Building and En i onmen 263 (2024) 111861
A ailable online 20 July 2024
0360-1323/© 2024 The Au ho s. Published by Else ie L d. This is an open access a icle unde he CC BY-NC license (h p://c ea i ecommons.o g/licenses/by-
nc/4.0/).
Ca ego isa ion o u ban open spaces o hea adap a ion: A clus e
based app oach
Ane Villa e de
a,*
, I an zu ´
Al a ez
b
, Edua do Rojí
a
, Lei e Ga mendia
a
a
Uni e si y o he Basque Coun y (UPV/EHU), Mechanical Enginee ing Depa men , Plaza Ingenie o To es Que edo, 48013, Bilbao, Spain
b
Uni e si y o he Basque Coun y (UPV/EHU), G aphic Design and Enginee ing P ojec s Depa men , Plaza Ingenie o To es Que edo, 48013, Bilbao, Spain
ARTICLE INFO
Keywo ds:
Clima e change
U ban ca ego isa ion
Hea wa e adap a ion
Clus e ing
Buil en i onmen
ABSTRACT
In he con ex o clima e change, adap ing a ci y’s open spaces o hea wa es and ex eme hea is c ucial o
mi iga ing he U ban Hea Island e ec and ensu ing he wellbeing o i s inhabi an s. Howe e , he hea adap-
a ion po en ial o open spaces a ies wi hin a ci y. This s udy de elops an objec i e and eplicable me hod o
ca ego ise u ban a eas based on hei open spaces’adap i e capaci y and imp o emen po en ial o hea wa es a
he mic o-scale. This is achie ed using a clus e ing app oach, elimina ing manual ope a ions, using openly
a ailable da a, employing indica o s sui able o e e y u ban ab ic ype, and using he u ban block as he
analysis uni . The alida ion o he me hod in he case s udy o Bilbao (Spain) demons a es i s capaci y o ob ain
meaning ul insigh s ega ding open space hea adap a ion po en ial. An analysis o he esul ing ca ego ies e-
eals signi ican open space imp o emen po en ial h oughou he ci y and a a ying adap i e capaci y ha
depends on u ban densi y. Ca ego ies wi h he lowes adap i e capaci y ha e median open space a ios be ween
0.35 and 0.65 and median ee co e a ios be ween 0.07 and 0.15. Those wi h he highes adap i e capaci y
p esen median open space a ios be ween 0.8 and 1 and median ee co e a ios be ween 0.05 and 0.25. The
me hod can aid local policymake s in iden i ying oppo uni y spo s o hos ing adap a ion solu ions and un-
de s anding he challenges he ci y may ace in planning adap a ion ac ion. The me hod’s eplicabili y enables i
o be applied in o he ci ies, con ibu ing o a b oade explo a ion o clima e change adap a ion po en ial.
1. In oduc ion
The e ec s o human-induced global wa ming a e al eady no able.
Vicedo-Cab e a e al. [1] ecen ly ound ha 37 % o hea - ela ed
mo ali y du ing wa m seasons be ween 1991 and 2018 could be
a ibu ed o human-induced clima e change. The summe hea wa e o
2019 was es ima ed o be he deadlies ex eme e en o he yea , wi h
app oxima ely 2050 excess dea hs in he empe a e coun ies o F ance,
he Ne he lands, he UK, and Belgium [2]. Un o una ely, he si ua ion is
expec ed o ge wo se. The In e go e nmen al Panel on Clima e Change
(IPCC) p edic s ha in he nex 20 yea s, a e age global empe a u es
will ise by a leas 1.5 ◦C abo e p e-indus ial le els [3]. Tempe a u es
ac oss Eu ope a e p edic ed o ise con inuously mo e han he es o
he wo ld o e his cen u y. P ojec ions indica e ha by 2071, a e age
empe a u es in Eu ope will inc ease by 1.2–3.4 ◦C (SSP1-2.6 scena io)
and 4.1–8.5 ◦C (SSP5-8.5 scena io) [4]. This ise in empe a u e will be
combined wi h mo e equen , in ense, and longe hea wa es [3]. Ci ies
a e especially ulne able o hea wa es and high empe a u es due o he
U ban Hea Island (UHI) phenomenon, which makes ci ies wa me han
hei u al su oundings, hus in ensi ying he hea in u ban a eas [5].
The UHI depends on ac o s such as he p esence o ege a ion, he Sky
View Fac o (SVF), he p opo ions o s ee canyons, impe ious su -
aces, o su ace albedo [6]. Tempe a u e exposu es can also be in lu-
enced by mic o-scale condi ions, c ea ing a a ie y o mic oclima es
wi hin he local scale [7,8].
Highe u ban empe a u es inc ease he use o ene gy o cooling,
educe ou doo and indoo he mal com o , inc ease he concen a ion
o pollu an s such as oposphe ic ozone, and impac human heal h and
Abb e ia ions: IPCC, In e go e nmen al Panel on Clima e Change; UHI, u ban hea island; NDVI, no malised di e ence ege a ion index; LCZ, Local Clima e Zone;
GIS, Geog aphical In o ma ion Sys em; TCC, ee canopy co e ; DSMV, digi al su ace model o ege a ion; OSA, open space ampli ude; OSR, open space a io; RSR,
oad su ace a io; A, albedo; SVF, sky iew ac o ; DSM, Digi al Su ace Model; I, impe iousness; SR, sola adia ion; PCA, p incipal componen analysis.
