SEARCH

SEARCH BY CITATION

Keywords:

  • urban heat island;
  • CFD;
  • Tokyo;
  • urban planning;
  • wind path

Abstract

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Analysis method
  5. 3. Input data for the simulation
  6. 4. Simulation results for the 23 wards of Tokyo
  7. 5. Case studies on increasing the sea breeze in urban spaces
  8. 6. Result: case with urban redevelopment
  9. 7. Summary
  10. Acknowledgements
  11. References

Recently, countermeasures against the urban heat island effect have become increasingly important in Tokyo. Such countermeasures include reduction of anthropogenic heat release and enhancement of urban ventilation. Evaluations of urban ventilation require construction of a high-resolution computational fluid dynamics (CFD) model, which takes into account complex urban morphology. The morphological complexity arises from multi-scale geometry consisting of buildings, forests, and rivers, which is superimposed on varying topography. Given this background, airflow and temperature fields over the 23 wards of Tokyo were simulated with a CFD technique using a total of approximately 5 billion computational grid cells with a horizontal grid spacing of 5 m. The root mean square (RMS) error of the air temperature between the simulation and observation results at 127 points was 1.1 °C. Using the developed model, an urban redevelopment plan for two districts in metropolitan Tokyo was examined from the viewpoint of air temperature mitigation. Numerical results showed that a reduction by 1 ha in the area covered by buildings increases the area with temperatures below 30 °C by 12 ha. Copyright © 2010 Royal Meteorological Society


1. Introduction

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Analysis method
  5. 3. Input data for the simulation
  6. 4. Simulation results for the 23 wards of Tokyo
  7. 5. Case studies on increasing the sea breeze in urban spaces
  8. 6. Result: case with urban redevelopment
  9. 7. Summary
  10. Acknowledgements
  11. References

Air temperature is an important environmental factor for buildings and human activities. The urban heat island (UHI) effect is caused by human activities in urban regions with a high population density. Thus, the UHI effect has significant consequences for humans. Furthermore, long-term variations of air temperature in urban regions are likely subject to the influence of global warming as well as the UHI effect. This circumstance calls for measures to mitigate the air temperature increase in urban regions. There exist two major approaches to mitigate UHI phenomena. The first approach is to reduce the amount of heat generated in urban areas. Various heat sources exist in urban regions, such as the sensible heat released from the ground and building surfaces that result from their increased temperatures. Urban heat sources also include anthropogenic heat which is released into the atmosphere as a result of fossil fuel consumption by automobiles, factories, and the heat source systems of buildings. In Japan, a number of measures have been addressed to mitigate the air temperature increase in urban regions. The Ministry of Land, Infrastructure, Transport and Tourism (MLIT) of Japan has established guidelines for building owners to mitigate the UHI effect and has also created an evaluation system to rate architectural designs in terms of their contributions for mitigating the UHI effect (IBEC, 2006). The factors that are assessed by the evaluation system include implementation of green roofs, use of high reflectivity materials, enhancement of ventilation, and reduction of anthropogenic heat. For the purpose of mitigating the UHI effect, Tokyo Prefecture also mandates implementation of green roofs on new buildings over a certain size (Tokyo Metropolitan Government, 2005). The Ministry of the Environment (MOE) has been promoting projects which numerically and objectively evaluate the performance of UHI countermeasure technologies. As a part of this effort, the Ministry has publicized the effects of sprinkling water on individual outdoor air conditioning units in terms of the conversion rate of anthropogenic sensible heat into latent heat (MOE, 2005). The abovementioned measures have been funded by the national and local governments and have resulted in actual changes in the design of urban structures. For example, the annual production of high reflectivity materials in Japan has increased threefold over the past few years, and 10 ha of green roofs have also been newly added in Tokyo over the same time period.

The second approach to mitigating UHI phenomena is to enhance ventilation effects in urban spaces. This approach calls for effective building arrangements so that heat generated in an urban region can rapidly diffuse into the surroundings. While the first approach addresses the reduction of heat from individual buildings and machinery, the second approach focuses on the interaction between the atmosphere and the heat released from urban structures. The interaction between the atmosphere and the heat released from urban structures is affected by multiple open spaces with various scales such as roads, rivers, and parks. In addition, because most of the major cities in Japan are located in coastal areas, using the passive cooling effects of the land/sea breeze circulation is considered very effective for reducing summertime air temperatures in these cities (AIJ, 2000).

To date, many computational fluid dynamics (CFD) studies that resolve individual buildings in urban areas have been conducted to predict the urban atmospheric environment (Table I). However, there have been very few CFD studies that investigate the interaction between heat released from individual buildings and atmospheric circulations with spatial scales much larger than individual buildings such as the land/sea breeze circulation and circulations in the open spaces of rivers and parks.

Table I. Examples of recent CFD simulations that resolve individual buildings in urban areas
LiteratureApplication (turbulence model)Horizontal domain size (m2)Vertical domain height (m)
The present paperThermal environment (RANS)33 000 × 33 000500
Baik et al. (2009)Pollutant dispersion (RANS)980 × 1140400
Blocken and Persoon (2009)Wind environment (RANS)3000 × 3000900
Bou-Zeid et al. (2009)Wind environment (LES)1500 × 1500400
Hanna et al. (2009)Pollutant dispersion (RANS)Approximately 3200 × 900Approximately 650
Xie and Castro (2009)Pollutant dispersion (LES)1200 × 800200
Nozu et al. (2008)Wind load (LES)2048 × 1024800
Oguro et al. (2008)Wind environment (RANS)1 0000 × 1 0000400
Tamura (2008)Wind load (LES)2900 × 12001000
Burrows et al. (2007)Wind environment (RANS)2100 × 2100300
Chan and Leach (2007)Pollutant dispersion (RANS)1030 × 3010425
Flaherty et al. (2007)Pollutant dispersion (RANS)900 × 1200300
Hendricks et al. (2007)Pollutant dispersion (RANS)1400 × 1400200
Huang et al. (2005)Thermal environment (RANS)400 × 400450

In this study, the thermal environment of the 23 wards of Tokyo is investigated by conducting a CFD simulation that resolves individual buildings. The 23 wards of Tokyo are specially administered wards of Tokyo Prefecture and are the centre of the Japanese economy and government with a population of 8.8 million. The land area of the wards is 621 km2; approximately 56, 20, and 10% of the land area are covered by residential area, roads, and green spaces, respectively. The rest of the land area is covered by water and other surfaces. For the numerical simulation, a large-scale numerical analysis system is developed and is then executed on a large-scale vector parallel computer. The thermal environment of the 23 wards of Tokyo is simulated with 5-m grid spacing, enough high resolution to study the urban thermal environment on scales from that of an individual building to that of the entire urban area. The validity of the system has been confirmed for an individual building and for groups of buildings through comparisons with wind tunnel tests (Ichinose et al., 2005, 2007). In the present study, the simulated thermal environment of the 23 wards of Tokyo is compared to observed meteorological data. Finally, the developed system is applied to the evaluation of a redevelopment plan for urban Tokyo, which aims to incorporate a climate-sensitive design that takes into account urban ventilation effects.

