POST OCCUPANCY DESIGN INERVENTION TO IMPROVE COMFORT AND ENERGY PERFORMANCE IN
A DESERT HOUSE
Arizona State University
P O Box 871605, Tempe, AZ, 85287-1605, USA
Venkata Ramana Koti
Arizona State University
P O Box 871605, Tempe, AZ, 85287-1605, USA
This paper reports on the methods used to analyze the
effect on thermal comfort and energy performance as a
result of incremental changes applied to the building
envelope. The object of investigation was the “House of
Earth and Light”, a residence built in the hot and dry
climate of the low Arizona desert. As new owners
occupied the house in 2003, they experienced severe
overheating in the living room. A simple single path
method was developed to determine the effect on surface
temperatures as a function of changes in floor, glass
walls, and roof R-values as well as changes in the indoor
air temperature. The mean radiant temperature (MRT)
was then calculated for each incremental improvement
while the indoor air temperature was changed in order to
achieve a target operative temperature (OT) of 78°F. The
effect of the incremental improvements on the energy
performance of the building was analyzed by means of
hour-by-hour calculations using the Energy-10 simulation
Thermal comfort is related to both the physiological
contentment of the inhabitant with a space and the
necessary energy used to condition the space. In 1999,
when the Building Owners and Management Association
(BOMA) in partnership with the Urban Land Institute
(ULI) surveyed 1,829 office tenants in which they were
asked to rate the importance of 53 building features and
amenities, 99% of them felt thermal comfort was an
important feature (1).
Fig. 1: House of Earth and Light: night view of the fabric
Despite the best intentions of the designer, nature has its
own ways of reminding us of the potential thermal
shortcomings of an assembly. Sometimes, in an attempt to
address a wider universal ideal, formal and spatial
considerations take priority over environmental concerns.
Though applying post occupancy corrective measures are
not the best approach, it sometimes becomes imperative.
The single story residential building under consideration
here has been widely published and commended for its
transparency to the sky during day and night (2). The
analysis described in this paper was limited to the living
and dining room, a 36 by 24 foot steel and glass “bridge”
with a concrete floor spanning across a dry wash (arroyo).
As built by the original owners, a translucent fabric roof
covered the entire house, including the “bridge”. The
outer perforated fabric layer was stretched over three-
dimensional steel trusses extending six feet to form an
overhang on the south side. Except for the north and south
facing glass walls of the bridge, all exterior walls and
most of the interior walls were made of 18-inch thick
slabs of concrete poured in plywood forms.
The building is situated in Phoenix, which has a hot and
dry desert climate with the highest outdoor air
temperature measured in 2004 at 112°F. When the
original owner built the house, they made several
deviations from the architect’s specifications: The living
room floor slab, extending across the dry wash, was not
insulated. Single pane glass was used in the floor-to-
ceiling north and south walls, except for a few smaller
double pane glass panels. The third (inner) roof fabric
layer was not installed.
Fig. 2: Typical cross-section showing the three-layered
fabric roof assembly, as designed. Note; this section is not
cut at the living room (bridge).
Since the house had not been built entirely according to
specifications, the goal of this research was to investigate
potential incremental improvements in thermal comfort
and energy performance as a function of bringing the
floor, glass walls, and roof up the standard that the
architect had specified. In accordance with claims
presented by the architect, a hypothesis was formulated as
Indoor thermal conditions in the living room could be
brought into the comfort zone on a summer design day (or
design hour) if the specified incremental improvements
were put in place.
2.1. Data acquisition
Surface and air temperature data were recorded in the
living room (the “bridge”) on May 27th, 2003 from before
sunrise until after sunset. Outdoor air temperatures were
recorded near the entrance on the south side, in the shade.
A complete data set is available upon request.
