Content uploaded by Timur Dogan
Author content
All content in this area was uploaded by Timur Dogan on Apr 06, 2020
Content may be subject to copyright.
Towards High-Resolution Annual Outdoor Thermal Comfort Mapping In Urban Design
Patrick Kastner1, Timur Dogan1
1Environmental Systems Lab, Cornell University, Ithaca, New York, USA
Abstract
Global warming and increasingly dense cities lead to poor
outdoor thermal comfort that may not only be detrimental
to our health and well-being but also decreases social and
commercial activities. Although workflows for the anal-
ysis of thermal comfort exist, they have yet transitioned
into the quotidian architectural design process. Our work-
flow allows for annual outdoor comfort analyses that are
seamlessly integrated into a commonly-used CAD environ-
ment. We simulated the annual outdoor thermal comfort
on a university campus and discuss which simplifications
seem appropriate by means of preliminary on-site measure-
ments. The results exemplify the possibility to conduct
such analyses within reasonable time and accuracy if some
simplifications to the UTCI estimation are acceptable.
Introduction
With rapid urbanization, outdoor spaces begin to compete
for natural resources such as access to sun, wind, air, and
daylight with other driving forces like space use efficiency.
Poor outdoor thermal comfort leads to many problems
related to the health and well-being of humans and also
decreases social and commercial outdoor activities. Mu-
nicipalities, universities, and organizations with large plots
of land are making efforts toward significantly reducing
their carbon footprint and focusing on sustainability. New
York, for example, plans to reduce carbon emissions by 80
% by 2050 (NYC.gov,2019) and Cornell University aims
to be carbon neutral by 2030 (Collins et al.,2016). Holistic
strategies are thus necessary to facilitate an urban design
that fosters passively comfortable urban microclimates in
light of future environmental risk and global warming.
In the last decade, the interest in outdoor thermal comfort
analysis increased both in the scientific community and in
practice. Some commercial tools exist which aim to sim-
plify the input of complex physical boundary conditions
by step-by-step guides. Those tools include but are not lim-
ited to ENVI-met (Huttner and Bruse,2009), SOLWEIG
(Solar and Long-Wave Environmental Irradiance Geom-
etry (Lindberg et al.,2008), RayMan (Matzarakis et al.,
2010), CitySim (Walter and K
¨
ampf,2015), and the Lady-
bug Grasshopper plugins (Mostapha Sadeghipour Roudsari,
2013). They quantify outdoor environmental conditions
by estimating either the Mean Radiant Temperature (MRT)
or the Universal Thermal Climate Index (UTCI). While
designers, planners, and municipalities are usually commit-
ted to creating pleasant outdoor spaces, it remains difficult
to assess the impact of the urban form on outdoor ther-
mal comfort. For fast-paced urban design processes, in
particular, none of the state-of-the-art approaches seem rea-
sonable. Generating feedback about the urban design from
such simulations is a computationally expensive modeling
task that is cumbersome to apply in an iterative manner.
Naboni et al. postulate that the acquisition of building
geometry is not streamlined enough to allow for usage
in an integrated architectural design environment. Hence,
microclimate studies are still not widely used in the urban
planning process, although studies have shown that geomet-
ric interventions are able to improve the outdoor thermal
comfort in urban areas (Ebrahimabadi,2015;Thorsson
et al.,2011).
This study aims to facilitate the annual simulation of out-
door comfort and proposes an easy-to-use and reliable
methodology using OpenFOAM, Radiance, and Energy-
Plus to simulate the annual outdoor comfort of a university
campus within a reasonable time. Further, we validate
the results by means of on-site measurements and discuss
possible simplifications.
Case study: campus of Cornell University
The university campus of Cornell University located in
upstate NY and overlooking Lake Cayuga consists of 608
buildings and covers an area of
9.3 km2
. The campus is
surrounded by forest and mid-sized townhouses, see fig-
ure 1. The section of interest in this study is the vicinity
around a library building which is also used for on-site
measurements, see figure 1b.
