Even though impervious surfaces only cover a small percentage of the Earth’s land surface, their representation in land-surface (LS) models and subsequently numerical weather prediction and climate models is of great importance. And even though the number of available LS schemes is high, the number of available urban parameterizations embedded in those schemes - as revealed by the intercomparison study by e.g. Grimmond et al. (2010, 2011) – might be even higher. The intercomparison study also outlined that models with various complexities have various strengths and weaknesses, while all models have a varying performance across the energy balance components. Besides the varying model complexities (physics), LS schemes also differ in the amount and characteristics of their external parameters describing the surface characteristics. The amount of parameters is generally associated with a models' complexity.

For this study, four models with varying complexity (CLMU, TEB, SUEWS and TERRA) are used in combination with three different external parameter namelists. The Jackson et al. (2010) and Eoclimap (Masson et al., 2003) namelist – which are native for CLMU and TEB respectively - and a reference set of parameters describing the geometrical, radiative and thermal characteristics of the urban surface. All model and namelist combinations are run offline for the tropical site of Telok Kurau in Singapore. In a first phase, a general evaluation of the capacity of all four models to model the surface energy balance is presented. In a second phase, multi-temporal scale analysis are performed to indicate to role of and interaction between the models' complexity and the external parameter settings.

The simulation of urban fluxes (e.g. radiative and sensible or latent heat fluxes) is of great importance for fundamental or applied climatic studies. Several canopy models have been developed to simulate these processes into towns and some of them can be directly coupled with numerical weather models working at mesoscale. These canopy models greatly improve the results of weather simulations over town, because they are able to reproduce the interactions between the boundary layer and the underlying urban canopy layer. This is possible only because the urban environment is parameterised to preserve the simulation time. In some cases however, these parameterisations are not welcome. When higher resolutions or more precise results are required, it is necessary to take the real geometry of the town into account to reproduce the buildings layouts and their projected shadows. This means that the shape of each town element (buildings, roads and trees) must be realistically reproduced in three dimensions. In this type of canopy models, the buildings can be depicted with their inclined roofs, balconies and windows and the artificial or vegetated areas can be introduced at their real places... In principle, with such a detailed description level, all the fluxes can be better reproduced for each element (like facades, roofs, roads…). But, as a consequence of the use of 3D geometry, the radiative algorithms must be simulated with specialized procedures to obtain the shadows, the anisotropic sky, and the interreflections between all the objects. This dramatically increases the simulation time. LASER/F (LAtent, SEnsible, Radiation fluxes) is a 3D urban canopy model specially designed when high resolution is required for urban climatic studies. After more than ten years of development the latest version is able to simulate most of the physical processes. A simulation has been undertaken with this model to test it on a real urban environment (a district of Strasbourg, France) and under clear sky conditions (to allow the maximum thermal contrasts between the surface elements). The simulation was forced with data acquired over the mean roof level by several sensors. The processes simulated into the canopy layer have been compared with the measurements obtained during an experimental campaign specially designed to this aim by our team.

Three different urban parameterization schemes (UCM, BEP and BEP+BEM) were researched and compared using high resolution land use data to replace USGS default data with three hot, sunny days in July, 2013 in the coastal city of Dalian that were used as examples. The simulations were compared to climate stations and in situ measurements. The results for the BEP+BEM scheme were the most consistent with the measurements, while the BEP and UCM schemes were the second and third best, respectively. The wind directions for all three schemes were largely identical to the measurements.

A 1-D Canopy Interface Model (CIM) was developed in order to better simulate the effect of urban obstacles on the atmosphere in the boundary layer. The model solves the Navier-Stokes equations on a high-resolved gridded vertical column. The effect of the surface is simulated testing a set of theories and urban parameterizations. The final proposition guarantees its coherence with past theories in any atmospheric stability and terrain configuration.

Obstacle characteristics are computed using surface and volume porosities in each cell of the model domain. These porosities are used to weight several terms in the Navier-Stokes equations. A 1.5-order turbulence closure is used in order to compute the turbulent coefficients with the TKE. The mixing length takes into account the density of the obstacles and their height. The turbulent scheme is designed in order to keep CIM coherent with the Prandtl theory in neutral atmospheric conditions and with the MOST in stratified atmospheric stability when CIM is used over plane surfaces. The modifications brought to the main governing equations are discussed following theoretical analysis and experiences with CIM, simulating the averaged meteorological variables (wind speed, turbulent kinetic energy (TKE), temperature and humidity). Simulations are compared with analytical solutions, when possible, and also simulations issued from a computational fluid dynamics (CFD) model.

