See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/237581710

Building morphology, transparency, and energy performance

Article · January 2003

CITATIONS READS

64 344

2 authors, including:

Ardeshir Mahdavi
TU Wien
631 PUBLICATIONS 4,288 CITATIONS

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

EVA - Evaluierung Visionärer Architekturkonzepte: Prüfung bauphysikalisch und energetisch innovativer Gebäudekonzepte auf ihre Machbarkeit unter Monitoring und
Evaluierung eines Mock-Ups View project

EDEN - Entwicklung einer strukturierten und fehlerminimierten Datenaufbereitung und Dokumentation für Energieausweise View project

All content following this page was uploaded by Ardeshir Mahdavi on 29 January 2021.

The user has requested enhancement of the downloaded file.

 Eighth International IBPSA Conference
Eindhoven, Netherlands
August 11-14, 2003

BUILDING MORPHOLOGY, TRANSPARENCE, AND ENERGY PERFORMANCE

Werner Pessenlehner and Ardeshir Mahdavi

Department of Building Physics and Human Ecology
Vienna University of Technology
Vienna, 1040 - Austria

ABSTRACT orientation of a building (e.g., south orientation versus

Certain energy-related building standards make use of
simple numeric indicators to describe a building's
geometric compactness. Typically, such indicators
make use of the relation between the volume of a built
form and its surface area. The indicators are then used
along with information on the thermal transmittance of
the building enclosure elements to evaluate the degree
to which a building design meets the relevant thermal
insulation criteria. Using extensive parametric thermal
simulations, this paper examines the reliability of such
simple compactness indicators for energy-related
evaluative assessments given that buildings with the
same compactness attribute could differ in enclosure
transparence, orientation, and morphology.
1. INTRODUCTION
west orientation) does not change its compactness, but
may affect thermal performance given changes in
insolation and shading conditions.
Given these critical considerations, the present study
examines the reliability of geometric compactness
indicators for energy-related evaluative assessments
based on extensive parametric thermal simulation
studies. We explore, via variations of building
morphology and transparence (the size and distribution
of transparent enclosure components), the limitations
of exclusive reference to shapes compactness in
thermal performance assessment guidelines and
standards. Specifically, we demonstrate the thermally
relevant interdependencies between compactness and
transparence for a specific climatic context and for a
morphologically varied class of residential building
shapes.

Prescriptive building energy codes often set minimum

requirements concerning thermal properties of
building components. To account for the geometry of
buildings in a simple manner, some energy-related
building standards make use of simple numeric
indicators that focus on building's geometric
compactness (Heindl and Grilli 1991, Mahdavi et al.
1996, ÖNORM 2002). Typically, such indicators are
derived based on the relationship between the volume
of a built form and the surface area of its enclosure.
The indicators are then used along with information on
the thermal transmittance of the building enclosure
elements to evaluate the degree to which a building
design meets the relevant minimum thermal
requirements. However, the usage of geometric
compactness for such evaluative purposes could be
criticized on multiple grounds. First, compactness does
not capture the specific morphology (or the unique
three-dimensional formal articulation and massing) of
a building's shape, even though it could influence the
thermal performance (e.g. via self-shading). Second,
compactness does not capture the amount and
distribution of the transparent components of the
enclosure. Thus, corresponding radiative gains and
losses are not accounted for. Third, changing the
2. APPROACH
2.1 Overview
The research design for the present study involves the
following steps: i) A sample of different building
shapes is selected, providing morphological variance;
ii) Different glazing scenarios are generated through
variance in glazing area and orientation; iii) The
resulting set is thermally analyzed via simulation; iv)
The simulation results (energy load, overheating
index) are discussed in the context of the sample's
variance in morphology and transparence.
2.2 Shapes
A modular geometry system was derived based on an
elementary cube (3.5×3.5×3.5 m). To generate
different building shapes, 18 such elements were
used (see Figure 1). These elements were aggregated
in different ways to create 54 morphological
variations. Figure 2 illustrates this set according to
their compactness.

- 1025
1017 -
 interchangeably to characterize shape compactness.
We prefer to use RC, though, as previous studies
have indicated that it better describes the subjective
18 ELEMENTS (perception-based) categorization of shape
MODULE = 3,5 x 3,5 x 3,5 m
VOLUME = 771,75 m3 compactness by designers (Mahdavi u. Gurtekin
2002a). Moreover, lc can be easily derived from RC
if necessary:
Figure 1 – Generation of shapes based on 18 cubical
elements lc = RC . V0.66 . 6-1 (eq. 4).