Simulation framework for annual outdoor
comfort analyses
The simulation framework consists of several simulation
engines, namely: OpenFOAM, Daysim, and Radiance, see
figure 2. All engines are centered around a toolkit called
Eddy3D that is implemented in Rhinoceros and Grasshop-
per which handles pre-, post-processing, and the data han-
dling between those engines. We use Eddy3D to create
the simulation domain, the specification of boundary con-
ditions, and the processing of the weather data based on
the building geometry in Rhinoceros. Similarly, we use
Rhinoceros’ meshing capabilities to export building and
terrain meshes for both OpenFOAM and Radiance. We use
Radiance to simulate irradiation and compute view factors
for each sensor point. From an EnergyPlus simulation, we
estimate surface temperatures from which we calculate the
mean radiant temperature (MRT) in combination with the
respective view factor for each hour of the year and sensor
point. For the annual outdoor thermal comfort evaluation,
we simulated 8 wind directions in a 45
◦
interval, that serve
________________________________________________________________________________________________
________________________________________________________________________________________________
Proceedings of the 16th IBPSA Conference
Rome, Italy, Sept. 2-4, 2019
621
https://doi.org/10.26868/25222708.2019.210458
a) b)
Figure 1: a) Aerial view of the university campus with the simulation domain highlighted in yellow, the library building of
interest (M), and 3 local weather stations (S1-S3). Image source: Google Earth. b) 3D model of library building showing
the terrace that was used for on-site measurements.
as the nearest neighbor lookup table for the annual wind
velocity data. On a number of sensor points (see figure 7),
we calculate the dimensionless wind velocity by dividing
the probes from the CFD simulation by the scaled inlet
velocity according to equation (1).
Uscale =
κ·U
logzre f +z0
z0
κ
·log re f .height +z0
z0(1)
Then, we multiply the dimensionless wind velocity of
each sensor point from the nearest simulated wind direc-
tion with the corresponding velocity and wind direction
from the weather data for every hour of the year. This
yields a matrix with wind reduction factors of the size
[8760 h x number of sensor points]
from which the wind
velocities for the UTCI calculation are retrieved. Those
wind velocities and mean radiant temperatures serve as
input for the UTCI calculation for every hour and sen-
sor point for which both ambient temperature and relative
humidity are retrieved from the hourly weather data.
Weather data
To set up the initial simulation in a way that it facilitates
later validation, seasonal analysis of historical weather
data is required. We parsed weather data from the
nearby Weather Underground stations KNYITHAC1 (S1),
KNYITHAC52 (S2), and KNYITHAC70 (S3), and one uni-
versity weather station (S4) from 2016 until 2018 — see
figure 1for their locations. We sanity-checked the data
and found the same prevalent wind direction across all 4
weather stations was SSE (
≈160◦
) for that period. From
the off-campus station S4, we extracted one day that ex-
hibited cold/sunny weather conditions with a relatively
constant wind direction for which on-campus measure-
ments were also available. In early 2019, such conditions
were present on March 27 for Ithaca, upstate NY.
Universal Thermal Climate Index (UTCI)
Like other outdoor comfort metrics such as the physiolog-
ical equivalent temperature (PET), the wind chill index
(WC), or the SET temperature, the UTCI was developed
conceptually as an equivalent temperature measure. It is
based on the multi-node thermo-physical “Fiala” model
which was coupled with an adaptive clothing model to
take clothing habits into account by the urban population
and their behavioral changes in clothing in relation to the
surrounding temperature (Fiala et al.,2012). Thus, for
any combination of air temperature, wind, radiation, and
humidity, UTCI is defined as the air temperature in the
reference condition which would elicit the same dynamic
response of the physiological model (Br
¨
ode et al.,2013).
We chose the UTCI as a performance metric, as a number
of studies have shown the UTCI to be a superior metric for
potentially hazardous weather in winter while also achiev-
ing good performance in tropical climates (Proven
c¸
al et al.,
2016). Mathematically, the UTCI is a polynomial approxi-
mation that, in simple terms, may be described as (Broede,
2009;Hardy,1998):
UTCI =Ta+f(Ta,TMRT ,U,pva pour )(2)
The resulting temperatures may then be classified and re-
ported as thermal stress categories, see table 1. Ideally, one
would strive for maximizing the annual urban climate for
the ”No thermal stress” condition. Undoubtedly, the
TMRT
and
U
are the two input variables that exhibit significant
spatial variation in an urban neighborhood, whereas
Ta
and
pvapour remain relatively spatially constant.
Table 1: Thermal stress categories of the UTCI.