The results show how constant values, usually prescribed, can be theoretically estimated and how the buoyancy term of the turbulent kinetic energy balance equation should be adjusted accordingly. After modifications, it is shown that CIM is coherent with past propositions in any case of atmospheric stabilities over plane surfaces. The use of CIM in presence of obstacles is based on the extension of the 1.5 order turbulence closure to compute the turbulent coefficients with the TKE. CIM shows simulations in good agreement with the CFD simulations in the presence of obstacles. It is able to reproduce an inertial sub-layer as described by the Prandlt and constant-flux layer theory above a displacement height over a homogeneous canopy.

In order to account for urban climate at the regional scales, a new efficient urban land-surface parametrization TERRA_URB has been developed and coupled to the atmospheric numerical model COSMO-CLM. Hereby, several new advancements for urban land-surface models are introduced which are crucial for capturing the urban surface-energy balance and its seasonal dependency in the mid-latitudes. This includes a new PDF-based water-storage parametrization for impervious land, the representation of radiative absorption and emission by greenhouse gases in the infra-red spectrum in the urban canopy layer, and the inclusion of heat emission from human activity. TERRA_URB has been applied in offline urban-climate studies during European observation campaigns at Basel (BUBBLE), Toulouse (CAPITOUL), and Singapore, and in online studies for urban areas in Belgium, Germany and Switzerland.

Because of its computational efficiency, high accuracy and its to-the-point conceptual easiness, TERRA_URB has been selected to become the standard urban parametrization of the atmospheric numerical model COSMO(-CLM). This allows for better weather forecasts for temperature and precipitation in cities with COSMO, and an improved assessment of urban outdoor hazards in the context of global climate change and urban expansion with COSMO-CLM. In this work, we propose additional developments in TERRA_URB towards improved urban climate modelling. On the one hand, global datasets are constructed for describing the physical land-surface properties of cities over the world. Hereby, global datasets in the EXTernal PARameters (EXTPAR) tool of COSMO are used to derive the 'sealed' surface Fraction referring to the presence of buildings and streets, but also the Surface Area Index (SAI) referring to the surface morphology based on the Floar Area Index (FAI). On the other hand, it is focussed on the urban/rural contrast in terms of turbulent transport in the surface layer by means of model sensivity experiments. On the theoretical basis of the TKE-based surface-layer transfer scheme of COSMO, we investigate the consistency between empirical functions for thermal roughness lengths and the morphological parameters of urban/rural canopies.

The world's population will increase in the next decades especially in urban areas. Additionally, the living conditions are affected largely by the local urban climate. Due to the projected higher frequency of heat waves and the related heat stress, high resolution spatio-temporal distribution of air temperature is an important key for urban planning and development. In this study we evaluate simulations of the mesoscale weather and climate model COSMO-CLM using the dense 2m temperature mobile measurements of Berlin (Germany) for several summer days in 2012. COSMO-CLM, driven by ERA-interim re-analysis data, is either used with its default bulk formulation (no DCEP) of the urban surfaces or online coupled with the Double Canyon Effect Parameterization (DCEP) urban canopy scheme.

The results show, that the simulated 2m temperature using the new scheme is consistent with the averaged 2m temperature measurements for Berlin. The urban heat island has been successfully simulated by DCEP scheme, while in the mesoscale model (no DCEP) the city center is warmer than surrounding during the day and cooler during the night. In the bulk scheme (no DCEP) the urban areas are characterized as natural surfaces with higher roughness length and reduced vegetation. This may lead to warmer center during day time. In the case, that the mesoscale model overestimates the 2m temperature, DCEP scheme shows no improvement and leads to higher errors. The dependency of mean absolute error and the land use classes are still under study.

