Progress on an urban surface energy balance model comparison study Acknowledge:  UK Met Office, Vasilis Pappas (KCL), Rob Mullen (KCL) COST-728 Exeter meeting, 3-4 May 2007 Sue Grimmond, Martin Best, Janet Barlow King's College London, UK Met Office, University of Reading With (people participating so far): J-J Baik (Korea), M Best (UK), M Bruse (Germany), I Calmet (France), A Dandou (Greece), K Fortuniak (Poland), R Hamdi (Belgium), M Kanda (Japan), H Kondo (Japan), S Krayenhoff (Canada), S-B Limor (Israel), A Martilli (Spain), V Masson (France), K Oleson (USA), A Porson (UK), U Sievers (Germany), H Thompson (UK)

 For example:  Meso-scale modelling  Global climate modelling  Air quality  View factor determinations  Heat island studies  Upper boundary conditions for other models  Weather forecasting  Energy assessments  Emergency response  This Study  Suite of different models  Range of complexity  Range of applications  Range of data needs  Range of computer needs  Common: all run offline Variety of Applications for Urban Energy Balance Models Meso-scale UrbanVegetationWater

Available models Computational Requirements Number of Parameters Parameters difficult to get? Too expensive to run? Globally more applicable?

Past Model Evaluations TEB Vancouver: Masson et al Mexico City: Masson et al Marseille: Lemonsu et al Lodz: Offerle 2003 MOSES Vancouver: Best et al Mexico City: Best et al LUMPS Lodz: Offerle 2003 CLMU Vancouver: Oleson et al Mexico City: Oleson et al. 2007

Distinct Features of Comparison Models run offline Following the methodology used by PILPS Project for Intercomparison of Land-Surface Parameterization Schemes Henderson-Sellers et al. (1993) Common Forcing Data Set All fluxes evaluated Canyon variables: Temperature, Wind speed Increasing levels of information provided Forcing data only Easily obtained urban morphology Urban fabric properties Evaluation data (back calculate parameters)

Key Questions What are the main physical processes controlling the urban energy balance which need to be resolved? How complex does a model need to be in order to produce a realistic simulation of urban surface fluxes and temperatures? Which input parameter information is required by an urban model to perform realistically? Are we measuring the correct variables at the correct scales for model evaluation?

Current Status Determine level of interest and identify participants Inventory of models Collect details of all participating models Set up project infrastructure How to distribute data Website / Staged distribution of data Data formats Sample forcing dataset preparation Sent out to people Waiting for data runs to come back Obtain suitable observational datasets Comprehensive set of observations Ideally dataset unused previously for model testing

Immediate Next steps Ensure quality of observational datasets Ensure dataset fulfils requirements of comparison Identify limitations of experiment from observational dataset Analysis for suitability of observational dataset Waiting to hear from NERC Other alternatives Met Office has funded the initial stages Funding Different levels of input data are released to the modellers At each stage more information is released about the morphology and physical properties of the site enables determination of model parameters with more accuracy Multi-step model runs

 Simulation of the components of the surface energy balance (net radiation, storage, sensible and latent heat fluxes) for the location(s) of the evaluation dataset  Four stages  Different levels of input data are released to the modellers  At each stage more information is released about the morphology and physical properties of the site  enables determination of model parameters with more accuracy  Staged approach to establish the required accuracy for each model parameter by comparing the quality of the simulation at each stage.

Models run with no prior knowledge of the urban surface i.e. model default values for all parameters only the main forcing data supplied, e.g. winds, temperature, solar radiation. Forcing data only: Morphological information provided, e.g. building density, mean building height, vegetation fraction. More easily obtained data sets Add urban morphology: Urban building materials information would be given e.g. thermal properties, albedo information specific to each city/site not known in general on a global basis. Reliance on these types of data makes a scheme difficult to use for global applications Add urban fabric properties: Evaluation dataset released optimisation of model parameters for best fit to observations Optimised parameters returned as well as the standard outputs. requested limit parameter values between observational limits encouraged to undertake analysis of their results if the optimal solution required unrealistic parameter values. Add evaluation data: Multi-step model runs