* Co esponding au ho .
E-mail add esses: [email p o ec ed] (A. Villa e de), [email p o ec ed] (I. ´
Al a ez), [email p o ec ed] (E. Rojí), [email p o ec ed]
(L. Ga mendia).
Con en s lis s a ailable a ScienceDi ec
Building and En i onmen
jou nal homepage: www.else ie .com/loca e/builden
h ps://doi.o g/10.1016/j.builden .2024.111861
Recei ed 25 Ap il 2024; Recei ed in e ised o m 29 June 2024; Accep ed 18 July 2024
Building and En i onmen 263 (2024) 111861
2
wellbeing [5].To ensu e he wellbeing o ci izens in u u e hea wa e
e en s, i is essen ial o success ully adap ou ci ies o his incoming
clima e [3]. This is especially ele an o empe a e ci ies, whe e he
buil en i onmen is no p epa ed o ex eme high empe a u es. Fo
ha , e ec i e adap a ion ools a e needed [9,10]. Changing exis ing
u ban open spaces o adap o clima e change is an e ec i e s a egy o
imp o e ou doo he mal com o du ing ex eme hea e en s. These
adap a ion solu ions also con ibu e o he egula ion o he u ban
clima e a he local scale (by educing he UHI in ensi y) and a he
mic o-scale (by egula ing he mic oclima e) [11–14].
Recen s udies in u ban clima e and clima e change adap a ion ha e
signi ican ly imp o ed he abili y o quan i y he e ec i eness o hea
adap a ion solu ions [15]. Inc easing ege a ion, a p o en solu ion o
educe he mal s ess and mi iga e he UHI e ec [16], has been
ex ensi ely s udied. While ee plan ing is one o he mos s udied so-
lu ions [17–20], sh ub plan ing and inc easing g ass su ace ha e
demons a ed e ec i eness, oo [14,21]. O he success ul solu ions a e
inc easing cool pa emen s and educing impe ious su aces in u ban
open spaces [5,13,21,22]. Blue in as uc u e, such as wa e ponds o
lakes, has also shown e ec i eness [23,24] e en hough i s esea ch is
mo e limi ed han o he solu ions [15], and i is challenging o in eg a e
hem in o exis ing u ban a eas. U ban a i icial shading is also a
well-documen ed adap a ion solu ion [25,26]. Fo ins ance, s ee sun
sails, a non-in usi e s ee shading solu ion adi ional o sou h Spanish
ci ies, subs an ially educe g ound and açade empe a u es [27].
Finally, u ban in e en ions o ien ed o libe a e u ban space o accom-
moda e adap a ion solu ions can be conside ed adap a ion solu ions,
oo. Fo ins ance, pedes ianisa ion wo ks ha educe he su ace
des ined o mo o ised a ic ee space ha can be used o implemen
adap a ion solu ions. An example o his kind o in e en ion is he
Ba celona supe block model [28].
In ecen yea s, e o s ha e been ocused on es ima ing he e ec-
i eness o u ban hea adap a ion solu ions [15]. These e o s a e
essen ial o unde s anding he po en ial e ec s o clima e change
adap a ion solu ions and, hus, mo i a ing ac ion and alloca ing e-
sou ces [29]. Unde s anding he bene i s is also c ucial o suppo pol-
icymake s’decision-making and p o ide in o ma ion on esilience
moni o ing. Howe e , among a ious cons ain s in adap a ion, he
easibili y o implemen ing e ec i e solu ions, including cons ain s
ela ed o physical cha ac e is ics, is a signi ican challenge. This
cons ain is a ely add essed in u ban hea adap a ion li e a u e. Fo
ins ance, Zhang e al. [16] quan i ied he impac o UHI educ ion by
ans o ming low ege a ion a eas in o ege a ion a eas wi h a high
No malised Di e ence Vege a ion Index (NDVI). Thei wo k o e s
essen ial insigh s in o he po en ial o ege a ion solu ions. Ne e he-
less, i lacks he mic o-scale app oach needed o conside he easibili y
o ans o ming densely buil a eas in o high NDVI a eas.
No e e y open space can accommoda e iden ical solu ions, and he
e ec i eness o solu ions may change om one open space o ano he .
Bassolino e al. [30] showed ha exis ing dense u ban ab ics ha e
limi ed space o accommoda e adap a ion solu ions, and as a esul , hei
capaci y o adap is educed. Mo eo e , o he s udies ha e shown ha
he cha ac e is ics o u ban a eas can inc ease o hinde he e ec i e-
ness o solu ions. Fo example, Koloko sa e al. [31] p o ed ha in s ee
canyons wi h a high aspec a io (heigh o wid h a io), adap a ion
solu ion e ec i eness is lowe because ai empe a u e in hose s ee s is
highly in luenced by ad ec ion. Dimoudi and Nikolopoulou [18]
demons a ed ha he highe he ege a ed- o-buil a ea a io, he
g ea e he e ec i eness in ai empe a u e educ ion o ee plan ing.