The thermal comfort of urban residents is mainly affected by the Urban Heat Island (UHI) effect (difference in air temperature between urban and surrounding area). This effect is caused by anthropogenic sources, low vegetated areas and heat stored in buildings and roads released during the night. In case of heat wave, strong UHIs occur and mortality rates can increase in towns. Indeed, the air temperature difference between the city center and the suburbs can reach 10K for the biggest urban areas. A tool to simulate UHI fast and with only a few meteorological information can be useful for the community studying the thermal comfort. Such a model will be presented. In our study, the surface is modeled by the SURFEX model. The SURFEX surface model contains an urban canopy model, TEB, a soil-vegetation-atmospheric transfer model, ISBA and a water surface models. SURFEX can be used on-line (coupled with a mesoscale atmospheric model) or off-line (forced with meteorological conditions above the canopy layer). The off-line approach has a low computational cost but needs meteorological information that only specialized professional community can obtain through mesoscale simulations or short- time experiments. In order to make urban climate predictions accessible to other communities such as building engineers or urban planner, a method to calculate meteorological forcing for surfaces models with weather data files from an operational measurement stations outside the city is presented. This method has multiple advantages. First, these files can easily be found for a lot of cities, in most case in airports. Second, the forcing temperatures calculated can be influenced by urban planning scenarios.

The methodology presented integrates day-time and night-time boundary layers and impact of the wind. An iterative method is used to extrapolate measure information at the forcing level above the canopy. The iterative method has bean validated with data of a 30 meters mast at the airport of Roissy, France, which provides temperature, humidity at 2m and 30m and wind at 10m and 30m.

At each time-step, the UHI is calculated on a two dimensional grid with a 2km mesh-size. The energy budget of the whole boundary layer for each 4km2 grid mesh is computed. Note that the boundary layer height also evolves with the wind speed, precipitation or cloud fraction. The mean temperature of a column with a height corresponding to the boundary layer height is calculated with the surface heat flux given by the surface model. In order to take into account wind effects, the 2D field of temperature is also advected with a 2D lagrangian advection model. . The UHI modeling has been simulated over the city of Paris, France and its suburbs.

The method has been validated with an operational weather station network giving temperatures over the region of Paris during the years 2010 to 2011 and by comparing the UHI simulations to a complete high resolution atmospheric simulation (MesoNH model) done over the Paris region at 2km of resolution during years 2010 and 2011 .

QUIC EnvSim (QES) is a building resolving (2-5m) urban microclimate modeling tool capable of rapidly computing time-averaged fields of velocity, turbulence variables, temperature, moisture, and scalar concentrations in around the built environment. The tool has been developed to take advantage of computer graphics hardware and ray-tracing techniques to compute radiation balances in complex urban geometries. Additional use of graphics processing units (GPUs) facilitates the rapid solution of other transport quantities in the atmosphere and through urban surfaces. With this system, simulation domains on the order several squre kilometers can be simulated on a single high-end workstation. Previous work and presentations on QES have focused on verification and validation of components of the system. In this presentation, we show the power of QES to help explore different real world design scenarios and investigate the effects small-scale place based design changes on various urban scales. In particular, simple surface flux scenarios of urban greening (e.g. albedo changes, surface moisture changes, impermeable surface fraction) at the building and neighborhood scales for Salt Lake City, UT USA will be presented and evaluated.

QUIC EnvSim (QES) is a modular framework for efficiently simulating and visualizing building scale environmental interactions in urban domains. QES leverages the computational resources available on both modern graphics hardware cores (GPUs) and CPUs to efficiently simulate complex interactions between the built environment and environment scalars such as heat and moisture. Visualizing simulation results enables users to analyze the data, to understand the interactions with the environment, and to communicate processes to urban planners, stakeholders, and decision makers.

The speed and efficiency of QES comes from its utilization of GPUs to perform simulation computations in parallel. A synergistic side-effect of performing computation on the GPU is that much of the data resides directly in GPU memory and can therefore be visualized immediately upon completion of a simulation timestep. QES' visualization interface provides control over the simulation and allows for direct presentation of environmental scalars, including sky view factor, solar and longwave radiation, and temperature across the simulation domain.

