impact of urban microclimate on summertime building ...building – – – – – – – – –...

Impact of urban microclimate on summertime building coolingdemandCitation for published version (APA):Toparlar, Y., Blocken, B., Maiheu, B., & van Heijst, G. J. F. (2018). Impact of urban microclimate on summertimebuilding cooling demand: a parametric analysis for Antwerp, Belgium. Applied Energy, 228, 852-872.https://doi.org/10.1016/j.apenergy.2018.06.110

DOI:10.1016/j.apenergy.2018.06.110

Document status and date:Published: 15/10/2018

Document Version:Publisher’s PDF, also known as Version of Record (includes final page, issue and volume numbers)

Please check the document version of this publication:

• A submitted manuscript is the version of the article upon submission and before peer-review. There can beimportant differences between the submitted version and the official published version of record. Peopleinterested in the research are advised to contact the author for the final version of the publication, or visit theDOI to the publisher's website.• The final author version and the galley proof are versions of the publication after peer review.• The final published version features the final layout of the paper including the volume, issue and pagenumbers.Link to publication

General rightsCopyright and moral rights for the publications made accessible in the public portal are retained by the authors and/or other copyright ownersand it is a condition of accessing publications that users recognise and abide by the legal requirements associated with these rights.

• Users may download and print one copy of any publication from the public portal for the purpose of private study or research. • You may not further distribute the material or use it for any profit-making activity or commercial gain • You may freely distribute the URL identifying the publication in the public portal.

If the publication is distributed under the terms of Article 25fa of the Dutch Copyright Act, indicated by the “Taverne” license above, pleasefollow below link for the End User Agreement:www.tue.nl/taverne

Take down policyIf you believe that this document breaches copyright please contact us at:[email protected] details and we will investigate your claim.

Download date: 01. Apr. 2021

https://doi.org/10.1016/j.apenergy.2018.06.110https://doi.org/10.1016/j.apenergy.2018.06.110https://research.tue.nl/en/publications/impact-of-urban-microclimate-on-summertime-building-cooling-demand(9f1da7ff-86e7-421a-88a6-804881da07ae).html

Contents lists available at ScienceDirect

Applied Energy

journal homepage: www.elsevier.com/locate/apenergy

Impact of urban microclimate on summertime building cooling demand: Aparametric analysis for Antwerp, Belgium

Y. Toparlara,b,⁎, B. Blockena,c, B. Maiheub, G.J.F. van Heijstd

a Building Physics and Services, Department of the Built Environment, Eindhoven University of Technology, P.O. Box 513, 5600 MB Eindhoven, The Netherlandsb Environmental Modeling, Flemish Institute for Technological Research, Boeretang 2400 Mol, Belgiumc Building Physics Section, Department of Civil Engineering, KU Leuven, Bus 2447, 3001 Leuven, Belgiumd Fluid Dynamics Laboratory, Department of Applied Physics, Eindhoven University of Technology, P.O. Box 513, 5600 MB Eindhoven, the Netherlands

H I G H L I G H T S

• One rural and two urban microclimatic conditions are specified with CFD simulations.• The urban locations are positioned 80m and 400m away from a park.• BES are performed considering buildings with varying characteristics and use types.• The effect of urban microclimate is studied based on monthly Cooling Demand (CD)• Buildings near the park have 13.9% less CD than the buildings away from the park.

A R T I C L E I N F O

Keywords:Computational Fluid Dynamics (CFD)Building Energy Simulations (BES)Urban heat island effectUrban parkBuilding characteristicsBuilding type

A B S T R A C T

Meteorological measurements are conducted in Antwerp, Belgium in July 2013, followed by CFD urban mi-croclimate simulations considering the same city and time period. The simulations are found to be able toreproduce measured air temperatures inside central Antwerp with an average absolute difference of 0.88 °C. Thesimulation results supplemented with measurements are used to generate location-specific MicroclimaticConditions (MCs) in three locations: (1) a rural location outside Antwerp; (2) an urban location inside Antwerp,away from an urban park; and (3) another urban location, close to the same park. Building Energy Simulations(BES) are performed for 36 cases based on three different MCs, two building use types and six sets of constructioncharacteristics, ranging from pre-1946 buildings to new, low-energy buildings. Monthly Cooling Demands (CDs)are extracted for each case and compared with each other. The results demonstrate that compared to the airtemperatures in the rural area, on average, air temperatures at the urban sites away and close to the park are3.3 °C and 2.4 °C higher, respectively. This leads to an additional monthly CD of up to 90%. CDs of buildings withbetter thermal insulation and lower infiltration rates can increase by 48% once moved from the rural location toan urban location, which may lead to the reconsideration of design guidelines of low-energy buildings exposed toan urban MC. Although the proximity of an urban park cannot fully compensate the increased CD by an urbanMC, residential buildings close to the park are found to have on average 13.9% less CD during July 2013,compared with buildings away from the same park. The influence of the urban park on the CDs of buildings in itsvicinity is strongly linked to the meteorological wind direction. Professionals focusing on energy-efficientbuildings in cities are advised to conduct energy predictions with location-specific MC data, instead of only usingcity-averaged meteorological data.

1. Introduction

According to the European Commission and the United States (US)Energy Information Administration, buildings are responsible for ap-proximately 40% of the total energy demand in the European Union

(EU) and the US [1,2]. Among the energy used in buildings within theEU, space heating has the largest share, but with the new buildingregulations that demand well-insulated building envelopes, the share ofspace heating is expected to decrease in the future [3,4]. In contrast,space cooling has a lower share among the total building energy

https://doi.org/10.1016/j.apenergy.2018.06.110Received 3 December 2017; Received in revised form 5 April 2018; Accepted 21 June 2018

⁎ Corresponding author at: Building Physics and Services, Department of the Built Environment, Eindhoven University of Technology, P.O. Box 513, 5600 MB Eindhoven, TheNetherlands.

E-mail address: [email protected] (Y. Toparlar).

Applied Energy 228 (2018) 852–872

0306-2619/ © 2018 Elsevier Ltd. All rights reserved.

T

http://www.sciencedirect.com/science/journal/03062619https://www.elsevier.com/locate/apenergyhttps://doi.org/10.1016/j.apenergy.2018.06.110https://doi.org/10.1016/j.apenergy.2018.06.110mailto:[email protected]://doi.org/10.1016/j.apenergy.2018.06.110http://crossmark.crossref.org/dialog/?doi=10.1016/j.apenergy.2018.06.110&domain=pdf

demand but due to climate change [5–7], heat waves [8–11] and rapidurbanization [12,13], the share of space cooling is expected to rise[4,14]. Building energy demand can be affected by different aspects ondifferent scales, such as occupant behavior [15,16], building installa-tions [17,18], building envelopes [19–21] and urban microclimate[22–24].

Urban microclimate can be defined as the local climate observed inurban areas, which can be significantly different from the climate ofsurrounding rural areas [25]. With the rapid urbanization of the worldpopulation [26,27], research on urban microclimate has gained popu-larity in the past years [22,28,29]. From the building energy perspec-tive, urban microclimate can be considered as the set of meteorologicalconditions to which buildings in cities are subjected. The effect of urbanmicroclimate on building energy demand is mostly researched withinthe theme “Urban Heat Island (UHI) effect” [30–33]. Studies on thetopic typically follow two steps: (1) Analysis of urban microclimatewith specific target microclimate parameter(s) (mostly air tempera-ture); (2) and then calculating building energy demand, mostly withBuilding Energy Simulations (BES) [34,35] while using target micro-climatic parameter(s) as an input.

For the urban microclimate part, the majority of the studies in theliterature collected measurement data from different parts of the sameurban area [12,30,31,33–42]. The collected data were subsequentlyused to define a series of location-specific Microclimatic Conditions(MCs). Measurement studies typically focus only on air temperature asthe target microclimate parameter, which can be considered reasonableas several statistical studies, such as Sailor and Munoz [43] and Funget al. [44], demonstrated that deviations in air temperature are re-sponsible for almost all the deviations in building energy demand.

Urban microclimate can be investigated also with computationalapproaches. The main advantage of computational approaches overmeasurements is the ability to generate spatially explicit informationfor the target microclimatic parameters [45,46]. In addition, a provenand appropriate computational methodology for urban microclimateanalysis can be used to investigate different urban design scenarios,which would be very challenging to investigate with measurementcampaigns [47–49]. In Table 1, an overview of studies investigating theeffect of urban microclimate on building energy demand with a com-putational approach on urban microclimate analysis is provided. Thefollowing entries are used: authors and publication year; computationalmethodology; microclimate cases; target parameter(s); building typesinvestigated.

Some studies in Table 1 used data morphing techniques, which re-fers to the addition of pre-measured UHI intensities (°C) (UHI scenarios)to the measured rural air temperatures [50,51]. Several studies usedUHI predictor tools such as the Urban Weather Generator [52], whichpredicts the UHI intensity (°C) based on location-specific morphologicalparameters (i.e. canyon aspect ratio) [53–57]. Kolokotroni et al. [58]used an artificial neural network model to predict urban microclimaticparameters. Another methodology for urban microclimate analysis isEnergy Balance Models (EBM) [59], which are employed in severalstudies [60–64]. Some studies performed Computational Fluid Dy-namics (CFD) simulations to analyze urban microclimate [65–67]whereas others employed a coupled CFD-EBM approach [68–71]. CFDanalysis of urban microclimate is a developing research field [29,46,49]and the coupling of velocity and temperature fields with a high spatialresolution is an advantage over other computational approaches [45].However, CFD simulations can be computationally demanding[45,46,68].

