ARTICLE IN PRESS

Building and Environment 43 (2008) 480493 www.elsevier.com/locate/buildenv

Temperature decreases in an urban canyon due to green walls and green roofs in diverse climates
Eleftheria Alexandria, Phil Jonesb,
b

Mantzakou 2-6, 114 73 Athens, Greece Welsh School of Architecture, Cardiff University, King Edward VII Avenue, Cardiff CF10 3NB, UK Received 23 February 2006; received in revised form 24 July 2006; accepted 31 October 2006

Abstract This paper discusses the thermal effect of covering the building envelope with vegetation on the microclimate in the built environment, for various climates and urban canyon geometries. A two-dimensional, prognostic, micro scale model has been used, developed for the purposes of this study. The climatic characteristics of nine cities, three urban canyon geometries, two canyon orientations and two wind directions are examined. The thermal effect of green roofs and green walls on the built environment is examined in both inside the canyon and at roof level. The effects of this temperature decrease on outdoors thermal comfort and energy savings are examined. Conclusions are drawn on whether plants on the building envelope can be used to tackle the heat island effect, depending on all these parameters taken into consideration. r 2006 Elsevier Ltd. All rights reserved.
Keywords: Green roofs; Green walls; Urban canyon; Plants; Built environment

1. Introduction Since the beginning of human existence man has clearly intended to alter his microclimate, to a more humanfriendly one, protecting himself from extreme climatic conditions. Even from the rst evidence of Neolithic houses and settlements, it is obvious that they were not sited in a purely natural environment, but in a part of nature transformed according to a human plan [1]. With the evolution of human societies, settlements were transformed, evolved into villages, towns or cities, developed or faded away, according to the geographical, economic, social and cultural transformations taking place throughout time. With the Industrial Revolution, urban spaces expanded dramatically, much faster and with much more signicant changes than in their previous evolutionary periods. The large areas modern cities occupy, their structure, materials and the general lack of vegetation cannot but have altered the climatic characteristics of urban spaces.
Corresponding author.

These changes have a direct effect on the local climate of urban spaces, especially the central parts of the city, causing a signicant rise of the urban temperature and other alterations, known as the heat island effect. This may cause serious local climatic unpleasant conditions and even imperil human health, especially for cities in climates with a distinctively hot season [2,3]. The moderation of extreme heat in the local environment of such climates could mean not only their sustainability, but also the potential of occupying them without the morbidity and mortality risks caused by excessive heat [4,5]. On prima facie evidence, the general lack of vegetation in existing cities is one of the factors affecting the formation of raised urban temperatures. In most urban spaces, appreciable amounts of vegetation exist mostly concentrated in parks or recreational spaces. Although parks manage to lower temperatures within their vicinity [69], they are incapable of thermally affecting the concentrated built spaces where people live, work and spend most of their urban lives. By placing vegetation within the built space of the urban fabric, raised urban temperatures can decrease within the human habitats themselves and not only in the detached spaces of parks. Urban surfaces which

0360-1323/$ - see front matter r 2006 Elsevier Ltd. All rights reserved. doi:10.1016/j.buildenv.2006.10.055

ARTICLE IN PRESS
E. Alexandri, P. Jones / Building and Environment 43 (2008) 480493 481

Nomenclature Latin Letters C c 1: c 2: cc cpa cp D I1 KHx KHz KEx KEz q net sensible heat loss on leaf tissue (W/m2) isobaric specic heat of component 1 (moisture) of the mixture (J/kg K) isobaric specic heat of component 2 (air) of the mixture (J/kg K) building material specic heat capacity (J/kg K) isobaric specic heat capacity of air (J/kg K) specic heat capacity of the leaf tissue (J/kg K) binary diffusion coefcient (m2/s) any source of mass of moisture (kg/m3s) eddy diffusion coefcient of energy in x-axis (m2/s) eddy diffusion coefcient of energy in z-axis (m2/s) eddy diffusion coefcient of water vapour in xaxis (m2/s) eddy diffusion coefcient of water vapour in zaxis (m2/s) relative concentration of water vapour, expressed as specic humidity (kg/kg)

qE t T Tin T Tout U u w

heat gains/losses from the buildings fabric (W/m2) time (s) air temperature (K) indoors air temperature (K) leaf surface temperature (K) outdoors air temperature (K) average buildings fabric U-value (W/m2 K) air velocity in x axis (m/s) air velocity in z axis (m/s)

Greek Letters aa ac am e l lE ra r Fn coefcient of thermal diffusivity of air (m2/s) building material thermal diffusion coefcient (m2/s) diffusion coefcient of moisture in the building material (m2/s) evaporation number of the building material latent heat of vaporisation (J/kg) net latent heat loss on leaf tissue (W/m2) density of moist air (kg/m3) density of the leaf tissue (kg/m3) net heat gain on leaf tissue from radiation (W/m2)

are not used, such as the building envelope (walls and roofs), could easily be covered with vegetation and alter the microclimate of the built environment, as well as the local climate of the city. The magnitude of temperature decreases due to this transformation depends on the climatic characteristics, the amount of vegetation and urban geometry. This paper presents the results of a quantitative research on how the heat island effect can be tackled by covering the envelope of urban buildings with vegetation. The aim of this research has been to assess the potential of mitigating raised urban temperatures through vegetation, for different urban geometries and climates.

urban

10m

urban

urban

in ground W a1 a2 in

in

Fig. 1. Two-dimensional canyon model.

heat transfer [11,12]: 2. Methodology A two-dimensional, prognostic (dynamic) micro-scale model has been developed and programmed in C++, describing heat and mass transfer in a typical urban canyon (Fig. 1). The differential equations describing heat and mass transfer in the air, building materials (considered as capillaryporous bodies), soil and vegetation have been solved with nite differences approximations, where surface elements are replaced by nodes [10]. The effect of vapour gradients on temperature gradients has been described analytically in the air nodes near the surfaces as the effect of diffusion on   dT l c1 c2 raa rT I1 DrqrT, dt cpa ra cpa dq 1 rDrq I 1 , dt ra (1)

(2)

where T is the air temperature (in K), t is time (in s), aa the coefcient of thermal diffusivity of air (in m2/s), D the binary diffusion coefcient (in m2/s), cpa the isobaric specic heat capacity of air (in J/kg K), ra is the density of moist air (in kg/m3), l the latent heat of vaporisation (in J/kg), I1 is any source of mass of moisture (in kg/m3s), q is the relative concentration of water vapour, expressed as

