Proceedings of Building Simulation 2011:

12th Conference of International Building Performance Simulation Association, Sydney, 14-16 November.

A MODELLING METHOD TO ASSESS THE EFFECT OF TREE SHADING FOR

BUILDING PERFORMANCE SIMULATION

Dominique Hes1, Anthony Dawkins1, Chris Jensen1 & Lu Aye2

1
Faculty of Architecture, Building and Planning, The University of Melbourne, Australia
2
Renewable Energy and Energy Efficiency Group, Department of Infrastructure Engineering,
Melbourne School of Engineering, The University of Melbourne, Australia

McPherson reported that shading on the west side of

ABSTRACT houses showed the highest reduction in cooling
Increasing urban tree numbers is a simple but energy demand and that adding two shade trees on
effective means to provide climate change adaptation the west and one on the east of a house would
to the urban environment by reducing the thermal reduced annual cooling costs for the house by
load on buildings. To better communicate and value between 10-50% (Simpson 1996).
the importance of urban trees it is necessary to Donavan and Butry provide data that supports
quantify these benefits and to understand the simulations and previous studies in this area. A
properties and processes that influence the magnitude simulated application of their method to the Akbari et
of these benefits. For this we need verified and al. study in California shows forecasted savings to be
effective ways of modelling the trees in modelling consistent with measured savings. They stated that
software. morning shade is of little value as it is provided
This paper presents the results and problems during hours when air-conditioning is not typically
encountered when trying to model trees effectively. required while western shade offers the greatest value
The aim is to present our approach which was to treat (Donovan and Butry 2009). This assumption
the shade as a shading co-efficient on the wall. This discounts the value of eastern surface temperature
allows for the consideration of the benefits of reductions, which may contribute significantly to the
deciduous versus evergreen species. A modelling daily reductions. Given the nature of our study we
method to assess the effect of tree shading was will be able to quantify the value of east façade
developed and presented in this paper. shading in relation to the total daily reductions and to
Key words: thermal simulation, solar thermal load, peak energy demand. Donavan’s study also considers
modelling of tree shading crown size and distances of trees from the buildings
in order to determine their impact on energy
reduction. A system was required to normalize the
INTRODUCTION classification of buildings, sites and tree cover as
they measured the benefits of existing tress at some
There are three studies that provide significant
460 locations. As this study was intended to close the
precedents with specific regard to the cooling energy
gap between large-scale simulations and small scale
savings achieved through tree shade. These include
empirical testing it is understandable that the
the 1996 simulated study by Simpson and
complexity of the study required much assumption
McPherson, the Akbari et al. 1997 small-scale study
and extrapolation of collected data. It presents good
of two houses in California, and the 2009 study of
evidence that simulated results provide realistic
460 single family homes in California by Donovan
forecasts for actual energy savings achieved through
and Butry. Nikoofard et al. 2011 recommended that
strategic planting of shade trees. We will take a step
external shading effects need to be accounted for in
down in scale in addition to removing a large number
modelling residential energy consumption. These
of assumptions in order to focus on indoor
studies document the impact of tree shade on the
environment conditions without the impact of varied
cooling energy demand through theoretical and
internal gains, size, construction and orientation of
empirical data across small, medium and large scales.
buildings, occupancy profile and microclimate. The
The results correlate well and provide evidence that
value of our study is that there will be very little
data obtained from both controlled studies and from
forecasting and extrapolation due to the length of
simulations is relatively consistent and can
time that data will be collected.
potentially provide the basis for reliable forecasting
of the value of strategic tree planting within the urban Akbari et al. conducted an experiment where 16
environment. deciduous trees, eight at 6 meters tall and eight at 2.4
meters tall, were placed on the east and west of two
Akbari, Bretz and Hanford showed seasonal cooling
similar houses in California. Data included air
energy savings of 30% (4 kWh per day) and
conditioning electricity use, indoor and outdoor dry
estimated peak demand savings of 0.7 kW per house
bulb temp and humidity, roof and ceiling surface
in their study (Akbari, 1997). Simpson and
temperatures, inside and outside wall temperatures,

- 161 -
 Proceedings of Building Simulation 2011:
12th Conference of International Building Performance Simulation Association, Sydney, 14-16 November.