2. Analysis method

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Analysis method
  5. 3. Input data for the simulation
  6. 4. Simulation results for the 23 wards of Tokyo
  7. 5. Case studies on increasing the sea breeze in urban spaces
  8. 6. Result: case with urban redevelopment
  9. 7. Summary
  10. Acknowledgements
  11. References

2.1. Outline

Figure 1 shows the flow of the analysis performed in this study. The inputs of the simulation system include the boundary conditions that either represent the present urban surface or the urban surface set forth by the redevelopment plan of the area to be analysed. The output of the simulation system includes environmental variables such as air temperature and wind speed and direction. The simulation is performed on a high-speed supercomputer system, named the Earth Simulator (Japan Agency for Marine-Earth Science and Technology, Yokosuka, Japan).

thumbnail image

Figure 1. Diagram of UHI analysis. This figure is available in colour online at wileyonlinelibrary.com/journal/joc

Download figure to PowerPoint

2.2. Governing equations and numerical schemes

Table II summarizes the governing equations and numerical schemes that are used in the present study. The governing equations are compressible Reynolds averaged equations for the conservation of mass, momentum, energy, turbulence kinetic energy (k), and dissipation rate of turbulence kinetic energy (ε). As for the turbulence model, a standard k − ε model is modified so that the influences of the Coriolis force on flow motions and adiabatic process on air temperature are taken into consideration. Moreover, the buoyancy effect due to water vapour is considered as a source of turbulence kinetic energy. In order to take into account the aerodynamic influences of the configurations of objects that are smaller than the grid resolution, all the equations are formulated with the fractional-area-volume obstacle representation (FAVOR) technique (Hirt, 1993). The drag of trees is modelled using the parameter values given in Mochida et al. (2006).

Table II. Governing equations and numerical schemes for the CFD model
ComponentContent
  1. Note that the model is intended for low Mach number flow of a well-mixed gas with variable density.

  2. In order to take into account objects smaller than the grid resolution, all the governing equations are formulated with the FAVOR technique.

Governing equationsMass conservation equation
 Momentum transport equation (buoyancy, Coriolis force, and drag of trees are considered)
 Energy transport equation (formulated in terms of potential temperature; sensible heat released from walls and anthropogenic heat are considered)
 Water vapour transport equation (formulated in terms of specific humidity; latent heat released from walls and anthropogenic heat are considered)
 Transport equation for turbulence kinetic energy (k) (turbulence generated by buoyancy, humidity, and trees is considered)
 Equation for rate of dissipation of turbulence kinetic energy (ε) (dissipation by buoyancy, humidity, and trees is considered)
Turbulence modelStandard k − ε model
Coordinate systemThree-dimensional orthogonal coordinate system
Computational gridStaggered grid
Discretization methodFinite difference method
Spatial discretization methodFirst-order upwind scheme (advection term), Second-order central difference scheme (all other terms)
Time discretization methodFully implicit scheme
Calculation algorithmAMG-CG solver, BiCGSTAB solver

The governing equations are discretized by the finite difference method on a Cartesian staggered grid system. A first-order upwind differencing scheme and a second-order central differencing scheme are applied for the advective and diffusion terms, respectively. For time integration, a fully implicit method is utilized. The algebraic multigrid (AMG) method (Stuben, 2001) is used to solve Poisson's equation for pressure and the backward difference formulas.

2.3. Boundary conditions

In order to determine the values for the boundary conditions, a mesoscale model that includes the CFD domain is run prior to the CFD simulation; then the computed flow quantities are interpolated onto the boundaries of the CFD domain. However, all the interpolated values cannot be used at all the boundaries of the CFD domain because mass would not be conserved over the entire CFD domain due to interpolation errors. Several approaches have been proposed to conserve mass when linking a mesoscale meteorological model and a CFD model. For example, Baik et al. (2009) applied a zero-gradient boundary condition at the outflow boundaries and set the vertical velocity at the inflow boundaries to zero. Coirier et al. (2007) used spatially interpolated flow quantities from mesoscale simulations only in the lateral directions. In the present simulation, the boundary conditions of the CFD domain are set as shown in Table III and are as follows. From the results of the mesoscale meteorological simulation, it is determined whether each grid cell on the boundary of the CFD domain is experiencing an inflow or an outflow. In the course of the CFD simulation, each boundary grid cell is always either an inflow grid cell or an outflow grid cell unlike in typical CFD simulations for which the entire side of the domain is assigned to be inflow or outflow. For the upper boundary, the pressure and the horizontal wind velocities, u and v, are provided from the mesoscale simulation, and the vertical wind velocity, w, is provided by the CFD analysis to satisfy mass conservation over the entire CFD domain. Because the magnitudes of the vertical wind velocities calculated by the mesoscale meteorological simulation are generally much smaller than those of the horizontal wind velocities, the effect of not adopting the vertical velocities calculated by the mesoscale simulation as the upper boundary conditions of the CFD is expected to be much smaller than the effect of not adopting one of the other components of the wind velocity at the boundary of the CFD domain. In order to set k and ε, at the upper and side boundaries, the friction velocity is calculated using the logarithmic law, which has been frequently applied for airflows above an urban canopy. Alternatively, the friction velocity may be calculated from a mesoscale analysis which takes into account the effect of the urban canopy, and the calculated friction velocity may then be used as a boundary condition for the CFD simulation. The application of this approach is beyond the scope of the present paper and will be addressed in our future research. With the use of the calculated friction velocity and the assumption of local equilibrium at the upper and side boundaries, k and ε, are calculated.

