Coupling of WRF and building-resolving urban CFD models for analysis of strong winds over an urban area

Size: px

Start display at page:

Download "Coupling of WRF and building-resolving urban CFD models for analysis of strong winds over an urban area"

Andrew Scott
5 years ago
Views:

1 .6 Couplng of WRF and buldng-reolvng urban CFD model for analy of trong wnd over an urban area Hromaa Nakayama *, Tetuya Takem 2 and Haruyau Naga Japan Atomc Energy Agency, Ibarak, Japan 2 Dater Preventon Reearch Inttute, Kyoto Unverty, Kyoto, Japan. INTRODUCTION The hghly varable and complex phenomena of atmopherc flow are characterzed by manly two factor: large-cale meteorologcal dturbance and mall-cale wnd fluctuaton produced by urface terran and roughne element. Accordng to the obervatonal tudy of Van der Hoven (957), the pectrum of horzontal wnd peed near the ground urface how a large peak near perod of 4day and a mall peak near a perod of mnute. The former due to large-cale atmopherc moton and the latter due to turbulence. Partcularly, for denely bult urban area covered wth hghly rough ground urface, hgh-re buldng have gnfcant nfluence on the mall-cale fluctuaton of atmopherc flow. In cae of trong wnd nduced by meteorologcal dturbance, the occurrence of guty wnd over uch urban area hould be condered n term of dater preventon and urban plannng. For undertandng the wnd ytem over urban area, a numercal modelng a ueful tool. To mulate and forecat atmopherc flow n real meteorologcal ettng, numercal weather predcton (NWP) model are commonly ued. Although the accuracy of NWP model for daly weather contnuouly mprovng, t dffcult to reproduce mall-cale fluctuaton nduced by urban roughne element that are not explctly repreented n the NWP model. For mulatng wnd flow accountng for the urban urface geometre, a computatonal flud dynamc (CFD) technque one of the commonly *Correpondng author addre: Hromaa Nakayama, Japan Atomc Energy Agency, Ibarak, Japan; e-mal: nakayama.hromaa@aea.go.p ued approache. In the CFD model, urban urface geometre can be explctly repreented at hgh reoluton. In partcular, LES-baed CFD model ha been regarded a an effectve tool wth the rapd development of computatonal technology. LES model can reproduce mall-cale wnd fluctuaton uch a turbulent behavor around obtacle. Therefore, an approach to couple the LES-baed CFD and the NWP model hould be promng to mulate trong wnd over actual urban area under real meteorologcal condton. In order to couple NWP and CFD model, the NWP output can be ued a the ntal and boundary condton of a CFD model. Here, a erou ue encountered when mpong unteady turbulent nflow data for LES from the NWP output, becaue the NWP model are not able to reproduce mall-cale turbulent fluctuaton. Therefore, a proper couplng technque hould be appled, conderng the generaton of turbulent nflow for LES. To generate effectvely turbulent nflow, the recalng technque of Lund et al. (998) ueful. They produced realtc turbulent fluctuaton by recalng the velocty feld at a downtream taton and re-ntroducng at the nlet n the pecal doman. In th tudy, we extend the extng turbulent nflow technque to couple the CFD and NWP model. We conduct a buldng-reolvng LES of trong wnd over the central dtrct of Tokyo durng the paage of a maor typhoon and examne the uefulne of our approach by comparng the LES reult wth the obervaton.

2. NUMERICAL MODEL 2.2 MESOSCALE METEOLOROGICAL SIMULATION MODEL The Weather Reearch and Forecatng (WRF) model, the Advanced Reearch WRF Veron 3.. (Skamarock et al.