UTCI [◦C] Stress category
>46 Extreme heat stress
+38 to +46 Very strong heat stress
+32 to +38 Strong heat stress
+26 to +32 Moderate heat stress
+9 to +26 No thermal stress
+9 to 0 Slight cold stress
0 to -13 Moderate cold stress
-13 to -27 Strong cold stress
-27 to -40 Very strong cold stress
<-40 Extreme cold stress
________________________________________________________________________________________________
________________________________________________________________________________________________
Proceedings of the 16th IBPSA Conference
Rome, Italy, Sept. 2-4, 2019
622
Building geometry
Boundary conditions
Wind velocity [m/s]
.epw weather data
Meshing with
blockMesh and
snappyHexMesh
Inputs Toolkit
CFD for 16 wind
directions
Output
Pre-processing Simulation Post-processing
Raytracing with
Daysim/Radiance
Specification
of sensor
points
Mean radiant
temperature [°C]
Relative humidity [%]
Amb. Temp. [°C]
Domain-
builder
Meshing
UTCI [°C]
scaling with respect to epw. wind velocity for that hour
Figure 2: Flow chart that depicts the pre-processing, simulation, and post-processing workflows developed in this study.
CFD simulations
Computational model and simulation domain
Ithaca’s elevation varies to the extent that it is necessary to
make use of a digital elevation model. The model we used
incorporates the building footprints which we extruded to
their actual height reported in the campus GIS repository.
Computational mesh
The mesh was created by OpenFOAM’s blockMesh utility
for the background mesh and snappyHexMesh to subse-
quently snap the background mesh to the building geome-
try. For the background mesh, we used a cylindrical simu-
lation domain approach discussed in (Kastner and Dogan,
2018). This meshing approach allows for reusing the same
computational mesh for every subsequent wind direction.
The simulation domain was set up with a radius of 500 m
around the library building and a height of 350 m while tak-
ing into account all relevant surrounding buildings which
resulted in 11.6 ×106cells. A side view of the cylindrical
mesh with the terrain cutout is given in figure 3.
Figure 3: Side view of mesh with terrain cutout and the
silhouette of the campus buildings.
Boundary conditions
The Grasshopper plugin called Eddy3D was used to au-
tomate the pre-processing, including the assignment of
boundary conditions. Depending on the wind direction,
we mapped the inlets to a one-half circle of the simulation
domain and the outlet on the opposite side as described
in (Kastner and Dogan,2018). The half circular domain
inlet was set to a uniform profile for
U
,
k
, and
ω
, and a
roughness length
z0=1
that corresponds to “regular cov-
erage with large size obstacles with open spaces roughly
equal to obstacle heights, suburban houses” (Wallace and
Hobbs,2006), according to equations 3-5. The particu-
lar wind direction used for the validation study was 310
◦
,
based on the measured weather data. At the outlet of the
computational domain, constant pressure is assumed, while
the other variables are imposed to be zero-gradient. The
ground and the building geometry use the same boundary
conditions, a no-slip condition for velocity, a zero-gradient
condition for the pressure and wall functions for
k
and
ω
.
For the field turbulence eddy viscosity
νt
, the intelligent
wall function called nutUSpaldingWallFunction was used,
given its universal applicability across wide ranges of
y+
values (De Villiers,2006). The front, back, and top faces
are set to a symmetry boundary conditions for all vari-
ables. The kinematic viscosity,
ν
, was set to
1.5 ×10−5
.
The turbulence inlet parameters were calculated using the
following equations:
k=1.5·T2
u·U2
re f (3)
ε=Cmu ·k2
ν·µt
µ
(4)
ω=ε
Cmu ·k(5)
Other computational parameters
Since CFD is currently a bottleneck in this simulation pro-
cedure, we used an incompressible, isothermal, steady-
state solver from OpenFOAM in combination with a
k−ω−SST
RANS turbulence model to calculate the
wind velocities that are needed for the UTCI calculation.
We chose the
k−ω−SST
based on its superior accuracy
while only being slightly more computationally expensive
(Ramponi and Blocken,2012). The pressure-velocity cou-
pling was established with the SIMPLE algorithm using
three non-orthogonal correctors. Buoyancy effects were
neglected due to air velocities that are well above
1.8 m s−1
(Tecle et al.,2013;Boulard et al.,1996;Magnusson et al.,
________________________________________________________________________________________________
________________________________________________________________________________________________
Proceedings of the 16th IBPSA Conference
Rome, Italy, Sept. 2-4, 2019
623
2014). We ran all the simulations for 10000 iterations,
for which the simulations reached the following residuals:
1×10−4 f
or
ω
,
1×10−4 f
or
k
,
5×10−2 f
or
p
,
1×10−4 f
or
Ux
,
Uz
, and
Uy
. The relaxation factors were chosen to be
0.3
for
p
and
0.7
for
U
,
k
and
ω
. All simulations ran on
an AMD Ryzen Threadripper 1950X 16-Core Processor
running Windows 10. We used the Docker Version 2.0.1.0
(30090) to run OpenFOAM 4.1.