Table III. Boundary conditions of the CFD model
VariableUpper boundarySideLower boundary (soil and building surfaces)
  Inflow boundaryOutflow boundary 
PressureConstantZero gradientZero gradientZero gradient
WindConstant u, vConstant u, v, wConstant u, v, wGeneralized log law
HeatConstant potential temperatureConstant potential temperatureZero potential temperature gradientHeat transfer by bulk formula
VapourConstant specific humidityConstant specific humidityZero specific humidity gradientMoisture transfer by bulk formula
k and εConstantConstantZero gradientGeneralized log law

2.4. Calculation of surface temperatures

Although it is desirable to estimate the surface temperatures of buildings and other objects on the ground by performing a fully coupled conductive, radiative, and convective heat transfer simulation in three dimensions, the following simplified method is adopted in the present study. First, an unsteady heat conduction analysis is performed for individual cover types, such as asphalt and grassland. In the analysis, one-dimensional vertical heat conduction is assumed and evaluated according to the meteorological condition of the day of the simulation. The surface temperature of a cover type is calculated for the case with no shading from the sun and the case with shading from the sun. Subsequently, from the solar position for the hour for analysis, the sunlit and shaded conditions of the urban surfaces are determined in every grid cell of the three-dimensional analysis domain. The sunlit and shaded conditions of the urban surfaces are determined by taking into account the building arrangement within the city to be investigated by the CFD analysis. Finally, the surface temperature of the urban surface within an individual grid cell is assigned according to the surface cover types and the sunlit and shaded conditions.

When the surface heat energy budget is solved, parameters such as albedo, emissivity, and evaporation efficiency are required. Ichinose et al. (1999) evaluated values for five parameters relevant to the surface heat energy budget (i.e. albedo, evaporation efficiency, density, specific heat, and thermal diffusion coefficient) of ten land cover types. Ihara et al. (2003) evaluated values for a different set of five parameters relevant to the surface heat energy budget (i.e. albedo, emissivity, thermal diffusion coefficient, heat capacity, and thermal conductivity) of eight land cover types. The relevant parameters and the corresponding values from Ichinose et al. (1999) and Ihara et al. (2003) which are used in the present study are summarized in Table IV Although the value of the thermal diffusivity is influenced by the wind speed, it is set constant at 11.6 W m−2 K−1 (Yoshida et al., 2000) in the present study. The mass transfer coefficient is estimated using Lewis's law.

Table IV. Values of parameters relevant to the surface energy budget
Land use classification adopted for the present studyBuildingsAsphaltGrasslandWaterTrees
Albedo [−] (Ichinose et al., 1999)0.180.180.160.080.16
Surface emissivity [−] (Ihara et al., 2003)0.960.910.950.930.95
Evaporation efficiency [−] (Ichinose et al., 1999)0.050.050.301.000.30
Density [kg m−3] (Ichinose et al., 1999)2.4 × 1032.1 × 1031.8 × 1031.0 × 1031.8 × 103
Specific heat [J kg−1 K−1] (Ichinose et al., 1999)882882117642001176

In the present study, a simple method is adopted for determining sunlit and shaded areas on the ground, building walls, and building roof surfaces in individual grid cells (Figure 2(a)). For each vertical column of grid cells, the lowest height reached by the solar radiation is determined. The lowest height reached by the solar radiation and the relevant building heights are taken into consideration for determining the sunlit and shaded areas for each vertical column of grid cells. The vector normal to the building wall, Si, j, k, and the solar-radiation vector, es, are defined as:

  • equation image(1)
  • equation image(2)

where Δxi, Δyj, and Δzk are the widths of grid cells in the east–west, north–south, and vertical directions, respectively; GA, i+1/2, GA, i−1/2, GA, j+1/2, and GA, j−1/2 are the open area ratios of the east, west, north, and south sides of an individual grid cell in the three-dimensional grid system (i, j, k), respectively; θs is the solar elevation angle (0 ≤ θs ≤ π/2), and ϕs is the solar azimuth angle (−π ≤ ϕs ≤ π). Figure 2(b) illustrates geometrical patterns between a building wall and the solar-radiation vector. When Si, j, k· es ≥ 0, the building wall is determined to be sunlit. In contrast, when Si, j, k· es ≥ 0, the building wall is determined to be shaded.

thumbnail image

Figure 2. Simple method for determining shaded region. This figure is available in colour online at wileyonlinelibrary.com/journal/joc

Download figure to PowerPoint

3. Input data for the simulation

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Analysis method
  5. 3. Input data for the simulation
  6. 4. Simulation results for the 23 wards of Tokyo
  7. 5. Case studies on increasing the sea breeze in urban spaces
  8. 6. Result: case with urban redevelopment
  9. 7. Summary
  10. Acknowledgements
  11. References

3.1. Location of analysis

A square area of 33 × 33 km2 is considered for analysis (Figure 3). This area includes the entire 23 wards of Tokyo. The target area is surrounded by a 1.5 km wide buffer zone. The buffer zone is included for the following reason: when buildings are present near the boundary of the target area, the area fraction occupied by the air on the boundary plane and the effective volume fraction of grid cells near the boundary become small. In this case, if the wind speeds at the boundary are set using the results from the mesoscale analysis, the calculated values of wind speeds may be overestimated near the boundary because of the use of the continuity equation in the calculation procedure. The elevation within the buffer zone is set using the values of the elevation in the grid cells situated at the boundary of the 33 × 33 km2 area so that the ground surface becomes continuous with no change of elevation across the boundary between the target area and the buffer zone. Furthermore, there is no heat or moisture transfer between the ground surface and the atmosphere in the buffer zone. The computational domain extends vertically to a height of 500 m and is divided into grid cells that vary in height from 1 to 10 m with the smallest grid cells closest to the ground. The domain is horizontally divided into equally sized 5 m grid cells. The horizontal dimensions of the grid cells are set to 5 m so that airflows within and over the major highways and open spaces including green spaces can be examined on the district scale. The entire computational grid, including the buffer zone, is comprised of approximately 5 billion grid cells. The computation is performed using 300 nodes of the Earth Simulator and requires 16 h to complete. The residuals associated with the momentum, heat transfer, and other equations are examined at each calculation step. Achievement of steady state was confirmed by checking the time-integrated values of the residuals. The spatial scale of the domain in the present simulation is one of the largest ever investigated in an urban CFD study (Table I).

thumbnail image

Figure 3. Simulation domain for the 23 wards of Tokyo. (a) The computational domain of the mesoscale model. The areas defined by the dashed and solid rectangles correspond to the first and second layers of nesting, respectively. (b) An enlarged view of the Tokyo area and its surrounding. The border of the figure is the location of the second layer of nesting of the mesoscale model. The smaller area defined by thick lines is the CFD analysis domain. (c) The CFD analysis domain which includes the 23 wards of Tokyo. The surface elevation of the analysis area is indicated by the gray-scale gradient. Urban redevelopment is currently planned for the area marked with the bold rectangle (see sections on case studies). This figure is available in colour online at wileyonlinelibrary.com/journal/joc