This conclusion sugges s ha u ban a eas wi h a highe a io o open
space o buil a eas ha e a highe adap i e capaci y. Zhao e al. [32]
ound a co ela ion be ween ee cooling e iciency and cha ac e is ics
such as albedo and sola adia ion. P e ious s udies p o ided insigh s
in o adap a ion solu ion e ec i eness and iden i ied cons ain s, ye
he e is a lack o sys ema ic analyses conside ing he adap i e capaci y
o open spaces unde going adap a ion.
The IPCC de ines he e m adap i e capaci y as he capaci y o
communi ies and ins i u ions o adop adap a ion solu ions [33], and
commonly in li e a u e is linked o he abili y o indi iduals o collec-
i es o adap [34,35]. In his esea ch wo k, we ex end he meaning o
he physical capaci y ha open spaces ha e o accommoda e e ec i e
adap a ion solu ions. Toge he wi h adap i e capaci y, his pape de-
ines imp o emen po en ial as he deg ee o which a gi en space can be
u he enhanced o add ess ex eme hea h ough adap a ion solu ions.
This e m cap u es he idea ha open spaces al eady well-equipped wi h
adap a ion solu ions may exhibi limi ed po en ial o addi ional
imp o emen . A he same ime, hose wi h ewe in e en ions p esen
a highe scope o enhancemen . Fo ins ance, highly ege a ed
pe meable open spaces may ha e low imp o emen po en ial, deno ing
ha hei adap a ion limi has been eached.
Knowing he adap i e capaci y and imp o emen po en ial o he
di e en open spaces o a ci y would allow decision-make s, on he one
hand, o ha e an o e iew o he adap a ion oppo uni ies and adap a-
ion limi s o a ci y and, on he o he hand, o de e mine which adap-
a ion solu ion is sui able o each open space. Howe e , assessing one
by one he adap i e capaci y and imp o emen po en ial o u ban open
spaces is ime-consuming and cos ly. The e o e, an u ban ca ego isa ion
app oach is undamen al o add essing his issue.
U ban ca ego isa ions o classi ica ions aim o so a gi en ci y’s
elemen s in o g oups ha sha e simila i ies. These g oups a e he ca e-
go ies, and o en, an a che ype ha ep esen s he ypical alues o i s
ca ego y is iden i ied [7,36,37]. By assessing he a che ype, he a i-
ables a e educed, and esul s can la e be ex apola ed, allowing
localised analyses o be expanded o he ci y le el [38]. E en hough
ca ego isa ions suppo ed by da a mining echniques a e widely used o
educe simula ion o assessmen ime a he building scale [39,40], hei
applica ion o p edic ing he sui abili y o adap a ion solu ions is s ill
limi ed among u ban clima e o u ban adap a ion s udies.
U ban ca ego isa ions ha e been mainly used in u ban clima e
s udies o assess UHI in ensi y o p edic mic oclima e beha iou .
Nume ous s udies in his ield use he amewo k o S ewa d and Oke’s
Local Clima e Zones (LCZ). LCZ is a classi ica ion ha desc ibes clima e
condi ions on a local o neighbou hood scale, sui able o a esolu ion o
1 km ×1 km [41]. Recen wo ks ha e de eloped me hods o nuance
u he his classi ica ion. Some o he wo ks ha e ocused on assessing
u ban clima e on a mo e de ailed scale han LCZ by adding block le el,
s ee le el, o building le el cha ac e is ics [7,37,42,43]. O he s ha e
supplemen ed LCZ wi h pa ame e s beyond he physical ealm, such as
communi y coping capaci y pa ame e s [44]. L´
opez-Mo eno e al. [42]
de eloped a ca ego isa ion o aid ene gy pe o mance assessmen ha
could be sui able o an adap i e capaci y and imp o emen po en ial
assessmen due o he di e si y indica o s used. Howe e , his ca ego-
isa ion is based on a local classi ica ion, so he me hod is no easily
eplicable. E en hough hese app oaches demons a e ha da a mining
and u ban ca ego isa ion a e p omising app oaches o assessing he
issue o u ban hea , hey a e no pu pose ully de eloped o explo e he
possibili ies o adap a ion.
A ew wo ks ha e de eloped u ban classi ica ion me hodologies o
iden i y u ban a ea ypes ha ha e simila u ban clima e beha iou and
adap a ion op ions. Howe e , some o hese me hods ely on p e-
exis ing classi ica ions, such as Land Co e and Land Use classi ica-
ions [31,45] o local p e-exis ing bespoke u ban ab ic classi ica ions
[30]. In he ecen wo k by Sü zl e al. [46], hey conduc ed a 250 ×250
m esolu ion g id classi ica ion based on k-means clus e ing and used
indica o s om he u ban clima e li e a u e. The me hod e ec i ely
iden i ies u ban ypes wi h simila su ace empe a u es; la e , hese
ypes a e used o analysing adap a ion op ions. While being e y use ul,
his wo k lacks he mic o-scale app oach needed o closely examine he
adap a ion op ions ha he u ban open spaces ha e. The assessmen uni
used, he g id, does no adap o he mo phology o he ci y, which could
o e look mic o-scale a iables. This hind ance is e lec ed in he wo k
by Re . [42], whe e hey concluded ha only 37 % o pixels o he size
A. Villa e de e al.
Building and En i onmen 263 (2024) 111861
3
200 ×200 m had homogeneous cha ac e is ics. The classi ica ion esul s
using he g id a e also mo e challenging o u ban planne s o designe s
o in e p e , as hey do no e lec eal u ban s uc u es. On he o he
hand, he me hod elies on commonly used u ban clima e indica o s and
does no add addi ional pa ame e s in he classi ica ion ha can a ec
adap a ion op ions, such as s ee wid h o oad su ace a io.