From the sample shown in Figure 2 a subset of 12

shapes with distinct RC values was selected for the
simulations (see Figure 3).
0,98

0,90

0,86
RELATIVE COMPACTNESS (RC)

0,82

SHAPE 01 SHAPE 02 SHAPE 03 SHAPE 04

0,79 RC = 0,98 RC = 0,90 RC = 0,86 RC = 0,82
lc = 1,50 lc = 1,37 lc = 1,31 lc = 1,26

0,76

0,74

SHAPE 05 SHAPE 06 SHAPE 07 SHAPE 08

0,71 RC = 0,79 RC = 0,76 RC = 0,74 RC = 0,714
lc = 1,21 lc = 1,17 lc = 1,13 lc = 1,09

0,69

0,66
SHAPE 09 SHAPE 10 SHAPE 11 SHAPE 12
RC = 0,69 RC = 0,66 RC = 0,64 RC = 0,62
0,64 lc = 1,05 lc1,02 lc = 0,98 lc = 0,95

0,62
Figure 3 – The 12 shapes selected for simulation

Figure 2 – All generated shapes

To specify compactness, we used the "Relative
Compactness" (RC) indicator (Mahdavi u. Gurtekin
2002a). The RC of a shape is derived in that its
volume to surface ration is compared to that of the
most compact shape with the same volume. For
sphere as the reference, it is given by:
RC = 4.84 . V0.66 . A-1
Since most buildings have orthogonal polyhedronal
shapes, we use cube as the reference shape, thus
arriving at the following definition of RC:
To achieve variation in enclosure transparence, the
amount of glazing and its distribution across the
enclosure walls was changed. Concerning glazing
area, three levels were considered, namely 10%, 25%,
and 40% glazing, expressed as fraction of the gross
floor area (Figure 4). Subsequently, the glazing area
was distributed across the enclosure in five different
ways, as per Table 1. In addition, the shapes were
rotated four times in intervals of 90 degrees (Figure 5).
(eq. 1). Given 12 shapes, 3 glazing area options, 5 glazing
distribution options, and 4 orientations, a total of 720
variations were thus generated for simulations.

RC = 6 . V0.66 . A-1 (eq. 2). 10% 25% 40%

RC is purely shape-dependent, in contrast to

conventional compactness indicators such as
characteristic length (lc) which depend on the shape's
size (i.e., absolute value of the volume):
lc = V . A-1
Figure 4 – Illustration of the variation of the glazing
percentage (fraction of gross floor area)
(eq. 3).

Since our main morphological sample involves

shapes of equal volumina, RC and lc could be used

- 1026
1018 -
Table 1 – Variations of transparence 2.4 Simulations

Percentage of glazing facing: Simulations were performed using the application

NODEM (Mahdavi u. Mathew 1995). The simulation
Variation North East South West
results were expressed in terms of two performance
Uniform 25 25 25 25
indicators, namely annual heating load (in kWh.m-2.a-1)
North 55 15 15 15
and overheating index (in Kh.a-1). The latter was
East 15 55 15 15
defined as the sum of hourly temperature differences
South 15 15 55 15 between the room temperature and an overheating
West 15 15 15 55 reference temperature (27o during day and 25o in the
night). This reference temperature is currently applied
in Austria for residential buildings (ÖNORM 1999).
UNIFORM NORTH EAST
N/E/S/W = 25/25/25/25 % N/E/S/W =55/15/15/15 % N/E/S/W =15/55/15/15 %

3. RESULTS

Figure 6 shows (for all simulated instances) the

SOUTH
N N/E/S/W =15/15/55/15 % N/E/S/W =15/15/15/55 %
W O
S Larger glazing areas result in slightly lower heating
ROTATION
IN INTERVALS
OF 90 DEGREES
Figure 5 – Glazing area distribution
2.3 Invariant assumptions
A number of parameters were kept constant
throughout the simulations, namely location (Vienna,
Austria), building use (residential), building volume
(772 m3), building construction, internal heat gains,
relationship between heating load and RC. The
respective correlation is fairly high (R2=0.88). If the
WEST
results are sorted according the glazing percentage (see
Figure 7), even higher correlations emerge (R2=0.95
for 10% glazing, 0.94 for 25%, and 0.90 for 40%).
loads (increased solar gains apparently outweigh
increased transmission losses). A further distinction of
the results in terms of the distribution of glazing (see
Figure 8 for 25% glazing area option) reveals still
higher correlations (0.97 for uniform and north, 0.95
for south, 0.96 for west and east). These results
confirm the expectation that dominantly south-oriented
glazing results in the lowest heating load, whereas
dominantly north-oriented glazing results in the
highest heating load.
35,00