ARTICLE IN PRESS
482 E. Alexandri, P. Jones / Building and Environment 43 (2008) 480493

specic humidity (in kg/kg), c1 is the isobaric specic heat of component 1 (moisture) of the mixture (in J/kg K) and c2 the isobaric specic heat of component 2 (air) of the mixture (in J/kg K). According to Eckert and Drake [12], the effect of thermal diffusion on heat transfer (fourth term of Eq. (1)) can be neglected in normal engineering masstransfer processes. However, they point out [12] that it contributes essentially when temperature gradients are extremely large. In the description of the thermal exchanges in the built environment, where temperature gradients in the boundary layer of surfaces exposed to direct solar radiation are relatively large, the expression of thermal diffusion is essential for the accurate description of the phenomenon. When air velocity is considered, eddy diffusion is much stronger than molecular diffusion (conduction) in the air in the atmosphere away from the boundary layer of the surface. Despite the fact that molecular diffusion always takes place in the air, it is omitted from both heat and mass transfer in the air at these levels, as it is 104105 smaller than eddy diffusion [13]. The effect of vapour gradients onto temperature in the air nodes well above the ground is expressed through eddy diffusion coefcients [14,15]. Heat and mass transfer in the two-dimensional model of the binary airwater vapour mixture thus becomes:       q q q q qT q qT u w K Hz K Hx T , qt qx qz qz qz qx qx (3)       q q q q qq q qq u w K Ez K Ex q , qt qx qz qz qz qx qx (4)

Regarding plants, they are considered to be a layer consisting of canopy leaves and the air among them. Equations describing heat and mass transfer in the air are the ones given by Eqs. (1) and (2), while heat transfer in the leaf is given by dT Fn C lE, (7) dt where r is the density of the leaf tissue (in kg/m3), cp is the specic heat capacity of the leaf tissue (in J/kg K), T is the leaf surface temperature (in K), Fn is the net heat gain from radiation (in W/m2), C is the net sensible heat loss (in W/m2) and lE is the net latent heat loss (in W/m2). Radiative heat exchanges between the canyon surfaces have been described analytically, according to the radiative heat transfer theory [16] and not with the use of a combined convection and radiative heat transfer coefcient. Thus the radiative heat exchanges between surfaces with different emissivities in closed enclosures is expressed by  N  qi X 1 1 F ij qj H 0i j i j1 r cp
N X j1

F ij E bi E bj

for i 1; 2; . . . ; N,

where KHz and KEz are the eddy diffusion coefcient of energy and water vapour, respectively, in the vertical axis and KHx and KEx are the respective diffusion coefcients in x-axis. The expression of these diffusion coefcients is given by the MoninObukhov similarity theory [14,15]. The water vapour gradients are taken into consideration in the calculation of the eddy diffusion coefcients of energy. Regarding solid materials, they are considered as a system, consisting of a capillary-porous building material in the medium of wet air and in a region of positive temperatures (no ice). The equations describing the onedimensional heat and mass transfer can be expressed by [11] dT q2 T l qq ac 2 , dt qz cc qt dq q2 q am 2 , dt qz (5)

(6)

where ac is the building material thermal diffusion coefcient (in m2/s), cc is the building material specic heat capacity (in J/kg K), e is the evaporation number of the building material, and am is the diffusion coefcient of moisture in the building material (in m2/s).

where qi is the radiation emitted from the surface i (W/m2), ei is the i surfaces emmisivity and ej is the emmisivity of the rest of the surfaces, Fij is the view factor of surface i towards surface j, H0i is any external radiation arriving at surface i, and Ebi equals to sTi4, where s is the StefanBoltzmann constant (5.67 108 W/m2 K4) and Ti is the temperature of the ith surface. Climatic characteristics, such as air temperature, relative humidity and wind speed, are set as the boundary nodes of the model, placed 10 m above the upper part of roofs. These climatic characteristics, as well as solar radiation derive from meteorological data from METEONORM [17]. Solar radiation is input onto the surfaces, according to their orientation, inclination and shading pattern. The shading pattern, determined by the canyon geometry and the geographic latitude, was calculated with the software ECOTECT [18], where the same canyon geometries, as the ones described below were input, for the different latitudes and longitudes examined. Air velocities in the vicinity of the canyon were calculated with the CFD code WinAir4 [19]. WinAir4 is an in-house code, which uses the xed viscosity models, with a variation on simple solution scheme. The canyon geometries in the CFD model are the same as for the heat and mass transfer model. As the CFD model is three-dimensional, the length of the canyon was 40 m and the rest of the canyon dimensions (height, width) varied, according to the canyon geometry, as discussed below. The CFD code mesh was also the same as for the two-dimensional heat and mass transfer model, as described in Fig. 1. The grid is not uniform; near the building and road surfaces the grid is 0.30 m, while two

ARTICLE IN PRESS
E. Alexandri, P. Jones / Building and Environment 43 (2008) 480493 483

nodes away from the surface, the grid varies according to the canyon geometry, from 1 to 3 m for the canyon width, and from 1 to 2 m for the canyon height. For the length, it is constantly set at 2 m. At the boundary surface are set 10 m away from the roof of the buildings, 10 m away from the windward surface and 100 m away from the leeward surface when the building height is 5 m, and 10, 20 and 200 m, respectively, when the building height is 10 m. The input climatic data of wind speed, air temperature and relative humidity are input as the boundary conditions for each hour of the diurnal prole. The heat gains on buildings and the street are calculated according to the solar radiation absorbed by the surface, which depends on its orientation, and shading pattern, the latter having been dened by [18]. The output air velocities from the CFD code in the middle of the canyon are input at the respective nodes of the two-dimensional heat and mass transfer model. Four types of vegetation covering the building envelope are examined for each canyon geometry: (a) a base case, where no green is placed in and around the canyon, referred to as the no-green case, (b) the green-roofs case, where both roofs are covered with vegetation (ground-covering grasses) (c) the green-walls case, where both walls inside the canyon are covered with vegetation (ivies) and (d) the green-all case, where both roofs and walls are covered with vegetation. Three types of canyon geometries are examined, according to the wind ow developed in each: (a) a canyon with height (H) 10 m and width (W) 5 m, referred to as H10W5 canyon, where, according to Santamouris [6], skimming ow is developed, with very low air velocities, and sun shaded, (b) a canyon with 5 m height and 10 m width, referred to as H5W10 canyon, where wake interference ow is developed, with bigger air velocities and more exposed surfaces to direct solar radiation and (c) the H5W15 canyon, with 5 m height and 15 m width, where isolated roughness ow is developed, with much larger air velocities, and greater exposure to solar radiation. The canyons are examined with two orientations: (a) one where the canyons axis was parallel to the EastWest axis (referred to as the EW canyon) and (b) one where the canyons axis was parallel to the NorthSouth axis (referred to as the NS canyon). Two directions of wind ow are considered: (a) one aligned to (referred as x) and (b) one parallel to the canyons axis (referred as y).