Download figure to PowerPoint

3.2. Date and time of analysis

The CFD analysis is performed for 1400 local standard time (LST) on 31 July 2005. This time is selected for analysis for two reasons. First, the UHI effect becomes the most severe during daytime in summer. Second, the results of the current analysis can be compared to two sets of simultaneous large-scale meteorological observations: those made by the Metropolitan Environmental Temperature and Rainfall Observation System (METROS) which is managed by the Tokyo Metropolitan Government, and those from an observational campaign conducted by the National Institute for Land and Infrastructure Management (NILIM) in collaboration with the Waseda University, the Tokyo Metropolitan University, and the Nippon Institute of Technology. Specifically, the simulated air temperatures are compared to the METROS air temperature data collected at 127 observation shelters (Stevenson screens) located at elementary schools across the 23 wards of Tokyo and to the NILIM campaign air temperature data collected at 173 locations in the coastal area of Tokyo Bay.

The present simulations are steady-state simulations using a Reynolds averaged Navier–Stokes (RANS) model and cannot reproduce unsteady flow behaviour unlike a large-eddy simulation (LES). The present study focuses on the analysis of the time-averaged wind and thermal fields over a wide area and not on that of unsteady flow behaviour, an analysis which would require much more extensive calculations than that from the present study. For example, Xie and Castro (2006) reported that in order to obtain a time-averaged solution, an LES simulation of flow over an array of cubes on a flat surface required 40 times as much computation time as that required for a RANS simulation with a k − ε model using the same mesh.

3.3. Meteorological data

In order to obtain the boundary values of the CFD computational domain, the LOCALS mesoscale model of ITOCHU Techno-Solutions Corporation, Tokyo, Japan (Matheson and Ashie, 2008) is run prior to the CFD analysis. The LOCALS model is run for two domains with one-way nesting. The horizontal lengths of the sides of the domain are 300 and 4.5 km, respectively, for a coarse grid simulation, and 100 and 1 km, respectively, for a fine grid simulation. The coarse grid simulation is run using initial and boundary values based on data from the Regional Spectral Model-Grid Point Value provided by the Japan Meteorological Agency. The period of analysis is from 2100 LST on 30 July 2005 to 2400 LST on 31 July 2005.

3.4. Effective volume fraction and open area ratio

As mentioned earlier, the FAVOR method (Hirt, 1993) is applied to account for the aerodynamic effects of objects that are smaller than the grid size. In this method, the geometric information of the buildings and the ground surface is converted into effective volume fraction and open area ratio. Here, the effective volume fraction is defined as the ratio of the volume occupied by fluid to the total volume of the grid cell. The open area ratio is defined for each of the six boundary faces of a grid cell as the ratio of the area occupied by fluid on a boundary face to the entire area of the boundary face. Using a similar approach that accounted for the porosity effects of grid cells, Hanna et al. (2004) performed CFD simulations of dispersion in urban areas and confirmed the validity of their model through comparisons with multiple observations. In this study, polygon data for each building are extracted from the geographic information system (GIS) data of Tokyo to calculate the effective volume fraction and the open area ratio within each grid cell using an accurate building height obtained from the aerial laser measurement data of the Geographical Survey Institute of Japan.

3.5. Water surfaces and trees

The water surfaces in the analysis region are identified using the GIS data and the 5-m digital elevation model data of the Geographical Survey Institute of Japan. The water surfaces identified are further classified into seawater and freshwater areas that are present inland. The latter includes water under bridges, water surfaces in parks, and green spaces in the urban area. The surface elevation of freshwater areas is set to 3 m below the elevation of the nearest grid point that represents the land surface. For the surface elevation of covered rivers, the elevation of the ground surface above the rivers is used.

The GIS data of Tokyo are limited for identifying areas covered by trees in the city. The GIS data do not specifically classify natural surface types. For example, the green space in temples and shrines is classified as common land, and urban green space is not further classified into woodland, lawn, or grassland. Therefore, areas containing trees are identified using digital data of vegetation in the 1990s prepared by the Tokyo Metropolitan Government.

3.6. Anthropogenic heat

The anthropogenic heat sources included in the analysis are buildings, automobiles, and factories. In order to estimate the sensible and latent heat released from buildings (1) the energy demand in August is estimated from the literature for each class of buildings according to their usage (e.g. office and housing) and (2) the heat source systems of buildings are deduced from construction records. A database of exhaust heat from automobiles is created using highway traffic census data and fuel consumption figures classified by types of vehicles. The heat discharge from factories into the atmosphere is calculated from the power and oil consumption, the heat rejection temperature of smoke stacks, and data on the operation status of power plants and incineration plants. The anthropogenic heat is classified into sensible heat and latent heat, and they are treated separately as source terms in the governing equations of the CFD model. The amount of anthropogenic heat in individual grid cells for the time of the simulation is estimated using the reported values in MLIT/MOE (2004). The estimation takes into consideration the location of anthropogenic heat release (Table V) as well as the total area of rooftops, building walls, and road surfaces in each grid cell.

Table V. Location of anthropogenic heat release (summer)
CategoryUseHeat release sourceLocation of release
BuildingsHousesAir conditioning systems for air cooling and ventilation; hot-water supply systems for baths and showersWall surface
 ApartmentsAir conditioning systems for air cooling and ventilation; hot-water supply systems for baths and showersWall surface
 Offices, business buildingsAir conditioning systems for air cooling and ventilation; hot-water supply systems for baths and showersRooftop surface
 HotelsAir conditioning systems for air cooling and ventilation; hot-water supply systems for baths and showersRooftop surface
Factories and plantsDistrict cooling facilitiesCooling towersRooftop surface
 FactoriesBoilersRooftop surface
  ChimneysHeight of building + 10 m
 Garbage and sewage plants and power plantsBoilersRooftop surface
  ChimneysHeight of chimney
TrafficTrunk roadsCars on expresswayHighest road surface
  Cars on bridge over waterRoad surface on bridge
  OthersGround surface
 Non-trunk roadsCars on bridge over waterRoad surface on bridge
  OthersGround surface

4. Simulation results for the 23 wards of Tokyo

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Analysis method
  5. 3. Input data for the simulation
  6. 4. Simulation results for the 23 wards of Tokyo
  7. 5. Case studies on increasing the sea breeze in urban spaces
  8. 6. Result: case with urban redevelopment
  9. 7. Summary
  10. Acknowledgements
  11. References