The esea ch wo k p esen ed he e adds o he con ibu ions o hese
wo ks by including in he ca ego isa ion pa ame e s ha di ec ly anal-
yse he possibili y o applying adap a ion solu ions by including in-
dica o s ha : 1) assess he capaci y o accommoda e solu ions, 2)
conside he e ec i eness o adap a ion solu ions, 3) e alua e
imp o emen capaci y h ough adap a ion solu ions. Addi ionally, he
analysis uni used will be sui able o he mic o-scale assessmen o
adap a ion op ions. This way, he p esen esea ch ocuses on iden i-
ying u ban open spaces ha ha e simila adap a ion po en ial a he
han simila UHI magni ude. The e o e, he need o a ca ego isa ion
me hod o aid hea wa e adap a ion ac ions in open spaces is s ill
e iden .
Hence, his esea ch aims o de elop an objec i e, eplicable, mic o-
scale ca ego isa ion me hod o iden i ying he adap i e capaci y and
imp o emen po en ial o indi idual open spaces wi hin a ci y loca ed in
a empe a e clima e. To demons a e he applicabili y and eplica ion
po en ial o he me hod, i is applied in a case s udy ci y and he esul s
ob ained analysed.
The esea ch wo k is o ganised as ollows: Sec ion 2desc ibes he
s eps ollowed in he de elopmen o he u ban ca ego isa ion me h-
odology; Sec ion 3p esen s he applica ion o he me hod on a case
s udy; Sec ion 4includes he esul s o he s udy and discussion; Sec ion
5summa ises he s udy’s conclusions.
2. U ban ca ego isa ion me hodology based on adap i e
capaci y and imp o emen po en ial
The me hod de eloped o he u ban ca ego isa ion is based on he
da a mining echnique o clus e ing. Da a mining e e s o he p ocess o
ex ac ing om da a p e iously unknown, implici , po en ially help ul
in o ma ion. This in o ma ion can be used o decision-making o p e-
dic ion, among o he po en ial uses [47]. The clus e ing echnique is a
ype o da a mining whe e he goal is o g oup he da a ins ances based
on mul iple a iables. No classes a e de ined in he clus e ing, and he
esul s usually p o ide an a che ypical ins ance o each clus e . Since
clus e analysis was in oduced o he s udy o he buil en i onmen ,
esea che s ha e elied on i o sys ema ically ca ego ise u ban a eas [7,
37,44,48]. One o he i s wo ks in his ield [48] p o ed he capaci y o
clus e analysis o di e en ia e wo neighbou hoods based on mo pho-
logical indica o s. K-means is a widely used clus e ing algo i hm known
o i s compu a ional e iciency and use -de ined clus e numbe , he k
alue. Unlike o he algo i hms, such as hie a chical clus e ing, i does
no impose a high compu a ional bu den. The low compu a ional e-
qui emen s and he lexibili y o adjus he k alue make i sui able o
explo a i e me hods such as he one p esen ed in his pape . In his
wo k, adap i e capaci y and imp o emen po en ial indica o s con o m
o he clus e ing a iables, and he esul ing clus e s a e he u ban ca -
ego ies. The clus e ed ins ances a e de i ed om he spa ial analysis
uni chosen and he analysis ins ance closes o he clus e cen e is he
a che ype, which p esen s he mos ypical alues o he clus e .
The analysis uni has o be adequa e o he pu poses o he ca e-
go isa ion. Analysis uni s such as he neighbou hood o census le els
would o e look mic o-scale cha ac e is ics, making hem oo la ge o
he me hod. The g id is unsui able o small he e ogeneous ci ies
because i would ail o cap u e he cha ac e is ics o homogeneous
u ban zones wi hin he p ede ined squa e-shaped pa cels [42]. De ining
he open space bounda ies one by one would be excessi ely
ime-consuming, hinde ing he eplica ion o he me hod on a la ge a ea.
Consequen ly, he chosen analysis uni is he u ban block de ined as he
land pa cel delimi ed by he s ee axis. The choice o he u ban block
i s he mic o-scale, uses commonly publicly a ailable da a o i s de i-
ni ion, and can be adap ed o di e en u ban ab ics. Mo eo e , he
u ban block uni is easily unde s ood by u ban planne s and u ban de-
signe s. These cha ac e is ics make i sui able o he p esen me hod.
The ollowing sec ions show he s eps ollowed, which ha e been guided
by he wo k ca ied ou by Re s. [39,47,48].
2.1. Da a p epa a ion
Fi s o all, he u ban blocks should be de ined and ca e ully il e ed.