This particular monitoring day turned out to be hot and
sunny with the ambient temperature reaching 111 °F at 2
PM. Since the summer design temperature for Phoenix is
107 °F and the thirty-year average maximum dry bulb
temperature is 115°F, the conditions at 2 PM on May 27th
of 2003 may be seen as a “design hour” used to size the
cooling system and to assess summertime thermal
comfort. The part of the analysis that pertains to thermal
comfort was therefore focused on the 2 PM hour on May
27th, while the energy performance simulations were
carried out using an 8760-hour TMY2 weather file for
2.3. Surface temperatures
A single path steady state heat flow method was
developed to simulate changes in surface temperatures as
a function of lowering the indoor air temperature and
increasing the R-values of the floor, glass walls, and roof
assemblies. Since the R-values of the individual
components were known and each layer of the assembly
would have a temperature difference based on the
gradient of R-values, the R-values were tweaked until the
calculated surface temperature matched the measured.
This straightforward approach proved valid for the floor
assembly, as seen in Fig. 3.
Since the single path heat flow method did not account for
thermal bridging represented by the steel framing of the
glass walls, and similarly did not account for heat gain
from radiation represented by ground reflection, the
outdoor air temperature was elevated 10°F as a means of
calibrating the glass wall module of the model. Similarly,
the outdoor air temperature was elevated 29°F in order to
calibrate the roof module of the model. This elevated
outdoor air temperature for the roof calculations may be
seen as an “attic” temperature of the air mass trapped
between the outer perforated fabric (shade cloth) and the
inner weatherproofing fabric.
Fig. 3: Calibration of R-value of the existing floor
assembly using measured surface and air temperatures.
All temperature values are in °F.
With the single path method calibrated, we could now
simulate changes in the surface temperatures as a function
of lowering the indoor air temperature. Fig. 4 shows how
the surface temperature of the floor would be lowered to
87°F if the indoor air temperature was lowered to 71°F.
The temperature difference between the indoor air and the
floor surface would increase from 9°F at 88°F indoor air
temperature to 16°F at 71°F indoor air temperature.
Fig. 4: Calculated effect on surface temperature as a
function of reduced indoor air temperature. All
temperature values are in °F.
The next step was to introduce an improvement in the R-
value of the assembly. Fig. 5 shows how the temperature
difference between the indoor air and the floor surface
could be reduced to 1°F by adding 13 inches of fiberglass
insulation to the underside of the floor slab.
Fig. 5: Calculated effect on surface temperature of
reduced indoor air temperature and 13 inches of fiberglass
insulation added to the underside of the concrete slab. All
temperature values are in •F.
2.4. Mean radiant and operative temperatures
Based on the measured and calculated surface
temperature data, mean radiant temperature (MRT) and
the resultant operative temperature (OT) were calculated
using an MRT calculator that allows for simulating step
changes in the indoor air temperature, see Fig. 6.
Fig. 6: The MRT calculator was designed to allow for
simulating the effect of variable indoor air temperature.
Mean radiant temperature as experienced by a seated user,
see ASHRAE Handbook of Fundamentals, equation 53.
The value cells for floor, glass walls, and roof surfaces
temperatures were linked to the single path calculations in
order to take advantage of the dynamic relationship
between changing air temperatures and the temperatures
of all interior surfaces. The surface temperatures of the
interior (thermally neutral) east and west mass walls were
assumed to be 1°F above the indoor air temperature, as
observed on the day the space was monitored.
As can be seen in Fig. 6, the data at 2 pm shows a
temperature difference of 9.7o F between the calculated
MRT and the maintained indoor air temperature, and a
resultant operative temperature of 92.8o F. Note that the
interior surface temperature of the fabric roof was
measured as 119°F while the outdoor air temperature near
the ground was 111°F. The asymmetry in the space was
30°F between the warmest surface (ceiling) and the
relatively cooler interior mass walls. Studies have shown
that a lower percentage of people express dissatisfaction
with a cool ceiling and a warm wall, and a higher
percentage of people express dissatisfaction with a warm
ceiling and cool walls (3), see Fig. 9.
Operative temperature (OT), which is one of the factors
primarily responsible for thermal comfort, is an average
of the MRT and the indoor air temperature (IAT). Hence
thermal comfort requires that OT falls within the comfort
zone on the psychrometric chart and also that the MRT
and the IAT are fairly close to each other. A high MRT
will require lower IAT in order to bring the operative
temperature within the comfort zone. This in turn requires
lower supply air temperature (from the AC unit) and
higher air velocity (since more heat needs to be removed
per unit of time).