In Table 1, the table entry “(micro)climate cases” lists the different(micro)climatic conditions each study has evaluated, such as urban vs.rural climates. The entry “target parameters” lists the calculated mi-croclimatic parameters that are used as input for calculating buildingenergy demand. The entry “Building types” denotes whether differentbuilding types were considered within the same study. Compared withprevious studies, the present study provides two clear distinctions.

The first distinction concerns the compared microclimate cases. Tothe best of our knowledge, no prior study on the topic considered theimpact of meteorological wind direction on the energy demand of dif-ferent buildings in the same urban area, which can be influential[34,72]. The measurement study by Ca et al. [34] demonstrated that inthe presence of a local cooling source (e.g. an urban park), buildings inthe same urban area can have varying energy demands, depending ontheir locations with respect to the prevailing wind direction. To in-vestigate this, in addition to a typical rural vs urban microclimatecomparison, the present study considers an urban-urban comparisonwhere one of the target buildings is chosen close to an urban park andthe other is chosen away from the same urban park.

The second distinction concerns the uncertainties related to thebuilding under study. The hypothesis of this paper is that the effect ofurban microclimate on building energy demand can vary significantlyfor buildings with different construction characteristics (e.g. U-values ofthe construction components) and with different use types (e.g. re-sidential vs office buildings). Even though some previous studies fo-cused on similar considerations [35,54,55,67,71,73,74], the aim of thispaper is to demonstrate this complexity by focusing on a wider group ofbuildings ranging from pre-1946 buildings to modern low-energybuildings and to challenge the common considerations on energy effi-cient building design.

In this study, CFD urban microclimate simulations are performed forthe Antwerp central region and the resulting air temperatures arecompared with measurement data obtained during July 2013.Measurement data and CFD simulation results for air temperature (°C),wind speed (m/s) and wind direction (°) are extracted at three loca-tions: (1) a rural area outside of Antwerp; (2) an urban area insidecentral Antwerp, away from an urban park and (3) another urban areainside central Antwerp, close to the same urban park. Based on the CFDsimulation results, three location-specific MCs are defined. The re-sulting MCs are used as input for the building energy simulations (BES)of a building with the same form and orientation but with different usetypes and with different construction characteristics. The resultingbuilding cooling demands are reported for every individual case andcompared with each other.

2. Description of the study area, buildings and measurementcampaign

This study focuses on Antwerp, a city located in the North ofBelgium (Fig. 1a). The area of interest in this study is the central An-twerp area (Fig. 1b), specifically the area surrounding the urban parknamed “Stadspark” (Fig. 1c). Municipal drawings and the GIS databasecorresponding to the area of interest are acquired from the city ofAntwerp GIS service. The data contains the locations of all the buildingsand the trees taller than 2m. 365 buildings are specified in the area ofinterest and in Fig. 2a, the height distribution of these buildings isprovided. The highest building is 60 m and the buildings within the9–11m height interval are the most common (75/365, or 21%). Theaverage building height in the area of interest is 13m. The Entranzedatabase [75] reports the distribution of the buildings in Belgium withrespect to their construction dates. This distribution is demonstrated inFig. 2b, which shows that the majority of the buildings in Belgium areconstructed prior to 1970 (62% of all buildings). Note that the data onFig. 2b do not pertain to the study area itself but to all the buildings inBelgium.

Meteorological measurement data used in this study is collected bythe Flemish Institute for Technological Research (VITO) [76]. Themeasurements were conducted by stations at two locations: (1) On therooftop of a high-school building in central Antwerp (Fig. 1c); and (2) ina rural area located 8 km away from the central Antwerp area (Fig. 1b).The urban measurement station was positioned 2m above the ap-proximately 5m high building rooftop (approximately at 7m heightfrom the ground level) and the rural measurement station was

Y. Toparlar et al. Applied Energy 228 (2018) 852–872

853

Table1

Anov

erview

ofstud

iesinve

stigatingtheeff

ectof

simulated

urba

nmicroclim

ateon

build

ingen

ergy

deman

dan

dthepo

sition

ingof

thepresen

tstud

y.

#Autho

rs(year)

[Ref.]

Metho

dology

(microclim

ate)

(micro)clim

atecases

Target

parameter(s)

Build

ingtype

s

1de

laFlor

andDom

ingu

ez(200

4)[60]

EBM

Differen

tcities

inSp

ain

AT,

SR,S

TSa

meforallthecases

2Kikeg

awaet

al.(20

06)[61]

EBM

Differen

turba

nmorph

olog

ieswithch

anging

SVFs;

differen

talbe

dova

lues

onou

tdoo

rsurfaces

AT

3differen

tco

nstruc

tion

characteristics

3Heet

al.(20

09)[62]

EBM

Differen

turba

nde

sign

caseswithch

anging

build

ingco

verage

ratio

SR,S

TSa

meforallthecases

4Kolok

otroni

etal.(20

10)[58]

ANN

Urban

vs.rural;differen

tregion

sin

thesameurba

narea

AT

Sameforallthecases

5Bo

uyer

etal.(20

11)[68]

CFD

-EB

MSimulations

only

withEB

Mvs.w

ithco

upledCFD

-EBM

approa

chAT,

CHTC

,MRM,S

TSa

meforallthecases

6Won

get

al.(20

11)[53]

UHIpred

ictor

Differen

turba

nmorph

olog

ieswithch

anging

greenarea

ratio,

build

inghe

ight

variationan

dbu

ilding

density

AT

Sameforallthecases

7Yan

get

al.(20

12)[65]

CFD

Differen

tve

getation

scen

arios;

differen

tbu

ildingpo

sition

ing(stand

-alone

vs.inage

nericurba

nsetting)

AT,

MRM,S

T,WS

Sameforallthecases

8Yag

hoob

ianan

dKleissl

(201

2)[63]

EBM

Differen

tou

tdoo

rsurfacematerials

inage

nericurba

nsetting

AT,

DPT

,SR,W

SSa

meforallthecases

9Rad

hian

dSh

arples

(201

3)[51]

UHIscen

arios

Urban

vs.rural;

differen

tregion

sin

thesameurba

narea

AT

Sameforallthecases

10Groset

al.(20

14)[64]

EBM

Base

case

vscase

withdifferen

tsurfacereflectivity

(sho

rt-w

ave)

foron

ebu

ilding

AT,

CHTC

,MRM,S

TSa

meforallthecases

11Rod

eet

al.(20

14)[55]

UHIpred

ictor

Differen

turba

nmorph

olog

ytype

sof

differen

tcities;d

ifferen

tbu

ildingform

san

dfaçade

insulation

SR5bu

ildingform

s;2differen

tco

nstruc

tion

characteristics

12Su

nan

dAug

enbroe

(201

4)[54]

UHIpred

ictor

Differen

tcities

inUSA

;Rep

resentativeclim

ates

from

ruralareas,

subu

rban

areasan

dcity

centers

AT

3differen

tco

nstruc

tion

characteristics

13Liuet

al.(20

15)[66]

CFD

Differen

tclim

ates

withincrease

inATby

1%,5

%an

d10

%;

Differen

tbu

ildingpo

sition

ing(stand

-alone

vs.inage

nericurba

nsetting)

AT,

WS

Sameforallthecases

14Liao

etal.(20

15)[50]

UHIscen

arios

Differen

tregion

sin

thesameurba

narea

AT,

RH

Sameforallthecases

15Sk

elho

rnet

al.(20

16)[67]

CFD

Stan

d-alon

ebu

ildingan

dva

ryingco

nditions

(base,

base

withshad

ing,

base

anurba

nsetting,

base

inan

urba

nsettingwithsurrou

ndingtrees)

AT,

SR,W

S3differen

tbu

ildingform

s

16Morilleet

al.(20

16)[71]

CFD

-EB

MBa

secase

(abu

ildingin

astreet

cany

on);Differen

tve

getation

scen

arios(green

wall,greenroof,street

trees)

AT,

CHTC

,ST,

WS

2differen

tco

nstruc

tion

characteristics

17Presen

tstud

yCFD

Urban

vs.rural;d

ifferen

tregion

sin

thesameurba

narea

(close

toan

urba

npa

rkan

daw

ayfrom

the

sameurba

npa

rk)

AT,

WD,W

S6differen

tco

nstruc

tion

characteristics;

2bu

ilding

usetype

s

Abb

reviations:A

NN:A

rtificial

neural

netw

ork;

AT:

Airtempe

rature

(°C);CFD

:Com

putation

alfluiddy

namics;CHTC

:Con

vectivehe

attran

sfer

coeffi

cien

t(W/m

2K);DPT

:Dew

-point

tempe

rature

(°C);EB

M:E

nergyba

lanc

emod

el;H

FB:H

eatfl

uxba

lanc

e;MRM:M

assrate

ofmoisture(m

3/s)RH:R

elativehu

midity(%

);SR

:Solar

radiation(W

/m2);ST

:Surface

tempe

rature

(°C);SV

F:Sk

yview

factor;U

HI:Urban

heat

island

;WD:W

inddirection

(°);WS:

Windspeed(m

/s)


854

Fig. 1. (a) Location of Belgium in Europe and the location of Antwerp and Brussels (capital region) in Belgium (map source: freevectormaps.com); (b) location of thecentral Antwerp area and the rural measurement station; (c) view of the central Antwerp area with the Stadspark, the urban measurement station and the two urbanlocations under study; (d) view from the Breughelstraat where urban point #1 is located (source: Google Maps); (e) view from the Quellinstraat where urban point #2is located (source: Google Maps).