Process-oriented statistical analysis flux by flux hour by hour evaluation as well as central tendency of the mean assessment will be done at a series of time-scales (hourly, daily, monthly, annual etc.) to determine any biases within the model performance. statistics will consist of a range of metrics mean, standard deviation, probability distribution function, linear regression, root mean square error (systematic, unsystematic), index of agreement, mean absolute error, mean bias error, correlation coefficient, coefficient of determination, etc. Statistical analysis of the model performance relative to the observations e.g. positive sensible heat flux at night storage heat flux magnitude and timing latent heat flux - often neglected term Analysis to assess urban climatological phenomena explicitly e.g. air temperature, surface temperature Evaluate performance Many of the models also predict variables beyond the SEB terms

Urban Energy Balance Models participating so far CODEAuthorsContact PersonVersion usedCountry BEP02MartilliAlberto Martilliolder versionSpain BEP0XMartilliHeather ThompsonLinked to METRASUK CLMUOleson et alKeith OlesonUSA CTTCLimor & HoffmanS-B LimorGreen CTTC modelIsrael ENVIBruseMichael BruseGermany LUMPSGrimmond & OkeSue GrimmondUK/USA MCBMKondo, HiroakiHiroaki Kondov.1.0Japan MM5uDandou & Tombrou Aggeliki Dandou, Maria Tombrou MM5V3-6-1Greece MOSES1TM. Best One tile versionUK MOSES2TM. Best Two tile versionUK MUKLIMOSiewers, UweU. SieversThermodynamicGermany SM2UDupont & MestayerIsabelle CalmetFrance SRUMPorson, Harman, Clark, Best, Belcher A. Porson Under development UK SUMMKanda, T.Kawai, R Moriwaki Manabu Kanda, Toru Kawai, Ryo Moriwaki Coupled with 1D-vegetation model Japan TEBMasson, ValeryValery MassonSingle-layerFrance TEB07Masson, ValeryRafiq Hamdilast versionBelgium TUF2d Krayenhoff & Voogt Scott Krayenhoff2-d versionCanada TUF3d Krayenhoff & Voogt Scott Krayenhoff3-d versionCanada TUFopt Krayenhoff & Voogt Scott KrayenhoffOptimized 3-d verCanada TVM_BEP05Martilli, AlbertoRafiq Hamdilast versionBelgium ULEBFortuniak, KrzysztofK. FortuniakPoland VUCMLee, S-H & Park, S-UJong-Jin BaikKorea

Multiple versions CODEAuthorsContact PersonVersion usedCountry BEP02MartilliAlberto Martilliolder versionSpain BEP0XMartilliHeather ThompsonLinked to METRASUK CLMUOleson et alKeith OlesonUSA CTTCLimor & HoffmanS-B LimorGreen CTTC modelIsrael ENVIBruseMichael BruseGermany LUMPSGrimmond & OkeSue GrimmondUK/USA MCBMKondo, HiroakiHiroaki Kondov.1.0Japan MM5uDandou & Tombrou Aggeliki Dandou, Maria Tombrou MM5V3-6-1Greece MOSES1TM. Best One tile versionUK MOSES2TM. Best Two tile versionUK MUKLIMOSiewers, UweU. SieversThermodynamicGermany SM2UDupont & MestayerIsabelle CalmetFrance SRUMPorson, Harman, Clark, Best, Belcher A. Porson Under development UK SUMMKanda, T.Kawai, R Moriwaki Manabu Kanda, Toru Kawai, Ryo Moriwaki Coupled with 1D-vegetation model Japan TEBMasson, ValeryValery MassonSingle-layerFrance TEB07Masson, ValeryRafiq Hamdilast versionBelgium TUF2d Krayenhoff & Voogt Scott Krayenhoff2-d versionCanada TUF3d Krayenhoff & Voogt Scott Krayenhoff3-d versionCanada TUFopt Krayenhoff & Voogt Scott KrayenhoffOptimized 3-d verCanada TVM_BEP05Martilli, AlbertoRafiq Hamdilast versionBelgium ULEBFortuniak, KrzysztofK. FortuniakPoland VUCMLee, S-H & Park, S-UJong-Jin BaikKorea

Methods used to model outgoing shortwave radiation CODE # reflections albedo MUKLIMO TEBinfinitecanyon, roof TEB07infinitecanyon, roof BEP02multiplecanyon SRUMmultiplebulk/effective CLMUmultipleby facet TVM_BEP05multiplecanyon BEP0Xmultiple TUF3dmultiple (min 2)patches /facet TUF2dmultiple (min 2)patches /facet TUFoptmultiple (min 2)patches /facet VUCMthree MCBMtwoby facet MOSES2Tonecanyon, roof MOSES1Tonebulk SM2Uonebulk/effective MM5uonebulk/town ENVIoneby facet CTTConeby facet ULEBonebulk/town