and air change rates. The assumptions regarding

32,50
thermal transmittance and surface density values of the
Heating Load (kWh.m -3.a-1)

primary building components of the model are 30,00

summarized in Table 2. Lighting and equipment loads,
as well as the number of people were assumed to be 27,50
dependent on the time of day, amounting to an average
value of approximately 5 W.m-2. Air change rates were 25,00

varied from 0.5 to 2 h-1 according to the time of the

22,50
year.
20,00

17,50
Table 2 – Building components properties
15,00
0,60 0,65 0,70 0,75 0,80 0,85 0,90 0,95 1,00
Building Surface density U
component [kg.m-2] [W.m-2.K-1] Relative Compactness (RC)
Floor 850 0.33
External wall 310 0.21
Roof 860 0.16
External floor 620 0.20 Figure 6 – Simulated heating loads as a function of RC
Window 20 1.10 (all instances)

- 1027
1019 -
 35,00 based on the orientation of glazing (see Figure 10),
higher correlations emerge (0.93 for uniform, 0.88 for
32,50 south, 0.87 for west, 0.84 for north, and 0.86 for east).
Heating Load (kWh.m -3.a-1)

The higher occurrence of overheating for the south

30,00
exposure (as compared to east and west) is mainly due
27,50
to the longer duration of the south façade's solar
exposure.
25,00

22,50 3000

2750
20,00
2500
17,50
2250

Overheating (Kh.a-1)
2000
15,00
0,60 0,65 0,70 0,75 0,80 0,85 0,90 0,95 1,00
1750
Relative Compactness (RC) 1500
10% 25% 40%
1250
10% 25% 40%
1000

Figure 7 – Simulated heating loads as a function of RC 750

and glazing percentage (10%, 25%, and 40%) 500

250

0
35,00
0,60 0,65 0,70 0,75 0,80 0,85 0,90 0,95 1,00

32,50
Relative Compactness (RC)
Heating Load (kWh.m -3.a-1)

30,00

Figure 9 – Simulated overheating as a function of RC

27,50
(all instances with 40% glazing)
25,00

3000
22,50
2750

20,00 2500

2250
Overheating (Kh.a-1)

17,50
2000

15,00 1750
0,60 0,65 0,70 0,75 0,80 0,85 0,90 0,95 1,00
1500
Relative Compactness (RC)
1250
uniform south west north
1000
east uniform south west
north east 750

500

250

Figure 8 – Simulated heating loads as a function of 0

RC and glazing distribution options (for 25% glazing 0,60 0,65 0,70 0,75 0,80 0,85 0,90 0,95 1,00

option) Relative Compactness (RC)

uniform south west north
east uniform south west
While from the simulation results a significant north east
association between heating load and RC can be
inferred, the association between overheating and RC
is relatively week. Since simulations for 10% and 25% Figure 10 – Simulated overheating as a function of RC
glazing areas did not result in noteworthy overheating and glazing distribution options (all instances with
occurrences, only the results for 40% glazing are given 40% glazing)
in Figure 9 (R2=0.59). If these results are grouped

- 1028
1020 -
4. DISCUSSION 20%

A solid association between RC and heating load can 15%

Deviation from regression in %

be inferred from the sample of shapes considered in
10%
this study. More compact shapes result indeed in

(Heating Load)
somewhat smaller heating loads. RC seems to capture 5%
geometry well, despite its negligence of the
morphological variance of the sample. Distinctions 0%

regarding transparence (amount and orientation of

-5%
glazing) allow to further refine this association, but do
not change its general trend. Moreover, given low U- -10%
value glazing systems, increased glazing area can –
contrary to the conventional wisdom – reduce heating -15%

load, whereby increased transmission losses through

-20%
the enclosure are more than compensated by increased 0,60 0,65 0,70 0,75 0,80 0,85 0,90 0,95 1,00
solar heat gains. Relative Compactness (RC)