855

http://freevectormaps.com

positioned at 2m above ground level. 15-min averaged meteorologicaldata for air temperature (°C), relative humidity (%), solar radiation (W/m2), wind speed (m/s) and wind direction (°) were collected from June2012 until September 2013. In this study, the measurement data used isfrom the month of July 2013 as the data for this month is acquiredwithout any intermittencies and can be considered as representative ofsummertime conditions in Antwerp.

3. CFD simulations: settings and parameters

To study the urban microclimate of the area of interest in Antwerp,CFD simulations are performed with the commercial CFD softwareANSYS Fluent 15.1 [77]. Computational settings and parameters usedin the simulations are described briefly as most of the settings are thesame as reported in Toparlar et al. [78].

3.1. Computational domain and grid

The computational domain is divided into 2 areas: (1) the areawhere buildings and trees are modeled explicitly, i.e. with their actualshape and size, and (2) the area surrounding the explicitly modeledobjects where the objects are modeled implicitly, i.e. by imposing theequivalent sand-grain roughness height associated with the properaerodynamic roughness length (z0) to the ground surface, in accordancewith the guidelines by Blocken [46]. The explicitly modeled area in-cludes the Stadspark and its surroundings, covering an area of2027×1972m2 (Fig. 3a). The z0 value is estimated based on the up-dated Davenport roughness classification [79] by investigating theroughness characteristics of the terrain surrounding the explicitlymodeled area [78]. The estimated z0 is 0.25m for East and Southeastdirections and 0.5 m for the other directions.

The dimensions of the computational domain are determined fol-lowing the CFD best practice guidelines by Franke et al. [80], Tominagaet al. [81] and Blocken [46]. The guidelines recommend domain di-mensions to be imposed based on the height of the highest building (H)(m) in the area of interest, which is 60m for this study (Fig. 3c). The 3Dcomputational domain generated has the dimensions of6500×6500×420m3, which satisfies the guidelines by Franke et al.[80] and Tominaga et al. [81]. The domain dimensioning satisfies alsothe conservative recommendations by Blocken [46] on the so-calleddirectional blockage ratio, i.e. the vertical blockage ratio and the lateralhorizontal blockage ratio. The grid is generated by following the surfacegrid extrusion technique introduced by van Hooff and Blocken [82].The horizontal grid resolution is decreased from the Stadspark towardsthe regions further away from the park (Fig. 3b and d). The height ofthe center of the ground wall adjacent cell (zp) (m) is 0.11m and 0.71mfor the explicitly and implicitly modeled areas, respectively. The re-sulting computational grid contains 9,078,916 hexahedral cells.

3.2. Boundary conditions

The four vertical outer faces of the domain are specified as flowboundaries (Fig. 3c) and depending on the hourly wind directionmeasured at the rural measurement station, two of them are modeled asinlets and the other two as outlets. The top boundary is modeled as afree-slip wall with zero normal gradients for all the variables. Streetsurfaces (streets below trees and open streets), building roofs and wallsand the ground boundary are specified as wall type boundaries.

At the inlets, the profiles of mean wind speed U(z) (m/s), turbulentkinetic energy k(z) (m2/s2) and turbulence dissipation rat e ε(z) (m2/s3)are imposed [83]:

⎜ ⎟= ⎛⎝

+ ⎞⎠

∗U y u

κy y

y( ) ln 0

0 (1)

=∗

k y uC

( )μ

2

(2)

=+

∗ε y u

κ y y( )

( )0

3

(3)

where u∗ (m/s) is the atmospheric boundary layer friction velocity, κ(–)(=0.41) is the von Karman constant, z (m) is the height coordinate,z0 (m) is the aerodynamic roughness length and Cµ (–)(=0.09) is amodel constant. Reference wind speed, wind direction and air tem-perature at the inlets are imposed based on the data from the ruralmeasurement station. The air temperature data is imposed at the inletswith a uniform profile. The outlets of the domain have zero static gaugepressure.

To resolve the fluid-wall interaction at the wall type boundaries, theStandard Wall Functions [84] are employed with the sand-grain basedroughness modification [85]. The z0 value of the street surfaces is0.03m [86] whereas the z0 value of the ground wall of the implicitlymodeled area can be either 0.25m or 0.5m, depending on the winddirection. To ensure a horizontally homogeneous approach flow,roughness parameters are specified by satisfying the following re-lationship between the roughness height (kS) (m), roughness constant(CS) (–) and z0 [87]: kS= 9.793 z0/CS. Based on this equation, for the z0values of 0.5 m, 0.25m and 0.03m, the values of kS are 0.7 m, 0.7 mand 0.1m, respectively, whereas the values of CS are 7, 3.5 and 2.9,respectively.

The wall type boundaries can be categorized in four groups: (1)ground boundary for streets without trees; (2) ground boundary forstreets with trees; (3) ground boundary in the implicitly modeled regionand (4) building roofs and walls. The entire ground boundary is mod-eled as a layer with 10m thickness and with zero heat flux at 10mdepth. The buildings are modeled as air-conditioned spaces during

Fig. 2. (a) Distribution of the building heights in the area of interest and (b) Distribution of the buildings in Belgium with respect to construction date.Source: http://www.entranze.enerdata.eu/.


856

http://www.entranze.enerdata.eu/

summertime with a constant 24 °C interior temperature. The materialsand components used for the wall type boundaries are specified inTable 2. Building walls and roofs have the construction characteristicsof a 1971–1990 building as defined in the “Typology Approach forBuilding Stock Energy Assessment” (TABULA) report [88], whichdocuments the construction characteristics of the Belgian buildingstock. The value of the short-wave absorptivity of the street surfacesbelow the trees is reduced to represent shaded areas as described in thevegetation model presented by Toparlar et al. [78].

3.3. Other computational settings

The 3D Unsteady Reynolds-Averaged Navier-Stokes (URANS)equations are solved in combination with the realizable k-ε turbulencemodel [89] for closure. The time dependent sun direction vector andthe diffuse fraction of the global horizontal radiation is calculated with

the Solar Calculator of ANSYS Fluent [77], which uses the Solar Posi-tion and Intensity Code (Solpos) of the National Renewable EnergyLaboratory (NREL) [90]. Radiation is handled with the P-1 radiationmodel [77] and natural convection is modeled with the Boussinesqapproximation.

The trees within the area of interest are modeled as volumetricporous zones from 3m to 9m high and for the computational cells ofthese volumetric zones, source/sink terms are added for momentum(Eq. (4) from Green [91]), turbulent kinetic energy (Eq. (5) from Liuet al. [92]), turbulence dissipation rate (Eq. (6) from Sanz [93]) andheat transfer (Eq. (7) from Huang [94] and Toparlar [78]) as:

= −S ρC LADU UU d ii (4)

= −S ρC LAD U Uk(β β )k d p d3 (5)

Fig. 3. (a) Top view of the computational domain; (b) top-view of the grid near the Stadspark (number of cells: 9,078,916) and top view of the corresponding region(source: bing.com/maps); (c) section view (A-A) of the computational domain denoting the height of the highest building (H) and the height of the domain; d)Isometric view of the grid on the surfaces of the buildings, trees and streets in the explicitly modeled area (view from South) (number of cells: 9,078,916).


857

http://bing.com/maps

= −S ρC LAD εk

U β Uk(C β C )ε d ε p ε d4 3 5 (6)

=P ETPρλ ADc L (7)

where ρ (kg/m3) is the density of air, LAD (m2/m3) is the Leaf AreaDensity, Cd (–) is the leaf drag coefficient, Ui is the velocity componentin direction i and βp, βd, Cε4 and Cε5 are model coefficients with thevalues 1.0, 5.1, 0.9 and 0.9, respectively. Details of these model coef-ficients are described in the study by Gromke et al. [95]. In Eq. (7), thevolumetric cooling power associated with the three crowns is specified.In this equation, ETP (m/h) denotes the potential evapotranspirationand λ (Wh/kg) is the latent heat of vaporization. The performance ofEq. (7) in predicting the cooling power from trees was evaluated with aset of experimental data in the study by Toparlar et al. [78].

Hourly data for air temperature, wind speed and wind directionimposed at the inlets are obtained from the rural measurement stationwhereas the data for solar radiation area taken from the urban mea-surement station. The hourly rural measurement data also reported thehourly standard deviations in the wind direction readings for each re-spective hour. The average standard deviation of all the individualhourly wind direction readings is 32° and it can be as high as 105°. Thisindicates that on average, the hourly meteorological wind direction canfluctuate within a 32° interval. To take into account this uncertainty,the hourly wind direction imposed at the inlets is averaged over 30°intervals, ranging from 0° (North) to 330° (Northnorthwest).