A s ee o oad axis laye can be used o de ine he u ban blocks. This
laye is hen con e ed om a ec o line laye o a ec o polygonal
laye using Geog aphical In o ma ion Sys em (GIS) so wa e. Addi ional
laye s like wa e body bounda ies can be used o imp o e he delimi-
a ion o u ban blocks. A e wa ds, he esul ing polygons should be
ca e ully il e ed, as including ins ances ha a e no o in e es o he
analysis can al e he clus e analysis esul s, making hem unusable. Fo
ins ance, he polygons ha lay ou side he u ban a eas should be
emo ed, as well as o he polygons ha lie in a eas ha a e no o in-
e es o he s udy and a e expec ed o ha e ou lie alues, such as i e
basins.
An induc i e p ocess is ollowed o iden i y he indica o s sui able o
he ca ego isa ion. The p ocess is explained in Fig. 1. Fi s , a se o
possible indica o s is c ea ed based on a li e a u e e iew. On he one
hand, a lis o indica o s o open space cha ac e is ics linked wi h he
adap i e capaci y o he open spaces is added. This lis includes in-
dica o s ela ed o he possibili y o accommoda ing solu ions and in-
dica o s ela ed o cha ac e is ics ha condi ion he e ec i eness o
solu ions. Fo ins ance, “s ee wid h”is added because i indica es he
possibili y o plan ing ees in he s ee , and “s ee heigh o wid h
a io”is added because i condi ions he e ec i eness o speci ic solu-
ions. On he o he hand, indica o s linked wi h he po en ial o
imp o emen in open spaces a e included. Fo ins ance, “ ege a ion
co e a io”is added because i indica es ha an open space al eady has
enough ege a ion, indica ing ha ee plan ing canno imp o e ha
a ea. Finally, addi ional indica o s ound in scien i ic publica ions ha
a e ele an o he opic a e added, such as li e a u e abou u ban
clima e, hea ulne abili y and hea coping capaci y. This addi ion is
done o ensu e ha he ini ial lis is no o e ly na owed. A e wa ds,
wi h he lis o indica o s in hand (57 in o al), il e ing is done o ensu e
ha he chosen indica o s a e aligned wi h he objec i es o he me hod
(Appendix A). Fo his aim, he ollowing ques ions a e pos ed.
•Is his indica o linked wi h u ban open spaces’physical adap i e
capaci y o imp o emen po en ial? This ques ion mainly il e s in-
dica o s ha a e ela ed o he capaci y o indi iduals o commu-
ni ies o adap .
•Is applying his indica o objec i e, eplicable, o sui able o mic o-
scale assessmen ? This ques ion p ima ily excludes indica o s ha do
no apply o di e se u ban ab ics o mus be manually calcula ed.
•Is open-access da a a ailable o eed he indica o ? This ques ion
excludes indica o s wi h una ailable da a o indica o s wi h a ail-
able da a ha a e no openly accessible. This s ep is conduc ed o
ensu e he me hod is eplicable o di e en case s udies and use s.
Tha is, he me hod is lexible and scalable enough o be imple-
men ed ega dless o he size and ype o u ban ab ic.
E e y indica o wi h a “no”answe o any o hose ques ions is
emo ed. As a esul , 12 indica o s a e ob ained. Finally, he indica o s
ha e e o o e lapping cha ac e is ics (i.e. ege a ion co e and NDVI)
a e uni ied o a oid duplica e in o ma ion. The inal lis includes 8 in-
dica o s. Once he indica o lis is decided, each indica o is calcula ed
using GIS so wa e and he da abase o he clus e ing is c ea ed.
The indica o ee canopy co e a io (TCC) measu es he pe -
cen age o open space o an u ban block co e ed by ees. This indica o
is linked wi h imp o emen po en ial. I e lec s he amoun o ee
A. Villa e de e al.
Building and En i onmen 263 (2024) 111861
4
densi y in he open space and, he e o e, he amoun o addi ional ees
ha can be plan ed. Fo i s calcula ion, a Digi al Su ace Model o
ege a ion (DSMV) as e image can be used. F om he as e image,
pixels wi h a heigh alue below 2 m a e emo ed. Then, he oo p in s
o buildings a e emo ed. TCC is calcula ed as he a io o emaining
ege a ion pixels on each u ban block.
The indica o open space ampli ude (OSA) quan i ies he
spaciousness o ex en o open a eas wi hin an u ban block. This indi-
ca o is linked wi h adap i e capaci y. In he i s indica o lis , s ee
wid h was iden i ied as a c i ical indica o ha e lec s he possibili y o
accommoda ing adap a ion solu ions such as ees. Howe e , only some
u ban a eas ha e de ined s ee canyons, and no e e y open space is a
s ee . Squa es o esidual spaces a e po en ial a eas o accommoda ing
adap a ion solu ions oo [49]. In o de o so his obs acle, open space
ampli ude is used ins ead o s ee wid h. A laye wi h he building
oo p in is needed o calcula e he indica o . Then, a p oximi y as e
laye is c ea ed on a GIS so wa e using he building laye as a base.
E e y poin in he open space is hen assigned he dis ance alue o he
closes building. Finally, he indica o is calcula ed as he a e age dis-
ance alue o e e y open space pixel wi hin each u ban block.
The indica o open space a io (OSR) measu es he pe cen age o
open space su ace in an u ban block conce ning he space co e ed by
buildings. This indica o is linked wi h adap i e capaci y. A highe OSR
is linked wi h a highe adap i e capaci y o an a ea since he e is mo e
space o accommoda ing open space adap a ion solu ions; hence, hei
e ec i eness is mo e signi ican [7,18,31].