insulation, and wind speed and direction. They computer simulations provide the opportunity to
reported up to 25% reductions in outside surface develop effective predictive modelling and a tool for
temperatures as a result of shading and a reduction in strategic planting of urban environments.
wind speeds of up to 16%. The experiment produced The investigation into effective methods of tree shade
96 days of complete data out of the 129 days of simulation shown here have been undertaken in the
measuring. Air conditioners were continuously on, context of a field test where a comprehensive data set
windows were always closed, and lighting electricity has been, and continues to be collected for the
use was strictly controlled. They identified two major purpose of long term studies of the effect of shade on
flaws in their data collection. Firstly, thermocouples annual heating and cooling energy use. A basic
would lose adhesion with external surfaces, which explanation of this experiment and its methodology is
meant that temperature readings were not of the presented here to define the context within which the
surface but of the pocket of air that formed between field test data is produced. The approach of
the thermal couple and surface material. The second simulating trees as three dimensional objects has
flaw was that ceiling mounted sensors indicated been used in an effort to simplify the input process
rising temperatures even after the outside while allowing the software to account for the
temperature was falling or the air conditioning was shading effects as well as the reduction of wind
operational. This was due to direct solar gain and/or velocities due to the presence of trees, which can
convection, depending on the placement of the have a significant effect on building surface
sensors. This study is very close in nature to our own temperatures. An alternative method of adjusting
and therefore will provide a valuable point of shading coefficients on the walls to represent the
reference against which to compare results and shading effect is also presented, and forms the basis
findings. for this paper. The results, methods and limitations of
The precedents highlight the importance of isolating each approach are discussed below.
the effects of tree shade in order to quantify the
energy savings. A possible flaw in existing data is METHOD
human behavior as the actual energy consumption The site for this study located in Eastern metropolitan
was measured. As we have removed the human Melbourne with coordinates 37°48′49″S,
element there is an opportunity to have significantly 144°57′47″E. The moderate oceanic climate results
more control over the test environment and in significant residential heating and cooling loads
consequently, to obtain more specific information throughout the year and is an ideal climate for
about the internal temperatures and energy demands. passive design principles. The site context was
The value of shade during summer was relatively modelled with test buildings being unshaded in order
consistent between the studies, with all reporting or to compare simulation results against field test data
forecasting an average annual energy saving of for the purpose of calibrating the digital model and to
around 30%. These studies took place in Sacramento, assess the impact of local shade on each building’s
California, which is a Mediterranean climate internal temperature. It was found that the small
(Koppen Csa) and typically produces hotter summers difference in amounts of early morning shade had a
and colder winters when compared to the climate of negligible impact on the internal temperatures of the
Melbourne. It can be assumed that the same study buildings. For the purpose of simplifying the
taken place in a Melbourne would produce lower computer model and eliminating the influence of
energy savings due to lower cooling and heating varied environmental factors on simulated results it
demands. It is commonly understood that energy was deemed appropriate to exclude the surrounding
savings range significantly between tree species, with trees and buildings from the simulations (Figure 1a
deciduous trees providing the greatest annual benefit, and 1b).
permitting solar gain during the winter months. The
assumption is made that reported benefits of tree
shade in these studies refer to the use of deciduous
species when referring to annual benefits.
Simulations by Akbari et al. performed in parallel to
their field tests indicated that energy savings were
consistently underestimated by up to 50% when
compared to measured data. This is an indication of
the complexity of modelling environmental factors
such as tree shade and how assumptions and
approximations can significantly affect the
simulation results. It is unclear as to the expected
margin of error in other simulated studies, however Figure 1a Site model
value of simulations in our context lies primarily in
the verification of trends within measured data. Used
in conjunction with data from controlled field tests,

- 162 -
 Proceedings of Building Simulation 2011:
12th Conference of International Building Performance Simulation Association, Sydney, 14-16 November.

Figure 2a Figure 2b
Figure 2a shows a two dimensional plane with
perforations of approximately 50% of the surface
area. Figure 2b shows the intersecting of two planes
to represent a tree canopy and approximated
shading.