Buildings are made of concrete, and the street is covered with asphalt. A summary of the hydrothermal properties of the materials and vegetation considered in the canyons is made in Table 1. All these cases are examined for nine cities in nine different types of climates, where cities and evapotranspiring vegetation can be found. Based on Koeppens climatic classication [20], the nine cities studied, and the climatic type in which they belong, are summarised in Table 2. All cases are examined for a typical day of their hottest month. Their climatic data have derived from hourly data from METEONORM [17]. The effect of vegetation on the urban texture of each city is examined for its hottest month. For Athens, Hong Kong, London, Montreal, Moscow and Riyadh, July is chosen as the hottest month, while for Mumbai May is used, for Beijing June, and for Bras lia September. The typical day
Table 1 Hydrothermal properties of plants, soil, building materials (concrete) and street materials Characteristic Specic thermal capacity (MJ/m3K) Thermal conductivity (W/mK) Vapour diffusivity (106 m2/s) Ratio of vapour diffusion coefcient to total moisture diffusion coefcient Emissivity Albedo Hydraulic conductivity (104 m/s) Moisture potential, when soil is saturated (cm) Maximum volumetric water content (m3/m3) Coefcient b Convective heat resistance (s/m) Resistance expressing the plant type (s/m) Canopy extinction coefcient Level of soil moisture below which permanent wilting of the plant occurs (m3/m3) Concrete 1.60 1.70 0.55 0.20 Asphalt 2.00 1.30 1.58 0.10 Soil 1.15 Plants 2.60

0.94 0.23

0.81 0.10

0.94 0.23 0.01 49.0 0.492 10.40

0.94 0.30 200 100 1.4 0.25

Table 2 Table of cities studied City London, UK Montreal, Canada Moscow, Russia Athens, Greece Beijing, China Riyadh, Saudi Arabia Hong Kong, China Mumbai, India Bras lia, Brazil Climate Temperate Subarctic Continental cool summer Mediterranean Steppe Desert Humid subtropical Rain forest Savanna Location 51.32N, 0 45.31N, 73.34W 55.45N, 37.37E 37.59N, 23.43E 39.48N, 116.23E 24.38N, 46.43E 22.16N, 114.12E 18.54N, 72.5E 15.48S, 47.54W

ARTICLE IN PRESS
484 Table 3 Maximum, minimum, average and day-time average values of climatic characteristics (total solar radiation on a horizontal plane, air temperature, relative humidity, wind speed) for the 24-h prole of the typical day of the hottest month of each city Daytime average E. Alexandri, P. Jones / Building and Environment 43 (2008) 480493

Wind speed

Max

Day-time average

6.4 4.3 4.7 4.9 4.4 5.5 5.2 4.6 5.9

of the hottest month is a 24-h prole whose climatic data are averaged for each hour of the month examined, to make a diurnal typical climatic prole for the month studied. The maximum, minimum, average and daytime average quantities of temperature, relative humidity, wind speed and solar radiation on a horizontal plane are presented in Table 3. In Figs. 25 the diurnal prole of air temperature, relative humidity, solar radiation on a horizontal plane and wind speed are presented for the typical day of the hottest month of all nine cities. 3. Discussion and analysis

Average

Min

45.9 57.9 52.1 70.9 70.5 62.1 71.6 62.4 32.4

3.5 3.3 3.2 3.0 3.9 3.3 3.5 2.3 3.8

4.8 3.8 3.8 3.8 4.2 4.3 4.3 3.3 4.8

5.2 4.0 4.0 4.2 4.2 4.6 4.5 3.8 5.2

3.1. Direct cooling effects Air and surface temperatures lower signicantly in all climates examined, when walls and roofs are covered with vegetation, as can be observed in Figs. 710. The heat uxes on the vegetated and on the non-vegetated surfaces are very different. As can be observed in Fig. 6 for a green roof and for a concrete roof in Montreal, the 24-h prole of
Total Solar Radiation on a Horizontal Plane 900 800 Solar Radiation (W/m2) 700 600 500 400 300 200 100 0 1 3 5 7 9 11 13 15 Time (Hours) 17 19 21 23
Athens Beijing Brasilia Hong Kong London Montreal Moscow Mumbai Riyadh

Average

Relative humidity

Min

Max

Daytime average Average Air temperature Min Max Daytime average

30.1 29.1 28.0 33.1 19.7 25.9 20.9 33.5 42.8

22.3 19.6 20.4 26.7 14.5 17.1 13.9 25.3 31.2

25.7 24.2 23.5 29.5 17.2 21.2 17.3 28.9 36.5

27.0 25.6 25.4 30.9 18.0 22.3 18.0 30.8 38.3

57.4 75.0 65.7 85.1 77.3 77.7 86.3 79.1 45.8

37.3 47.1 44.0 62.9 63.5 49.7 60.3 53.1 23.5

50.2 62.7 58.6 77.1 74.3 66.1 74.7 69.5 36.2

Total solar radiation on a horizontal plane

Fig. 2. Twenty four-hour prole of the total solar radiation on a horizontal plane, for the hottest month of each city, which is input on the unshaded, horizontal surfaces of the heat and mass transfer model.

Average

457.5 331.9 437.8 417.2 286.5 376.5 297.9 502.1 466.9

298.4 216.4 228.4 235.8 199.3 261.9 220.2 283.8 284.2

Air Temperature 45 40 Temperature (C) 35 30 25 20 15 10 5 0 1 3 5 7 9

Athens Hong Kong Moscow Beijing London Mumbai Brasilia Montreal Riyadh

Min

Athens, July Beijing, June Bras lia, September Hong Kong, July London, July Montreal, July Moscow, July Mumbai, May Riyadh, July

809.0 589.2 726.0 704.3 497.3 692.2 527.8 846.9 850.0

Max

0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0

11 13 15 Time (Hours)

17

19

21

23

Fig. 3. Twenty four-hour prole of air temperature for the hottest month of each city, which is input at the boundary nodes of the heat and mass transfer model.

City

ARTICLE IN PRESS
E. Alexandri, P. Jones / Building and Environment 43 (2008) 480493 485

Relative Humidity 100 90 80 70 60 50 40 30 20 10 0 1 3 5 7 9

Athens Hong Kong Moscow

Beijing London Mumbai

Brasilia Montreal Riyadh

11 13 15 Time (Hours)

17

19

21

23

Fig. 4. Twenty four-hour prole of relative humidity for the hottest month of each city, which is input at the boundary nodes of the heat and mass transfer model.

Wind Speed 7 6 Wind Speed (m/s) 5 4 3 2 1 0 1 3 5 7 9 11

Athens Hong Kong Moscow Beijing London Mumbai Brasilia Montreal Riyadh

13

15

17

19

21

23

Time (Hours)
Fig. 5. Twenty four-hour prole of wind speed for the hottest month of each city, which is input at the boundary nodes of the heat and mass transfer model.
400 300 Heat Flux Density (W/m2) 200 100 0 -100 -200 -300 -400 -500 -600 Time (Hours)
Conv, rf-gr Evap, rf-gr Rad, rf-gr Cond, rf-con Conv, rf-con Evap, rf-con Rad, rf-con 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24