The indica o oad su ace a io (RSR) quan i ies he pe cen age o
he open space wi hin he block ha is des ined o oads. This indica o
is linked o bo h imp o emen po en ial and adap i e capaci y. An open
space wi h many oads has less space o accommoda e solu ions such as
ees, g ass, o impe ious pa emen s. Simul aneously, i has imp o e-
men po en ial h ough pedes ianisa ion, which has been iden i ied as a
c ucial ac ion o u ban ans o ma ion [28]. A ec o oad laye can be
used o calcula e his indica o . Then, he laye is as e ised and he
building oo p in s emo ed. RSR is calcula ed as he a io o emaining
oad pixels on each u ban block.
The indica o albedo (A) quan i ies he sola adia ion ha is
e lec ed by a su ace. I is one o he pa ame e s ha S ewa and Oke
[6] iden i ied o measu ing UHI in ensi y. In his case, his indica o is
linked wi h he imp o emen po en ial, as a lowe albedo indica es ha
a block has he po en ial o imp o emen by inc easing he albedo o i s
su ace ma e ials. The albedo is calcula ed using Landsa 8–9 OLI/TIRS
C2 L2 e lec ance images. The images chosen o he calcula ion should
be aken du ing summe , wi h low ae osol concen a ion and wi hou
clouds. The de ailed p ocedu e ollowed o ob ain he albedo alues
om he e lec ance images can be ound in Re . [50]. The inal alue is
calcula ed as he a e age albedo alue o each block.
The sky iew ac o (SVF) indica o measu es he isible sky ac-
ion a a gi en loca ion. On a shallow s ee canyon, SVF alues a e
close o 1; as he p opo ions o he s ee canyon na ow, he alues
dec ease. This indica o is linked wi h adap i e capaci y. In li e a u e,
na ow s ee canyons ha e been ela ed o lowe e ec i eness o
adap a ion solu ions [18,31,51]. Using a Digi al Su ace Model (DSM), a
as e laye is c ea ed by assigning an SVF alue o each pixel wi hin a
block. Fo ha , he UMEP plugin o QGIS can be used. Then, he
a e age SVF o he pixels ha lay on an open space wi hin a block is
calcula ed. Finally, he indica o is calcula ed as he a e age SVF o
e e y open space poin wi hin each u ban block.
The indica o impe iousness (I) e e s o he deg ee o which a
su ace is impe meable o esis an o wa e in il a ion. S ewa and
Oke [6] ela e his pa ame e o he UHI e ec , and educing he
impe ious su ace is a known adap a ion solu ion o hea wa e
adap a ion [21,52]. This indica o is linked wi h he imp o emen po-
en ial, as i indica es he po en ial o dec easing impe ious su aces as
an adap ion measu e. A sa elli e as e image o he su ace impe i-
ousness should be ob ained o he calcula ion, such as a Cope nicus
impe iousness image. The indica o is calcula ed as he a e age alue
o impe iousness wi hin each u ban block.
The indica o sola adia ion (SR) e e s o he amoun o sola a-
dia ion ha eaches he open space su ace. The in ensi y o SR in an
open space indica es he po en ial o he open space o imp o e by
in oducing shading solu ions; hus, i is linked wi h imp o emen po-
en ial. The alues can be calcula ed o each pixel o an u ban block
using an DSM and he plugin UMEP o QGIS. Then, he a e age SR o
Fig. 1. The p ocess ollowed o he selec ion o he indica o s.
A. Villa e de e al.
Building and En i onmen 263 (2024) 111861
5
he pixels ha lay on open spaces is calcula ed.
Ca e ully p epa ing he da a is an essen ial s ep o ensu e ha he
clu e ing esul s a e help ul o he pu poses o he analysis. Mo eo e ,
he k-means algo i hm is sensi i e o ou lie alues, which can dis o
he esul s and lead o unp oduc i e ca ego ies. Consequen ly, da a has
o be cleaned by emo ing missing and ou lie alues. Then, a iables
a e no malised o p e en undesi ed weigh ing. Fo ha , alues a e
escaled, ans o ming hem in o he [0–1] ange by applying he
ollowing equa ion:
x
ij,no m
=(x
ij
−x
j,min
)/(x
j,max
−x
j,min
) (1)
Whe e x
ij,no m
is he no malised alue o a iable j o block i; x
ij
is he
ac ual alue o a iable j o block i; x
j
,
min
is he minimum alue o
a iable j o all blocks and x
j,max
is he maximum alue o a iable j.
A e wa ds, a co ela ion analysis is pe o med o exclude dependen
indica o s ha could bias he ca ego isa ion. Because he dis ibu ion o
he da a o mos o he indica o s is non-no mal, a Spea man co ela ion
analysis is conduc ed, and (co ela ion coe icien ) and p(s a is ical
signi icance) alues a e ob ained.