Figure 1b sensor set up

The next challenge was to effectively model the

proposed shade trees within Integrated
Environmental Solutions’ Virtual Environment
software (IES-VE) and to generate realistic results Figure 3a 100% Shading model for IES-VE
that would be confirmed by measured data. It should
be noted that the IES-VE software, while powerful
and intuitive, has not been used for detailed studies
on external shading at the time of this paper so
significant experimentation was necessary in order to
establish the appropriate method required to produce
the required output. Given the complex geometry of a
single tree, and the inherent variations even within
trees of the same size and species, it is necessary to
abstract their physical form. Limitations within the Figure 3b 50% Shading model for IES-VE
software required the vector count for modelled
‘trees’ to be as low as possible while our experiment The properties of building materials, building
demanded the most accurate representation envelope and operation factors such as blinds,
achievable. It was also necessary to ensure that the infiltration, ventilation and internal heat gains where
effect of indirect solar radiation reflecting off leaf modelled to replicate those of the field test with the
surfaces was included. The solution was to represent greatest degree of accuracy possible. Figure 4
a tree canopy with two single circular planes illustrates the geometry and materials of the field test
intersecting each other at right angles (refer to buildings while Table 1 lists the critical properties
Figures 2a, 2b and 3a, 3b). that have been measured, predetermined or assumed,
and are consistent between the field test and
Solid planes would represent full shade similar to the simulations.
Golden Ash trees during summer while small
perforations amounting to approximately 50% of the
surface area of the circular planes would represent
the Red Ironbark trees throughout the year. While the
perforation areas of 50% and 0% appear arbitrary,
they are a reasonable approximation of the shading
effect of our test trees during summer and provide a
good point of reference upon which conclusion can
be drawn and recommendations made.

Figure 4 measured drawings for typical shed

- 163 -
 Proceedings of Building Simulation 2011:
12th Conference of International Building Performance Simulation Association, Sydney, 14-16 November.

method is the lack of accuracy with which the

Table 1 Summary of building and environment amount of shading is determined. Factors such as leaf
attributes. reflectivity and emissivity, leaf positions,
evapotranspiration and canopy density are not
scientifically calculated and represented leading to
potential error due to over or under estimation. This
is why the values of 0%, 50% and 100% shading
were used for the ‘trees’ within the IES model.
Shading coefficients of trees
The second method was to consider the effect of
shading as a shading coefficient for the wall
receiving shade. The use of shading coefficients for
the purpose of simulating the effect of shade has been
tested by Jim and He (2011) in their study of
estimating the heat flux transmission of green walls
and by McPherson and Simpson (1996) in their broad
As environmental properties between the field test simulated study of the energy saving benefits of
and computer simulations needed to be consistent it urban tree shade. It appears that McPherson and
was necessary to construct a weather file from the Simpson simulated tree shade by modelling
collected data that could be used within IES-VE. abstracted trees and applying an appropriate shading
Climate data for the IES-VE software is based on a coefficient to each tree canopy. While there are clear
typical weather year (TWY) however this did not advantages of this approach, an alternative method of
correlate well with the actual weather experienced simulating shading effect is presented here in the
during the data collection period (an unusually cold interests of establishing a predictive modelling tool
and wet summer). This made direct comparisons that can be applied to a broader range of situations.
between simulated results and collected data difficult By applying a similar approach to Jim and He,
and often misleading. Once the weather file was average hourly shading coefficients of both the
constructed calibration of the digital model was Ironbark and Golden Ash trees from the field test
confirmed and the necessary adjustments made. data can be used to establish an adjusted solar
The calibration process is critical as it provides a absorptance value for external surfaces. This adjusted
reference point from which to compare various solar absorptance value mimics the effect of tree
shading conditions as well as providing the link shade on external wall surfaces. In order to account
between empirical and simulated data. Previous for the variation of incident solar radiation
studies of this nature have had the element of human throughout the day hourly average data is applied. In
behaviour and the variations between subject the context of IES this is a simple procedure
buildings to consider, and both impact on the ability however, separate simulations need to be run for each
to quantify the value of shade.(Simpson 1996, Akbari adjusted solar absorptance value in order to reflect
1997) As these factors have essentially been the changes in incident radiation to wall surfaces.
excluded from our experiment it is reasonable to The simulated data provides a relatively accurate
suggest that with a well-calibrated model meaningful representation of daily cooling energy use and an
and relevant data will be produced. effective comparison between species and location of
trees is possible.
Limitations of IES-VE
Shading coefficients of two species of tree were
The IES-VE is a building energy modelling software estimated by using ‘Sol-Air Temperature’ (Te) for
package intended to inform the design of both vertical surfaces equation (ASHRAE Handook
buildings and environmental systems across a range 2009).
of project scales. As its focus is on controlling
internal building conditions the use of this software
Eq (1)
to predict effects of an external and dynamic element
such as shade is problematic and requires
Where:
considerable simplification in the modelling process.
To = outdoor air temperature
The value of the software for our application is in the
α = absorptance of surface for solar radiation
forecasting the energy savings achieved through a
Et = total solar radiation incident on surface
range of shading conditions, extending the empirical
ho = outside air convective heat transfer coefficient
research to different scales, densities and contexts.
There are two reasonable approaches that can be Equation (1) can be rewritten as follows for the wall
used. The first, described above, consisted of with tree shade.
abstracting the ‘trees’ producing approximated
shading of external surfaces. The shortfall of this