external surface of the concrete roof is much larger than the convective heat ux density in the upper part of the green roofs canopy. For the concrete roof its 24-h prole ranges from 345.1 to 128.6 W/m2, while for the green roofs upper surface, it only ranges from 51.3 to 99.9 W/m2. The convective heat exchanges between the grass foliage and the air are milder than those between the solid concrete roof and the air. The total radiative heat ux density (both short and long wave radiation) on the external surface of the concrete roof is also larger than that on the upper part of the canopy layer. It ranges from 158.2 to 355.1 W/m2 on the concrete roof and from 38.8 to 229.5 W/m2 on the green roof. Due to the redistribution of radiation within the vegetated layer, the total radiative heat exchanges are smaller on the vegetated surface, when compared with the concrete roof. As can be observed in Eq. (7), the conductive heat component is omitted in the relationship governing heat transfer in plants as too small [1315,21,22], while it is an important factor in the heat transfer of a concrete roof, with a range from 444.5 to 154.5 W/m2 on the external part of the roof. Nonetheless, the greatest differences are observed at the evaporative heat uxes, which range from 46.3 to 170.6 W/m2 for the concrete roof and from 593.2 to 26.4 W/m2 for the green roof. As the evaporative heat transfer on the green roof acts constantly as a heat sink and the radiative energy absorbed by the green roof is smaller than that absorbed by the concrete roof, the energy uxes on a green surface can only offer lower surface and air temperatures, when compared to those produced by concrete surfaces. Because of these energy distributions, canyon air temperature lowers the most when both walls and roofs are covered with vegetation in all climates examined. This can be explained by the fact that when roofs are covered with vegetation, air masses enter the canyon much cooler, from the vegetated roofs. On the other hand, when only walls are covered with vegetation, air masses enter the canyon heated by the plain roofs, which absorb the quite

Relative Humidity (%)

Decrease of canyon air temperature, green-all case 0 4 Temperature Decrease (C) -2 -4 -6 -8 -10 -12 Time (Hours)
Fig. 7. Air canyon temperature decrease in the EW, H5W10 canyon, with parallel wind ow, when both roofs and walls are covered with vegetation, for all climates examined.

12

16

20

24

Athens Beijing Brasilia HongKong London Montreal Moscow Mumbai Riyadh

Fig. 6. Convective (Conv), evaporative (Evap), long and short-wave radiative (Rad) and conductive (Cond) heat uxes on a concrete roof (rf con) and on a green roof (rf-gr) in Montreal.

the convective, conductive, evaporative and radiative heat uxes on the green and the concrete roofs differ signicantly. The convective heat ux density at the

ARTICLE IN PRESS
486 E. Alexandri, P. Jones / Building and Environment 43 (2008) 480493

Temperature Decrease (C)

great amounts of summer insolation. For this reason, in the instance of Hong Kong, canyon air temperature decrease reaches 8.4 1C maximum and 6.9 1C daytime average in the green-all case (Fig. 7), while for the green-wall case these numbers become only 3.9 and 2.5 1C, respectively (Fig. 8). In general, the temperature decrease is quite signicant for both surface and air temperatures both inside the canyon and at roof level. Regarding surface temperature decrease of the south-oriented wall, it reaches from 18.7 1C maximum and 14.3 1C daytime average for Riyadh to 9.8 and 5.6 1C, respectively, for Moscow (Fig. 9). Roof surface temperatures lower even more, due to the greatest amounts of solar radiation horizontal surfaces receive in summer; the greatest day-time average temperature decrease is noted for Riyadh (12.8 1C) and the greatest maximum for Mumbai (26.1 1C), while the smallest decreases are noted for Moscow and London (Fig. 10). Moscow reaches the smallest daytime average surface temperature decrease (9.1 1C), while London the smallest maximum (19.3 1C). In the subject of air temperature decrease inside the canyon for the green-all case, it reaches its peak for Riyadh (11.3 1C maximum and 9.1 1C daytime average), while its
Decrease of canyon air temperature, green-wall case 0 Temperature Decrease (C) 4 -1 -2 -3 -4 -5 -6 Time (Hours)
Fig. 8. Air canyon temperature decrease in the EW, H5W10 canyon, with parallel wind ow, when only walls are covered with vegetation, for all climates examined.

Decrease of roof surface temperature 5 Temperature Decrease (C) -5 4 -15 -25 -35 Time (Hours)
Fig. 10. Roof surface temperature decrease when covered with vegetation, for all climates examined.

12

16

20

24

Athens Beijing Brasilia HongKong London Montreal Moscow Mumbai Riyadh

Decrease of asphalt surface temperature 0.0 4 -0.5 -1.0 -1.5 -2.0 -2.5 8 12 16 20 24 Athens Beijing Brasilia HongKong London Montreal Moscow Mumbai Riyadh

12

16

20

24

Athens Beijing Brasilia HongKong London Montreal Moscow Mumbai Riyadh

Time (Hours)
Fig. 11. Asphalt temperature decrease when walls are vegetated in the H5W10 canyon for all the nine climates examined.

smallest decreases are noted in Moscow (3.6 and 3.0 1C, respectively) (Fig. 7). For the green-wall case, air temperature decrease reaches its maximum again for Riyadh (5.1 1C maximum and 3.4 1C daytime average) and its lowest decreases for Moscow (2.6 and 1.7 1C, respectively) (Fig. 8). 3.2. Indirect radiative cooling effects On prima facie evidence, the air inside a canyon with vegetated walls is reduced due to the evapotranspirational rate from plants and the lower surface temperatures of vegetated surfaces. The latter are responsible not only for lowering the air temperature but also for lowering surface temperatures of surfaces not covered with vegetation. As the radiative heat exchanges between the urban canyon surfaces have been modelled analytically, a decrease is observed in the asphalt surface temperature when walls are covered with vegetation. In Fig. 11 the decrease of the asphalt surface temperature is presented for the H5W10 canyon for all the climates examined. As can be observed, the greatest decreases occur for hot and with high solar radiation Riyadh, with a maximum decrease of 2.0 1C and a daytime average 1.3 1C. The lowest surface asphalt temperature decreases take place in much colder and with lower insolation Moscow (maximum 0.9 1C, daytime average 0.6 1C). As the air temperature near the

Decrease of surface temperature of south-oriented wall 0 Temperature Decrease (C) 4 -5 8 12 16 20 24 Athens Beijing Brasilia HongKong London Montreal Moscow Mumbai Riyadh Time (Hours)
Fig. 9. Surface temperature decrease of the south-oriented wall, when covered with vegetation, in the EW, H5W10 canyon, for all climates examined.