4.1. Air temperature distribution over the 23 wards of Tokyo

Figure 4 illustrates the simulated air temperature field at 10 m above the ground surface. The prevailing wind is nearly southerly at this hour. While the air temperature over Tokyo Bay is approximately 28 °C, that over the land is higher than 30 °C. In some locations, the air temperature over the land is more than 5 °C higher than the air temperature over the Bay. Areas farther north in the domain, that is farther downwind are characterized by higher air temperatures. The air temperature over Nerima is particularly high. In contrast, the air temperatures are relatively low in the coastal area at the north of Tokyo Bay.

thumbnail image

Figure 4. Simulated air temperature distribution at 10 m height above the ground surface. Line A shows the location of the cross-section in Figure 7

Download figure to PowerPoint

Figure 5 compares the horizontal distribution of air temperatures observed at the 127 observation sites operated by METROS to the air temperatures simulated at the same locations by the model. The simulated air temperatures in this figure represent those from 2 m above the ground. The figure also includes data observed at one point over the sea (35°27′52.09″ N, 139°52′28.35″ E) by MOE. The simulation and observation results show the same tendency of low air temperatures in the coastal area and of high air temperatures in the central metropolitan area and its leeward areas. The RMS error between the simulation results and the METROS observations from the 127 points is 1.1 °C.

thumbnail image

Figure 5. Air temperature distribution at the 127 METROS observation points

Download figure to PowerPoint

Figure 6 compares the horizontal distribution of air temperatures observed at the 173 sites by the NILIM observation campaign to the air temperatures simulated at the same locations by the model. The simulation results shown in the figure are from 5 m above the ground, while the NILIM campaign observations were conducted from 3 to 5 m above the ground. Although the observations were conducted mainly along streets, both the simulation results and observations clearly show the same tendency of low air temperatures near rivers and green spaces and of high air temperatures in built-up areas. The RMS error between the simulation results and the NILIM campaign observations from the 173 points is 1.1 °C.

thumbnail image

Figure 6. Air temperature distribution at the 173 points of NILIM observation campaign

Download figure to PowerPoint

4.2. Streak-patterned air temperature distribution

Close examination of the simulated air temperature distribution in Figure 4 reveals that locally within the computational domain, areas of high and low air temperatures occur alternately in a streak pattern. The simulated air temperature varies by 2–3 °C within a relatively small geographical area, which has implications for habitability and land use planning. In the coastal areas, the high-temperature streaks originate from the buildings. Cool sea breezes flow in from the bay, and heat released from the buildings generates streaky high-temperature zones downstream of the buildings. However, the width of individual high-temperature streaks is much larger and the spacing of the high-temperature streaks is wider in inland areas than in coastal areas.

4.3. Small-scale circulations over urban Tokyo

Figure 7 illustrates the simulated results of the air temperature and wind velocity distributions in the vertical cross-section along Line A in Figure 4. For clarity, the vertical component of the wind velocity is multiplied by ten in Figure 7. In the figure, counter-rotating vortices are shown, and up- and downward flows appear continuously along the cross section. The updraft and downdraft regions are characterized by high and low air temperatures, respectively, suggesting that the air temperature and wind velocity fields affect each other. For example, close to the surface in the vicinity of x = 12 200 m (marked by an ellipse in Figure 7), downward flow is shown. This flow transports cold air from aloft down to the surface, which then flows out into the surrounding area, cooling the air near the surface and transporting the heat released from the surface to the regions of upward flow.

thumbnail image

Figure 7. Wind velocity and air temperature in the vertical cross-section along Line A in Figure 4

Download figure to PowerPoint

Counter-rotating vortex rolls that are nearly aligned with the mean wind of the convective boundary layer are called horizontal convective rolls (HCRs), or horizontal-roll vortices, and are occasionally observed in satellite photographs of cloud streets (Etling and Brown, 1993). The depth of HCRs is usually the depth of the boundary layer. The aspect ratio (the ratio of the roll wavelength λ to the boundary layer depth H) of HCRs over natural landscapes ranges from 2:1 to 15:1 (Etling and Brown, 1993; Atkinson and Zhang, 1996). Miao and Chen (2008) confirmed the existence of HCRs over the Beijing urban area from observations and also simulated HCRs over the urban area in their mesoscale meteorological simulations with 500-m horizontal grid resolution. In their simulation results, the aspect ratio of HCRs over the urban area was approximately 1.5:1, with λ being approximately 3 km and H being approximately 2 km. In addition, they found that an increase in the urban building height decreased λ, while an increase in the anthropogenic heat increased H; hence, the aspect ratio of HCRs over urban areas is smaller than the typical value over natural landscapes.

In our simulation results, the aspect ratios of HCRs over the inland areas are between 2:1 and 4:1 with λ being 1–2 km and H being 500 m. If the computational domain height is made higher than that in the present setting, it is likely that the values of the aspect ratios will be smaller and close to the values of Miao and Chen (2008). In order to accurately simulate HCRs, it is necessary to use a computational domain height higher than the boundary layer height.

In mesoscale meteorological simulations using a grid cell spacing of a few kilometres, subgrid-scale circulations, such as the relatively small-scale circulations generated around buildings, are governed by the diffusion coefficients of the model. On the other hand, the CFD analysis in the present study uses 5-m horizontal grid spacing which enables a fine-scale analysis. The present analysis also employs an analysis area with horizontal dimensions of a few tens of km. Therefore, the CFD analysis is able to directly simulate atmospheric circulations and the influences of heat transfer occurring on fine scales up to a few km, the direct simulations of which were previously not feasible. Regional CFD analysis as in the present study is beneficial for investigating the details of the structure of heat transport above a city and in its overlying atmospheric layers.

5. Case studies on increasing the sea breeze in urban spaces

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Analysis method
  5. 3. Input data for the simulation
  6. 4. Simulation results for the 23 wards of Tokyo
  7. 5. Case studies on increasing the sea breeze in urban spaces
  8. 6. Result: case with urban redevelopment
  9. 7. Summary
  10. Acknowledgements
  11. References

Generally, urban development is accompanied by an increase in the total floor area over an entire district. Furthermore, the increase in building density as a result of urban development tends to reduce ventilation while creation of open spaces enhances ventilation. Accordingly, verticalization of urban structures which ensures an increase in floor area needs to be examined in terms of whether it can create open spaces and contribute to the enhancement of urban ventilation. Given this background, the CFD analysis method developed in the present study is applied to actual issues of urban redevelopment as case studies. The districts considered for the case studies are the Nihonbashi district and Tokyo Station and its surrounding district. In the former district, hereafter referred to as Area I, an elevated highway is scheduled to be removed. In the latter district, hereafter referred to as Area II, a shopping centre building at Tokyo Station is soon to be removed. With this urban redevelopment, continuous open space may be created in these districts. Accordingly, the influence of the open space on the urban thermal environment is investigated numerically. The present simulation provides a detailed and refined analysis of the relationship between the enhancement of ventilation and air temperature reduction.