10-5
10-4
10-3
10-2
10-1
100
0 2000 4000 6000 8000
Residual
Iteration
Ux
Uy
Uz
p
omega
k
Figure 4: Residuals of the CFD simulation.
MRT calculation
Studies have analyzed several methods to estimate the MRT
that range from simple curve fits more holistic view factor
analysis methods (Kessling et al.,2013;Thorsson et al.,
2007). In this study, the MRT calculation was conducted
with the help of EnergyPlus with which we calculated ex-
terior surface temperatures. We then used Radiance to
calculate view factors with respect to all building surfaces,
ground patches and sky for all sensor points. From those
values, we calculated the solar-adjusted mean radiant tem-
perature according to:
MRT ="cs·αs·I·1
σ+
nv f
∑
i=1
Fp,i·T4
sr f ,i#1
4
−273 (6)
where
MRT
is Mean Radiant Temperature in
◦C
,
I
is the
hourly irradiation in
W m−2
,
σ
is the Stefan-Boltzmann
constant,
cs=0.25
is a projection factor assumed for seat-
ing,
αs=0.7
is an assumed skin absorption coefficient,
Fp,i
is the view factor between a sensor point and a surface
“i”, Tsrf ,iis the surface temperature in K.
Preliminary validation with on-site measure-
ments
Validation with on-site measurements is essential for a
complex metric with multiple inputs such as the UTCI. To
validate simulation results, we meter weather data includ-
ing temperature, humidity, and radiation as well as wind
speed and direction at an undisturbed and well-exposed
location ca. 1 km from campus, see S4 in figure 1. In
parallel, we monitor relevant UTCI components such as
Tamb
,
MRT
, wind speed and radiation exposure at the ob-
servation point shown in figure 6, using a Thermal Micro
Climate Data Logger shown figure 5. We then translate
the raw weather data into the EPW format utilizing psy-
chometric calculations and an irradiation split into indirect
and diffuse solar radiation using the HDKR/Reindl model
(Reindl and Beckman,1990). This EPW file serves as
weather input for the simulation model that we used to
approximate UTCI parameters for later comparison against
measurements at the observation point. The measurements
were taken on the paved terrace of the library building
to minimize the influence of evaporative cooling in this
area, which is not considered in our model, see figure 5.
The measurement frequencies were set to 15 s which was
averaged with a moving average over 240 instances. Dur-
ing the measurements, the approaching velocity at S4 was
observed, in order to wait for periods of time in which
constant 60-minute averaged wind speeds and directions
were present, which was the case between 13:00 and 16:00
h on March 27, see table 2. Here, we treated winds to come
from a “constant” direction if they lie within a 22
◦
angle
sector. With those 60-minute intervals, we were able to
extract time windows that were used as inputs for our simu-
lations. In sum, we used wind velocity, wind direction, and
radiation measurements from the undisturbed, unshaded
weather station as input for our simulation whereas ambi-
ent temperature, and relative humidity are taken from the
measurements on site.
Figure 5: HD32.1 - Thermal Microclimate Data Logger
used as measurement setup with sensors for ambient tem-
perature, relative humidity, global irradiation, and globe
thermometer. The photo was taken in the direction of the
approaching flow.
Results
With respect to validation of the CFD results, we report
that the moving average of the wind velocity at the mea-
surement location yields
2.5 m s−1
. The average probed
velocity from the red rectangle in figure 6shows a value of
2.6 m s−1, probing the larger rectangle yields 3.2 m s−1.
________________________________________________________________________________________________
________________________________________________________________________________________________
Proceedings of the 16th IBPSA Conference
Rome, Italy, Sept. 2-4, 2019
624
Table 2: Hourly average off-site measurements from the weather station
S4
on March 27, 2019. The on-site measurements
shown in figure 6were taken from 10:15 h to 18:30 h. From 13:00 to 16:00 h, both wind velocity and direction were fairly
constant. The wind direction 310◦was later used for the validation study.