-10

-15

-20

ARTICLE IN PRESS
E. Alexandri, P. Jones / Building and Environment 43 (2008) 480493 487

Temperature Decrease (C)

ground is primarily affected by the horizontal surface temperature [14,15,21], the effect of vertical surface temperatures is not as crucial as that of the horizontal surface. This radiative cooling of the street asphalt by green walls, reaching up to 2.0 1C in hot Riyadh has an additional effect on lowering air temperatures, apart from evapotranspirational and convective cooling effects. 3.3. Climatic characteristics It can be said with certainty that, the hotter and drier a climate is, the more important the effect of green walls and green roofs on mitigating urban temperatures is (Figs. 7 and 8). As can be observed in Figs. 3 and 4 and Table 3, Riyadh is the hottest and most arid of all the cases examined, with urban temperatures reaching 42.8 1C maximum and 31.2 1C minimum, with a daytime average of 38.3 1C, while relative humidity spans from 45.8% to 23.5%, with a daytime average of only 32.4%. These extreme climatic inputs, benet the most from green walls and green roofs and, as has been mentioned in Section 3.1, the green-all case reaches temperature decreases of the magnitude of 11.3 1C maximum and 9.1 1C daytime average and the green-walls case 5.1 and 3.4 1C, respectively. Much more humid Mumbai (Fig. 4) reaches smaller decreases, of the magnitude of 6.6 1C daytime average and 8.0 1C maximum for the green-all case and 2.7 and 4.4 1C, respectively, for the green-walls case. The colder climates of London, Moscow and Montreal benet the least, reaching daytime average decreases from 1.7 to 2.1 1C and maxima from 2.6 to 3.2 1C for the green-walls case and from 3.0 to 3.8 1C and from 3.6 to 4.5 1C, respectively, for the green-all case. 3.4. Roof versus canyon Temperatures at roof level decrease more than inside the canyon, when the building envelope is covered with vegetation. This is because the roof, being more exposed to the much larger amounts of summer solar radiation on the horizontal plane, raises its temperatures even more when plain, low albedo-building materials are exposed to direct solar gains. However, the canyon, due to its geometry, is generally more shaded, not reaching the peak temperatures roof surfaces do. By covering the roof with a vegetated medium, which regulates its temperature so as not exceed some crucial levels, roof temperatures decrease more than the temperature inside the canyon, when both roofs and walls are covered with vegetation. For all the nine climates examined, the maximum temperature decrease at the air layer 1 m above the roof reaches from 26.0 1C for Riyadh to 15.5 1C for London and daytime average temperatures from 12.8 1C for Riyadh to 5.8 1C for Moscow (Fig. 12). The air inside the canyon reaches lower decreases; for the green-all case the air temperature decrease reaches a maximum from 11.3 1C for Riyadh to 3.6 1C for Moscow and daytime average from 9.1 1C for

Decrease of air temperature 1m above the roof 5 0 -5 4 -10 -15 -20 -25 -30 -35 Time (Hours)
Fig. 12. Air temperature decrease 1 m above the roof, for all climates examined.

12

16

20

24

Athens Beijing Brasilia HongKong London Montreal Moscow Mumbai Riyadh

Riyadh to 3.0 1C for Moscow (Fig. 7). However, temperatures at roof level start falling after 12:00, while for the more stable conditions inside the canyon, temperatures due to vegetation on walls start decreasing from early in the morning, as can be observed by comparing Figs. 7 and 12. 3.5. Canyon orientation Canyon orientation determines the shading pattern on both the horizontal and the vertical parts of the canyon geometry. It determines the amount of insolation received, especially for the vertical planes, depending on their orientation. During the summer months examined, the amount of irradiation received on vertical planes is much smaller than the horizontal one, for all orientations. Thus, the orientation, despite the fact that it plays an important role in temperature distributions in and around the canyon, it does not affect temperature decreases so signicantly when vegetation covers its vertical surfaces and roofs. The magnitude of the effect strongly depends on the geographic latitude. The examples of Hong Kong (22.16N) and Athens (37.59N) are discussed below. For all the climates examined, it has been observed that the amount and geometry of vegetation is more important than the canyons orientation. In the instance of Hong Kong, solar radiation in all vertical orientations is not so high, reaching a maximum of only 185 W/m2 for the south orientation and 427 W/m2 for the west orientation (Fig. 13). The green-walls case of the EW and NS oriented H5W10 canyons result in 2.4 and 2.0 1C daytime average temperature decrease, respectively, inside the canyon, with an only 0.4 1C difference between the two orientations. For the maximum, this difference becomes 0.7 1C (3.8 1C maximum temperature decrease for the EW oriented canyon and 3.1 1C for NS one). For the green-all case, the differences between the two orientations become even smaller, reaching 0.2 1C for the daytime average (temperature decrease being 6.8 1C for EW and 6.6 1C for NS), and 0.0 1C for the maximum (maximum temperature decrease being 8.5 1C for both EW and NS orientation). It can be observed in Fig. 14 that the amount and geometry of

ARTICLE IN PRESS
488
Global rad, horiz. Globalrad, south Global rad, north Global rad, east Global rad, west

E. Alexandri, P. Jones / Building and Environment 43 (2008) 480493

800 700 Solar Radiation (W/m2) 600 500 400 300 200 100 0 1.00

900 800 Solar Radiation (W/m2) 700 600 500 400 300 200 100

Global rad, horiz. Global rad, south Global rad, north Global rad, east Global rad, west

4.00

7.00

10.00

13.00

16.00

19.00

22.00

0 1 4 7 10 13 Time (Hours)
Fig. 15. Athens global solar radiation on horizontal and vertical planes of east, west, south and north orientation in July.

16

19

22

Time (Hours)

Fig. 13. Hong Kong global solar radiation on horizontal and vertical planes of east, west, south and north orientation in July.

35 30 Temperature (C) 25 20 15 10 5 0 12 15 18 21 Time (Hours)

0 -1 -2 -3 -4 -5 -6 -7 -8 -9

Temperature Decrease (C)

Ta,can [no gr]NS-HgKg

35 Temperature (C) 30 25 20 15 10 5 0 12 15 18 21 Time (Hours)

-1 -2 -3 -4 -5 -6 -7 -8

Temperature Decrease (C)

Ta,can[no gr]EW-HgKg

40

Ta,can[no gr]Ath-EW Ta,can[no gr]Ath-NS Ta,can[gr a]Ath-EW Ta,can[gr a]Ath-NS Ta,can[gr w]Ath-EW Ta,can[gr w]Ath-NS DTa,can[gr a]Ath-EW DTa,can[gr a]Ath-NS DTa,can[gr w]Ath-EW DTa,can[gr w]Ath-NS

Ta,can[gr a]EW-HgKg Ta,can[gr a]NS-HgKg Ta,can[gr w]EW-HgKg Ta,can[gr w]NS-HgKg DTa,can[gr a]EW-HgKg DTa,can[gr a]NS-HgKg DTa,can[gr w]EW-HgKg DTa,can[gr w]NS-HgKg

Fig. 14. Hong Kong temperature distributions and decreases inside the canyon for no-green [no-gr], green-all [gr-a], green-walls [gr-w], for EW and NS oriented H5W10 canyon.