2 2. NUMERICAL MODEL 2.2 MESOSCALE METEOLOROGICAL SIMULATION MODEL The Weather Reearch and Forecatng (WRF) model, the Advanced Reearch WRF Veron 3.. (Skamarock et al. 2008) ued for a meocale meteorologcal mulaton. We ue a netng capablty to reolve the Tokyo regon at a fne grd pacng by ettng one-way neted, four computatonal doman (wth the top beng at the level of 50 hpa). The four doman cover area of 800 km by 900 km at 4.5-km grd, 270 km by 300 km at.5-km grd, 93 km by 93 km at 300-m grd, and 25 km by 30 km at 60-m grd, repectvely (Fg. a-d). The number of vertcal level 43, wth 5 level n the lowet -km depth. The terran data for the modeled topography are the global 30-econd data (GTOPO30) from the U.S. Geologcal Survey for the outer 2 doman and the 50-m meh dgtal elevaton dataet by the Geographcal Survey land-ue and land cover nformaton obtaned from the 00-meh dataet from the Mntry of Land, Infratructure, Tranport and Tourm of Japan. A the ntal and boundary condton, we ue 6-hourly Meocale Analy (MANAL) data of Japan Meteorologcal Agency (JMA), 6-hourly Fnal Analy data of the U.S. Natonal Center for Envronmental Predcton, and daly Merged Sea Surface Temperature (MGDSST) analye of JMA. The horzontal reoluton of MANAL and MGDSST are 0 km and 0.25 degree, repectvely, whch are ueful for hgh-reoluton meocale mulaton. Full phyc procee are ncluded n the preent mulaton n order to reproduce real meteorologcal phenomena. A phyc parameterzaton that cloely relevant to the mulaton of wnd feld a PBL mxng parameterzaton. We chooe a Mellor-Yamada Level 2.5 cheme of Janc (2002) n whch the ued only for the outermot doman, and a vertcal A B C Fgure : Computatonal area of the neted WRF model for (a) the 4.5-km grd, (b) the.5-km grd, (c), the 300-m grd, and (d) the 60-m grd doman and of the CFD model for (e) the 20-m grd and (f) the 5-m grd. The nflow boundary of the CFD model on the left. The color hadng n (a)-(d) ndcate the urface elevaton caled by the maxmum heght n each doman (2409 m n (a); 3285 m n (b); 255 m n (c); and 54 m n (d)). The whte rectangular n (a)-(d) ndcate the area of the chld doman. The color hadng n (e) and (f) ndcate the heght of the buldng and tructure. The pont A, B, and C n (e) repreent the locaton ued n Fg. 2. The yellow crcle n (f) ndcate the locaton of the wnd Inttute obervaton of Japan te n for Fg. the 3. nner 2 doman. The mxng done vertcally between the adacent