The SIMPLEC algorithm [96] is employed for pressure-velocitycoupling and second-order discretization schemes are used for theconvective, viscous and temporal terms. The simulations are performedwith hourly time steps and for each time step, 600 iterations are per-formed, based on a sensitivity analysis. The conditions during 30 June2013 are imposed as the first day and then simulations are performedfor the entire month of July 2013. This includes 768 time steps, solvedwith 460,800 iterations. At the end of each time step, the minimumvalues for the scaled residuals are 10−5 for velocity components, 10−4

for k, ε and continuity and 10−7 for energy and radiation.The hourly simulation results for air temperature (°C) are extracted

in the point that corresponds to the position of the urban measurementstation. In addition, the hourly simulation results for air temperature(°C), wind speed (m/s) and wind direction (°) are extracted in twourban points (Fig. 1c): (1) in the street named Breughelstraat (urbanpoint #1), located 400m south of the Stadspark and (2) in the streetnamed Quellinstraat (urban point #2), located 80m north of theStadspark. The urban points of interest have different proximities to theStadspark to evaluate the impact of the urban park on the micro-climatological conditions and consequently on the building energydemand. Both points are located at 2m above the ground level, which isthe recommended height by the World Meteorological Organization(WMO) to acquire microclimatological data in urban areas [97]. Post-processing of the location-specific microclimatological data is explainedin Section 4.3.

4. Building energy simulations: settings and parameters

BES are performed using the EnergyPlus 8.6 software [98] devel-oped by the US Department of Energy. Simulations with this softwarewere validated extensively in the past [99] and the software has beenused in various studies [11,100]. In total, 36 BES cases are generatedbased on six building construction characteristics, three MCs and twobuilding use types.

4.1. Building form and orientation

The simulation cases consider the same building form and orienta-tion to remove these parameters from the comparative analysis. Thebuilding height interval of 9–11m, which is the most common amongthe buildings in Belgium, corresponds to a typical terraced house [88]Ta

ble2

Thespecification

sof

materials

andco

mpo

nentsused

fortheCFD

simulations

inthis

stud

y.

Materials

Thermal

cond

uctivity

(W/m

.K)

Den

sity

(kg/

m3)

Specifiche

at(J/k

g.K)

Absorptivity(–)

(sho

rt-w

ave)

Emissivity

(–)

(lon

g-wav

e)

Earth(non

-sha

dedareas)

1.3

1400

1000

0.75

0.93

Earth(sha

dedareas)

1.3

1400

1000

0.15

0.93

Brick

0.8

2050

900

0.78

0.91

Insulation

0.04

750

840

(not

asurfacematerial)

Con

crete

0.6

2300

880

0.73

0.88

Com

pone

nts

Laye

r1

Laye

r2

Laye

r3

Material

d(m

)Material

d(m

)Material

d(m

)

Groun

d(for

streetswitho

uttrees)

Con

crete

0.5

Earth

9.5

––

Groun

d(for

streetswithtrees)

Earth

10–

––

–Groun

d(implicitly

mod

eled

area)

Con

crete

0.5

Earth

9.5

––

Build

ingroofs/walls

(197

1–19

90bu

ilding)

[88]

Brick

0.18

Insulation

0.02

Brick

0.18


858

(Fig. 4a and b). The terraced building modeled in BES is shown inFig. 4c. The front façade of the building is parallel to the two streets ofinterest (Fig. 1c), i.e. 285° clockwise from north (Fig. 4e). The modeledterraced building is composed of three floors, each with a floor area of66m2 (198m2 in total). Front and rear facades of the building areidentical. Both of the facades and the roof are exposed to the ambientconditions whereas the two lateral sides of the building are consideredto be adiabatic.

4.2. Construction characteristics

The Belgian building stock is categorized in the TABULA report [88]based on individual construction characteristics as follows [88]: (1) pre-1946, (2) 1946–1970, (3) 1971–1990, (4) 1991–2005, (5) post-2005and (6) Low Energy (LE) buildings. Typical thermal transmittance va-lues or U-values (W/m2K) for the walls, roofs, floors, windows anddoors and typical ex/infiltration rates (m3/hm2) are specified for everygroup as shown in Table 3. The table demonstrates that newer con-structions have lower U-values, indicative of higher thermal insulationlevels compared to older constructions. In addition, newer construc-tions are more airtight as shown by the lower ex/infiltration rates.

The BES simulations consider buildings with different constructioncharacteristics by imposing the specific U-values and ex/infiltrationrates mentioned in Table 3. Note that the TABULA report specifies onlythe U-values and insulation thicknesses of components. Therefore, inthe present study, the thicknesses of other materials are determinedbased on the material specifications to match the corresponding U-

values. To compare the heat transfer characteristics of componentsamong different construction groups, the thermal time constants (τ) (h)are calculated (Table 3) using the following equations:

∑=R dλcomponent

material

material (8)

∑=C d c ρcomponent material material material (9)

=τR C

3600componentcomponent component

(10)

where Rcomponent (m2 K/W) is the thermal resistance of individual com-ponents and Ccomponent (J/m2 K) is the heat capacity per unit area ofindividual components [101]. In Eqs. (4)–(6), dmaterial (m), λmaterial (W/mK), ρmaterial (kg/m

3) denote the thickness, thermal conductivity anddensity of every individual material in a respective component. Sincethe material specifications of the building windows are not known, theyare modeled as so-called “no mass” components and only their thermalresistance values are specified in the BES (as shown in Table 3).Therefore, the thermal time constants of building windows are notcalculated.

4.3. Microclimatic conditions

One set of rural microclimatic data and two sets of urban micro-climatic data are generated by editing the EnergyPlus weather files(.epw). An annual climate file for Belgium is updated based on thegeographical conditions of Antwerp (e.g. latitude, longitude, elevation)

Fig. 4. (a) A terraced building in the Breughelstraat, where the urban point #1 is located; (b) a terraced building in the Quellinstraat, where the urban point #2 islocated; (c) the front façade of the terraced building modeled in BES and its dimensions; (d) an isometric view of the modeled building; and (e) top view of themodeled building with the orientation indicator.


859

Table3

Materialan

dco

mpo

nent

specification

sin

theBE

Ssimulations

andco

nstruc

tion

characteristicsof

build

ings

inBe

lgium.Informationon

construc

tion

characteristicsis

obtained

from

Ref.[

88].

Materials

Thermal

cond

uctivity

(W/m

K)

Den

sity

(kg/

m3)

Specifiche

at(J/k

gK)

Absorptivity(–)(sho

rt-w

ave)

Emissivity

(–)

(lon

g-wav

e)

Brick

0.8

2050

900

0.78

0.91

Insulation

0.04

750

840

(not

asurfacematerial)

Woo

d0.14

800

1300

0.82

0.94

Con

crete

0.6

2300

880

0.73

0.88

Material(n

omass)

Thermal

resistan

ce(R

-value

)(m

2K/W

)Finishing

0.03

Air

cavity

0.15

Com

pone

ntCon

structiongrou

pDetails

(from

exterior

tointerior)

U-value

(W/m

2K)

Thermal

timeco

nstant

(τ)(h)

Build

ingwalls

Pre-19

46Brick

2.2

83.0

1946

–197

0Brick/

aircavity/b

rick

1.7

83.0

1971

–199

0Brick/

aircavity/insulation(2cm

)/brick

1.0

161.7

1991

–200

5Brick/

aircavity/insulation(6

cm)/brick

0.6

234.6

Post-200

5Brick/

aircavity/insulation(8

cm)/brick

0.4

399.0

Low

energy

Brick/

aircavity/insulation(15cm

)/brick

0.25

678.2

Build

ingroof

Pre-19

46Woo

d/finishing

1.7

1319

46–1

970

Woo

d/finishing

1.9

1019

71–1

990

Woo

d/insulation

(4cm

)/finishing

0.85

12.7

1991

–200

5Woo

d/insulation

(8cm

)/finishing

0.5

24.8

Post-200

5Woo

d/insulation

(15cm

)/finishing

0.3

35.5

Low

energy

Woo

d/insulation

(30cm

)/finishing

0.15

80.3

Build

ingfloo

rPre-19

46Con

crete

0.85

459.1

1946

–197

0Con

crete

0.85

459.1

1971

–199

0Con

crete

0.85

459.1

1991

–200

5Con

crete/insulation

(4cm

)0.7

341.9

Post-200

5Con

crete//insulation(5

cm)

0.4

879.1

Low

energy

Con

crete/insulation

(10cm

)0.25

2137

.7

Build

ingwindo

ws

Pre-19

46Sing

leglazing

5.0

–19

46–1

970

Sing

leglazing

5.0

–19

71–1

990

Dou

bleglazing

3.5

–19

91–2

005

Dou

bleglazing

3.5

–Po

st-200

5Highpe

rforman

ceglazing

2.0

–Lo

wen

ergy

Highpe

rforman

ceprofi

lesan

dglazing

1.6

–

Build

ingdo

ors

Pre-19

46Woo

d4.0

2.6

1946

– 197

0Woo

d4.0

2.6

1971

–199

0Woo

d4.0

2.6

1991

–200

5Woo

d(thicker)

3.5

3.3

Post-200

5Woo

d/Insulation

2.8

2.1

Low

energy

Woo

d/Insulation

1.5

4.0

Con

structiongrou

pEx

/infi

ltration

rates(for

thewho

lebu

ilding

)

Ex/infi

ltration

rates(m

3/h

m2)

Pre-19

4614

.9m

3/h

m2

1946

–197

014

.9m

3/h

m2

1971

–199

014

.1m

3/h

m2

1991

–200

510

.0m

3/h

m2

Post-200

56.0m

3/h

m2

Low

energy

2.5m

3/h

m2


860

and then the climate file is edited for air temperature, wind speed, winddirection and solar radiation based on the input sources specified inTable 4. The editing is conducted only for the month of July 2013,which is the month considered in the CFD simulations.

