Pergamon

www.elsevier.com / locate / solener

PII:

Solar Energy Vol. 70, No. 3, pp. 237244, 2001 2001 Elsevier Science Ltd S 0 0 3 8 0 9 2 X ( 0 0 ) 0 0 0 9 2 X All rights reserved. Printed in Great Britain 0038-092X / 01 / $ - see front matter

AIRFLOW IN COURTYARD AND ATRIUM BUILDINGS IN THE URBAN ENVIRONMENT: A WIND TUNNEL STUDY
S. SHARPLES* , and R. BENSALEM**
*Centre of the Built Environment, School of Environment and Development, Shefeld Hallam University, Unit 9, Science Park, Howard Street, Shefeld S1 1WB, UK **School of Architecture, University of Shefeld, Shefeld S10 2TN, UK

AbstractA wind tunnel study was carried out to investigate the airow through courtyard and atrium building models located within an urban setting and exposed to an urban atmospheric boundary layer. Ventilation strategies resulting from the use of different courtyard and atrium pressure regimes (positive pressure and suction) were examined. The model buildings were monitored both in isolation and in idealised urban environments of varying group layout densities. The effect of wind direction was also observed. The results from the study suggest that the open courtyard in an urban environment had a poor ventilation performance whilst an atrium roof with many openings operating under a negative (suction) pressure regime was the most effective. Changing the wind direction from perpendicular to the building fac ades to a 458 incidence angle had the effect of making the differences in the observed ows between all the models much smaller. 2001 Elsevier Science Ltd. All rights reserved.

1. INTRODUCTION

A key feature of urban sustainability is the design and use of energy efcient buildings within the urban context. Many urban buildings are large, non-domestic structures used for ofce, commercial and industrial purposes. Such buildings have frequently used mechanical ventilation and / or air conditioning systems to create their internal environments, and the use of such systems has increased dramatically throughout Europe during the last decade (Santamouris and Asimakopolous, 1996). Apart from the global consequences of energy-intensive, highly serviced buildings there has also been concern related to the health of the occupants of such buildings (Turner and Binnie, 1990). In response to these issues there has been a revival of interest in the use of natural ventilation for non-domestic buildings to provide an internal environment with satisfactory thermal comfort and good indoor air quality. This revival has incorporated both major research projects (Limam et al., 1997; Kukadia, 1998) and the production of design guidance and design tools (CIBSE, 1997; Allard, 1998) to help engineers and architects integrate natural ventilation solutions into their buildings. One building type that has been thought to offer great potential for utilising natural ventila-

Author to whom correspondence should be addressed. Tel.: 1 44-0114-225-4794; fax: 1 44-0114-225-3206; e-mail: s.sharples@shu.ac.uk 237

tion and reducing energy consumption is the atrium. The atrium form was originally used in ancient Greece and Rome, where it was an uncovered internal roof opening to allow daylight in and smoke from res out. This uncovered form was extended in traditional Arab architecture to create a larger courtyard space. Courtyard spaces were covered with glass in wrought iron frames during the 19 th century when these materials became widely available as a result of the Industrial Revolution. The modern origins of the atrium can be traced back to the construction of the Hyatt Regency Hotel in Atlanta in 1967 (Saxon, 1994). The atrium building has proved so popular that most large commercial buildings completed during the last 25 years contain some form of atrium, often as a central core circulation space. Although this has other environmental implications, such as for solar gain, daylight and thermal comfort, this study has focussed on the ventilation and airow performance of the atrium. The atrium building utilises both stack and wind forces to generate air ows through the room spaces adjoining the atrium well. In summer the stack force can be dominant and the atrium well acts as a chimney to vent warm air out of rooftop openings. For other times of the year wind forces may create pressure gradients between the outer fac ades of the buildings and the inner fac ades facing the atrium well. A complex mix of parameters will determine the magnitude and direction of the wind-induced internal airow. These include the magnitude and distribution of the pressure coefcients around and inside the