- 164 -
 Proceedings of Building Simulation 2011:
12th Conference of International Building Performance Simulation Association, Sydney, 14-16 November.

of 2.1% for the 50% shading and a 4.5% increase for

Eq (2) the 100% shading. During the cooling months, from
November to March, we saw a reduction of cooling
Where: energy by 8.5% for the 50% shading and 11.2% for
Tes = sol-air temperature for a wall with shading the 100% shading.
SC = shading coefficient Using the shading co-efficient approach, the
SC equation can be obtained by combining Equations simulated data shows the effect on cooling loads of
(1) and (2). each adjusted absorptance value with reference to the
unshaded control building, illustrated by the Figure 6.
Eq (3)
It is apparent that shade provided to the west faced
has almost double the effect on cooling energy
The measured outside wall temperatures were reduction than shading to the North and East façade.
assumed to be same as Sol-Air Temperatures and the Surprisingly there is only a marginal reduction in
time varying shading coefficients were calculated cooling energy use with the denser Golden Ash
(see Table 2). canopy as compared to the Ironbark trees (Figure 5).
The following table lists the adjusted solar However, in the context of annual energy
absorptance values that were applied to the external consumption the presence of shade through the use of
walls of the simulated buildings. The unshaded an evergreen species such as the Ironbark tree has a
control building (Shed B) has a solar absorptance significant effect on heating energy during winter
value of 0.6 for external walls and adjusted values months as the passive heating effects of solar gain are
were calculated using the following method: reduced.

Adjusted Solar Absorptance = 0.6 x (1 – Shading

Coefficient) Eq (4)

Table 2
Shading coefficients derived from 4 – 7 January 2011
data - Adjusted solar absorptance

Time (Hour) Iron bark Golden Ash

Shed A Shed C
9:00 0.378 0.360 Figure 5 – showing the decreased cooling
10:00 0.426 0.408 requirements for the shaded buildings (7 January
11:00 0.462 0.438 2011)
12:00 0.486 0.450 The summary of field test results below are achieved
13:00 0.498 0.450 form a culmination of both methods. They represent
14:00 0.498 0.438 a starting point for further investigation and are
15:00 0.480 0.408 assumed to be conservative average energy saving
16:00 0.450 0.360 reductions. The dynamic nature of the ‘real world’
17:00 0.402 0.300 result in greater maximum saving and this will
18:00 0.348 0.222 undoubtedly be supported by future field test data.
RESULTS Calibration of the model
The complexity of accurately representing shading In order to simulate the impact of tree shade on
effects through computer modelling is evident in this cooling and heating energy use throughout the year a
investigation and it is clear that more work is reliable predictive model is required. The approach
required in order to produce definitive results across taken in this research was to run consecutive studies
the year. Because of the limitations of the field of both computer simulations and of empirical data
testing and simulation it is assumed that, like collected from controlled field tests. It was assumed
previous studies, the energy saving benefits are that the two studies would validate one another while
underestimated to a some degree. Without the providing the basis for a predictive model that could
support of extended field test data it is not possible to be applied to a broader context. However, the
speculate as to the extent that savings are complexity of modelling the effects of trees on
underestimated here. The research is continuting for internal temperatures is significant and output is
the next 2 years to try to reduce these uncertainties. often only useful in the support of a notional ideas
and concepts. Since the study of tree shade is largely
IES - VE Simulations an exercise in proving the obvious, it is necessary to
Using the percentage coverage approach the data develop a specific and relatively accurate predictive
shows an increase in winter time heating energy use modelling tool that can be used to determine actual