Six variants (conditions) were defined relative to the
existing condition as monitored (variant 1) and
incremental improvements represented by lower IAT and
higher R-values (variants 2-6). Surface temperatures,
MRT, and OT were calculated for variants 2 through 6.
The target OT was set at 78°F for these variants.
2.5. Simulating energy performance
The annual energy performance of the 36 by 24 foot space
with thermally neutral end walls was simulated in Energy-
10 for all variants 1 through 6.
Though initially eQUEST was considered for simulation,
it was realized that modified surface temperatures could
be determined using the single path method explained
earlier. Since only a single zone was considered for this
analysis, Energy-10 could be used to perform hour-by-
hour load calculations and simulations of the annual
A 36 feet long and 24 feet wide space with a height of 10
feet was defined in Energy-10. The 10 feet height
accounted for the 9-foot high glazing on the north and
south with a 1-foot high bond beam above. The concrete
masonry walls on the east and west were considered
thermally neutral with an R-value of 1000.
The “auto-build” function in Energy-10 sets up a
“reference case” roughly representing standard practices
and a “low-energy” case roughly representing best
practice energy efficient design. While simulating the six
variants of the incremental change method, the low-
energy case can be seen as a benchmark.
As compared to the low-energy case (benchmark) and the
space as monitored (v-1), incremental improvements were
defined as follows: Lowering IAT to 71°F in order to
obtain an OT of 78°F (v-2), adding insulation to the floor
(v-3), modifying the existing single glazing to a double
low-E glazing (v-4), and adding a second fabric layer with
an intermediate air space to the roof (v-5). These variants
are all defined within the basic premise of maintaining the
visual appearance and architectural quality of the
building. Variant 6 represents the actual situation today
(2005) after the new owners decided to replace the fabric
roof with sections of solid standing seem covered opaque
roof panels intersected by stripes of low-E glazed
skylights at the steel trusses, see Fig. 7.
Fig. 7: Solid roof with skylights, as built 2005.
The results from the Energy-10 simulations were kept as a
series of bar charts as exemplified in Fig. 8.
Fig. 8: Annual cooling energy use in MMBtu, as
simulated for the auto-build reference case, the low-
energy (benchmark) case, and the six variants of the
incremental change analysis.
A summary of the results from the single path method, the
MRT-calculations, and the Energy-10 simulations is
included in Table 1at the end of this paper.
3.1. Conditions in the space as monitored
The indoor air temperature measured at 2 pm was 88oF.
This condition reflects the fact that the AC unit was
unable to maintain a lower IAT due to the excessive heat
gain. The OT was estimated at almost 93°F and the
asymmetry was 30°F between the interior mass wall
surfaces and the warm ceiling surface. Obviously, this
was not a condition of thermal comfort.
3.2. Lowering the IAT
A target value for the OT of 78°F could theoretically be
achieved by lowering the IAT to 71°F with no
improvements to the building envelope. This would lower
the MRT to approximately 85°F, but the asymmetry
would increase to 40°F since the ceiling fabric
temperature would still be around 112°F. The simulated
annual energy requirement for heating, cooling, and fans
would increase to more than the double, from 58 MMBtu
to 117 MMBtu. For comparison, the similar value for the
benchmark “low-energy” case was only 13 MMBtu.
3.3. Insulating the floor
A layer of 13” fiberglass insulation was added to the
underside of the floor slab in order to reduce the surface
temperature and decrease the heat loss. This would allow
us to maintain the OT at 78°F with a slightly increased
IAT at 72°F, while lowering the MRT only about 1°F.
The simulated annual energy requirement for heating,
cooling, and fans was reduced to about 93 MMBtu.