The following two cases are considered for analysis:

  • Case 1: Case with present conditions (building area: 97.5 ha, floor area: 1153.3 ha)

  • Case 2: Case with urban redevelopment (building area: 96.6 ha, floor area: 1212.1 ha)

Figure 8 shows bird's-eye view of the computational domains for two cases. In the case studies, the horizontal grid spacing is set to 1 m in order to reproduce the urban configuration in detail. The digitalized data of the three-dimensional configuration of the elevated highway are used as input data for the numerical calculation. The heat released by railroads is also included in the case studies. Otherwise, the method of computation and that of input data creation are identical to those adopted for the simulation for the entire 23 wards of Tokyo. The case studies are performed for 1200 LST, 31 July 2005.

thumbnail image

Figure 8. Computational domains for Case 1 and Case 2

Download figure to PowerPoint

6. Result: case with urban redevelopment

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Analysis method
  5. 3. Input data for the simulation
  6. 4. Simulation results for the 23 wards of Tokyo
  7. 5. Case studies on increasing the sea breeze in urban spaces
  8. 6. Result: case with urban redevelopment
  9. 7. Summary
  10. Acknowledgements
  11. References

Figure 9 illustrates the simulated air temperature at 2 m above the ground and difference in air temperature and wind speed in Case 2 as compared to those in Case 1. The present figure is discussed in the following subsections.

thumbnail image

Figure 9. Simulated air temperature at 2 m above the ground in Case 2 and difference in air temperature and wind speed in Case 2, as compared to those in Case 1

Download figure to PowerPoint

6.1. Simulation results on air temperature distribution after urban redevelopment

With the implementation of the urban redevelopment, the elevated highway above the Nihonbashi River to the left of the Edobashi Junction in Area I has been removed and replaced by an underground highway (upper panel in Figure 9(a)). With this change, the air temperatures at and in the vicinity of the Nihonbashi River become approximately 1 °C lower than that in the surrounding areas. Particularly, the air temperature decrease is large in the areas marked by A and B probably because of the cool air that flows into these areas from the river. With the influence of the prevailing wind, air flows upriver, that is from right to left on the figure, and may be flowing into the regions marked A and B due to the inertia of the wind at the bends of the river.

In Area II, the air temperature is lower in the area marked by ‘a’ (Outer Garden of the Imperial Palace) than its surroundings (lower panel in Figure 9(a)). This result is consistent with the fact that a large portion of the district of the Outer Garden of the Imperial Palace is covered by water and greenery. In areas b and c, which are situated downwind of high rise buildings, the air temperatures are relatively low probably because of downward flow generated along the building walls. Finally, the air temperature in Area II is higher on average than that in Area I.

6.2. Simulation results on air temperature change resulting from urban redevelopment

The air temperature change in Area I that results from the urban redevelopment is illustrated in the upper panel of Figure 9(b). Air temperatures decrease in both areas A and B, however, the magnitude of the air temperature decrease is larger in area B than in area A. The spatial extent of air temperature decrease is also larger in area B than in area A. That is, the effects of the removal of the elevated highway, reduction in the building-to-land ratio, and reduction in the volume of buildings on the air temperature field is more evident downwind than upwind. In contrast, the air temperature increases in area C probably because of the change in the airflow as discussed below.

The lower panel of Figure 9(b) shows the simulated air temperature change in Area II that accompanies the implementation of the urban redevelopment. In areas a, b, c, d, and their surroundings, air temperatures decreased by more than 0.5 °C. The magnitude of the air temperature decrease in area a is particularly large, i.e. 1–2 °C. Because a large portion of area a is covered by green space, the area was initially characterized by low air temperatures even prior to the urban redevelopment. After the urban redevelopment, air with low temperature is transported down to the ground surface, resulting in further reduction of the air temperature near the ground surface in area a. Outside these four areas, air temperature increases of more than 0.5 °C are shown.

6.3. Simulation results on wind speed change resulting from urban redevelopment

Subsequent to the urban redevelopment, the wind speed generally increases over the Nihonbashi River area in Area I (upper panel of Figure 9(c)). The wind speed increase is attributable to the removal of obstacles to the wind such as the elevated highway and buildings. In area C, in which an air temperature increase is shown, a wind speed decrease results from the urban redevelopment. Upwind of area C, the removal of buildings along the river may reduce the impinged and diverted flows from the building walls on the opposite shore. As a result, the wind that flows into area C from the bend of the river may be modified, and the wind speed decreases in that area.

In Area II, increases in wind speed are shown in areas b, c, and d (lower panel of Figure 9(c)). These areas roughly coincide with areas in which air temperatures are lower in Case 2 than in Case 1 (lower panel of Figure 9(b)). In an area with large scalar wind speed, the air temperature decreases probably because of the enhanced exchange of near-surface air with colder air aloft, that is, because of enhanced ventilation effects. Wind speed changes occur in a highly complex manner. With the urban redevelopment, the shopping centre building at Tokyo Station is removed, and twin towers are added, one on either side of the vacated lot. The wind flow is diverted around the twin towers (arrows in Figure 9(c)), and a part of the flow enters Gyoko-Dori Avenue and contributes to the increase in wind speed in area d. Another part of the flow that is diverted around the twin towers streams through the former site of the shopping centre building and Tokyo Station. Part of this flow is funnelled into area b and splits further into northwestward (to the left in the figure) and northeastward (upward in the figure) flows along the streets. In area a, the wind speed decreases with the urban redevelopment despite the decreased air temperature shown there. The decrease in air temperature in this area is attributable to factors other than the scalar wind speed, such as the vertical heat diffusion.