- 165 -
 Proceedings of Building Simulation 2011:
12th Conference of International Building Performance Simulation Association, Sydney, 14-16 November.

savings to space conditioning energy use through the • Indoor air temperature [oC] (Sensor: Type T
presence of tree shade. thermocouple, 30 AWG, Neoflon PFA
Calibrating the model raised additional questions insulated)
about methodology and the appropriate tools for The total thermal resistance of layers 1 to 5 (Table 3)
measuring shading effects in the context of this was estimated by using Equation 5.
experiment, and without long-term field test data an
accurate assessment of the software and inputs Eq (5)
cannot be made.
Field test inside temperature data was the control Where:
used to calibrate the computer model. The graph in
Figure 6 shows inside temperatures of building B
from the field test and simulation and indicates some
variation between data sources with the IES
simulation producing warmer temperatures during
warm times and cooler temperatures during cool
times. An explanation for this is that while the field
test temperature represents a single point in the room
the simulation temperature is an average of the whole
room. As this includes air adjacent to windows and Table 3 Thermal resistance of wall (weatherboard
doors we see a greater variation in overall shed), Summer
temperature. The fields test data showed that the Thickness Conductivity Conductance Resistance

location of the sensor was very influential on the No Construction [m] [Wm-1K-1] [Wm-2K-1] [m2KW-1]

reported inside temperature and so it is reasonable to x K C=k/x R=1/C

accept these simulated results as rational. Again, it 1 Indoor air - - 8.5 0.118
will be necessary to review this against future data film
and particularly through varied climatic conditions in 2 Masonite 0.006 0.060 10.0 0.100
order to verify its accuracy. board
3 Fibreglass 0.050 0.049 0.98 1.020
batts
4 Air space 0.030 - - 0.030
5 Weather 0.020 0.220 11.0 0.091
boards
6 Outside - - 24.0 0.042
air film
Total resistance 1.401

The total thermal resistance calculated for the wall is

1.4m2K1W-1. If we assumed the temperature
difference 10°C between the inside room and the
Figure 6. A comparison between simulated and field outside ambient temperature, the calculated steady
test inside air temperatures of the unshaded building state heat flux across the wall is 7.1 Wm-2. The
B. It has been calibrated and variations are due to resultant temperature differences across the wall
average overall room temperatures of the layers are shown in Table 4. It should be noted that
simulations. the estimation of properties for the bulk insulation in
the wall is critical since it provides about 73 % of the
Sensitivity analysis of increasing the insulation thermal resistance.
levels
Thermal resistance for the fiberglass batt applied in Table 4 Temperature differences across the wall
the test sheds were unknown. It was estimated by
layers (To-Ti=10oC)
using the on-site measured data. The following
Resistance Temperature
hourly average measured data were used for the
No Construction [m2KW-1] difference
estimations. [C]
• Heat flux across the wall [Wm-2] measured 1 Indoor air film 0.118 0.84
at the interface between Fiberglass batt and 2 Masonite board 0.100 0.71
Masonite board (Sensor: Omega Thin Film 3 Fibreglass batts 1.020 7.28
Heat Flux Sensor, HFS 4) 4 Air space 0.030 0.21
• Surface temperature of the outside wall [oC] 5 Weather boards 0.091 0.65
(Sensor: Omega Infrared Thermocouple, 6 Outside air film 0.042 0.30
Type K, OS36)

- 166 -
 Proceedings of Building Simulation 2011:
12th Conference of International Building Performance Simulation Association, Sydney, 14-16 November.