2.2. Clus e analysis and alida ion
Be o e conduc ing he clus e analysis, he k alue needs o be cho-
sen. The k alue selec ion and he esul s’ alida ion a e made in an
i e a i e p ocess. Fi s , se e al k- alue analyses a e conduc ed, and a -
e wa ds, he esul s o each analysis a e e alua ed. Since quan i a i e
me hods a e no always conclusi e o selec he op imal k alue [46],
quan i a i e and quali a i e me hods a e combined. Among he exis ing
alida ion me hods, in his esea ch wo k in e nal alida ion, isual
alida ion, and ex e nal alida ion a e blended [53].
Fi s , a gap analysis is conduc ed as pa o he quan i a i e in e nal
alida ion. The gap analysis indica es he op imal k alue by compa ing
he dispe sion o da a poin s wi hin clus e s o wha would be an ici-
pa ed unde a ele an null e e ence dis ibu ion [54]. Taking he k
alue sugges ed by he gap analysis, he op imal k alue and o he close
k alues a e chosen o pe o m se e al clus e analyses. Then, he clus e
analysis esul s a e isually alida ed using he P incipal Componen
Analysis (PCA) sca e plo . The esul s co esponding o obse ed
meaning ul pa e ns in he plo a e selec ed o ex e nal alida ion.
Human judgmen is ecommended o e alua e he use ulness o he
clus e s ob ained ega ding he p oblem in ended o be sol ed [46],
([55], p. 105), ([56], p. 134).
To make su e ha he clus e analysis esul s a e aligned wi h he aim
o he me hod and he cha ac e is ics o he case s udy ci y, a quali a i e
assessmen o he esul s is sugges ed as ex e nal alida ion. The quali-
a i e assessmen is done by e alua ing in de ail, by expe judgmen ,
he i ness o he clus e esul s wi hin he e e ence neighbou hoods o
he case s udy ci y. These e e ence neighbou hoods a e ypi ied as hey
sha e homogeneous cha ac e is ics because hey we e buil unde he
same de elopmen p ojec , u ban plan o his o ical pe iod. The blocks
wi hin each e e ence neighbou hood a e expec ed o be ca ego ised in
he same clus e . To conduc his ex e nal clus e alida ion, p e ious
knowledge o he case s udy ci y is needed. Addi ionally, his o ical
li e a u e ha could aid he selec ion o he e e ence neighbou hoods
can be used. The aim o his alida ion is o assess he obus ness o he
esul s, by checking ha he ca ego ies a e no in luenced by anecdo al
a iabili ies, hus, he clus e s a e no o e - i ed. By e alua ing he i o
he clus e s among hese e e ence neighbou hoods, he mos app o-
p ia e k alue is selec ed.
3. The case s udy o bilbao
The u ban ca ego isa ion me hodology was es ed in Bilbao’s case
s udy (no he n Spain, 43.25◦N, 2.93◦W). The s udy a ea includes all
u ban neighbou hoods wi hin he municipali y o Bilbao. Bilbao co e s
an a ea o 40.59 km
2
and has 342,484 inhabi an s, making a popula ion
densi y o 8437.64 indi iduals/km
2
. Heal h isks caused by hea wa es
in Bilbao a e p ojec ed o inc ease by 7–12 % be ween 2011 and 2040 a
by 16–25 % be ween 2071 and 2100 (RCP 4.5) [57], unde sco ing he
need o e ec i e hea adap a ion ac ion in he ci y.
The me hod was applied wi h he aid o Ma lab and QGIS so wa e.
Ma lab is a p og amming and nume ical calcula ion so wa e, and in his
wo k, i was used o da a analysis and da a isualisa ion. QGIS is an
open sou ce GIS, and in his wo k, i was used o da a ga he ing, p o-
cessing and isualisa ion.
The laye s used o de ine he u ban blocks we e he oad axis and
i e basin bounda y p o ided by he egional go e nmen and he
municipali y, as well as he municipal bounda y a ailable a he
cadas e. Using he so wa e QGIS, he polygons laid in u al neigh-
bou hoods we e emo ed, as well as hose lying in he i e basin and
unde de eloped u ban a eas (i.e. Zo o zau e island). The u al
neighbou hoods we e iden i ied acco ding he land zoning o he local
Gene al U ban Planning Plan.
Finally, na ow blocks wi h an a ea- o-pe ime e a io smalle han
i e we e au oma ically emo ed. These ins ances we e mainly small
esidual su aces de i ed om he in e sec ion o oads and we e
gene a ed because oad axes we e used o de ine he u ban blocs. These
ins ances we e no ele an o he pu pose o he me hod bu we e la ge
in numbe , which could ha e al e ed he clus e ing esul s. By his
il e ing, 879 ins ances we e emo ed, emaining 1373. The included
and excluded ins ances a e shown in Fig. 2.
A e wa ds, he da a ha eeds he indica o s was impo ed in o
QGIS o p ocessing. In Table 1, indica o s, da a sou ce, dimension
(adap i e capaci y o imp o emen po en ial) and e e ences a e sum-
ma ised. Fo he case o Bilbao, no oad su ace da a was openly a ail-
able o eed he RSR indica o , so ins ead, i was ob ained by mul iplying
he s ee axis by 3 m ( he minimum lane wid h).