The solar radiation input for the urban and rural cases is directlyimposed as the measured values from urban and rural measurementstations, respectively. However, shading on building facades can besignificantly different for buildings in urban and rural areas. Therefore,using the shading algorithm of EnergyPlus [102], additional buildingblocks are placed 15m away from the front and rear facades, only forthe buildings subjected to an urban MC. The imposed street width of15m is typical for the streets in Antwerp, even though in reality theQuellinstraat (17m wide) is slightly wider than the Breughelstraat(13m wide). The building blocks placed for the shading algorithm are13m high, which is the average building height in the area of interest.

4.4. Building use types

This study evaluates two building use types: (1) residential and (2)office buildings. Typically, on working days, residential buildings areoccupied during early mornings and evenings whereas office buildingsare occupied during the daytime. The difference in daily occupancyschedule is reflected in the internal gains by lighting and electricequipment. Occupancy/operation schedules and internal heat gainsmodeled in BES for both building types are summarized in Table 5. BESare performed by considering the workday/weekend schedules duringthe month of July 2013.

For the residential building, a family of four people is consideredand for the office building, three offices on separate floors are con-sidered. The occupancy of the offices is specified with an occupancyratio of 0.1 person/m2 [103] (net area) and in the case of the modeledbuilding, this yields 6 people on every floor (Table 5).

4.5. Other computational settings

Based on the recommendations of the EnergyPlus reference book[107], simulations are performed with 10 time steps per hour andground temperature just below the building floor is specified as 18 °C,which is 2 °C less than the average indoor temperature. The surfaceconvection for the interior and exterior surfaces is handled with theTARP and DOE-2 algorithms, respectively [108]. The aim of this studyis to calculate the cooling demand under ideal conditions, wherethermal set points are met at every hour independent of the systemcapability. Therefore the air-conditioning system is specified as theideal loads air system [102], which calculates the cooling demandunder idealized system conditions. The system provides sufficientcooling to each zone, and this approach is commonly used for demandcalculations [107].

Simulations are performed for the entire year of 2013, starting from

1 January 00:00 h. However, only the simulation results from July 2013are evaluated. From each BES case, the resulting data for the monthlyand daily cooling demand (kWh) are extracted. Results are reported interms of the cooling demand per unit area (kWh/m2) and in terms ofpercentage differences in cooling demand among different cases.

5. Results

5.1. CFD simulations: Comparison with the measurement data

In Fig. 5a, the resulting air temperatures from the CFD simulationsand the measurements from July 2013 are compared. Even though thediurnal trend is well reproduced, this repetition is largely governed bythe imposed measured air temperatures, which already have a similardiurnal trend. Fig. 5b demonstrates the hourly urban–rural air tem-perature differences based on the CFD simulation results and mea-surements. The figure shows that the CFD simulations generally un-derpredict the urban heat island intensity, which may be attributed toneglecting anthropogenic heat sources and simplifying surface char-acteristics (e.g. same albedo of all building surfaces).

Fig. 5c displays the hourly air temperature differences between theCFD simulations and the measurements at the urban measurementstation. The CFD simulations are able to reproduce the measured airtemperatures with an average absolute temperature difference of0.88 °C and a standard deviation of 0.61 °C. The maximum under-prediction by CFD is at 10 July, 12:00 h with −3.2 °C, and maximumoverprediction by CFD is at 7 July, 22:00 h with +2.4 °C. Given thefairly good agreement with the measurement data, the CFD output isconsidered suitable for generating the location-specific microclimaticdata for different parts in the central Antwerp area.

5.2. Location-specific microclimatic conditions

Location-specific MCs are generated for three locations: (1) the ruralarea, (2) the urban point of interest #1 (Breughelstraat) (urban#1) and(3) the urban point of interest #2 (Quellinstraat) (urban#2). Theseconditions are generated based on the CFD simulation results supple-mented with the meteorological measurements, as shown in Table 4.Fig. 6 shows the location-specific MCs with respect to air temperature(Fig. 6a), wind speed (Fig. 6b), wind direction (Fig. 6c) and solar ra-diation (Fig. 6d). Note that the urban MC also take into account theeffect of the park.

The air temperature data show that the warmest point of interest isurban#1 and the coolest point of interest is the rural location.Compared to the measurement data from the rural station, on average,urban#1 is 3.3 °C warmer whereas urban#2 is 2.4 °C warmer. Thedifference in air temperature between urban#1 and urban#2 is mostlydue to the close proximity of urban#2 to the Stadspark. Especially ondays when the wind direction is from southwast (between 180° - 270° in

Table 4Description of how different MCs are specified based on microclimate parameters and description of the shading imposed on BES cases in rural and urban areas.

Microclimatic condition Microclimate parameters and input sources Description

Air temperature (°C) Wind speed (m/s) Wind direction (°) Solar radiation (W/m2)

Rural Rural measurements Rural measurements Rural measurements Rural measurements Rural locationUrban #1 (Breughelstraat) CFD results

(urban point #1)CFD results(urban point #1)

CFD results(urban point #1)

Urban measurements Urban point of interest #1 (away fromthe park)

Urban #2(Quellinstraat)




Urban measurements Urban point of interest #2 (close to thepark)

Microclimatic condition Shading from surrounding buildings DescriptionRural Shading from other buildings is considered negligible –Urban #1 (Breughelstraat) Shading from other buildings is considered for front and rear facades Shading on both facades is modeled by positioning 13m high buildings

15m away from both facades.Urban #2 (Quellinstraat) Shading from other buildings is considered for front and rear facades Shading from both facades is modeled by positioning 13m high buildings

15m away from both facades.


861

Fig. 6c), the air temperature difference between the two urban points ofinterest increases up to a maximum of 5.3 °C. The maximum differenceoccurs on 22 July, where low wind speed leads to a concentratedcooling effect from the park [78]. In addition, since the modeling ap-proach used for the evapotranspirative cooling of the trees is dependenton air temperature and solar radiation, high air temperatures (> 30 °C)occurring on this particular day lead to an increased volumetric coolingpower from the trees.

Fig. 6b compares the CFD simulation results (urban #1 and urban#2) and the measurement data (rural) for wind speed. When normal-ized with respect to hourly averaged wind speeds at the rural area, thenormalized wind speed is 0.77 for the urban#1 and is 0.71 for theurban#2. This averaging indicates that urban#2 has relatively lowerwind speeds compared to urban#1. This might be because of the effectof tree foliage in the nearby park on wind speeds.

Fig. 6c compares the CFD simulation results (urban #1 and urban#2) and measurement data (rural) for wind direction. Most of the time,the urban wind direction is either southwest (180–270° interval) ornorthwest (0–90° interval). Typically, the wind direction at the urbanpoints is dominated by the orientation of their respective street can-yons.

Solar radiation data are based on the rural and urban measurementstations. In comparison, no significant difference is observed as theaverage differences in solar radiation measurements during whole Julyare around 11W/m2 (higher for the rural area).

5.3. BES: Basic sensitivity analysis

To understand the individual impact of each BES parameter on theJuly Cooling Demand (CD), first, a basic sensitivity analysis is con-ducted that does not yet focus on the differences induced by differentMCs. The base case selected for the sensitivity analysis is the building

with pre-1946 construction characteristics, subjected to the rural MCand with residential use. The CD is calculated for different cases whereBES parameters are changed one at a time as demonstrated in Table 6.

The impact of different U-values and different ex/infiltration rateson the CD is demonstrated in Fig. 7a and b, respectively. Fig. 7a showsthat compared to the base case, the added air cavity layer in 1946–1970constructions and the added thin insulation layer in 1971–1990 con-structions decrease the CD. However, the decrease in CD with de-creasing U-value is reversed once the building insulation is furtherimproved, as seen in the case of 1991–2005 constructions and onwards.Compared with the base case building, low-energy buildings can have15.1% more CD due to high insulation. This non-linear impact can beassociated with the changing transient heat transfer characteristics ofbuilding components, such as the thermal time constant. A similar non-linear relationship is observed in several prior studies [109,110], in-dicating that there might be a potential optimum for building CDconsidering the component U-values and thicknesses.

The relationship between ex/infiltration rates and CD is monotonic.Fig. 7b shows that the building CD increases with decreasing ex/in-filtration rates (Fig. 7b). This is because decreasing ex/infiltration rateslead to decreased heat losses due to ventilation during the summerperiod. Compared with the base case building, low energy buildings canhave 80.8% more CD due to lower ex/infiltration rates.

The results of the sensitivity analysis based on changing microcli-matic parameters are demonstrated in Fig. 8. The results show that thebuilding CD is highly dependent on changing air temperatures. Whenexposed to the air temperatures from urban#1 and urban#2, the CDincreases by 122.2% and 79%, respectively.