238

S. Sharples and R. Bensalem

atrium building, the leakage characteristics of the fac ades and the shape of the roof over the atrium well. Another key factor is the sheltering effect of surrounding buildings in the urban environment. Recent work by Santamouris et al. (1998) on airow in urban canyons has indicated how great the drop in natural ventilation potential can be within urban areas. For atrium buildings sited within a congested urban area the sheltering effect may be so great that wind-induced pressures on the buildings external walls may be very small. Under these circumstances the atrium roof can be the key element for generating sufciently strong positive or negative pressure gradients to induce satisfactory natural ventilation ows. The use of roof elements to enhance the ventilation of buildings in built-up areas is not new. Wind towers and wind catchers have been prevalent in the Middle East and North Africa for hundreds of years, and have formed a key component of the indigenous architecture of those regions. Fig. 1 shows a wind tower from Iran [9]. Vents incorporated into the roof ridge and eaves provide a more integrated solution in modern buildings. Bauman et al. (1988) used a wind tunnel to investigate a jack roof conguration

for densely packed housing in hot humid climates, with vents placed in the sides of an elevated roof ridge. The jack roof was found to be effective in inducing internal air movement. A study by Riskowski et al. (1998) of the airow performance of a number of commercial and fabricated ridge vents provided quantitative data for a range of wind speeds and directions. The shape of the roof can also have a major role in inducing internal air movement. Kindangen et al. (1997) performed a CFD analysis of ten roof congurations to study their impact on airow velocities and distributions. For the isolated, cross-ventilated dwelling modelled the shape of the roof did have an effect on the airow patterns, and, in particular, air velocities. Wind direction, roof overhangs and roof heights were also important inuences on airow. Some assessments of airow in courtyard and atrium buildings have been carried out, mainly as either wind tunnel or computer simulations. Walker et al. (1993) and Shao et al. (1993) performed CFD and wind tunnel studies of courtyards. Their work suggested that courtyards could provide adequate ventilation, but that local sheltering effects in urban areas might create serious prob-

Fig. 1. Wind towers of Yazd, Iran (from Gallo, 1995).

Airow in courtyard and atrium buildings in the urban environment: a wind tunnel study

239

lems. Alvarez et al. (1998) modelled airow patterns in courtyards as a function of their depth to width ratios. Andersen (1995) provided a set of formulae for estimating atrium ventilation parameters such as air velocity, temperatures and ventilation capacities in terms of openings, heat loads and building geometries. Guidelines for CFD modelling of atria have been suggested by Schild et al. (1995). Schaelin et al. (1996) compared several airow models of varying complexity with measured temperature and velocity data from an 8000 m 3 volume atrium, and obtained satisfactory agreement. Many of the above studies have tended to investigate either isolated building models or have been trying to apply the results to a particular actual full-scale building. Little work has been done on a parametric analysis of the interactions between several factors, such as roof shape, atrium ventilation mode, wind direction and surrounding buildings. Such a parametric analysis would help identify the best combination of design parameters to maximise the benets of wind-driven natural ventilation in atrium buildings in urban areas. The experimental details of just such a parametric study are given in this paper.
2. EXPERIMENTAL PROCEDURE

structed, instrumented and tested in a boundary layer wind tunnel.

2.1. Building models

The models represented four storey courtyard and atrium buildings at a scale of 1:100. They measured 339 3 339 externally 3 130 mm high. The central courtyard / atrium was square in plan with the sides being equal to the height of the building (i.e. 130 3 130 3 130 mm). The model walls representing room depth were 104 mm deep. Monopitch roofs were placed over the courtyard opening to produce a range of ventilation strategies. Each roof was 52 mm high, giving a roof pitch of just under 228. The models were constructed from Perspex and consisted of rectangular building block modules that could be xed together to create a range of model types. The models walls and roofs were perforated with 10 mm diameter holes to simulate building leakiness. The porosity of the walls (hole area to total fac ade area) was 11.4%. Table 1 gives details of the various courtyard and atrium roof ventilation strategies investigated in the study, while Fig. 2 shows schematically the types of ow produced by these strategies. Airow rates through the models were measured directly with an orice plate incorporated into one of the Perspex building block modules. This device was a square edged plate of 17 mm diameter inserted between two short brass pipes of 25 mm diameter, tted with two corner pressure tappings. The pressure drop across the tap-

In order to evaluate the wind-driven natural ventilation in courtyards and atria in an urban setting a range of model buildings were con-