While overheating tendency increases with increasing

RC, the association is much weaker in this case.
Figure 11 – Deviation of simulated heating loads from
Overheating is significantly affected by the amount of
regression-based predictions (for 40% glazing option)
the glazing (in the present study, it did not occur for
glazing areas under 30%). Moreover, the orientation
of glazing has a clear influence on the resulting
overheating. As expected, a south-dominant glazing
orientation results in significantly higher overheating 125%
than the north-dominated glazing orientation. The
deviation of the individual results from the general 100%

trend are generally large, implying that RC does not

Deviation from regression in %

75%
sufficiently capture those morphological properties of
50%
the design (such as self-shading) that could be relevant
(Overheating)

to the occurrence of overheating. 25%

0%
To further contrast the performance of RC in the case
-25%
of heating load to its performance in the case of
overheating, we consider the errors that occur, when -50%

heating load and overheating predictions are made

-75%
based on regression equations. Figure 11 illustrates the
relative deviation of individual simulation results for -100%

heating load (for 40% glazing area) from the -125%

corresponding predictions made based on linear 0,60 0,65 0,70 0,75 0,80 0,85 0,90 0,95 1,00

regression (cp. Figure 7). Deviations lie in this case Relative Compactness (RC)
between -10% and +10%. Figure 12 illustrates the
relative deviation of individual simulation results for
overheating from the corresponding predictions made
based on linear regression (cp. Figure 9). Deviations Figure 12 – Deviation of simulated overheating results
are in this case much larger and lie between -75% and from regression-based predictions (for 40% glazing
+125%. option)

- 1029
1021 -
5. ADDITIONAL EXPLORATIONS 20%

5.1 Different shapes 15%

Deviation from regression in %

In the previous discussion, we dealt with a sample of 10%

12 shapes with distinct RC values. To explore the

(Heating Load)
potential effect of morphological variance not 5%

accounted for by RC, we selected (from figure 2) five

0%
morphologically distinct shapes with the same RC
value of 0.86 (Figure 13) for further simulation
-5%
studies. Heating loads and overheating were computed
for these shapes, whereby the same glazing options -10%
and distributions were considered as in the previous
study (cp. Sections 2.2, 2.3, 2.4). Figure 14 illustrates -15%

the deviation of the simulated heating loads for these A B C D E

five shapes from predictions based on the regression -20%

equation of the original sample (cp. Figure 6). Figure Relative Compactness = 0,86
15 illustrates the deviations of the simulated
overheating values for the five shapes from predictions
based on the regression equation of the original sample Figure 14 – Deviation of simulated heating loads (for
(cp. Figure 9). A comparison of the error ranges of the the sample of five shapes with identical RC values)
sample of these five shapes with the error ranges of the from regression-based predictions of the original
original 12 shapes sample is presented in Table 3. The sample (cp. Figure 6)
deviations resulting from shape variance are in the
case of heating load in the same order of magnitude as
the deviations of the original sample, and in the case of
overheating somewhat larger than the deviations of the
150%
original sample. These results imply that regression-
based heating load predictions can reasonably rely on 125%

RC as geometry indicator, despite morphological 100%

Deviation from regression in %

variance. However, such predictions are not reliable in 75%

the case of overheating and are further compromised
50%
due to morphological attributes not captured by RC.
(Overheating)

25%

0%

-25%

-50%

-75%

-100%

A B C D E
-125%
A B C D E
-150%

Figure 13 – Sample of five distinct shapes with the Relative Compactness = 0,86
same RC value (RC = 0.86)

Figure 15 – Deviation of simulated overheating results

(for the sample of five shapes with identical RC
values) from regression-based predictions of the
original sample (cp. Figure 9)

- 1030
1022 -
Table 3 – Comparison of regression-based prediction 35,00

errors: five shapes sample versus original sample

32,50

30,00

Heating Load (kWh.m -3.a-1)

27,50
Deviation range for:
25,00
5 shapes sample Original sample
22,50

Heating load -15 to +10 % -15 to +12%

20,00

Overheating -80 to +130% -75 to +125% 17,50

15,00

12,50

10,00
5.2 Different volumina 0,60 0,65 0,70 0,75 0,80 0,85 0,90 0,95 1,00

Relative Compactness (RC)