Hour
Air temp [◦C] RH [%] Wind speed (m/s) Wind direction
Irradiation [W/m2]
Max Avg Max Avg deg comp
18:00 2.8 1.7 32 6.9 4.2 341 NNW 277
17:00 2.8 2.2 34 8.4 4.6 321 NW 447
16:00 2.2 1.7 32 8.7 5.1 315 NW 599
15:00 2.2 1.1 35 9.0 5.1 304 NW 704
14:00 1.1 0.6 39 7.9 4.9 307 NW 748
13:00 1.1 0.0 42 8.1 4.8 317 NW 741
12:00 0.6 0.0 42 6.8 4.2 326 NNW 668
Figure 7a-b) show a top view of the campus with UTCI
map for cold and sunny weather conditions recorded on
March 27, 2019, by the station S4, assuming a constant
wind velocity from 310
◦
(NW). We can see that the overall
UTCI patterns follow the daily cycle prescribed by the
ambient temperature, which increases from 12:00 h – 16:00
h. It is, further, evident that campus buildings provide
shelter against the wind, resulting in a UTCI increase in
their wake region. Finally, Figure 7c) depicts the annual
percentage of hours within comfort (without thermal stress)
across campus.
a) Top-view of cropped CFD result
b) Probed velocities around the measurement location
Figure 6: Measurement location for the UTCI validation
highlighted in red in the figure at the top. a) Top-view of
cropped CFD result at 2 m above ground. b) The average
wind velocity probed from the red rectangle at the measure-
ment location yields
2.6 m s−1
; the average wind velocity
probed from the larger rectangle around the measurement
location yields 3.2 m s−1.
Discussion
General limitations
The results presented show that it is possible to simulate
the annual outdoor comfort in an urban environment with
reasonable effort. The simplified methodology presented,
however, comes with limitations with respect to its general
applicability.
First, this study relies on separately validated engines as op-
posed to one integrated engine such as for example ENVI-
met or CitySim. As the UTCI is a metric that is derived
from multiple input data, the accuracy of these simulation
results depends on both the accuracy of every individual
engine and the extent of transient behavior occurring in the
real world. In the interest of simulation time, the simula-
tion engines in this paper were linked by static, external,
one-dimensional coupling, see figure 2, which is the sim-
plest of the three possible coupling strategies (Barbason
and Reiter,2014;Odnevall Wallinder et al.,2002). This
brings both advantages and disadvantages. For one, every
simulation engine has been validated on its own and there
exists trust among the research community that Radiance,
EnergyPlus and OpenFOAM are in principle able to gen-
erate accurate results. What is not possible, however, is
the reverse data exchange due to two major differences
between the approaches, namely the temporal difference of
simulated time (months to years for BES vs. hours calcu-
lated per time step in seconds for CFD) and the difference
in computation time. As a result, no transient behavior
across the individual engines, such as thermal mass being
in an energy exchange relation through convective heat
transfer, is taken into account by this approach. It follows
that the more transient behavior the system exhibits, the
less accurate the presented model tends to be.
Second, the RANS CFD methodology we use in this study
is known to have flaws when it comes to predicting turbu-
lent behavior in wake regions of buildings. Further, the
amount of wind buffeting that can be modeled with more
advanced methods like LES cannot be estimated due to the
averaging nature of the RANS model. At the same time, the
sensitivity of the wind velocity for arbitrary UTCI values
increases with decreasing ambient temperature. In other
words, the colder the ambient temperature, the more dom-
inant the wind chill effect due to convective heat transfer
________________________________________________________________________________________________
________________________________________________________________________________________________
Proceedings of the 16th IBPSA Conference
Rome, Italy, Sept. 2-4, 2019
625
a) 14:00 h
b) 16:00 h
c) Annual hours within thermal comfort
Figure 7: UTCI and comfort hour maps. a-b) Shows the
simulated UTCI distribution of the campus on March 27,
2019, for 14:00 h and 16:00 h. c) Shows the calculated per-
centage of annual thermal comfort hours with “no thermal
stress”, see table 1.