Table 1. Courtyard and atrium roof ventilation models and strategies used in study Model Reference Model A1 Roof ventilation strategy Open courtyard, square in plan, no roof Ventilation strategy: positive driving pressures across courtyard Atrium monopitch roof with no openings in roof Ventilation strategy: positive driving pressures across atrium Atrium monopitch roof, small area of openings in leeward pitch of roof, equivalent to 11.4% of total fac ade area Ventilation strategy: negative suction pressures on atrium roof openings Atrium monopitch roof, large area of openings in leeward pitch of roof, equivalent to 30.4% of total fac ade area Ventilation strategy: negative suction pressures on atrium roof openings Atrium monopitch roof, small area of openings in windward vertical element of roof equivalent to 11.4% of total fac ade area Ventilation strategy: positive driving pressures on vertical roof openings Atrium monopitch roof, openings in leeward pitch of roof and openings in windward vertical element equivalent to 30.4% of total fac ade area Ventilation strategy: positive and negative pressures on atrium roof openings

Model A2

Model A3

Model A4

Model A5

Model A6

240

S. Sharples and R. Bensalem

Fig. 2. Courtyard and atrium roof ventilation strategies used in study.

pings was measured using a digital manometer. The orice plate was calibrated, in its Perspex container, against a precision commercially available owmeter with an accuracy traceable to national standards. The dynamic pressure in the tunnel at the top of the boundary layer, together with the internal pressure in the atrium well at mid height, was also recorded. The ows through each model were monitored on each oor and on each fac ade of the model at two locations positioned centrally. The error in ow observa-

tions was estimated at approximately 610%, reecting the uctuations of the manometer signals. Fig. 3 shows the orice plate located in one of the Perspex atrium models.

2.2. The wind tunnel

The instrumented model was placed at the centre of a 1.1 m diameter turntable in an atmospheric boundary layer wind tunnel. Two wind directions (08 and 458 ) were investigated. The tunnel had a working length of 7.2 m with a

Fig. 3. Orice plate located in atrium building model.

Airow in courtyard and atrium buildings in the urban environment: a wind tunnel study

241

cross-section of 1.2 3 1.2 m, and a maximum speed of 25 m s 2 1 . The wind tunnel was normally run at this speed, which produced a windspeed at eaves level of 16.4 m s 2 1 , to produce the maximum Reynolds number. A series of spires, castellated fence and roughness elements in the wind tunnel generated a suburban type velocity prole at the turntable with a boundary layer height of 800 mm and a power-law exponent of 0.245. The urban environment around the model was simulated by surrounding the model with rectangular wooden blocks of similar dimensions as the model, where the height of each block, H, was the same as the eaves height of the atrium and courtyard models. The blocks were arranged in either a uniform or staggered (checkerboard) arrangement. The wall-to-wall spacing between the blocks, Sc, was set at 1.5 and 2.3 times the building height H. The lateral spacing, L, was set to 0.5Sc for the staggered arrangement and to Sc for the in-line layout. These arrangements produced area densities of 0.28 and 0.40 for the uniform layout and 0.38 and 0.50 for the staggered layout. Area density is dened as the ratio of building plan area to building site area. A set of measurements was also made on all the models in isolation with no surrounding buildings. The blocks were laid out to a fetch radius of 15H (three rows of blocks upstream and three rows downstream) as test results showed no change in measured airows beyond this fetch. The blockage in the tunnel was up to 8% at normal wind incidence (08 ) and up to 11% for the 458 wind incidence direction. Although these values are a little high it was decided to apply no corrections to the results. Each model was secured to the wind tunnel turntable and the required layout put in position. The tunnel was run for one hour to allow ow and temperature conditions to stabilise. Each measurement consisted of logging the pressure drop Dp across the orice plate whilst simultaneously recording the dynamic pressure in the tunnel at the boundary layer height of 800 mm using a pitot-static tube. From this dynamic pressure

measurements it was possible to derive the reference wind speed at the top of the boundary layer, V800 . The standard orice ow equation for estimating the ow Q though an opening of ow area A is given by: Q 5 Cd A([2Dp /r ])0.5 (1)

where Cd is the discharge coefcient of the opening, Dp is the pressure difference across the opening and r the air density. The air velocity at the orice plate opening, Vo , is given by Q / A, and can be derived from eqn (1). In this study the orice plate relationship between Q and Dp was established by calibration against a precision, commercial owmeter with an accuracy traceable to national standards. Since Q and A were known then Vo could be established The results from the study were expressed as a non-dimensional ow coefcient CQ l:
velocity at an opening in the model, Vo CQ 1 5 ]]]]]]]]]] velocity at the top of the boundary layer, V800

(2) or CQ 1 5 Q /(A 3 V800 ) (3)

CQ t was used to represent the average of all the CQ l values measured for each model.