Fig. 16. Athens temperature distributions and decreases inside the canyon for no-green [no-gr], green-all [gr-a], green-walls [gr-w], for EW- and NSoriented H5W10 canyon.

vegetation mostly affect the temperature decrease and not so much the canyons orientation. The green-all case in the EW-oriented canyon results in 5.4 1C higher for the daytime average and 4.7 1C higher for the maximum temperature decrease than the green-walls case. For the NS orientation, these differences become 4.6 and 4.4 1C, respectively. However, in the example of Athens, the amount of irradiation received in the east and west oriented vertical planes is much larger than on the south and north orientations and proportionally larger to those of Hong Kongs. The maximum solar radiation received by the south-oriented vertical plane is 374.0 W/m2, while for the east-oriented plane reaches the magnitude of 616.7 W/m2 (Fig. 15). This has a direct effect on the way the canyon orientation affects temperature decreases due to vegetated surfaces. For the green-all case, the difference between the temperature decrease of the air inside the canyon remains small, of the magnitude of 0.1 1C for the daytime average (temperature decrease being 5.6 1C for EW and 5.5 1C for NS), and 0.2 1C for the maximum (temperature decrease

being 6.6 1C for EW and 6.8 1C for NS orientation). For the green-walls case these differences become larger, reaching 0.8 1C for the daytime average (temperature decrease being 3.0 1C for EW and 2.2 1C for NS) and 1.2 1C for the maximum (temperature decrease being 4.5 1C for EW and 3.3 1C for NS). Yet again, the difference between the two amounts of vegetation (green-all and green-walls) is more crucial than the difference between the decreases of different orientations. The difference between the temperature decrease of the green-all and the greenwalls cases of the EW oriented canyon reaches 2.6 1C for the daytime average and 2.1 1C for the maximum. For the NS orientation, these differences become larger, 3.2 and 3.4 1C, respectively (Fig. 16). In general, it can be concluded that the orientation may play a countable role in temperature decreases due to vegetation, only when the amounts of solar radiation received by the vertical planes differ signicantly. Yet again, concerning temperature decreases, the amount of vegetation placed on buildings is more crucial than the orientation of the canyon, with the green-all case, when

ARTICLE IN PRESS
E. Alexandri, P. Jones / Building and Environment 43 (2008) 480493 489

both roofs and walls are covered with vegetation, leading to much larger temperature decreases. 3.6. Canyon geometry For all geometries and cases examined, it can be concluded that the wider a canyon is the smaller the effect of green roofs and green walls on its temperature decreases. For wider canyons, temperatures inside the canyon are dominated by the proportionally larger street surface and the fact that it is more exposed to direct solar radiation. In Riyadh, temperature decreases in the wide H5W15 canyon are of the magnitude of 1.2 1C for the daytime average and 1.7 1C for the maximum of the green-wall case and 7.3 and 9.3 1C, respectively, for the green-all case. For the narrower H10W5 canyon these decreases reach 6.3 1C daytime average and 9.1 1C maximum for the green-walls case and 8.9 and 12.3 1C, respectively, for the green-all case (Figs. 17 and 18). It can be observed in Fig. 18 that the wider a canyon is the smaller the effect of green walls is on its temperatures. Nevertheless, when the combination of green roofs and green walls is implemented (green-all case), temperature decreases rise signicantly. In contrast, for the narrow H10W5 canyon, whose walls are proportionally more dominant than the street, the green-walls case has a
Air temperature decrease for different canyon geometries in Riyadh for the green-all case 0 Temperature Decrease (C) 7 -5 11 15 19

signicant effect on lowering urban temperatures. The combination of both green roofs and green walls does not lead to such signicant further decreases, as is the case in the wider H5W10 and H5W15 canyons. The differences between the temperature decreases of the green-all and the green-walls case reaches an average of 5.7 and 6.1 1C, respectively, for the H5W10 and the H5W15 canyons, while for the H10W5 canyon it is only 2.6 1C. For the maximum, the discrepancy between the two wider and the narrower canyon is even larger, reaching 6.2, 7.6 and 3.2 1C, respectively. 3.7. Wind direction Wind direction affects temperature decreases inside the canyon even less than orientation. Although it is a signicant factor for temperature distributions, for the decreases due to vegetation it is the vegetation itself that plays the most important role. The differences between temperature decreases in the same canyons for different wind directions are insignicant. In Fig. 19 temperature distributions and temperature decreases for the EWoriented H5W10 canyon in Mumbai are presented, for both parallel (y) and perpendicular (x) to the canyons axis wind directions. It can be observed, that for the low air velocities inside the canyon, temperature distributions are not so different for the two wind directions as they were for the canyon orientations (Figs. 14 and 16). Temperature differences between the temperature decreases of the two wind directions become quite insignicant, reaching a 0.1 1C difference for both green-walls and green-all cases. It can also be observed that for both wind directions the most important factor for temperature decreases is the amount of vegetation, with the green-all case reaching temperature decreases 3.7 1C higher than the green-walls case. For the wider H5W15 canyon, with its much larger air velocities, temperature decreases are similar for the two wind directions (Fig. 20). The differences between the temperature decreases of the two wind directions reach a maximum of 0.3 1C, with 0.2 1C daytime average for the
40 Temperature (C) 0 -1 -2 -3 -4 -5 -6 -7 -8 -9 12 15 18 21 Time (Hours)

-10

-15

DTa [gr a]H5W10 DTa [gr a]H10W5 DTa [gr a]H5W15 Time (Hours)

Fig. 17. Air temperature decrease during the day in the H5W10, H10W5 and H5W15 canyon, for the green-all case, Riyadh.

Air temperature decrease for different canyon geometries in Riyadh for the green-wall case 0 Temperature Decrease (C) 7 -2 -4 -6 -8 -10 DTa[gr w]H5W10 DTa[gr w]H10W5 DTa[gr w]H5W15 Time (Hours)
Fig. 18. Air temperature decrease during the day in the H5W10, H10W5 and H5W15 canyon, for the green-wall case, Riyadh.

35 30 25 20 15 10 5 0

Temperature Decrease (C)

Ta,can[no gr]Mumb-EW-x Ta,can[no gr]-Mumb-EW-y Ta,can[gr a]Mumb-EW-x Ta,can[gr a]-Mumb-EW-y Ta,can[gr w]Mumb-EW-x Ta,can[gr w]-Mumb-EW-y DTa,can[gr a]Mumb-EW-x DTa,can[gr a]-Mumb-EW-y DTa,can[gr w]Mumb-EW-x DTa,can[gr w]-Mumb-EW-y

11

15

19

Fig. 19. Mumbai temperature distributions and decreases inside the canyon for no-green [no-gr], green-all [gr-a], green-walls [gr-w], for a parallel (y) and perpendicular (x) to the canyons axis wind direction in the EW-oriented H5W10 canyon.

ARTICLE IN PRESS
490 E. Alexandri, P. Jones / Building and Environment 43 (2008) 480493

45 40 35 30 25 20 15 10 5 0 12 15 18 21 Time (Hours)

1 0 -1 -2 -3 -4 -5 -6 -7 -8

Temperature Decrease (C)

Ta,can[no gr]MumbH5W15-EW-x Ta,can[no gr]-MumbH5W15-EW-y Ta,can[gr a]MumbH5W15-EW-x Ta,can[gr a]-MumbH5W15-EW-y Ta,can[gr w]MumbH5W15-EW-x Ta,can[gr w]-MumbH5W15-EW-y DTa,can[gr a]MumbH5W15-EW-x DTa,can[gr a]-MumbH5W15-EW-y

Fig. 20. Mumbai temperature distributions and decreases inside the canyon for no-green [no-gr], green-all [gr-a], green-walls [gr-w], for a parallel (y) and perpendicular (x) to the canyons axis wind direction in the EW-oriented H5W15 canyon. Fig. 22. PET for the EW-oriented H5W10 canyon, for the no-green [no gr] and green-all [gr-a] cases, inside the canyon (EW) and on the roof (rf), for Athens.