will be (Br
¨
ode et al.,2013). Consequently, when climates
with low ambient temperatures are studied, one should keep
in mind that such effects are not taken into account and
appropriate factors of safety should be considered. Mackey
et al. conducted extensive studies on the sensibility of
input variables for microclimate studies. They found that
the influence of the number of wind directions simulated is
almost negligible for a tropical climate such as that in Sin-
gapore. While this seems plausible for a tropical climate,
it is worth reiterating that microclimate maps will likely
suffer greatly from a reduced number of simulated wind
directions in colder climates (Br
¨
ode et al.,2013). Although
additional wind directions increase the simulation time for
the engine that is already the bottleneck in simulation pro-
cedure, the RANS method is state-of-the-art in terms of an
accuracy/efficiency/robustness-trade-off for outdoor com-
fort mappings, especially when additional measures are
taken into account (Kastner and Dogan,2018). However,
one area for further research might be to replace the RANS
method with an engine that is based on Lattice Boltzmann
methods which can be solved less expensively.
The comparison between the CFD simulation and the one
measurement location shows a fairly good agreement. We,
however, concede, that a single reference point in the up-
wind region of a building is not sufficient for a holistic
validation of the presented method. Further, we probed
and average from a larger area in front of the building the
deviation between both increases from
10 %
to
22 %
. This
clearly shows how highly sensitive the measurement lo-
cation is, which in turn shows the necessity for a higher
number of measurements for future studies both in terms of
the number of measurement locations themselves but also
the length and number of measurements for each location
over time.
Overall accuracy vs. level of detail necessary during
design process
There are unanswered questions with respect to micro-
climate maps that are created during the design process,
compared to maps that analyze an already existing urban
environment. Those questions usually revolve around a
trade-off between accuracy and computational efficiency.
First, the method to estimate the MRT in a reasonable
time without relying on the knowledge of all materials and
their corresponding emissivities is not yet agreed upon.
The overall goal for a tool such as the one presented is
computational efficiency as it ought to be used during the
design process. Thus, one might argue that it is sufficient
if only the thermal stress categories are correctly predicted
for a sufficiently high percentage of hours throughout the
year as those ultimately constitute the optimization goal
for the urban area, see table 1. To claim such accuracy,
validation studies for more hours in a year will have to be
carried out.
Second, how and when should evaporative cooling be
considered without introducing a host of input uncertain-
ties with respect to e.g. plants, rainfall, soil composition,
and pavement area? Integrated simulation tools such as
ENVI-met are capable of integrating evaporative cooling
by means of internal coupling. Such modeling approaches
require detailed knowledge about the additional simulation
input, namely the rainfall frequency, the type of plants and
trees to be used, their porosity, and the soil composition
(Manickathan et al.,2018). Unfortunately, in commonly
used EPW weather files there is rarely any rain data avail-
________________________________________________________________________________________________
________________________________________________________________________________________________
Proceedings of the 16th IBPSA Conference
Rome, Italy, Sept. 2-4, 2019
626
able, and if there is, its validity is questionable with respect
to a “typical” amount of rainfall per year. In case of a lack
of high-quality input data, the majority of inputs would
thus have to be substituted by default values which could
be counterproductive for the desired information gain. Re-
garding the external coupling used in our approach, it is
also questionable if incorporating vegetation is feasible
as it would pose great challenges on the data exchange
necessary due to shading by trees and the benefits from
evaporative cooling — the former of which has been shown
to be more beneficial by Manickathan et al..
Aside from that, Mackey et al. found that the sky heat
exchange is the most dominant variable in outdoor comfort
modeling, followed by wind patterns, and UHI/surface
temperature, producing errors that range from 2.4-0.5
◦
C
respectively. Although EP’s simulation time currently does
not constitute the bottleneck, one might argue in favor of
simplifying the MRT calculations by using an admittance
method or an even simpler linear regression going forward
(M.G,1994;Kessling et al.,2013).
We neglected buoyancy on the CFD side not only to gain
simplicity in selecting the boundary conditions but also
because it is less computationally expensive. As shown in
the validation study, this could be the reason for a large
discrepancy between the model and the measurements. Al-
legrini and Carmeliet showed a correlation between the
local air temperature, the volumetric flow rates in prede-
fined control volumes, and the local thermal diffusivity
in those volumes. In future studies of urban areas, this
approach could be used to predict the local heat island risk
of urban environments while utilizing an isothermal CFD
approach. Going forward, we hope to be able to define a
threshold for which the predicted air temperature from that
method would be used as an input for the UTCI.