The impact of wind speed on the building CD is much less than theimpact of air temperatures. The maximum change in CD of 2.8% whenthe base case building is subjected to the wind speed conditions fromurban#2. The comparison of wind speed conditions indicate that slower

Table 5Occupancy/operation schedules and internal gains modeled in BES for residential and office buildings. Values are obtained from Refs. [102,104–106]

Building type Occupancy/operation schedules

Cooling system People Lighting equipment Equipment

Time interval(h)

Setpoint Time interval(h)

Presence of people Time interval(h)

Lights(on/off)

Time interval(h)

Equipment(ON or standby)

Residential 00:00–08:00 24 °C 00:00–08:00 Present 00:00–18:00 OFF 00:00–18:00 Standby08:00–18:00 27 °C 08:00–18:00 Not present 18:00–23:00 ON 18:00–23:00 ON18:00–24:00 24 °C 18:00–24:00 Present 23:00–24:00 OFF 23:00–24:00 Standby

Office 00:00–08:00 27 °C 00:00–08:00 Not present 00:00–08:00 OFF 00:00–08:00 Standby08:00–18:00 24 °C 08:00–18:00 Present 08:00–18:00 ON 08:00–18:00 ON18:00–24:00 27 °C 18:00–24:00 Not present 18:00–24:00 OFF 18:00–24:00 Standby

Building type Heat gains

Floor People Lighting C

(W/m2)Equipment D

(W/m2)Number of people A Heat gain per person (W) B

Residential Floor 1 2 80 (from 23:00 till 07:00)120 (from 07:00 till 23:00) 10 10 (ON)4 (Standby)

Floor 2 1 80 (from 23:00 till 07:00)120 (from 07:00 till 23:00)

6 5 (ON)2 (Standby)

Floor 3 1 80 (from 23:00 till 07:00)120 (from 07:00 till 23:00)

6 5 (ON)2 (Standby)

Office Floor 1 6 120 12 16 (ON)6.4 (Standby)

Floor 2 6 120 12 16 (ON)6.4 (Standby)

Floor 3 6 120 12 16 (ON)6.4 (Standby)

A: For the residential building, a family of 4 people is considered. For the office building, an occupancy ratio of 0.1 person/m2 (net area) is considered. B: Heat gainper person is less for sleeping hours. C: Heat gains from lighting are less for the upper floors of the residential building due to lower occupancy. D: Heat gains fromequipment are less for the upper floors of the residential building due to lower occupancy.


862

wind speeds would yield to a higher energy demand, which is relatedeither to the changing outer convective heat transfer coefficient or tothe impact of wind speed on the air infiltration.

Imposing the wind direction data obtained from the urban #1 andurban #2 MC dataset yields to a 0.2% increase in the CDs of buildingscompared with the base case where the wind direction data obtainedfrom the rural MC is imposed. The similarity of the findings forurban#1 and urban#2 is expected, as the hourly averaged wind di-rection vectors, which are mostly dictated by the street canyon or-ientation, are similar for both of the urban points of interest. It shouldbe noted that the exposure to outer environment is somewhat limited inthe terraced buildings as the two sides of the base case building aremodeled as adiabatic and only the two short sides of the building areexposed to ambient conditions. Therefore, the impact of varying windspeed and wind direction on the results of BES can be different forbuildings with different form and orientation.

Solar radiation levels as measured in the urban measurement stationwas found to be on average 11W/m2 less than the levels measured inthe rural station. This difference might be caused due to the shading by

the buildings around the urban measurement station. Once imposed,the rural-urban differences in the solar radiation levels yield to a 6%reduction in the July CDs of buildings subjected to an urban MC. Lowersolar radiation measured in the urban measurement station leads to lesssolar heat gains from the façade, which is the reason for the decrease inthe CD of the base case building.

Another sensitivity analysis is conducted for the impact of the urbanshading algorithm and for the impact of the office type building occu-pancy/operation schedule, which are demonstrated in Fig. 9a and b.The urban shading algorithm reduces the building CD by 2.8%, com-pared to the base case scenario. Switching from a residential typebuilding use to an office type building increases the building CD sig-nificantly, by 200.8%. This increase is due to the additional internalheat gains inside the offices and due to the fact that the offices aremodeled as occupied during the times of the day when air temperaturesand solar radiation levels are higher (daytime) than the times whenresidential buildings are occupied (nighttime).

Fig. 5. (a) Comparison of hourly air temperatures from CFD simulations and measurement data obtained at the urban measurement station in July 2013; (b) hourlyurban-rural air temperature differences based on the CFD simulation results and measurement; and (c) differences in hourly air temperatures between CFD simu-lations and measurements obtained at the urban measurement station.


863

Fig. 6. Location-specific microclimatic data obtained at the rural point (black), urban #1 (orange) and urban #2 (green) for (a) air temperature; (b) wind speed; (c)wind direction (0 and 360 indicate north wind); and (d) global horizontal solar radiation. Note that the results for global horizontal solar radiation at both urban andrural locations are obtained directly from the measurements. (For interpretation of the references to color in this figure legend, the reader is referred to the webversion of this article.)


864

5.4. BES: Impact of Antwerp microclimate on building cooling demand

The impact of the local Antwerp microclimate on building energydemand is investigated for 36 simulation cases based on three location-

specific MCs, six construction characteristics and two building use typesas explained in section 4. CDs for the 36 simulation cases are calculatedand compared with each other in Fig. 10a and b. In both figures, basecase residential and office buildings are considered as those with thepre-1946 construction characteristics and subjected to the rural MC.The figure shows that buildings subjected to the rural MC have lowercooling demands than their counterparts located in urban#1 andurban#2. The CDs of the buildings in urban#1 and urban#2 show thefollowing percentage differences compared to the rural MC (thesefindings are for the cases where buildings with the same constructioncharacteristics are compared):

– For residential buildings

• Urban#1 MC compared to the rural MC: +90.0% (average) and108.2% (maximum)


– For office buildings



Larger impacts can be observed if buildings with different con-struction characteristics are compared. The results clearly demonstratethat the CD of residential buildings is more sensitive to changes in theMCs than the CD of office buildings.

Considering all the construction characteristics and building usetypes, being in the vicinity of an urban park cannot fully alleviate theextra CD caused from a warmer urban MC. Still, the average monthlyCD of buildings in urban #2 is noticeably less compared to the CD ofbuildings in urban #1, which is shown in Fig. 11. The maximum re-duction in CD caused only because of the urban location difference is20.3% (residential, 1946–1970 building). Averaged over the sameconstruction characteristics, residential buildings and office buildingsclose to the urban park can have 13.9% and 11.4% less cooling demand

Table 6An overview of the BES cases for the sensitivity analysis. The impact of simulation parameters is evaluated by changing the parameters one at a time compared to thebase case scenario.

Case Construction characteristics Microclimatic condition Building use type

U-value Ex/infiltration rate Air temperature Wind speed Wind direction Solar radiation Shading algorithm

Base case Pre-1946 Pre-1946 Rural Rural Rural Rural Rural Residential

Impact of changing U-valuesCase–1 1946–1970 Pre-1946 Rural Rural Rural Rural Rural ResidentialCase–2 1971–1990 Pre-1946 Rural Rural Rural Rural Rural ResidentialCase–3 1991–2005 Pre-1946 Rural Rural Rural Rural Rural ResidentialCase–4 Post-2005 Pre-1946 Rural Rural Rural Rural Rural ResidentialCase–5 Low energy Pre-1946 Rural Rural Rural Rural Rural Residential

Impact of changing ex/infiltration ratesCase–6 Pre-1946 1946–1970 Rural Rural Rural Rural Rural ResidentialCase–7 Pre-1946 1971–1990 Rural Rural Rural Rural Rural ResidentialCase–8 Pre-1946 1991–2005 Rural Rural Rural Rural Rural ResidentialCase–9 Pre-1946 Post-2005 Rural Rural Rural Rural Rural ResidentialCase–10 Pre-1946 Low energy Rural Rural Rural Rural Rural Residential

Impact of changing microclimatic parametersCase–11 Pre-1946 Pre-1946 Urban #1 Rural Rural Rural Rural ResidentialCase–12 Pre-1946 Pre-1946 Urban #2 Rural Rural Rural Rural ResidentialCase–13 Pre-1946 Pre-1946 Rural Urban #1 Rural Rural Rural ResidentialCase–14 Pre-1946 Pre-1946 Rural Urban #2 Rural Rural Rural ResidentialCase–15 Pre-1946 Pre-1946 Rural Rural Urban #1 Rural Rural ResidentialCase–16 Pre-1946 Pre-1946 Rural Rural Urban #2 Rural Rural ResidentialCase - 17 Pre-1946 Pre-1946 Rural Rural Rural Urban Rural ResidentialCase - 18 Pre-1946 Pre-1946 Rural Rural Rural Rural Urban Residential

Impact of building use typeCase - 19 Pre-1946 Pre-1946 Rural Rural Rural Rural Rural Office

Fig. 7. Sensitivity analysis of the impact of construction characteristics on theJuly CD of the base case building: (a) with different U-values and (b) withdifferent ex/infiltration rates.


865

compared with the buildings away from the urban park. The maximumabsolute change in the monthly CD caused by the urban park is1.9 kWh/m2 for office buildings with pre-1946 construction character-istics. Fig. 11 also demonstrates that older buildings can benefit morefrom the cooling benefit caused by the urban park.