Fig. 21. PET for the EW-oriented H5W10 canyon, for the no-green [no gr] and green-all [gr-a] cases, inside the canyon (EW) and on the roof (rf), for Moscow.

Temperature (C)

green-wall case and 0.4 and 0.3 1C, respectively, for the green-all case. Again, the amount of vegetation has a stronger effect than the wind ow direction, even in the wider canyon. It can thus be concluded, that for the generally low air velocities inside the urban canyons [6], the effect of wind direction is not so strong on temperature decreases due to vegetated roofs and walls, as is the amount and geometry of vegetation itself.

Fig. 23. PET for the EW-oriented H5W10 canyon, for the no-green [no gr] and green-all [gr-a] cases, inside the canyon (EW) and on the roof (rf), for Riyadh.

4. Thermal comfort In order to assess the thermal comfort improvements in outdoors spaces when walls and roofs are covered with vegetation, the physiological equivalent temperature (PET) is used, its expression deriving from Ref. [23] and its relationship with thermal sense from Ref. [24]. The results for the EW-oriented H5W10 green-all and no-green cases are presented here, for Moscow, Athens and Riyadh in Figs. 2123. Emphasis is given on thermal comfort, not only inside the canyon (symbolised with EW in the graphs), but also at the roof level (symbolised with rf).

It can be observed in Fig. 21, that for the much milder summer of Moscow, the greening of the building envelope does not lead to such major improvements of the outdoors thermal comfort. PET ranging from slightly warm and comfortable levels on the roof and inside the canyon for the no-green case, lowers to cooler levels, from comfortable to slightly cool, during daytime, when roofs and walls are covered with vegetation (green-all case). Although moving from slightly warm to comfortable might not be so spectacular, it could prove to be benecial for the thermal comfort and well being of populations used to cooler climatic conditions. For much hotter Athens (Fig. 22) and Riyadh (Fig. 23), the improvements of outdoors thermal comfort are more dramatic. For both climates, the bare concrete roof reaches the very hot level in the afternoon. When covered with vegetation, the sensation warm is reached only for 4 h in Athens and 5 in Riyadh. Most of the daytime, the exposed to direct solar radiation roof reaches the slightly warm and comfortable zone for both cities. For inside the

ARTICLE IN PRESS
E. Alexandri, P. Jones / Building and Environment 43 (2008) 480493 491

canyon, the thermal sensation improves even more, from hot, in both cases to slightly warm and comfortable, reaching even slightly cool for both cases in the early morning and late evening hours and even cool for Athens, in the early morning hours. In general, green roofs and walls can improve outdoors thermal conditions not only at street level, but also at roof level, turning these empty urban spaces into potentially usable ones, in the form of superterrestrial gardens. By covering roofs and walls with vegetation, thermal comfort in the built environment can improve signicantly, not only for hot climates, but for cooler ones, in which populations are acclimatised to lower temperatures. 5. Energy savings from green walls and roofs Apart from creating outdoor conditions, which are more human-friendly, from a thermal point of view, green roofs and green walls can also prove benecial for indoor thermal conditions. In addition to the fact that they add a further insulation layer to the buildings fabric, they can decrease cooling load demands inside the building quite signicantly due to the microclimatic modications discussed in this paper. In a simplied steady-state analysis, without taking into consideration internal thermal gains, heat gains/losses (qE) from the buildings fabric with an average U-value U, an indoors temperature Tin and an outdoors temperature Tout are given by the relationship: qE UT out T in . (9)

Fig. 24. Average cooling load decreases (%), with a 23 1C indoors temperature, for the green-all and green-walls cases of all the climates examined.

Similarly, for the green-walls case, if T[gr w] is the average air temperature inside the canyon with the green walls, the cooling load decrease becomes: DqEgr w T no gr T gr w . T no gr T in (13)

For the no-green base case [no gr], the cooling load for the non-vegetated canyon is given by the relationship: qEno gr U T no gr T in , (10) where T[no gr] is the averaged air temperature inside the canyon when no vegetation is placed either on walls or on roofs. For the green-all case, with an average air temperature inside the canyon T[gr a], heat gains are1: (11) qEgr a U T gr a T in . Thus, the decrease in the cooling load, when both walls and roofs are covered with green is given by DqEgr a DqEgr a qno gr qgr a ), qno gr T no gr T gr a . T no gr T in
gr]6Tin,

12 and T[gr a]4Tin.

For T[no

T[no

gr]4Tin

1 Despite the fact that the U-value is altered, when vegetation is placed on the buildings fabric, leading to further cooling load decreases, this is not taken into consideration here, as the aim is to directly compare between the effects of the microclimatic alterations on the buildings cooling load, without it being affected by alterations to the fabric.

For T[no gr]6Tin, T[no gr]4Tin and T[gr w]4Tin. Considering an indoor limit temperature for cooling of 23 1C for all climates studied, the cooling load decreases due to green-all and green-walls cases are given as a daytime average in Fig. 24 and for an hourly basis in Fig. 25. As can be observed in Fig. 24, the largest cooling load decreases in all climates examined, occur for the green-all case. For the geometries examined for Bras lia and Hong Kong, the cooling load decreases for the green-all case reach 100%; no cooling load is needed after covering roofs and walls with vegetation, while in both cities cooling load is needed in the afternoon and early evening hours for the no-green case (Fig. 25c and d). London and Moscow are not affected, regarding cooling loads, as no cooling load is needed for the typical day examined, even before vegetation was placed around the canyon. Riyadh experiences a quite high cooling load decrease, of the magnitude of 90%, as does Montreal (85%) for the green-all case, lowering their total hours of cooling demand from 12 to 5 and from 8 to 4, respectively (Fig. 25e and g). For the green-all case, Mumbai reaches a 72% decrease, lowering its cooling energy demand from 11 h to 6 (Fig. 25f), while for Athens and Beijing the decrease is 66% and 64%, respectively, lowering their energy demand by 4 and 3 h, respectively (Fig. 25a and b). For the green-walls cases, cooling load decreases are less dramatic. The largest one is noted for Bras lia (68%), with 6 h decrease in cooling demand (Fig. 25d). It is followed by a 66% and 2 h decrease for Hong Kong (Fig. 25e), 52% and 2 h for Montreal (Fig. 25e), 43% and 2 h for Athens (Fig. 25a), 37% and 2 h for Beijing (Fig. 25b), 37% and 3 h for Riyadh (Fig. 25(g) and 35% and 3 h for Mumbai (Fig. 25f). It can be noted that the differences between the green-all and green-walls cooling loads are smaller for