Summary and conclusions
The authors implemented an outdoor comfort model-
ing framework for urban design in C# as a plugin for
Rhinoceros and Grasshopper that enables seamless work-
flow integration. UTCI simulation results around a library
building on a university campus are available and seem
plausible. The results exemplify the possibility to conduct
such analyses within reasonable time and accuracy if some
simplifications to the UTCI estimation are acceptable. In
the future, we plan to increase the number of measurement
locations to be able to present a more holistic validation of
our simulation results.
Acknowledgements
The authors would like to thank the David R. Atkinson
Center for a Sustainable Future and The Center for Trans-
portation, Environment, and Community Health (CTECH)
for funding this research.
References
Allegrini, J. and J. Carmeliet (2017). Coupled cfd and
building energy simulations for studying the impacts of
building height topology and buoyancy on local urban
microclimates. Urban Climate 21, 278–305.
Barbason, M. and S. Reiter (2014). Coupling building
energy simulation and computational fluid dynamics:
Application to a two-storey house in a temperate climate.
Building and Environment 75, 30–39.
Boulard, T., J. Meneses, M. Mermier, and G. Papadakis
(1996). The mechanisms involved in the natural ventila-
tion of greenhouses. Agricultural and Forest Meteorol-
ogy 79(1-2), 61–77.
Br
¨
ode, P., K. Błazejczyk, D. Fiala, G. Havenith, I. Holm
´
er,
G. Jendritzky, K. Kuklane, and B. Kampmann (2013,
feb). The Universal Thermal Climate Index UTCI com-
pared to ergonomics standards for assessing the thermal
environment. Industrial health 51(1), 16–24.
Leibniz Research Centre for Working Environment and
Human Factors (2009). The Universal Thermal Climate
Index.
Cornell University (2016). Options for Achieving a Carbon
Neutral Campus by 2035.
De Villiers, E. (2006). The potential of large eddy sim-
ulation for the modeling of wall bounded flows. Ph.
D. thesis, Imperial College of Science, Technology and
Medicine.
Ebrahimabadi, S. (2015). Outdoor Comfort in Cold Cli-
mates: Integrating Microclimate Factors in Urban De-
sign. Ph. D. thesis, Lule ˚
a tekniska universitet.
Fiala, D., G. Havenith, P. Br
¨
ode, B. Kampmann, and G. Jen-
dritzky (2012). UTCI-Fiala multi-node model of human
heat transfer and temperature regulation. International
Journal of Biometeorology 56(3), 429–441.
Hardy, B. (1998). Its-90 formulations for vapor pressure,
frostpoint temperature, dewpoint temperature, and en-
hancement factors in the range–100 to+ 100 c. In Pro-
ceedings of the third international symposium on humid-
ity and moisture, Teddington, London, England.
Huttner, S. and M. Bruse (2009). Numerical modeling of
the urban climate–a preview on envi-met 4.0. In 7th
International Conference on Urban Climate ICUC-7,
Yokohama, Japan, Volume 29.
Kastner, P. and T. Dogan (2018). Streamlining meshing
methodologies for annual urban CFD simulations. In
eSim 2018.
Kessling, W., M. Engelhardt, and D. Kiehlmann (2013).
The human bio-meteorological chart. In PLEA.
Lindberg, F., B. Holmer, and S. Thorsson (2008). Solweig
1.0–modelling spatial variations of 3d radiant fluxes and
mean radiant temperature in complex urban settings.
International journal of biometeorology 52(7), 697–713.
________________________________________________________________________________________________
________________________________________________________________________________________________
Proceedings of the 16th IBPSA Conference
Rome, Italy, Sept. 2-4, 2019
627
Mackey, C., T. Galanos, L. Norford, M. S. Roudsari, and
N. S. Bhd (2017). Wind , Sun , Surface Temperature ,
and Heat Island : Critical Variables for High-Resolution
Outdoor Thermal Comfort. In Proceedings of the 15th
IBPSA Conference San Francisco.
Magnusson, S., A. Dallman, D. Entekhabi, R. Britter, H. J.
Fernando, and L. Norford (2014). On thermally forced
flows in urban street canyons. Environmental Fluid
Mechanics 14(6), 1427–1441.
Manickathan, L., T. Defraeye, J. Allegrini, D. Derome, and
J. Carmeliet (2018). Parametric study of the influence
of environmental factors and tree properties on the tran-
spirative cooling effect of trees. Agricultural and Forest
Meteorology 248(October 2017), 259–274.