The differences in the CDs of buildings exposed to the urban#1 andurban#2 MCs are in close relationship with the wind direction mea-sured at the rural measurement station. To illustrate this, the dailydominant wind direction measured at the rural station is summarizedbased on one of the four possible directions in an X-Y coordinate systemwhere +X denotes east and +Y denotes north direction. Wind flowapproaching from +X and +Y direction is denoted as from quadrant I,wind flow from +X and −Y is denoted as from quadrant II, wind flowfrom −X and −Y direction is denoted as from quadrant III and windflow from −X and +Y is denoted as from quadrant IV. For an officebuilding with the construction characteristics from 1946 to 1970, thedaily CD is compared when this building is subjected to the urban #1MC and to the urban#2 MC. In Fig. 12, the ratio of the daily CD of thisbuilding under these two MCs is plotted. The daily dominant wind di-rection is represented with different colors. Fig. 12 shows that higherdifferences in the daily CD occur when the wind is from the southwestdirection (quadrant III). This is a consequence of the positioning ofurban#2 location with respect to the urban park of interest. When thewind is approaching from the southwest direction (quadrant III), theurban#2 falls within the wake of the urban park and the cooling impact

caused from the park reduces air temperatures, and thus, the buildingCD. Compared to the daily CD of buildings in urban #1, the daily CD ofbuildings in urban #2 is reduced by a maximum of 43% (occurring on 2July). In terms of absolute values, the daily maximum reduction occursduring 25 July where the representative office building subjected to theMC of urban#2 has 0.181 kWh/m2 less daily CD than the same buildingsubjected to the MC of urban#1. Fig. 12 shows that at times when themeteorological wind is from the Northwest (NW) direction (quadrant4), CD in urban #2 is less than the CD in urban #1. This difference isnot due to the park cooling effect but actually because of the avenuetrees located NW of urban point #2. For NW wind direction, theseavenue trees create a cooling effect in their wake where the urban point#2 is located. Since there are no trees NW of urban point #1, thetemperatures in urban #2 become lower than in urban #1, leading todifferences in the daily CD as shown in Fig. 12.

To elaborate further on the impact of the meteorological wind di-rection on the CD of buildings in the urban area, a regression analysis isconducted. The analysis considers the daily CD (Wh/m2) of an officebuilding with the 1946–1970 construction characteristics. The in-dependent variables are specified as the daily averaged meteorologicalmeasurements (rural) of air temperature (°C), wind speed (m/s) andwind direction (°). The aim of this regression analysis is to produce aparametric relationship in the following form:

Fig. 8. Sensitivity analysis of the impact of microclimatic parameters from urban#1 and urban#2 on the CD of the base case building for July 2013.

Fig. 9. Sensitivity analysis of the impact of (a) the shading algorithm imposed on buildings subjected to an urban MC and (b) office type occupancy and operationschedule, on the July CD of the base case building.


866

⎛⎝

⎞⎠

= + −

−

Location specific daily cooling demand AT WS

WD

Whm

a a a

a

2 0 1 2

3

where AT, WS and WD refer to values obtained from the meteorologicalmeasurements of air temperature (°C), wind speed (m/s) and wind di-rection (°) and a0 ( )Whm2 , a1 °( )Whm C2 , a2 ( )Wh sm3 and a3 °( )Whm2 are the coef-ficients of these independent variables. The following parametricequations are determined for predicting the daily CD:

⎛⎝

⎞⎠

= − + −

− =

Rural cooling demand AT WS

WD R

Whm

664.67 51.54 32.89

0.13 ( 0.92)

2

2

⎛⎝

⎞⎠

= − + −

− =

Urban cooling demand AT WS

WD R

# 1 Whm

437.53 46.45 55.51

0.05 ( 0.80)

2

2

⎛⎝

⎞⎠

= − − −

− =

Urban cooling demand AT WS

WD R

#2 Whm

399.49 44.50 56.58

0.51 ( 0.81)

2

2

The analysis provided statistically meaningful predictions, with allR2 values higher than 0.8. According to the coefficients of the in-dependent variables, the WD has the highest impact on the CDs of

buildings subjected to the urban #2 MC. On the contrary, the WD hasless impact on the CDs of buildings subjected to the urban#1 MC. Itshould be noted these relationships are not established to estimate thedaily CD in the area of interest but rather to demonstrate the relativeimportance of the meteorological WD on the CD of buildings subject tothe MC of urban#2.

6. Discussion

6.1. Evaluation of the results

Although CFD for urban microclimate analysis is a computationallyexpensive approach [60], the capability of CFD for providing resultswith a high spatial resolution (i.e. around individual buildings) offersvarious opportunities. As shown in this study, studies measuring loca-tion-specific meteorological data can be used for validating CFD si-mulations. The CFD simulations performed in this study predict mea-sured air temperatures with fairly good accuracy. This comparison canbe strengthened further by using more measurement data obtained atdifferent locations in the central Antwerp area considered here.

The sensitivity analysis presented in Section 4.3 showed that the CDin the office building considered can be three times the CD in theconsidered residential building. This absolute difference in the CDwould mean that a unit change in the CD of offices and residential

Fig. 10. BES results for CD for (a) residential and (b) office buildings with different construction characteristics and MCs. Urban#1 is away from the urban park ofinterest whereas urban#2 is closer to the urban park of interest.


867

buildings would yield different percentage changes. This can be inter-preted as residential buildings being more sensitive to changing MCs(i.e. air temperature). A similar conclusion was found in the study byHirano and Fujita [74] which reported that the UHI effect has a largerimpact on the CD of residential buildings in Tokyo in terms of per-centage values but not necessarily in terms of absolute values. The largedifference between the CDs of the office building and the residentialbuilding presented in the sensitivity study (can in fact be lower ifcooling strategies such as night ventilation [111–113] or passivecooling [114,115] would be implemented in the office building con-sidered.

The sensitivity analysis demonstrated in Section 5.3 shows that theCD of buildings investigated in this study are mostly sensitive on lo-cation-specific air temperature whereas location-specific wind speedand wind direction have a lesser influence. Similar conclusions werereported in the studies by Sailor and Munoz [43] and Fung et al. [44].

Some prior studies that focused not only on the summertime CD buton the total energy demand of buildings for a whole year found thatdecreasing U-values can lead to savings in the annual building energydemand [55,71]. A similar whole-year analysis may produce similarfindings in the present research as well. Several prior studies docu-mented that decreasing U-values can lead to higher building CD whenthe buildings are exposed to high air temperatures caused by the UHIeffect [35] or by climate change [11]. Therefore, the design of lowenergy buildings with low U-values and low ex/infiltration rates shouldbe carefully evaluated with respect to CD especially if the building willbe exposed to an urban MC. According to Isaac and van Vuuren [4], theglobal annual energy demand for space cooling may become more thanthe demand for space heating within the mid-21st century. With suchrapid changes within a century, a building considered as low energy intoday’s conditions, might not be so throughout its lifetime. With therapid increase in air-conditioning devices also in the developing World[116], the CD of buildings in dense urban areas should be investigatedmore carefully and design regulations should focus on parameters suchas thermal mass and thermal time constant rather than only on U-va-lues.

In this study, the average differences in the CD of buildings sub-jected to the rural and urban MCs are in the range of 60.8–90.0% forresidential buildings and 17.3–30.6% for office buildings. Prior studiescomparing the CDs of buildings in rural and urban areas reported si-milar findings. In the study by Kolokotroni et al. [58], buildings inurban areas were found to have 32–42% more cooling demand than thebuildings in rural areas. Vardoulakis et al. [37] reported the maximumdifference in the CD of urban buildings compared to rural buildings as36.3% whereas Sun and Augenbroe [54] reported CD differences ofrural vs urban buildings averaged over 15 cities in the USA as 17.25%.According to Hassid et al. [30] urban/rural differences can lead to more

than 70% differences in the CD whereas Santamouris et al. [12] re-ported that the CD of buildings in urban areas can be doubled comparedto the CD of buildings in rural areas.

Even though the impact of an urban park on the microclimate in itsvicinity has been investigated in various studies in the past [78], to thebest of our knowledge, only the study by Ca et al. [34] investigated theimpact of an urban park on the building CD. Based on measured airtemperature differences in the wake of an urban park and in nearbylocations, Ca et al. [34] calculated that 15% of energy savings can beachieved during noon time for buildings located in the wake of the parkinvestigated. In the present study, the average decrease in the CDs ofbuildings close to the park compared to buildings far from the park isfound as 13.9% and 11.4% for residential and office buildings, re-spectively.

The present study demonstrates that the analysis of location-specificMCs at various locations in the same urban area should be conductedwith caution when localized cooling sources such as urban parks arepresent. Several studies in the past produced air temperature data atvarious locations in specific cities by using pre-measured UHI in-tensities or by using UHI predictor models which are based on mor-phological inputs. However, the present study demonstrates that loca-tion-specific air temperature data can be highly dependent on the winddirection for the urban area of interest. Even though the focus buildingsin this study are at locations with similar morphological characteristicsand positioned only 800m apart, differences in air temperature up to5.3 °C caused by an urban park are reported. Considering two buildingswith the same form, orientation, construction characteristics and usetypes, the differences in MCs can lead to daily CD differences of up to43% (Fig. 12). It is recommended for the UHI predictor models toconsider the impact of localized cooling sources as another inputparameter while specifying location-specific MCs.

6.2. Limitations

The CFD simulations performed in this study are subjected to sev-eral limitations. All building and street surfaces are assumed to haveidentical conditions with each other. Although CFD simulations can beperformed by imposing individual surface characteristics (i.e. albedo)for each part of the domain, such data is hard to acquire and wouldincrease the pre-processing of the simulations significantly.

Another limitation is neglecting humidity. While specifying loca-tion-specific MCs, differences in relative humidity between urban andrural areas could have been taken into account. However, to considermass transfer in CFD simulations appropriately, relevant boundaryconditions and source/sink terms for humidity sources, such as waterbodies and gardens, should be developed. BES performed in the presentstudy considers only the changes in temperature, wind speed, wind

Fig. 11. Comparison of the CD for different building types and with different proximities to the urban park of interest. Urban#1 is away from the urban park ofinterest whereas urban#2 is closer to the urban park of interest.