3 level. A Kan-Frtch cumulu cheme warm-ran and ce-phae mcrophyc cheme employed for cloud and precptaton procee n all the doman. The cae tuded here a hgh-wnd event n Tokyo durng the paage of Typhoon Melor (2009) that attaned the central preure of 90 hpa and the maxmum 0-mn averaged wnd of 55 m - at t maxmum ntenty on 4 October 2009 and made landfall on the Japan coat about 280 km wet of Tokyo at around 2000 UTC 7 October. The maxmum ntantaneou wnd peed recorded n Tokyo wa 30.2 m - at 2339 UTC 7 October. In order to mulate wnd feld for th event, the computaton for the outermot doman ntalzed at 0000 UTC 6 October 2009, whle the mulaton for the 2 nnermot doman are ntalzed at 800 UTC 7 October. The mulated output of the nnermot doman at -mn nterval are ued a the nput of a CFD model. 2.2 LES-BASED CFD MODEL The CFD model baed on the LES model developed by Nakayama et al. (20). The governng equaton are the fltered contnuty equaton and the fltered Naver-Stoke equaton, a follow: u = 0, () u u p u u u + = + ν + τ + f t x x x (2) ρ τ = uu uu (3) τ δ τ kk = ν 3 S = and Δ = SGS S ν ( u + u )/ 2 ( Δ Δ Δ ) 3 x y z SGS = 2 ( C f Δ) ( 2 S S ) 2 SGS (4) (5) (6) * where u, t, p, ρ, τ, δ, ν, ν, and u are wnd velocty, tme, preure, denty, ubgrd-cale Reynold tre, Kronecker delta, knematc vcoty, eddy vcoty coeffcent, and frcton velocty, repectvely. Subcrpt and tand for coordnate (treamwe drecton: x = x, panwe: x 2 = y, and vertcal: x 3 = z ). Varable wth an upper bar denote patally fltered one. In th tudy, the tandard Smagornky model (Smagornky 963) employed becaue of t mplcty and low computatonal cot. C et to 0.. f the Van Dret dampng functon (Van Dret 956). Δ denote grd-flter wdth. The body force, f, ncluded n the Naver-Stoke equaton n order to ncorporate the effect of buldng on flud flow. The feedback forcng formulaton by Goldten et al. (993) ued to repreent th body force and expreed a follow: t α f = u ( t ) dt + βu ( t), α < 0, β < 0 0 (7) where α and β are negatve contant. The tablty lmt gven by β ( β 2 2αk ) Δt < contant value of order. α and k a The buldng n the central dtrct of Tokyo are explctly repreented by a dgtal urface model dataet at 2-m reoluton. Two computatonal doman are ued. The ze of the doman (e) where the urban urface geometry explctly reolved 5.0 km (treamwe) by 2.0 km (panwe) wth the depth of.5 km (Fg. e), wth buffer zone wth a 500-m length beng placed at the up- and down-tream of the buldng-reolved area. Thu, the length of the man analy doman 6.0 km. The total meh number 300 by 00 by 80 node. The grd pacng 20 m n the horzontal drecton and m tretched n the vertcal drecton. The ze of the doman (f) where the ndvdual urban buldng are explctly reolved.0 km (treamwe) by.0 km (panwe) wth the depth of.5 km (Fg. f), wth buffer zone wth a 500-m length beng placed at the up- and down-tream of the buldng-reolved area. Thu, the length of the man analy doman

4 3.0 km. The total meh number 300 by 200 by 80 node. The grd pacng 5-m n the horzontal drecton. The grd pacng n the vertcal drecton the ame a the one n the LES model doman (e). The couplng algorthm of the velocty and preure feld baed on the MAC method (Chorn 967) wth the Adam-Bahforth cheme for tme ntegraton. The tme tep nterval 0.05 econd. The Poon equaton olved by the SOR method. For the patal dcretzaton, a econd-order-accurate central dfference ued. The boundary condton wthout applyng the WRF output are: a Sommerfeld radaton condton (Greho 992) at the outflow boundary; a free-lp condton for the horzontal velocty component and zero-peed condton for the vertcal velocty component at the upper boundary; a no-lp condton for each velocty component at the bottom urface; and a perodc condton at the panwe lateral boundare. On the other hand, the nflow boundary condton of the LES model doman (e) determned by the temporally and patally varyng wnd component nterpolated on the LES reoluton from the WRF model output at -mn nterval and 60-m grd pacng. The nflow boundary condton of the CFD model doman (f) determned by the temporally and patally varyng wnd component nterpolated on the LES reoluton from the CFD model (e) output at 0.05-ec nterval and 20-m grd pacng. 2.3 COUPLING WRF AND URBAN CFD MODELS The WRF model cannot reproduce turbulent fluctuaton becaue the ndvdual urban buldng and obtacle are not explctly reolved. To nget the WRF output for a buldng-reolvng LES, turbulent fluctuaton nduced by urban roughne element hould be added to the WRF wnd flow. The recalng approach of Lund et al. (998) condered to be ueful n term of both avng computatonal reource and phycal contency to boundary-layer dynamc. However, th technque not approprate to mulate atmopherc flow under real meteorologcal condton becaue the mean flow actually change wth tme owng to the meteorologcal varaton. Therefore, we extend the method of Lund et al. by takng nto account the temporal and patal varaton of the mean flow at the nflow boundary of the LES model. The WRF output durng 2300 UTC 7 October and 0000 UTC 8 October are ued for the preent LES. 3. RESULTS The mulated central preure ut before the landfall at around 2000 UTC 7 October wa 953 hpa, whch well agree wth the correpondng bet-track value of 955 hpa. In addton, the oberved track of Typhoon Melor (2009) wa well reproduced n the outermot doman of the WRF model. Thu, the WRF model condered to uccefully mulate the track and ntenty of Typhoon Melor (2009) before and durng the landfall on 7 October 2009, whch ndcate that the WRF output for ue n the preent LES reflect the overall feature of the trong wnd nduced by the typhoon. Fgure 2 the treamwe varaton of the vertcal profle of wnd peed from the WRF and the LES model. The pont A, B, and C are located at.0 km, 2.7 km, and 4.6 km dtance downtream of the uptream boundary the man analy doman (Fg. e). The LES wnd eem to fluctuate around the WRF wnd above about the 00-m heght. Below the 00-m heght, the LES wnd become gnfcantly weaker than the WRF wnd. Thee decreae of wnd peed are clearly nduced by reolvng the urban urface geometre n the LES model. Furthermore, the turbulent fluctuaton n the LES are well repreented at each downtream poton. Th fact ndcate that the preent approach effectve n producng urban boundary layer flow.