Matzarakis, A., F. Rutz, and H. Mayer (2010). Modelling
radiation fluxes in simple and complex environments:
basics of the rayman model. International journal of
biometeorology 54(2), 131–139.
M.G, D. (1994). The thermal response of an enclosure to
periodic excitation: The CIBSE approach. Building and
Environment 29(2), 217–235.
Mostapha Sadeghipour Roudsari, M. P. (2013). LADY-
BUG: A PARAMETRIC ENVIRONMENTAL PLU-
GIN FOR GRASSHOPPER TO HELP DESIGNERS
CREATE AN ENVIRONMENTALLY-CONSCIOUS
DESIGN. In 13th Conference of International Building
Performance Simulation Association, pp. 3128–3135.
Naboni, E., S. Coccolo, M. Meloni, and J.-L. Scartezzini
(2018). OUTDOOR COMFORT SIMULATION OF
COMPLEX ARCHITECTURAL DESIGNS A RE-
VIEW OF SIMULATION TOOLS FROM THE DE-
SIGNER PERSPECTIVE. In 2018 Building Perfor-
mance Analysis Conference and SimBuild, pp. 659–666.
NYC.gov (2019). The new york city carbon challenge
2019.
Odnevall Wallinder, I., J. Lu, S. Bertling, and C. Leygraf
(2002). On approaches to couple energy simulation and
computational fluid dynamics programs. Building and
Environment 37(8-9), 857–864.
Proven
c¸
al, S., O. Bergeron, R. Leduc, and N. Barrette
(2016). Thermal comfort in Quebec City, Canada: sen-
sitivity analysis of the UTCI and other popular thermal
comfort indices in a mid-latitude continental city. Inter-
national Journal of Biometeorology 60(4), 591–603.
Ramponi, R. and B. Blocken (2012). Cfd simulation of
cross-ventilation for a generic isolated building: impact
of computational parameters. Building and Environ-
ment 53, 34–48.
Reindl, D. and W. Beckman (1990). Diffuse fraction corre-
lations. Solar Energy 45(1), 1–7.
Tecle, A., G. T. Bitsuamlak, and T. E. Jiru (2013). Wind-
driven natural ventilation in a low-rise building: A
Boundary Layer Wind Tunnel study. Building and Envi-
ronment 59, 275–289.
Thorsson, S., F. Lindberg, J. Bj
¨
orklund, B. Holmer, and
D. Rayner (2011). Potential changes in outdoor thermal
comfort conditions in gothenburg, sweden due to climate
change: the influence of urban geometry. International
Journal of Climatology 31(2), 324–335.
Thorsson, S., F. Lindberg, I. Eliasson, and B. Holmer
(2007). Different methods for estimating the mean ra-
diant temperature in an outdoor urban setting. Interna-
tional journal of climatology 27(14), 1983–1993.
Wallace, J. M. and P. V. Hobbs (2006). Atmospheric Sci-
ence: An Introductory Survey (2 ed.). Academic Press.
Walter, E. and J. H. K
¨
ampf (2015). A verification of
CitySim results using the BESTEST and monitored con-
sumption values. In 2nd Building Simulation Applica-
tions conference, pp. 215–222.
Nomenclature
ABL Atmospheric boundary layer
BC Boundary condition
BES Building Energy Simulation
CFD Computational Fluid Dynamics
EPW EnergyPlus weather format
GIS Geographic information system
LES Large eddy simulation
MRT Mean-radiant temperature
RANS Reynolds-averaged Navier-Stokes
UHI Urban heat island, ◦C
UTCI Universal Thermal Climate Index , ◦C
SIMPLE Semi-Implicit Method for Pressure Linked Equations
SST Shear stress transport
kTurbulence kinetic energy, m2s−2
µt
µEddy viscosity ratio, -
Cmu Turbulence constant, 0.09
pvapour Vapour pressure, Nm−2
TaAmbient temperature, ◦C
TMRT Mean radiant temperature, ◦C
TuTurbulence intensity, %
UWind velocity, m s−1
u∗Friction velocity, m s−1
νtTurbulence eddy viscosity, m2s−1
y+Dimensionless wall distance, -
z0Aerodynamic roughness length, m
________________________________________________________________________________________________
________________________________________________________________________________________________
Proceedings of the 16th IBPSA Conference
Rome, Italy, Sept. 2-4, 2019
628
