Large Eddy Simulation Of Microburst Winds Flowing Around A Building

Journal of Wind Engineering and Industrial Aerodynamics,46 & 47 (1993)229-237Elseviert(229!;1Large eddy simulationaround a buildingof microburstwindsflowingM. NICHOLLSResearchAssociate, Department of Atmospheric Science,Colorado State University,Foothills Campus,Fort Collins, CO 80523,U.S.A.R. PIELKEProfessor, Department of Atmospheric Science,Colorado State University, FoothillsCampus,Fort Collins, CO 80523,U.S.A.R. MERONEYProfessor,Civil Engineering,ColoradoState University,Foothills Campus,Fort Collins,CO 80523,U.S.A.AbstractA large eddy simulation of a microburstproducing thunderstormis carried out. Thethunderstorm is initiated when a thermal within a developingmixed layer reachesthelifting condensationleveland strong latent heating occurs. A microburstis subsequentlyproducedas condensatefrom the thunderstormfalls beneaththe melting level. In thisstudy, the viability of using two-wayinteractive multiple nestedgrids to investigatetheinteraction of the outflow from the microburst with a much smallerscalearchitecturalstructure is investigated. The relationship of the fluctuating winds around the buildingto the microburst structure is described.1. INTRODUCTIONIt has been estimated by Thorn [1] that about one third of the extreme winds recordedin the U.S. are associated with thunderstorms. Outflows, which are often responsiblefor the extreme winds from thunderstorms, vary in scale. A small scale outflow whichproduces damaging winds near the surface has been termed a micro burst by Fujita [2].Larger scale outflows from thunderstorms can also cause damaging winds. For this preliminary study, an environment conducive to the formation of microbursts is considered.As discussed by Selvam [3], a downdraft wind profile is different from a developed boundary layer profile. Profiles in downdrafts often display a maximum close to the ground(",100 m) with lower velocities in the upper part of the outflowing layer of cold air.Moreover, downbursts are highly transient phenomena, and it is of some interest to determine the structure of the fluctuating winds associated with them and how this effectsthe wind loading on a building.Numerical simulations of microbursts using an axi-symmetric model have been madeby Proctor [4]. Downdrafts were initiated by specification of a distribution of precipitation at the top boundary of the model and allowing it to fall into the domain. In thispresent study, a microburst producing thunderstorm is initiated by a thermal within a0167-6105/93/ 06.00@ 1993 -Elsevier SciencePublishersB.V. All rights reserved.

230developing mixed layer, in order to obtain a more realistic fluctuating wind than mightbe obtained from a more idealized initialization procedure.It has been demonstrated by Murakami et al. [5] that a large eddy simulation modelcan successfully replicate many of the observed features of the flow around a building.Using a similar building model to Murakami et al., we investigate the wind flow around abuilding produced by a microburst. In order to simulate this phenomena which covers awide range of scales, two-way interactive multiple nested grids are utilized. This enablesthe formation of the thunderstorm and microburst to be simulated on a coarse grid,after which, successively finer grids are used to "telescope down" to the small scale flowaround a building.2. MODELThe model used in this study is the Colorado State University-Regional AtmosphericModeling System (CSU-RAMS). The model contains a full set of nonhydrostatic compressible dynamic equations for water and ice-phase clouds and precipitation. The modelhas two-way interactive multiple nested grid capability [6], which makes it particularlysuited to the simulation of phenomena which cover a wide range of scales. Acoustic termsare integrated with a small time step and low-frequency terms with a large time step.The equations are solved on a standard velocity staggered grid described by Tripoli andCotton [7]. A first order eddy viscosity subgrid scale parameterization is used. The upper boundary is a rigid lid. A Rayleigh friction layer was incorporated at the upper-mostlevels to prevent reflection of gravity waves from the top of the domain. The surfaceparameterization of momentum fluxes is based on the Louis scheme [8]. The reader isreferred to Nicholls et al. [9] for a description of the microphysical parameterizationsused in the model and to Pielke et al. [10] for a general discussion of model applications.For the building, the component of velocity normal to the surface is set to zero. At thecorners of the building a vorticity constraint is used for the calculation of the advectiveterms. The profiles of the tangential velocity components were assumed to obey a powerlaw expressed asU Q ,fi, as in Murakami et al. [5].3. EXPERIMENTAL DESIGNFor this preliminary experimentthe domainis two dimensional.Five grids are usedwith grid incrementsof 202.5,67.5,22.5,7.5 and 2.5 meters. Eachgrid has 90 horizontalgrid points and 60 vertical grid points, exceptfor the coarsestgrid which has 70 verticalgrid points. The Rayleigh friction layer is included at the uppermost10 levels of thecoarsestgrid. The lateral boundariesare periodic. A schematicof the nestedgrids isshownin Fig. 1. Each successivelyfiner grid is centeredwithin the next coarsestgrid.Eachgrid extendsto the surface.The building is centeredwithin the finest grid and has20x20 grid points.The sounding used in this experimentis shown in Fig. 2. It is based on the 2300UTa 2 August 1985,Dallas-Ft. Worth, TX microburst soundingused by Proctor [4].Small randomly distributed temperatureperturbations ( 0.2 k) are introduced at thelowest level abovethe surfaceto initiate thermals. In order to have some control overthe position of the thunderstorma slightly larger temperature perturbation of 0.4 K isintroduced at a specifiedlocation.