3. RESULTS AND DISCUSSION

The results of the experiments are discussed in terms of the average of all the CQ 1 model values measured on each model, CQ t, and the minimum value of CQ l observed in each model for the two wind directions. A comparison is also made between the CQ t values for the models surrounded by other building models and when the models were in isolation. The detailed results from the experiments are presented in Table 2.

Table 2. Average ow coefcients CQ t for uniform (U) and staggered (S) area densities (D ) and wind directions 08 and 458 Model Ref. 08 Model A1 Model A2 Model A3 Model A4 Model A5 Model A6 0.126 0.147 0.140 0.135 0.179 0.162 CQ t Isolation 458 0.179 0.188 0.166 0.159 0.189 0.162 08 0.074 0.093 0.085 0.134 0.108 0.126 CQ t U, D 5 0.28 458 0.144 0.153 0.145 0.137 0.159 0.126 08 0.071 0.086 0.087 0.136 0.119 0.100 CQ t U, D 5 0.40 458 0.118 0.124 0.114 0.127 0.127 0.100 08 0.065 0.080 0.088 0.139 0.086 0.104 CQ t S, D 5 0.38 458 0.126 0.119 0.122 0.135 0.117 0.104 08 0.057 0.062 0.083 0.131 0.098 0.070 CQ t S, D 5 0.50 458 0.084 0.091 0.095 0.124 0.094 0.070

242

S. Sharples and R. Bensalem

3.1. 08 wind direction

Among the structures tested, the courtyard model A1 had the poorest ventilation performance. For any of the urban layouts the CQ t values were typically between 0.065 and 0.071, compared to a value of 0.126 for the courtyard in isolation. The minimum CQ l was very low, being measured at one location at 0.02. The closed roof atrium model A2 showed a slightly improved performance over the courtyard, with CQ t values from 0.062 to 0.093 in the urban layouts (compared to 0.147 for the isolated case), and a minimum CQ l of 0.04. However, the distribution of the ows within the closed model was uneven, with weak ventilation ows identied on the leeward side of the atrium. This suggested that improved ventilation could be achieved by encouraging ows to enter via the roof. Atrium models A3 and A4 both operated under a roof suction regime, with A3 having a roof porosity of 11.4% and A4 a porosity of 30.4%. This roof suction mode was in conict with the negative pressure forces that were generated on the leeward walls of the model. As a consequence the CQ t values for the low porosity roof model A3 displayed little or no improvement over the other models, with a range of 0.083 to 0.088. The much greater roof porosity of model A4 created larger roof suction ows and removed the problem of the negative leeward wall pressures. CQ t values were now found to be from 0.131 to 0.139, with a minimum CQ l never falling below 0.10. It was observed that some of the CQ t values were slightly higher for model A4 in an urban layout than when in isolation. This is thought to indicate that the negative leeward wall pressures may have

been reduced in magnitude as the group layout closed up. Atrium models A5 (with a porosity of 11.4%) and A6 (with a porosity of 30.4%) operated under positive and near atmospheric roof pressure regimes respectively. Model A5 had CQ t values from 0.086 to 0.119 (0.179 in isolation), whilst model A6 had CQ t scores from 0.070 to 0.126 (0.162 in isolation). Minimum ow values for both models were raised to around 0.06. Fig. 4 shows, for the 08 wind direction, the ratio for each model and each area density of the CQ t value measured in the urban layout to the CQ t value with the model in isolation. It is apparent from Fig. 4 that the sheltering effect of the urban surroundings reduced the average ows through most of the models from 40% to 60% of the average ow values when the models were isolated. The exception for all area densities was model A4, which had a large number of roof openings operating under a suction ow regime. The combination of big openings and strong suction forces overcomes the effect of the surrounding buildings.

3.2. 458 wind direction

Two major changes were observed when the wind direction was altered to 458. Firstly, most of the CQ t values for the models increased that is, the ventilation performance of the models improved. Secondly, the range of CQ t values between the different atrium and courtyard models for a given group density became much narrower, being typically within 610% of the mean value for all models. The average CQ t values were approximately 0.15 at an area density of 0.28, 0.12 at 0.40 and 0.38 area densities and

Fig. 4. Ratio of sheltered to isolated average ow CQ t for the courtyard and atrium models: 08 wind direction.

Airow in courtyard and atrium buildings in the urban environment: a wind tunnel study

243

Fig. 5. Ratio of sheltered to isolated average ow CQ t for the courtyard and atrium models: 458 wind direction.