6.4. Simulation results on air temperature change over the entire analysis area

The total horizontal land area of the computational domain in the two case studies is 375 ha. The urban redevelopment plan for the two areas investigated calls for the slimming down of the buildings. Thus, with the implementation of the plan, the ground surface area not covered by buildings within the domain increases by a little less than 1 ha. The total area in which air temperatures at 5 m above the ground surface are below 30 °C, is 128 ha for Case 1 and 140 ha for Case 2. Therefore, the area with temperatures below 30 °C increases by 12 ha with the implementation of the plan. The domain-averaged air temperature at 5 m above the ground surface is 29.26 °C for Case 1 and 29.21 °C for Case 2; i.e. the domain-averaged air temperature at this height is lower in Case 2 by 0.043 °C than in Case 1. The air temperature averaged over the entire domain after the urban redevelopment becomes slightly lower than that before the urban redevelopment. Subsequently, the relationship is examined between the change in the building-to-land ratio and the change in the air temperature that results from the urban redevelopment. The relationship reveals that reduction of the gross building-to-land ratio by 1% corresponds to an air temperature reduction of 0.16 °C. In addition, the CFD analysis results for the entire 23 wards of Tokyo yield an air temperature reduction of 0.08 °C (r2 = 0.52, 4 356 500-m grid data values) for a reduction of the gross building-to-land ratio by 1%. Various factors such as site conditions differ between the areas analysed for the case studies and the entire 23 wards of Tokyo, thus, comparison of the rate of air temperature reduction for a given change in the gross building-to-land ratio is not straightforward. However, our analyses suggest that reduction in the gross building-to-land ratio has the effect of reducing the air temperature.

7. Summary

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Analysis method
  5. 3. Input data for the simulation
  6. 4. Simulation results for the 23 wards of Tokyo
  7. 5. Case studies on increasing the sea breeze in urban spaces
  8. 6. Result: case with urban redevelopment
  9. 7. Summary
  10. Acknowledgements
  11. References

With the use of a CFD model, airflow was simulated over a 33 × 33 km2 area. This area included the 23 wards of Tokyo and extended from the coastal area of Tokyo Bay to areas inland. The simulation was performed with a horizontal resolution of 5 m. The total number of grid points including the buffer zone was approximately 5 billion. The present simulation yielded detailed distributions of air temperature and wind speed over the area of analysis. The simulation results reproduced air temperature tendencies found in the observations such as lower air temperatures in coastal areas, rivers, and green spaces, and higher air temperatures in built-up areas. The RMS error between the simulation results and observations was 1.1 °C.

Furthermore, the CFD simulation showed that the air temperature distribution over urban Tokyo occurs in a streak pattern. This phenomenon was associated with the formation of small-scale atmospheric circulation cells. Streak-patterned distributions of air temperature may form in the atmosphere above an area with various land use types and/or above urban spaces on the spatial scale of a few km. Extensive air temperature observations over the areas of interest have not been easily feasible, thus no sufficient data are present to confirm the existence of streak-patterned distributions of air temperature over the urban area of Tokyo. However, in recent years, lidar observations have frequently been made to acquire detailed observational data of the urban boundary layer. The development of new lidar-based observational techniques can be anticipated, and they may be applicable to confirm a streak-patterned distribution of air temperature over an urban area.

The CFD-based simulation technique, which was developed for the urban thermal environment in the present study, was also applied to the redevelopment of urban Tokyo. The urban redevelopment plan calls for the removal of an elevated highway and a shopping centre located at a train station. According to the calculations in the present study, the removal of these urban structures would decrease the area covered by buildings by 1 ha, and increase the area with temperatures below 30 °C by 12 ha. Furthermore, the CFD simulation indicated that near-surface air temperatures decreased locally by as much as 1 to 2 °C with the implementation of the urban redevelopment plan. The air temperature averaged over the entire analysis area also decreased by a small amount.

It is well known that changes in urban morphology frequently become an issue in terms of the aesthetics of the urban landscape. The present study showed that changes in urban morphology influence the urban thermal environment in addition to the urban aesthetics. The simulation technique for the urban thermal environment that was developed in the present study allows the air temperature change caused by an urban redevelopment plan to be quantified. While the present study provided analysis results for a summer day, comprehensive urban planning that takes into account all seasons requires analysis results from seasons and times of day other than those considered in the present study. The results of the simulation can be used to discuss how optimally modify the urban morphology with urban development contractors, residents, and the local government so that an urban space of comfortable thermal environment can be created.

Acknowledgements

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Analysis method
  5. 3. Input data for the simulation
  6. 4. Simulation results for the 23 wards of Tokyo
  7. 5. Case studies on increasing the sea breeze in urban spaces
  8. 6. Result: case with urban redevelopment
  9. 7. Summary
  10. Acknowledgements
  11. References

The present research was conducted as a part of a research assignment funded by the ‘Grant for Operating Costs’ of the Building Research Institute (BRI) and also as a part of a collaborative project with the Earth Simulator Center of the Japan Agency for Marine-Earth Science and Technology (JAMSTEC). The authors are grateful to Dr Koji Kagiya (National Institute for Land and Infrastructure Management) for providing information on urban redevelopment in metropolitan Tokyo and NILIM campaign observation data. The authors also thank Dr Hitoshi Yokoyama (Tokyo Metropolitan Research Institute for Environmental Protection) for providing METROS observation data.