231Figure 1: Schematic of nested grids.Figure 2: Soundingusedto initialize model.

2324. RESULTSFigure 3a and b showthe vertical velocity and total condensate,respectively,at t 4800 s, for Grid 1. There is a deep convectivecell and shallowerthermals at lowlevels. The deep convectivecell developedwhenone of the thermals in the mixed layerreachedthe lifted condensationlevel. During the next 600 s the downdraft within thelower regionof the cloud developsinto a downburstproducinga strong outflow near thesurface.12.0 1 . 1 8.001N 6.0 a4.00-12.00-1 l00 1".-8.00-6.00-4.00-2.00.00x (km)bCondensate2.004.006.008.0 MRFigure 3: Fields at t 4800 s, for Grid 1: (a) Vertical velocity. The contourinterval is2 m S-l. Dashedisoplethsindicate negativevalues in this and subsequentfigures. (b)Total condensate.The contour interval is 1 g kg-1 and the label scaleis 10.

.,233Grids 2, 3, 4 and 5 were activated at t 4500 s. Figure 4a, b and c showstheperturbation temperature,perturbation pressureand horizontal velocity,respectively,att 5400 s, for Grid 2 (perturbations are from the initial state). A cold outflow canbe seennear the surface. An anti-clockwiserotating vortex occursbehind the leadingedgeof the outflow with a pressureminimum of -1.65 mb. The maximum horizontalwind velocities occur at -100 m above the surface. The building is centeredat x 0.3 km and is evidencedby negativehorizontal velocities which occur just behind it.The hemisphericregionof warm air at x 1.5 km, z 3 km is a cumuluscloud whichdevelopedat the top of an upward moving thermal. Figure 5 showsthe streamlinesat t 5400s, for Grid 3. This showsthe microburst vortex centeredat a height of 0.8 kmand a smallerclockwiserotating vortex which has beenshed from the building. Thereare two finer grids which further "telescopedown" in scale. Figure 6a and b showthestreamlinesand perturbation pressure,respectively,at t 5400 s, for the finest grid.The streamlinesshow separationhas occurred at the windward cornerof the building.The clockwiserotating eddy which formed behind the building, seenin Grid 3, has already advectedout of the fine grid. A smalleranticyclockwiserotating eddyhasformedjust behind the building. Smallereddiesoccur abovethe roof and at the front of thebuilding near the surface. A pressureminimum occurs at the windward corner. A lowpressureoccurs over most of the building which is mainly due to the pressureminimaassociatedwith the microburst vortex whichis positionedabovethe building at this time.For the simulated micro burst there are actually two wind maxima which occur nearthe surface. The first strong wind gust is at the leading edge of the outflow. The secondwind maxima is directly beneath the low pressure center aloft. This is consistent withthe modeling study by Droegemeier and Wilhelmson [11].5. CONCLUSIONSIn this study, it has beenshownthat two-wayinteractivemultiple nestedgrids shouldprove a valuable tool for investigatingthe interactionof . It can be seenhow the fluctuating winds around a building arerelated to the structure of the microburst. The fluctuating surfacewinds can dependonthe type of thunderstormoutflow. For instance,if the outflow is cold and deep KelvinHelmholtz instability is likely to occur. For a stably stratified lower atmosphere,gravitywavelikeperturbations may coexist with the density current and eventuallypropagateaheadof it, as discussedby Fulton et al. [12]. Further investigationof downburstsinteracting with buildings needsto be undertakenin a three-dimensionalframework. Thetime scalefor the developmentof a separationlayerand eddiesadjacentto the surfaceofthe building is dependenton the surfacefriction parameterization. Further refinementof this parameterizationand comparisonwith observationsneedsto be made. This typeof analysiscould alsobe extendedto study the interactionof buildings with a simulatedtornado (Pielke et al., [10]).

234Figure 4: Fields at t 5400 s, for Grid 2: (a) Perturbation temperature. The contourinterval is 0.4 k. (b) Perturbation pressure. The contour interval is 10 Pascals. (c)Horizontal velocity. The contourinterval is 1 m S-l,

r,,.""."'.".'235-'-j-",";'' : -s'3.5e1 .''\3. -12.5E 2.0N"';/"0.'rn 15 ,.'I1.52.50),.,''-.' - -,1.0c'.,-,".I'""-2. 1.00.00-1.00x2. 3(kmFigure 4: Continuede 6040200020(km40608000Figure 5: Streamlinesfor Grid 3, at t 5400s.

236-E a2025.3ax35(km)Figure 6: Fields at t 5400 s, for Grid 5: (a) Streamlines. (b) Perturbation pressure.The contour interval is 10 Pascals.