3.4. Replacing single pane glass with double pane low-E
The R-values of the north and south glass walls were
improved by replacing all single pane glazing with double
pane low-E glass. This would allow for an increase in the
IAT to 74°F while maintaining an OT of 78°F. Again, the
MRT would decrease only 1°F, but the simulated annual
energy demand for cooling, heating, and fans would
decrease to 80 MMBtu.
3.5. Adding a second layer of fabric with an intermediate
air space to the roof
An improvement of the fabric roof, as originally specified
by the architect, was now added to the already improved
floor and glass wall assemblies. The calculated values
show that this improvement would reduce the ceiling
fabric surface temperature to around 96°F. While keeping
the OT at 78°F, the MRT would decrease to 81°F and the
IAT could be increased to 75°F. The calculated
asymmetry between the warm ceiling surface and the
cooler interior mass walls would go down to 20°F and the
simulated annual energy requirement for cooling, heating,
and fans would decrease to about 61 MMBtu.
This condition represents an estimate of the thermal
comfort at the “summer design hour” of the “bridge”
space as specified by the architect, with a corresponding
annual energy requirement to condition the space
accordingly. With a simulated annual cooling energy
demand at about four times the benchmark “low-energy”
case in Energy-10, it is safe to assume that the airflow rate
and the forced air velocity would be quite high. One could
still argue that a reasonable level of thermal comfort
could be achieved in a space as specified, but this
marginal level of comfort would come at a cost of annual
energy for cooling, heating, and fans at about four times
the benchmark “low-energy” case.
3.6. Replacing the fabric roof with solid roof panels
intersected by stripes of skylights
The new owners decided to replace the fabric roof with
solid (opaque) standing seem covered roof panels
intersected by stripes of skylights at the steel trusses. The
exact specifications of this new roof assembly were not
known at the time of writing this report, but it is estimated
that the new roof assembly has an R-value of 24.
Keeping the OT at 78°F, the ceiling surface temperature
(opaque panels) was now estimated at 78°F. This is a
radical improvement from all the previous variants 1
through 5. The IAT could now be increased to 76°F. The
north and south glass wall surfaces would still be quite
warm, but the asymmetry would decrease to about 9°F
between the floor surface and the interior surface of the
south facing glass wall. The simulated annual energy
requirement for cooling, heating, and fans would roughly
be cut in half to 28 MMBtu.
In order to verify the hypothesis that thermal comfort
could be achieved by implementing the improvements
specified by the architect, one needs to assess the
estimated percentage of dissatisfied users. This approach
is widely accepted in the field (3).
Fig. 9 shows that the calculated surface temperatures of
variant 5, representing a radiant asymmetry of about 20°F,
would produce approximately 25% dissatisfied users.
This is a significant improvement over variant 1 and 2,
which could produce up to 80% dissatisfied users.
One could argue that the 2 PM “summer design hour” as
observed on May 27th, 2003 represents an extreme
condition and that a reasonable degree of comfort is
achieved for most times during a typical year. Until other
additional improvements have been explored, however,
we conclude that the hypothesis has not been verified.
The failure to verify the hypothesis with certainty, along
with an estimated high cost of installing a third layer of
fabric, lead the new users to make the decision to replace
the fabric roof.
The methods used in this analysis proved to be a set of
adequate, yet relatively simple and straightforward tools
for this type of task where both thermal comfort and
annual energy performance were investigated. Upon the
approval of the new owners, data acquisition will be set
up again for one day during early summer 2005. This new
data set, along with accurate specifications of the actual
improvements as built 2005, will be used to assess the
energy performance and thermal comfort of the space
with a higher degree of accuracy.
Fig. 9: Radiant temperature asymmetry. Reproduced
from Stein & Reynolds page 52 (3).
Table 1: Summary of surface temperatures (measured and calculated) and energy performance (simulated using Energy-10).
(1) The Regents of University of California. Thermal
Comfort of UFAD Systems. Sourced on 23rd November
(2) Several magazine articles: Dimensions, Vol.12;
Dwell, October 2000; Dwell, 2001; Architecture, May
2002; and others.
(3) Stein B, John S. Reynolds, Mechanical and Electrical
Equipment for Buildings, 9th Edition.