References

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Analysis method
  5. 3. Input data for the simulation
  6. 4. Simulation results for the 23 wards of Tokyo
  7. 5. Case studies on increasing the sea breeze in urban spaces
  8. 6. Result: case with urban redevelopment
  9. 7. Summary
  10. Acknowledgements
  11. References
  • AIJ (Architectural Institute of Japan). 2000. Klimaatlas of urban environment—Urban planning utilizing climatic information. Gyosei: Tokyo, Japan, ISBN978-4-3240-6278-4, (in Japanese).
  • Atkinson BW, Zhang JW. 1996. Mesoscale shallow convection in the atmosphere. Reviews of Geophysics 34: 403431.
  • Baik JJ, Park SB, Kim JJ. 2009. Urban flow and dispersion simulation using a CFD model coupled to a mesoscale model. Journal of Applied Meteorology and Climatology 48: 16671681.
  • Blocken B, Persoon J. 2009. Pedestrian wind comfort around a large football stadium in an urban environment: CFD simulation, validation and application of the new Dutch wind nuisance standard. Journal of Wind Engineering and Industrial Aerodynamics 97: 255270.
  • Bou-Zeid E, Overney J, Rogers BD, Parlange MB. 2009. The effects of building representation and clustering in large-eddy simulations of flows in urban canopies. Boundary-Layer Meteorology 132: 415436.
  • Burrows DA, Hendricks EA, Diehl SR, Keith R. 2007. Modeling turbulent flow in an urban central business district. Journal of Applied Meteorology and Climatology 46: 21472164.
  • Chan ST, Leach MJ. 2007. A validation of FEM3MP with Joint Urban 2003 data. Journal of Applied Meteorology and Climatology 46: 21272146.
  • Coirier W, Kim S, Marella S, Mayes J, Chen F, Michalakes J, Miao S. 2007. Progress towards a coupled mesoscale and microscale modeling capability. In The Seventh Conference on Coastal Atmospheric and Oceanic Prediction and Processes joint with the Seventh Symposium on the Urban Environment, San Diego, CA, 10–13 Sept 2007. Available from http://ams.confex.com/ams/7Coastal7Urban/techprogram/paper_127157.htm.
  • Etling D, Brown RA. 1993. Roll vortices in the planetary boundary layer: a review. Boundary-Layer Meteorology 65: 215248.
  • Flaherty J, Stock D, Lamb B. 2007. Computational fluid dynamics simulations of plume dispersion in urban Oklahoma city. Journal of Applied Meteorology and Climatology 46: 21102126.
  • Hanna SR, Hansen OR, Dharmavaram S. 2004. FLACS CFD air quality model performance evalumation with Kit Fox, MUST, Prairie Grass, and EMU observations. Atmospheric Environment 38: 46754687.
  • Hanna SR, Hansen OR, Ichard M, Strimaitis D. 2009. CFD model simulation of dispersion from chlorine railcar releases in industrial and urban areas. Atmospheric Environment 43: 262270.
  • Hendricks EA, Diehl SR, Burrows DA, Keith R. 2007. Evaluation of a fast-running urban dispersion modeling system using joint urban 2003 field data. Journal of Applied Meteorology and Climatology 46: 21652179.
  • Hirt CW. 1993. Volume-fraction techniques: powerful tools for wind engineering. Journal of Wind Engineering and Industrial Aerodynamics 46–47: 327338.
  • Huang H, Ooka R, Kato S. 2005. Urban thermal environment measurements and numerical simulation for an actual complex urban area covering a large district heating and cooling system in summer. Atmospheric Environment 39: 63626375.
  • Ichinose T, Shimodozono K, Hanaki K. 1999. Impact of anthropogenic heat on urban climate in Tokyo. Atmospheric Environment 33: 38973909.
  • Ichinose T, Ashie Y, Hirano Y, Kono T. 2005. The high-resolution numerical model of heat island phenomena. CGER's Supercomputer Activity Report 12–2003: 115121. Available from http://www-cger.nies.go.jp/publication/I061/I061.pdf.
  • Ichinose T, Ashie Y, Komatsu T, Kono T. 2007. Numerical simulation of thermal and airflow field around regularly arrayed buildings. CGER's Supercomputer Activity Report 14–2005: 3540. Available from http://www-cger.nies.go.jp/publication/I070/I070.pdf.
  • IBEC (Institute for Building Environment and Energy Conservation). 2006. Comprehensive assessment system for building environment efficiency CASBEE-HI (heat island): evaluation manual. Available from http://www.ibec.or.jp/CASBEE/english/index.htm (in Japanese).
  • Ihara T, Aida H, Yoshida Y, Handa T, Matsuhashi R, Ishitani H. 2003. Evaluation of CO2 emissions reduction effect from the building by introducing high light-reflective and high heat-emissive paint considering urban thermal environment. In Proceedings of the 19th Conference on Energy, Economy, and Environment Tokyo, Japan, 655660. Available from http://ci.nii.ac.jp/naid/10021967094/en/ (in Japanese).
  • Matheson MA, Ashie Y. 2008. The effect of changes of urban surfaces on rainfall phenomenon as determined by a non-hydrostatic mesoscale model. Journal of the Meteorological Society of Japan 86–5: 733751.
  • Miao S, Chen F. 2008. Formation of horizontal convective rolls in urban areas. Atmospheric Research 89: 298304.
  • MLIT/MOE (Ministry of Land Infrastructure, Transport and Tourism, Ministry of the Environment). 2004. FY 2003 survey reports on heat-island countermeasures by controlling anthropogenic heat release in urban areas. Available from http://www.mlit.go.jp/sogoseisaku/heat_island/01.pdf (in Japanese).
  • Mochida A, Yoshino H, Iwata T, Tabata Y. 2006. Optimization of tree canopy model for CFD prediction of wind environment at pedestrian level. In The 4th International Symposium on Computational Wind Engineering, Yokohama, Japan, 561564. Available from http://www.aij.or.jp/jpn/publish/cfdguide/TB2-01.pdf.
  • MOE (Ministry of the Environment). 2005. Technology for mitigating urban heat-island effect (Technology for reducing the sensible heat emitted from outdoor heat exchangers of air conditioners) Instructions for demonstration. Available from http://www.env.go.jp/press/file_view.php?serial = 6645&hou_id = 5902 (in Japanese).
  • Nozu T, Tamura T, Okuda Y, Sanada S. 2008. LES of the flow and building wall pressures in the center of Tokyo. Journal of Wind Engineering and Industrial Aerodynamics 96: 17621773.
  • Oguro M, Morikawa Y, Murakami S, Matsunawa K, Mochida A, Hayashi H. 2008. Development of a wind environment database in Tokyo for a comprehensive assessment system for heat island relaxation measures. Journal of Wind Engineering and Industrial Aerodynamics 96: 15911602.
  • Tamura T. 2008. Towards practical use of LES in wind engineering. Journal of Wind Engineering and Industrial Aerodynamics 96: 14511471.
  • Stuben K. 2001. A review of algebraic multigrid. Journal of Computational and Applied Mathematics 128: 281309.
  • Tokyo Metropolitan Government Bureau of Environment. 2005. Manual for the planning system of building environment in Tokyo metropolis. Available from http://www7.kankyo.metro.tokyo.jp/building/doc/060_h_2010.pdf (in Japanese).
  • Xie Z-T, Castro IP. 2006. LES and RANS for turbulent flow over arrays of wall-mounted obstacles. Flow Turbulence and Combustion 76: 291312.
  • Xie Z-T, Castro IP. 2009. Large-eddy simulation for flow and dispersion in urban streets. Atmospheric Environment 43: 21742185.
  • Yoshida S, Murakami S, Mochida A, Ooka R, Tominaga Y, Kim S. 2000. Influence of green area ratio on outdoor thermal environment with coupled simulation of convection, radiation and moisture transport. Journal of Architecture Planning and Environmental Engineering (Transactions of AIJ) 529: 7784, (in Japanese).