0.09 at 0.50 area density. The high porosity suction roof of model A4 still performed slightly better than the other arrangements, but the magnitude of the improvement was only signicant for the highest area density value. Fig. 5 shows, for the 458 wind direction, the ratio for each model and each area density of the CQ t value measured in the urban layout to the CQ t value with the model in isolation. It is apparent from Fig. 5 that the sheltering effect of the urban surroundings is not as pronounced as for the 08 wind direction. The average ows through most of the models were 60% to 80% of the average ow values when the models were in isolation.
4. CONCLUSION

ly, for the suction roofs the negative pressures on the leeward side of the building counteract the negative pressures on the roof. More efcient ventilation roof design may involve exploiting Venturi effects or vortex generation at roof leading edges where accelerated ows could be utilised.

REFERENCES
Natural ventilation in buildings: a design handbook, Allard F. (Ed.), (1998). James & James, London. Alvarez S., Sanchez F. and Molina J. L. (1998). Air ow pattern at courtyard. In Proceedings of PLEA 98, pp. 503 506, Lisbon. Andersen K. T. (1995) Natural ventilation in atria. ASHRAE Technical Data Bulletin 11(3), 3038. Bauman F., Ernest D. and Arens E. (1988) The effect of surrounding buildings on wind pressure distributions and natural ventilation in long building rows. ASHRAE Transactions 94(Part 2), 16701695. Chartered Institution of Building Services Engineers (1997). Natural ventilation in non-domestic buildings. CIBSE Application Manual AM10. Bioclimatic architecture, Gallo G. (Ed.), (1995). ENEA, Rome. Kindangen J., Krauss G. and Depecker P. (1997) Effects of roof shapes on wind-induced air motion inside buildings. Building and Environment 32(1), 111. Kukadia V. (1998) NatVentE: its aims and vision. In Proceedings of NatVentE Conference, Building Research Establishment, Watford. Limam K., Allard F. and Dascalaki E. (1997) Natural ventilation research activities undertaken in the framework of Pascool. Int. J. Solar Energy 19, 81119. Riskowski G. L., Ford S. E. and Mankell K. O. (1998) Laboratory measurements of wind effects on ridge vent performance. ASHRAE Transactions 104(Part 1A), 495 501. Passive cooling of buildings, Santamouris M. and Asimakopolous D. (Eds.), (1996). James & James, London. Santamouris M., Papanikolaou N., Koronakis I., Georgakis C. and Assimakopoulos D.N. (1998). Natural ventilation in urban environments. In Proceedings of 19 th AIVC Conference, 206213, Oslo. Saxon R. (1994). The atrium comes of age, Longman, London.

This study has investigated the ventilation performance of courtyards and atrium buildings in isolation and in urban group layouts. An open courtyard in urban areas was found to have a weak ventilation performance, particularly when the courtyard was perpendicular to the oncoming wind. Covering the courtyard with a porous roof to form an atrium enables the large pressure elds on the roof to provide stronger ventilation pressure differentials. Roofs positioned to experience positive or near atmospheric pressure conditions performed less well than roofs exposed to negative pressure forces (suction) when winds are perpendicular to the buildings. At an oblique wind direction (458 ) most of the atrium roofs performed to a similar standard. There are two major problems with using atrium roofs as ventilation devices in urban areas. Firstly, to use the weaker positive pressures requires large surface areas of roof and / or a great number of openings. Second-

244

S. Sharples and R. Bensalem Turner S. and Binnie P. (1990) An indoor air quality survey of twenty-six Swiss ofce buildings. Environmental Technology 11, 303314. Walker R. R., Shao L., and Woolliscroft M. (1993). Natural ventilation via courtyards: theory and measurement. In Proceedings 14 th AIVC Conference, 235250, Copenhagen.

Schaelin A., Moser A., van der Maas J. and Aiul D. (1996) Application of air ow models as design tools for atria. In ROOMVENT96, Proceedings 5 th Int. Conference on Air Distribution in Rooms, pp. 171178. Schild P. G., Tjelaat Per O. and Aiul D. (1995) Guidelines for CFD modelling of atria. ASHRAE Transactions 101(Part 2), 13111332. Shao L., Walker R. R. and Woolliscroft M. (1993). Natural ventilation via courtyards: the application of CFD. In Proceedings 14 th AIVC Conference, 393403, Copenhagen.