelling forregulatory purposes (Olesen, 1999 and subsequentpapers). Results from similar initiatives in related fieldswere also taken into account, for example from the

investigations carried out within the ‘Podbi’-model inter-comparison exercise (Lohmeyer et al., 2002), from theFP5 project EMU (Hall 1997), the thematic networkQNET-CFD or COST Action C14 which dealt with theindustrial application of CFD codes for engineeringapplications. Finally, the recommendations given bynational bodies, e.g., the Quality Assurance Guidelinesreleased by a task force of UKs 'Royal MeteorologicalSociety' (1995) and by Germanys 'VDI Commission onClean Air' (2002), were carefully considered. Withrespect to data the considerations outlined inSchatzmann et al (2002, 2003) were taken into accountand standards for validation data were defined whichcan generally only be met by data sets based on acombination of field and wind tunnel experiments.

Strategies for assuring the quality of a numerical

model can only be based on very generic scientificprinciples such as the principle of falsification (K. R.Popper, 1959). The decision about which particulartests should be performed and which particular datasets should be used for comparisons between modelresults and observations can ultimately be only basedon a consensus built up within and by the scientificcommunity. The impact of COST 732 is dependent onwhether the quality assurance procedures suggested bythe Action are accepted by the community of modeldevelopers and users or not. Therefore, the next logicalstep was to draft a first version of the evaluationprocedure and its underlying motivation in order toprovide the basis for subsequent discussions within thescientific community. This was done in form of tworelated documents: A rather lengthy

Background and justification document to

support the model evaluation guidance andprotocol document (Britter, R., and Schatzmann,M. 2007 a) and a much shorter

Model evaluation guidance and protocoldocument (Britter, R., and Schatzmann, M. 2007b).

The first document contains detailed explanationsconcerning the general model evaluation philosophy andthe sequence of tasks that should be completed. Thesetasks are

Model description: this should be a briefdescription of the characteristics of the model,the intended range of applicability, the theoreticalbackground on which the model developmentwas based, the software and hardwarerequirements, etc.

Database description: a complete description ofthe database that is to be employed for the

evaluation of the model, including the reasonswhy this specific database was chosen. Anestimation of the data variability is required.

Scientific Evaluation: this is a description of theequations employed to describe the physical andchemical processes that the model has beendesigned to include. If appropriate it should justify the choice of the numerical modellingprocedures and it should clearly state the limitswith respect to the intended applications.

(Code) verification: this process is to verify thatthe model produces results that are inaccordance with the actual physics andmathematics that have been employed. This is toidentify, quantify and reduce errors in thetranscription of the mathematical model into a

computational model and the solution (analyticalor numerical) of the model.

Model validation: this is a structured comparisonof model predictions with experimental data andis based on statistical analyses of selectedvariables. It seeks to identify and quantify thedifference between the model predictions and theevaluation datasets; it provides evidence as tohow well the model approximates to reality. Aquantification of the uncertainty of the modelpredictions should be produced.

User-oriented assessment: is there a readable,comprehensive documentation of the codeincluding technical description, user manual andevaluation documentation? The range ofapplicability of the model, the computingrequirements, installation procedures, and

troubleshooting advice should be available.Five of the steps of the evaluation procedure

described above are relatively straightforward but themodel validation is complex and requires more attention.Unfortunately this has led to the often-seen modelevaluation study that is no more than the validation step. At the heart of the complexity of the model validationprocess is the stochastic nature of atmospheric flows,whether real or physically modelled. For example, andprior to any comparison between mathematical modeland experimental results, the user or model evaluatorneeds to address issues such as:

Which quantities should be compared?

At which point within the area of interest shouldthe comparison take place?

Should the comparison take place on a point-to-point basis or on an area averaged basis?

Should the compared quantities be averagedover a specific period of time and if so what is thetime over which the averaging should take place?

Should the quantities be compared at the sametime or at different t imes?

The answers to these questions become clearerwhen the purpose of the model is precisely stated. Thevarious metrics to be used need to be carefully selectedand agreed upon. Experience has shown that there maybe some generally expected values for these metrics for


41/260

25

“state of the art/science” models when applied toparticular data sets subject to a specified protocol.

A special section is devoted to validation datarequirements. For the validation of micro-scalemeteorological models a suite of data sets withincreasing geometrical complexity is needed that allowssystematic testing of numerical codes. The data setsmust be ‘complete’, i.e. they must contain sufficient

information to set up a model run without furtherassumptions concerning the model input parametersand the uncertainty of the data must be known. It isexplained that the uncertainty of field data cannot easilybe quantified based on the results of fieldmeasurements alone. It is not just the accuracy of theinstrumentation used for field measurements thatdefines the reliability of field data. In addition, therepeatability of field measurements for similar boundaryconditions as well as the spatial representativeness ofindividual measurement locations with respect to aparticular flow and dispersion problem must beevaluated and quantified with respect to the measuredquantities before corresponding data can be used safelyfor model validation purposes. This is why COST 732suggests validation data sets that always comprisecombinations of field and laboratory experiments. Thebackground document closes with a glossary of termssince words like ‘validation’, ‘verification’, ‘evaluation’,‘quality assurance’ etc. are not unambiguously definedand used

Date post:	07-Aug-2018
Category:	Documents
Upload:	janos-kovacs
View:	218 times
Download:	0 times

Proceedings PHYSMOD 2007

Documents