The original study dealt with shapes of the same
volume (772 m3). To explore the implications of V 772 V 2680 V 6174
changes in volumina, a new sample was generated V 772 V 2680 V 6174

based on a subset of the shapes in the original sample

Figure 16 – Heating load of sample C as a function of
shown in Figure 3. We selected five shapes with
RC (for three volume ranges and 25% glazing option)
distinct RC values. We further assumed three possible
volumina for each shape (the original volume of 772
m3, as well as 2680 m3 and 6174 m3). The simulations
35,00
were performed for the 25% glazing option only.
Since RC is not volume-dependent, Figure 16 32,50

illustrates the simulated heating loads as a function of

30,00
Heating Load (kWh.m -3.a-1)

both RC and volume. Figure 17 illustrates the same

results as Figure 16 but instead of RC, it uses the 27,50

volume-dependent lc as the compactness indicator. 25,00

These results imply that simulated heating loads (HL)
22,50
can be reproduced via regression-based functions
fairly well, even when dealing with shapes of different 20,00
volumina. Accordingly, the association implied in
17,50
figure 17 can be represented via the following
(equivalent) equations: 15,00

12,50

10,00

HL = 27 . lc-0.75 (eq. 5) 0,50 1,00 1,50 2,00 2,50 3,00 3,50

Characteristic Length (lc)

HL = 105 . RC-0.75 . V-0.25 (eq. 6)

Figure 17 – Heating load of sample C as a function of
lc (for the 25% glazing option)

- 1031
1023 -
 6. CONCLUSION
Given the context and boundary conditions of the
present study, the reliability of simple indicators of
building geometry such as compactness indicators RC
and lc must be seen differentially. The association
between the values of such indicators and simulated
heating loads of buildings with various shapes,
orientation, glazing percentage, and glazing
distribution was found to be significant. Accordingly,
the use of such indicators in energy standards (for
heating load prediction and evaluation purposes) may
be justified. However, these indicators do not appear
to capture the geometry of a building to the extent
necessary for the predictive assessment of the
overheating risk.
Beyond their potential for relevant codes and
standards, reliable and intuitive numeric indicators of
geometry could be adopted as a design dimension of
the "design-performance space", which denotes a
virtual space defined by multiple design and
performance dimensions (Mahdavi u. Gurtekin
2002b). In this space, each design dimension
accommodates the range of possible values of a
discretized design variable. Design variables can
capture various geometric (volume, shape,
compactness) and semantic (thermal transmittance of
the building enclosure, thermal mass, internal loads)
features of design. Likewise, each performance
dimension accommodates the range of the values of a
specific performance indicator (e.g., energy use,
reverberation time in rooms, illuminance levels on
working planes). The design-performance space can
provide an effective context for the assessment and
comparative evaluation of the performance of
alternative building designs in the early stages of the
design process.
- 1032
1024 -
View publication stats
7. REFERENCES
Heindl, W., Grilli, P.V. 1991. On Establishing
Standards for the Overall Heat Transfer Coefficient
of Buildings, CIB-W67-Workshop 1991, Vienna.
Mahdavi, A., Gurtekin, B. 2002a. Shapes, Numbers,
and Perception: Aspects and Dimensions of the
Design Performance Space, Proceedings of the 6th
International Conference: Design and Decision
Support Systems in Architecture, The Netherlands,
ISBN 90-6814-141-4. pp 291-300.
Mahdavi, A., Gurtekin, B. 2002b. Adventures in the
design-performance space, Proceedings of the 16th
European Meeting on Cybernetics and System
Research. Vienna, Austria. Volume 1, pp. 269-274.
Mahdavi, A., Mathew, P. 1995. Synchronous
generation of homologous representations in an
active, multi-aspect design environment,
Proceedings. of IBPSA Conference. 31(5), pp 522-
528.
Mahdavi, A., Brahme, R., Mathew, P. 1996. The
"LEK"-Concept and its Applicability for the Energy
Analysis of Comercial Buildings, Building and
Environment, 31(5). pp 409-415.
ÖNORM 2002. ÖNORM B 8110-1: Wärmeschutz im
Hochbau – Anforderungen an den Wärmeschutz und
Nachweisverfahren, Österreichisches Normungs-
institut, Wien, 29. Januar 2002.
ÖNORM 1999. ÖNORM B 8110-3: Wärmeschutz im
Hochbau – Wärmespeicherung und Sonneneinflüsse,
Österreichisches Normungsinstitut, Wien, 1.
Dezember 1999.