ARTICLE IN PRESS
492 E. Alexandri, P. Jones / Building and Environment 43 (2008) 480493

Fig. 25. Cooling load decreases (%) for (a) Athens, (b) Beijing, (c) Hong Kong, (d) Bras lia, (e) Montreal, (f) Mumbai and (g) Riyadh for green-all and green-walls cases.

humid climates (of the magnitude of 3237%) and greater for arid climates (53% for Riyadh), due to the different humidity concentrations in the two climatic groups. In general, green roofs and green walls cool the microclimate around them, which can lead to quite important energy savings for cooling, depending on the climatic type, the amount and position of vegetation on the building. In cases where little cooling load is needed, cooling demand can be reduced to zero by covering building surfaces with vegetation. In other cases, energy savings can also be signicant, varying from 90% to 35%.2
2 These percentages can become even greater, when a higher than 23 1C limit temperature for cooling is considered. In general, inhabitants of hot climates are accustomed to higher temperatures (in the instance of Greek regulations, the limit temperature for cooling is set to 26 1C).

In addition to the energy savings themselves, this could lead to successful applications of further passive cooling techniques, especially ones employing ventilation, which are not easy to implement in the extremely hot urban conditions, in cases of large heat island densities. 6. Conclusions From this quantitative research, it has been shown that there is an important potential of lowering urban temperatures when the building envelope is covered with vegetation. Air temperature decreases at roof level can reach up to 26.0 1C maximum and 12.8 1C day-time average (Riyadh), while inside the canyon decreases reach up to 11.3 1C maximum and 9.1 1C daytime average, again for hot and arid Riyadh. It can be concluded that the

ARTICLE IN PRESS
E. Alexandri, P. Jones / Building and Environment 43 (2008) 480493 493

hotter and drier a climate is, the greater the effect of vegetation on urban temperatures. However, it has been pointed out that also humid climates can benet from green surfaces, especially when both walls and roofs are covered with vegetation, reaching up to 8.4 1C maximum temperature decrease for humid Hong Kong. Temperature decrease due to vegetation is primarily affected by the vegetation itself (amount and geometry), more than the canyon orientation in hot periods. In general, the larger amounts of solar radiation a surface receives, the larger its temperature decreases are when it is covered with vegetation. For the low air velocities inside the canyon, the wind direction does not have any signicant effect on temperature decreases due to vegetation. Regarding the urban geometry, the wider a canyon is, the weaker the effect green roofs and green walls have on temperature decrease. For all climates examined, green walls have a stronger effect than green roofs inside the canyon. Nonetheless, green roofs have a greater effect at roof level and, consequently, at the urban scale. The combination of both green roofs and green walls leads to the highest mitigation of temperatures inside the canyon. If applied to only one unit block, green roofs and green walls can create a small area of mitigated temperatures to the urban heat island effect, as has been shown in this microclimatic study. If applied to the whole city scale, they could mitigate raised urban temperatures, and, especially for hot climates, bring temperatures down to more human-friendly levels and achieve energy saving for cooling buildings from 32% to 100%. Acknowledgements This research has been funded by the State Scholarship Foundation of Greece (IKY) from 2001 to 2003. The authors are extremely grateful to Panagiotis Doussis for his guidance and contribution to computer modelling. References
[1] Benevolo L. The history of the city. London: Scholar Press; 1980. [2] Koppe C, Kovats S, Jendritzky G, Menne B. Health and global environmental change; heat-waves: risks and responses, series no. 2, energy, environment and sustainable development. Copenhagen: World Health Organization; 2004. [3] White KS, et al. Technical summary, climate change 2001: impacts, adaptation and vulnerability. In: McCarthy JJ, Canziani OF, Leary NA, Dokken DJ, White KS, editors. Climate change 2001: impacts,

[4]

[5]

[6] [7]

[8]

[9]

[10] [11] [12] [13]

[14] [15] [16] [17] [18] [19] [20] [21] [22] [23]

[24]

adaptation and vulnerability. Cambridge: United Nations Intergovernmental Panel on Climate Change (IPCC) Cambridge University Press; 2001. p. 1973. D az J, Ballester F, Lopez-Velez R. Impacts on human health. In: Moreno Rodr guez JM, editor. The preliminary assessment of the impacts in Spain due to effects of climate change, project ECCE. Madrid: Ministry of the Environment; 2005. Davies RE, Knappenberger PC, Novicoff WM, Michaels PJ. Decadal changes in summer mortality in US cities. International Journal of Biometeorology 2003;47:16675. Santamouris M, editor. Energy and climate in the urban built environment. London: James & James; 2001. Dimoudi A, Nikolopoulou M. Vegetation in the urban environment: microclimatic analysis and benets, PRECis project, the European Commission, Directorate General XII, Joule III, contract JOR3CT97-0192. Centre for Renewable Energy Sources, Pikermion, 2000. Bruse M, Fleer H. Simulating surfaceplantair interactions inside urban environments with a three dimensional numerical model. Environmental Modelling and Software 1998;13:37384. Giridharam R, Ganasan S, Lau SSY. Daytime urban heat island in high-rise and high-density residential developments in Hong Kong. Energy and Buildings 2004;36:52534. Versteeg HK, Malalasekera W. An introduction to computational uid dynamics. Harlow, London: Prentice-Hall; 1995. Luikov AV. Heat and mass transfer in capillary-porous bodies. Oxford: Pergamon Press; 1961. Eckert ERG, Drake RM. Heat and mass transfer. New York: McGraw-Hill Book Company Inc.; 1959. Noilhan J. Les facteurs physiques du microclimat au voisinage dun batiment. Etude bibliographique. Centre Scientique et Tecnique du Batiment, Etablissement de Nantes, 1979. Jacobson MZ. Fundamentals of atmospheric modeling. Cambridge: Cambridge University Press; 1999. Pielke RA. Mesoscale meteorological modeling. San Diego: Academic Press; 2002. Modest MF. Radiative heat transfer. New York: McGraw-Hill International Editions; 1993. Remund J, Lang R, Kunz S, Meteonorm, version 3.0, Meteotest, Bern, 1997. Marsh A. Ecotect, version 5.20(b), square one research PTY LTD, 2003. Welsh School of Architecture, Cardiff University, WinAir4. Cardiff, 2003. Critcheld HJ. General climatology. London: Prentice-Hall Internatioal Inc.; 1974. Geiger R. The climate near the ground. Cambridge, Massachusetts: Harvard University Press; 1965. Monteith JL, Unsworth M. Principles of environmental physics. London, New York: Edward Arnold; 1990. Matzarakis A, Mayer H, Rutz F. Radiation and thermal comfort. Sixth Hellenic conference in meteorology, climatology and atmospheric physics, Ioannina 2; 2002. p. 73944. Kuttler W. Stadklima, Teil 2: phanomene und wirkungen. In: Moller D. (Ed.), Beitragsserie klimaanderung und klimaschutz, UWSF, vol. 16(4); 2004. p. 26374.