5 (a) (b) (c) Fgure 2: The vertcal profle of wnd peed from the WRF (crcle) and the LES (old lne) model at the pont (a) A, (b) B, and (c) C (ee Fg. e) at 0000 UTC 8 October Fgure 3 compare the tme ere of 0-mnute averaged wnd peed obtaned at the JMA obervaton te (ee Fg. f) and the mulated wnd peed at the correpondng locaton n the WRF model durng 2300 UTC 7 October and 0000 UTC 8 October. The obervaton data are plotted wth one-mnute nterval. The wnd obervaton are conducted at the top of a buldng and at the 35-m heght from the ground urface, whle the WRF wnd are thoe at the 0-m heght a a repreentatve of the urface wnd. The WRF wnd generally agree well wth, but are a lttle tronger than, the obervaton. Note that the WRF wnd ncreaed wth heght and thoe at the 35-m heght were about two tme tronger than the oberved mean wnd, uggetng that the WRF model cannot reproduce urban-canopy flow. Although the LES averaged value overetmate the oberved mean wnd durng 2300 and 2340 UTC 7 October, the dcrepancy between the mulaton and the obervaton generally wthn 0 %. Fgure 4 compare the tme ere of the ntantaneou wnd peed of the LES and obervaton. The LES wnd are thoe obtaned at the 35-m heght at the obervaton locaton, and ther ntantaneou value are ndcated a runnng-mean for 3 econd. Although the large guty wnd at 2338 UTC 7 October are not captured by the LES, the ntantaneou LES wnd generally vary wthn the range of the oberved maxmum ntantaneou value. In order to evaluate the performance of the LES, gut factor from the LES are compared wth the oberved value. The gut factor computed a the rato of the maxmum ntantaneou wnd peed for every one mnute agant the 0-mn mean. Fgure 5 compare the frequency dtrbuton of gut factor of the LES wth the obervaton. There are ome dfference between the LES and the obervaton. Frt, the gut factor at the peak frequency of the LES dfferent from that of the obervaton. Second, the frequency of the large gut factor greater than 2.0 of the LES underetmated. Thee dfference may be partly due to the fact that the fluctuaton mulated n the WRF model, n pte of the hgh-frequency output, have maller varaton than thoe oberved (Fg. 3). If larger fluctuaton, whch hould be preent n the real ettng, could be mulated n the WRF model, gut factor repreented n the LES would be enhanced. Although ome dfference are oberved between the LES model and the obervaton, the frequency dtrbuton obtaned

October. Fgure 4: Tme ere of the horzontal wnd peed of the maxmum ntantaneou value (black lne) at the JMA obervaton te (the crcle pont n Fg.