Alternative visualization components of QES render the data external to the simulation but afford more advanced rendering of the domain data. In one such system, a head-mounted display gives users a first-person perspective on the simulation, allowing users to see scalar fields change around them over the course of a diurnal cycle. A virtual reality tracking system is also used to track user movements, gestures, and actions providing users with natural mechanisms to navigate through and interact with the data. Such vantage points and interactivity may be beneficial for communicating how differences in urban structure impact local changes within an urban domain. The viability of these interfaces over standard presentations is an open question that we are currently exploring. Our working hypothesis is that through compelling visuals and effective interaction techniques, urban environmental processes can be more directly communicated to users.

In our presentation, we will provide an overview of QES and highlight the capabilities of the visualization system. A live visual demonstration will be shown on a laptop to illustrate the efficiency of the QES simulations and rendering systems.

Landier L.1, Al Bitar A.1, Gregoire T.1, Lauret N.1, Yin T.1, Gastellu-Etchegorry J.P.1,

Aubert S.2, Mitraka Z.3, Chrysoulakis N.3

1 CESBIO, Toulouse University, 18 Av. Edouard Belin, 31401 - Toulouse, France. Email: landierl@cesbio.cnes.fr

2 Météo France, 42 Av. Gaspard Coriolis, 31057 - Toulouse, France. Email: sylvain.aubert@meteo.fr

3 FORTH - Hellas, N. Plastira 100, Vassilika Vouton, 70013 - Heraklion, Greece, Email: mitraka@iacm.forth.gr

DARTEB model is developed at CESBIO (www.cesbio.ups-tlse.fr) for simulating 3D (3 dimensional) energy budget of urban and natural scenes with topography and atmosphere. It uses DART model (Discrete Anisotropic Radiative Transfer: www.cesbio.ups-tlse.fr/dart/license/en/index.php) for simulating radiative mechanisms, which gives the 3D radiative budget and remote sensing acquisitions (LIDAR waveforms / photon counting, satellite/airborne images) of urban scenes, for any atmosphere, wavelength, sun/view direction, altitude and spatial resolution, over the whole optical domain (UV to TIR). DART uses innovative modeling approaches: multi-spectral discrete ordinate techniques with exact kernel, RayCarlo and Box methods for LIDAR, etc. It is developed since 1992. It was patented in 2003. Paul Sabatier University distributes it as free licenses to scientists. DARTEB simulates major energy mechanisms (heat conduction, turbulent momentum and heat fluxes, soil moisture, vegetation photosynthesis, evapotranspiration) that contribute to the energy budget. For urban canopies, it adapts equations from the TEB urban surface scheme (Masson, 2000). Each surface type (wall, soil, roof) is discretized into layers for simulating conduction fluxes to or from the ground and building interiors.

DARTEB assesses the 3D radiative budget distribution, and the 3D temperature distribution, with a prognostic approach. Temperature values at time k lead to the 3D TIR (Thermal Infra Red) and energy budgets at time k+1, which allows one to compute the 3D temperature distribution at time k+1, using the 3D visible and NIR radiation budget at time k+1. Although, it uses an actual 3D radiative budget and it applies TEB equations at each point of the 3D scene, DARTEB is not a full 3D model (e.g., 1D wind profile is used instead of 3D wind distribution). It was successfully tested in the frame of the CAPITOUL project (www.cnrm.meteo.fr/IMG/pdf/masson-capitoul.pdf) of Meteo France for simulating the time evolution of the temperature of walls in a street of Toulouse (France).

Here, we present four improvements that allow operational simulation of urban variables that are classically used in urban studies, with or without remote sensing data. They were applied to the new Toulouse urban data base, in the frame of EO4SEB and Neo-Campus projects.

- Sky-view factor. It represents the fraction of sky hemisphere visible from ground level. It varies with height and spacing of buildings and trees and it affects the radiational heating/cooling. DART can compute it using a DTM (Digital Terrain Model) and vegetation distribution.

- Time variation of surface albedo. DART computes the bi-hemispheric and direct-hemispheric albedo, for any time and date, using actual atmosphere data (e.g., ECMWF or Aeronet network).

- In-situ and airborne brightness temperature. It is usually acquired by large view radiometers. Thus, any acquisition corresponds to different view directions, which explains that measurements depend on the radiometer location. These two acquisition configurations are simulated with DART.

- LIDAR. This instrument provides topographic information. Data from a scanner LIDAR were simulated with DART and inverted with SPDlib software for testing the accuracy of the approach.