Figure 7 shows that the increased insulation levels loads. For climates such as Melbourne, these benifits
from R1.4 to R3 will reduce the impact of the are greatest using deciduous trees which allow sun to
shading by almost 20%, for the simulated day. contribute to passive winter heating.
Remembering that 0.6 was the solar absortpance of
the unshaded building and 0.36 that of the Golden
Ash were most shaded. REFERENCES
Akbari, H. 2002. "Shade trees reduce building energy
use and CO2 emissions from power plants."
Environmental Pollution 116(Supplement 1):
S119-S126.
Akbari, H. 2007. "Opportunities for Saving Energy
and Improving Air Quality in Urban Heat
Islands." Lawrence Berkeley National
Laboratory, LBNL Paper LBNL-751E
Akbari, H., Kurn, D.M., Bretz S. E., Hanford, J.W.
1997. "Peak power and cooling energy savings
of shade trees " Energy and Buildings 25: S119-
Figure 7 Showing the reduced effectiveness of the S126.
shading of the external wall with increased insulation
from R1.4 to R3 Akbari, H., United States. Dept. of Energy. 1992.
"Cooling our communities : a guidebook on tree
Thus, even though the impact is less it is still planting and light-colored surfacing.",
worthwhile to shade the walls with high solar
Washington, D.C., U.S. Environmental
incidence.
Protection Agency, Office of Policy Analysis.
Other benefits of trees Donovan, G.H., Butry, D.T. 2009. "The value of
Apart from potential energy savings from direct shade: Estimating the effect of urban trees on
reduction of heat load on the external surfaces of summertime electricity use." Energy and
buildings there are other benefits to trees. The first is Buildings 41(6): 662-668.
the impact on the microclimate. Trees reduce the
Jim, C.Y., He, H., 2011. Estimating heat flux
temperature of their general vicinity by
transmission of vertical greenery ecosystem.
evapotrasporation. Particularly relevant to the city
Ecol. Eng. (in press)
where increased temperatures have seen increased
physioplogical impact even death, this impact on the Kahn, P., Kellert, S.R. 2002. Children and nature:
reduction of the Urban Heat Island effect is part of psychological, sociocultural, and evolutionary
the ongoing research of this site (Rosenzweig et al., investigations. MIT Press
2006; Rosenfeld et al., 1997). Kellert, S.R. 1993. The Biophilia Hypothesis. Island
Current research also shows that having access and Press
contact with natural elements such as trees will affect Kellert, S.R., Heerwagen, J., Mador, M. 2008.
the inhabitant positively. That is, they are Biophilic Design: The Theory, Science and
psycologically, emotionally and physiologically Practice of Bringing Buildings to Life, New
healthier (Wilson, 1984; Kahn and Kellert
York: John Wiley.
2002;Kellert 1993; Kellert, Heerwagen, Mador,
2008; Sternberg, 2009). McPherson, E. G., Simpson J.R. 2004. "Potential
energy savings in buildings by an urban tree
CONCLUSION planting programme in California." Urban
Field tests and simulations have helped to better Forestry and Urban Greening: 73-86.
understand the impact of tree shade on heating and Nikoofard, S, Ugursal, VI & Beausoleil-Morrison, I
cooling energy, however the broader goal of this 2011, “Effect of external shading on household
research is to provide an effective predictive model energy requirement for heating and cooling in
with applications across a range of scales and Canada”, Energy and Buildings, vol. 43, no. 7,
shading conditions. The shading coefficient method pp. 1627-35.
presented here draws from empirical data and
Rosenfeld, A., Romm, J., Akbari, H., and Lloyd, A.
provides the foundations for such a predictive model,
1997. "Painting the Town White -- and Green".
with simulated results correlating well with collected
MIT's Technology Review 100: 52-59.
field data.
The research has also shown that there are potential Rosenzweig, C., Solecki, W., Parshall, L., Gaffin, S.,
energy savings using trees to shade the external Lynn, B., Goldberg, R., Cox,J., Hodges, S. 2006.
surfaces of buildings, although the spicies of shade Mitigating New York City's Heat Island With
trees impacts heavily on annual heating and cooling Urban Forestry, Living Roofs, And Light

- 167 -
 Proceedings of Building Simulation 2011:
12th Conference of International Building Performance Simulation Association, Sydney, 14-16 November.

Surfaces. New York State Energy Research and

Development Authority, accessed May 29th
2001,
<http://www.giss.nasa.gov/research/news/20060
130/103341.pdf>
Simpson, J.R. 2002. "Improved estimates of tree-
shade effects on residential energy use." Energy
and Buildings 34(10): 1067-1076.
Simpson, J. R., McPherson E.G. 1996. "Potential of
tree-shade for reducing residential energy use in
California." Journal of Aboriculture 22(1): 10-
18.
Sternberg E.M. 2009. Healing Spaces: The Science
of Place and Well-being. Cambridge: Harvard
University Press.
Wilson, E.O. 1984. Biophilia. Cambridge: Harvard
University Press.

- 168 -