Methods used to determine Anthropogenic Heat Flux CODE Anthropogenic heat flux Methods BEP0X MUKLIMOheat fluxes from the interior of the buildings TEBdomestic heating computed TEB07domestic heating computed BEP02Partially accounted for by imposing a fixed temp at the building interior BEP05Partially accounted for by imposing a fixed temp at the building interior TUF3dPrescribed bulk value TUF2d Prescribed bulk value TUFoptPrescribed bulk value VUCMPrescribed bulk value SM2UPrescribed CTTCPrescribed per vehicle (for vehicles only) CLMU prescribed traffic fluxes, parameterized waste heat fluxes from heating/ air conditioning MOSES2T not modelled itself but possible to be included for calculation of turbulent fluxes MOSES1T not modelled itself but possible to be included for calculation of turbulent fluxes SRUM not modelled itself but possible to be included for calculation of turbulent fluxes ULEB not modelled itself but possible to be included for calculation of turbulent fluxes MM5u calculated (offline) as a temporal & spatial function of the anthropogenic emissions ENVIfrom heat transfer ew through walls, no storage term MCBMModelled by Kikegawa et al. offline

CODE Methods to calculate turbulent sensible heat flux CTTC calculated by the model TEB07 From each surface BEP02 From each surface BEP05 From each surface SRUM Resistance network based on Harman et al. (2004) CLMU resistances between canyon surfaces and canyon air based on Rowley (1930), between canyon air and atmosphere depend on stability as in CLM3 BEP0X Resistances based on Clarke (1985) TUF3d Resistances based on flat-plate heat transfer coeffs (vertical patches) and based on MO similarity (horiz. patches) TUF2d TUFopt SM2U Resistance (Guilloteau, Zilitinkevich, 1995) TEB Resistance MOSES2T Standard resistance MOSES1T Standard resistance ENVI from turbulence model (wall function) and surface energy balance MM5u Parametric formulation VUCM Parametric formulation MCBM MO or Jurges MUKLIMO MO-laws ULEB M-O similarity: Louis (1979) modified by Mascart at al. (1995)

CODE Methods used to calculate Heat Storage Flux CTTCcalculated by the model BEP02 BEP0X TEBDiffusion TEB07diffusion CLMUDiffusion BEP05Diffusion TUF3dDiffusion TUF2dDiffusion TUFoptDiffusion MOSES2TDiffusion MOSES1TDiffusion VUCMDiffusion SM2UDifference + Diffusion + Force restore MM5uOHM scheme (Grimmond et al., 1991) ENVI soil: 1D model, fully resolved, walls/building system: no storage term ULEB As QG in urban slab (solution of multi layer thermal diffusion equation) MCBMFinite difference MUKLIMOWalls and roofs have a heat capacity

What is resolved in the model? CODE Resolved: CANYONS Resolved: Roof Resolved: walls Walls with orientation Walls sunlit/ shaded Road sunlit/ shaded Turbulence within canyon resolved BEP02 Yes NoYes BEP05 Yes NoYes BEP0X No CLMU NoYes NoYesNo CTTC Yes No Yes ENVI NoYes No Yes MCBM NoYes NoYes MM5u No MOSES1T No MOSES2T Yes No MUKLIMO No Yes SM2U No SRUM Yes No TEB NoYes No TEB07 No TUF Yes No ULEB No VUCM Yes NoYes No

Final Comments  Models that are already participating show a wide range of approaches  Need to follow up on some details  Multiple versions of some individual models are participating  Initial trial dataset now available  Data back from three groups  This is allowing us to iron out issues at both ends  People can still participate  Encouraged to do so! Contact me:  Participants will be co-authors in manuscripts etc  Waiting to hear if NERC will fund the next parts of this project

CODEType of Model Within Canyon processes modelled Canyons are resolved Above canyon modelled Canyon top modelled BEP02 Multiple layer No BEP05 Multiple layerNo BEP0X Multiple layerNo CLMU Single layerYesNoYes CTTC Single layerNo ENVI Yes MCBM Multiple layerNo MM5u Single layerNo MOSES1T Single LayerNo MOSES2T Single LayerNo MUKLIMO Yes No SM2U Single LayerNo SRUM Single LayerNo TEB No TEB07 No TUF NoYesNo BEP05 No ULEB Multiple layerNo VUCM Single layerYesNoYes