6 Fgure 3: Tme ere of the horzontal wnd peed of the 0-mn mean (black lne) at the JMA obervaton te (the crcle pont n Fg. f), the WRF mulaton obtaned at the 0-m heght (green lne), and the 0-mn averaged value from the LES obtaned at the 35-m heght (blue lne) durng the perod between 2300 UTC 7 October and 0000 UTC 8 October. Fgure 4: Tme ere of the horzontal wnd peed of the maxmum ntantaneou value (black lne) at the JMA obervaton te (the crcle pont n Fg. f), the ntantaneou (red lne) value from the LES obtaned at the 35-m heght durng the perod between 2300 UTC 7 October and 0000 UTC 8 October. from the LES generally mlar to that from the obervaton. The pont treed at th pont that the buldng effect are a gnfcant contrbutor n determnng the gut factor wthn the urban canopy. Fgure 5: Normalzed frequency dtrbuton of gut factor from the obervaton (black lne) and the LES (red lne). 4. CONCLUSION The extng turbulent nflow technque wa extended to couple between NWP and CFD model and predct a trong wnd event over the central dtrct of Tokyo durng the paage of Typhoon Melor (2009). The preent approach wa ued to conduct an LES of turbulent flow around

7 urban buldng n a real meteorologcal ettng. Frt, urban boundary layer flow from the WRF output wa well reproduced over the LES doman where the urban urface geometre (grd reoluton: 20m) are explctly reolved. Then, th tme-dependent boundary layer flow mpoed at the nflow boundary of the LES model where ndvdual urban buldng (grd reoluton: 5m) are explctly reolved. It found that gnfcant deceleraton of wnd peed wthn the urban canopy layer were reaonably repreented n the LES by reolvng the urban urface geometry. The range of wnd fluctuaton and gut factor were alo found to be well reproduced n the LES. From thee reult, t condered that the preent approach to couple an LES model wth a NWP model and predct guty wnd over urban area hould be effectve. varou geometre. Journal of Appled Meteorology and Clmatology, 50, 8, Skamarock, W. C., Klemp, J. B., Dudha, J., Gll, D. O., Barker, D. M., Duda, M. G., Huang, X.-Y., Wang, W., Power, J. G., A decrpton of the Advanced Reearch WRF Veron 3, NCAR Tech. Note, NCAR/TN-475+STR, pp. Smagornky J General crculaton experment wth the prmtve equaton. Monthly Weather Revew, 9, Van der Hoven Power pectrum of horzontal wnd peed n the frequency range from to 900 cycle per hour. Journal of Meteorology, 4, 2, REFERENCES Chorn, A. J., 967: A numercal method for olvng ncompreble vcou flow problem. Journal of Computatonal Phyc, 2, Goldten, D., R. Handler, and L. Srovch, 993: Modelng a no-lp flow boundary wth an external force feld. Journal of Computatonal Phyc, 05, Greho, P. M., 992: Some nteretng ue n ncompreble flud dynamc, both n the contnuum and n numercal mulaton. Advance n Appled Mechanc, 28, Janc, Z. I., 2002: Nonngular mplementaton of the Mellor-Yamada level 2.5 cheme n the NCEP Meo model. NCEP Offce Note, 437, 6 pp. Lund T. S., X. Wu, and K. D. Squre, 998: Generaton of turbulent nflow data for patally-developng boundary layer mulaton. Journal of Computatonal Phyc, 40, Nakayama, H., T. Takem, and H. Naga, 20: LES analy of the aerodynamc urface properte for turbulent flow over buldng array wth

Large-Eddy Simulation of Plume Dispersion within a Regular Array of Cubic Buildings

Large-Eddy Simulation of Plume Dispersion within a Regular Array of Cubic Buildings Progre n NUCLEAR SCIENCE and TECHNOLOGY, Vol., pp.463-469 () ARTICLE Large-Eddy Smulaton of Plume Dperon wthn a Regular Array of Cubc Buldng Hromaa NAKAYAMA * and Haruyau NAGAI Japan Atomc Energy Agency,