868

direction and solar radiation while assuming the same relative humidityconditions as measured in the rural measurement station.

The resulting wind speed and wind direction could have beencompared with the measurement data recorded at the urban measure-ment point, which is located on a rooftop. However, since the CFD si-mulations omitted some geometric details on the buildings, such aswindows, roof tiles and height differences on the roofs, the actual localflow field occurring in reality above the roof where the measurementstation is located might be different than the flow field predicted by theCFD simulations. Due to this limitation, the simulation results arecompared with the measurements only by considering air temperatureand not wind speed or wind direction.

The current approach for coupling urban microclimate with BES isbased on gathering location-specific microclimatological data at spe-cific points. The capability of these points in representing the clima-tological conditions around the buildings of interest can be questioned.The approach followed in this study offers the ease of adapting CFDsimulation results directly on the weather files commonly used in BES.More advanced approaches where the whole flow field data aroundindividual buildings are linked to BES can be adopted in the future.

To investigate the impact of the local Antwerp urban microclimateand the urban park on the net building energy demand, a full year studyshould be conducted where the heating demand is also evaluated [117].However, CFD simulations of urban microclimate focusing on an entireyear would be computationally very expensive. An idea to incorporate afull year investigation can be to focus only on representative days fromeach month and to combine results accordingly.

6.3. Future perspectives

In this study, the impact of an urban park on the CD of buildings inits wake is documented as a case study for an existing situation. Futurestudies can be conducted on how to spread this benefit to a larger area.Morphological conditions of an urban area can be influential inbuilding energy demand as argued by Golany et al. [118] and Rattiet al. [119]. Therefore, future studies can focus on morphologicalchanges aimed at carrying the cooled air in the wake of an urban parkto a larger area of influence.

Applying better thermal insulation to buildings in cities is con-sidered as a good approach in decreasing heating demands. However, asthe future of built environments is expected to be more cooling demanddominated, the design considerations for low energy buildings might

need some rethinking. Neglecting the potential rise in the CD ofbuildings might lead to problems in the future and the share of build-ings in the total global energy consumption may not decrease sig-nificantly, or not at all, if the CD continue to rise. It should be noted thatmore cooling demand in cities can lead to an increase in anthropogenicheat release due to air-conditioning devices and this may increaseurban air temperatures to an even higher level [120,121] – a viciouscircle. Future regulations on low energy building design should givemore attention to the cooling demand of buildings.

7. Summary and conclusions

The cooling demand of buildings is strongly linked to occupantbehavior, building installations, building envelope and urban micro-climate. Buildings in urban areas are generally subjected to higher airtemperatures than the buildings in the surrounding rural areas, andthus, they can exhibit higher cooling demands.

In this study, CFD simulations are performed to investigate theurban microclimate of central Antwerp area and to provide location-specific Microclimatic Condition (MC) to be used in Building EnergySimulation (BES). The CFD simulations use meteorological measure-ments conducted at a rural area outside of Antwerp and at an urbanarea within the city center. The 3D unsteady Reynolds-averaged Navier-Stokes equations are solved considering the climatic conditions of July2013. The simulated air temperatures are compared with the measuredvalues at the location of the urban measurement station and the CFDsimulations are found to reproduce the air temperatures with anaverage absolute difference of 0.88 °C. Given the fairly good agreement,the CFD simulations are used to provide climatic conditions for the BES.The CFD simulation results and measurements for air temperature (°C),wind speed (m/s) and wind direction (°) are extracted at three loca-tions: (1) a rural location outside of Antwerp; (2) an urban locationinside central Antwerp, away from an urban park and (urban#1) (3)another urban location inside central Antwerp, close to the same urbanpark (urban#2). Three location-specific MCs are defined based on theconditions at these locations. These MCs are used as inputs for the BESof a building with the same form and orientation but with two differentbuilding use types (residential and office) and with six different sets ofconstruction characteristics, ranging from pre-1946 buildings tomodern low energy buildings. The resulting Cooling Demands (CD)(kWh/m2) from the 36 simulation cases are reported and comparedwith each other. The following conclusions can be made:

Fig. 12. The ratio of daily CD during July 2013 for a representative building in urban#2 (close to the urban park) and urban#1 (away from the same urban park) andits relationship with the daily wind direction measured in the rural measurement station.


869

On average, air temperatures recorded at urban#1 and urban#2 are3.3 °C and 2.4 °C higher than the air temperatures recorded at the ruralarea considering the entire month of July 2013. The cooling effect fromthe urban park of interest leads to a temperature difference betweenurban#2 and urban#1 by 0.9 °C on average and up to a maximum of5.3 °C throughout the month of July 2013.

The differences in the MCs between the rural location, urban#1 andurban#2 lead to significant changes in the CD of buildings. Comparedwith the residential buildings with the same construction characteristicsin the rural area, buildings in urban#1 and urban#2 are found to haveon average 90% and 60.8% more CD, respectively. Compared with theoffice buildings with the same construction characteristics in the ruralarea, buildings in urban#1 and urban#2 are found to have on average30.6% and 17.3% more CD, respectively.

Newer buildings with better thermal insulation and airtight en-closures are found to have much more CD during July 2013. Themaximum difference in CD reported in this study is 209.3%, whichcorresponds to the case when the CD of a rural residential building withthe pre-1946 construction characteristics is compared with the CD of aresidential building in the urban area with low energy constructioncharacteristics.

Even though the vicinity of an urban park cannot completely alle-viate the additional CD caused from an urban MC, the park investigatedin this study can help in reducing building CD to some degree. Beingsubjected to the MC of urban#2 instead of urban#1 can lead to averageCD reductions of 13.9% for residential buildings and 11.9% for officebuildings during July 2013. The maximum absolute change in themonthly CD caused by the urban park is 1.9 kWh/m2 for office build-ings with pre-1946 construction characteristics.

The maximum difference in the daily CDs between urban#2 andurban#1 is highly dependent on the wind direction approaching to theurban area of interest and can reach up to 43% when the wind flow isfrom the southwest direction, which leaves the urban#2 in the wake ofthe urban park of interest. Such differences in the CDs of buildingswhich are only 450m apart indicate that location-specific micro-climatological information can be an important aspect in estimating theenergy demands of buildings in cities.

This study demonstrates that the impact of urban microclimate onbuilding cooling demand is quite complex as it can vary significantlywithin the same urban area depending on: the building location in theurban area of interest; building type; and the building’s constructionperiod. Considering this complexity, a building constructed with “low-energy” construction guidelines in a rural setting, may not hold itspromise if designed in the same way in city centers. For the futureguidelines on the design of low energy buildings and for the regulationson building energy efficiency, we would recommend taking into ac-count the impact of urban microclimates. Designers and engineers fo-cusing on low-energy building design should also take the findings ofthis study into account.

Acknowledgements

The authors gratefully acknowledge the partnership of the first twoauthors of this paper with ANSYS CFD. The authors would like to thankto the members of the Building Performance chair of EindhovenUniversity of Technology, led by Prof. Jan Hensen for providing valu-able comments and recommendations for the building energy simula-tions performed in this study.

References

[1] European Commission. Buildings. Energy efficiency in buildings 2017 [accessed January26, 2017].

[2] U.S. Energy Information Administration. Monthly energy review: December 2016.Washington, DC; 2016, doi:DOE/EIA-0035(2011/02).

[3] European Commission. An EU strategy on heating and cooling. Brussels; 2016.

[4] Isaac M, van Vuuren DP. Modeling global residential sector energy demand forheating and air conditioning in the context of climate change. Energy Pol2009;37:507–21. http://dx.doi.org/10.1016/j.enpol.2008.09.051.

[5] Holmes MJ, Hacker JN. Climate change, thermal comfort and energy: MEETINGthe design challenges of the 21st century. Energy Build 2007;39:802–14. http://dx.doi.org/10.1016/j.enbuild.2007.02.009.

[6] Watkiss P. The ClimateCOST Project, Final Report, Volume 1: Europe. Stockholm:Stockholm Environment Institute; 2011.

[7] Ruth M, Lin AC. Regional energy demand and adaptations to climate change:Methodology and application to the state of Maryland, USA. Energy Pol2006;34:2820–33. http://dx.doi.org/10.1016/j.enpol.2005.04.016.

[8] Changnon SA, Kunkel EK, Reinke BC. Impacts and responses to the 1995 heatwave: a call to action. Bull Am Meteorol Soc 1996;77:1497–506.

[9] Miller NL, Hayhoe K, Jin J, Auffhammer M. Climate, extreme heat, and electricitydemand in California. J Appl Meteorol Climatol 2008;47:1834–44. http://dx.doi.org/10.1175/2007JAMC1480.1.

[10] Pyrgou A, Castaldo VL, Pisello AL, Cotana F, Santamouris M. On the effect ofsummer heatwaves and urban overheating on building thermal-energy perfor-mance in central Italy. Sustain Cities Soc 2017;28:187–200. http://dx.doi.org/10.1016/j.scs.2016.09.012.

[11] van Hooff T, Blocken B, Timmermans HJP, Hensen JLM. Analysis of the predictedeffect of passive climate adaptation measures on energy demand for cooling andheating in a reside

impact of urban microclimate on summertime building ...building – – – – – – – – –...

Documents