Once he indica o s we e calcula ed o each u ban block, he case
s udy da abase was c ea ed. This da abase was cleaned, he alues
no malised and checked o co ela ions. The co ela ion analysis
esul ed in he exclusion o he a iable SR, due o i s signi ican co -
ela ion wi h OSR, SVF and OSA (Fig. 3).The a iables OSR and OSA,
and OSA and SVF showed a high posi i e co ela ion oo, bu o a lesse
ex en han SR. A e e alua ing he esul s o he clus e analysis wi h
Fig. 2. Included and excluded u ban blocks.
A. Villa e de e al.
Building and En i onmen 263 (2024) 111861
6
hese a iables included, he decision o keep hem was aken, since each
o hem ep esen s a di e en ele an cha ac e is ic o he adap a ion
po en ial o open spaces.
Quali a i e and quan i a i e me hods we e combined o selec he
op imal k alue. As a quan i a i e me hod, a gap analysis was con-
duc ed. The gap analysis was conduc ed using he Ma lab e alua ion
objec “GapE alua ion”[62] and he sea ch me hod “ i s MaxSE”. The
smalles k alue sa is ying he ollowing c i e ia (eq. (2)) is sugges ed as
he op imal k alue:
Gap(k)≥Gap(k+1)−SE(k+1) (2)
Whe e Gap(k) ep esen s he gap alue o he clus e ing solu ion o k,
and SE(k+1) is he s anda d e o o he clus e ing solu ion o k+1.
Following he gap analysis, he gap sugges ed ha he op imal k alue
was 7 (Fig. 4). Taking k=7 as app oxima e, a clus e analysis o k=5,
k=6, k=7, k=8 and k=9 was conduc ed o la e assessing quan-
i a i ely he i ness o he esul s by isual inspec ion o he PCA and
expe alida ion. The isual inspec ion o he PCA sca e plo s showed
a sa is ac o y i ing c i e ion o all k alues (Fig. 5). Fo he expe
alida ion, he e e ence neighbou hoods we e chosen based on he
au ho s’knowledge on he ci y and li e a u e abou he u ban de el-
opmen o Bilbao [63–66]. The chosen neighbou hoods we e checked
on-si e o con i m ha he blocks’cha ac e is ics wi hin each neigh-
bou hood we e uni o m. The neighbou hoods o he old qua e ,
mode nis housing de elopmen s om he 40’s and 60’s, low densi y
housing de elopmen s o ea ly XX
h
cen u y and u ban de elopmen s o
he XXI
s
cen u y we e chosen in he case o Bilbao (Fig. 6).
The k alues o 8 and 9 we e he i s disca ded, as hey ailed o
cap u e he homogenei y o he e e ence neighbou hoods “Medie al
old qua e ”,“G upo San u xu”and “San Inazio”. Be ween he emain-
ing k alues, k=6 was chosen as he inal, as i succeeded in cap u ing
bes he homogenei y o he e e ence neighbou hoods, e en “San Ina-
zio”and “O xa koaga.”The dis ibu ion o he clus e s o k=5, k =7, k
=8, and k=9 and he loca ion o he e e ence neighbou hoods a e
shown in Appendix B. The dis ibu ion o he clus e s o k=6 a e shown
in Fig. 7.
4. Resul s and discussion
4.1. Desc ip ion o he ca ego ies
The applica ion o he me hod in he case s udy ci y esul ed in a
ca ego isa ion o he u ban blocks o Bilbao in 6 ca ego ies. These ca -
ego ies a e dis ibu ed in he ci y as shown in Fig. 7. A c i ical inspec ion
o he s a is ical da a and he a che ype blocks (Fig. 8) allowed a
desc ip ion o he ca ego ies ega ding hei adap i e capaci y and
imp o emen po en ial. Each ca ego y was labelled wi h a desc ip i e
name o easie iden i ica ion.
Ca ego y 1, labelled “old high densi y”, can be ound in he whole
ci y, wi h a highe concen a ion in he olde a eas. I ep esen s
12.68 % o he u ban su ace included in he analysis. This ca ego y
is cha ac e ised by dense u ban blocks wi h high building densi y.
The adap i e capaci y o he blocks in his ca ego y is limi ed due o
hei low OSR and low OSA. The limi ed space educes he possibili y
o accommoda ing o ex ending solu ions. Mo eo e , he educed
SVF complica es he implemen a ion o adap a ion solu ions such as
Table 1
Indica o s used in he case s udy.
Indica o Dimension Da a sou ce Re e ences
T ee canopy co e
a io (TCC)
Imp o emen
po en ial
Digi al Su ace Model
o Vege a ion (DSMV).
Spanish Na ional
Cen e o Geog aphic
In o ma ion
[11,17,18,
21,58,59]
Open space
ampli ude
(OSA)
Adap i e capaci y Bizkaia Cadas e
( egional cadas e)
Sugges ed
by he
au ho s
Road su ace a io
(RSR)
Imp o emen
po en ial and
adap i e capaci y
Basque Go e nmen Sugges ed
by he
au ho s
Open space a io
(OSR)
Adap i e capaci y Bizkaia Cadas e
( egional cadas e)
[7,18,31]
Albedo (A) Imp o emen
po en ial
Landsa [6,60]
Sky iew ac o
(SVF)
Adap i e capaci y Digi al Su ace Model.
Spanish Na ional
Cen e o Geog aphic
In o ma ion
[18,31,51]
Impe iousness
(I)
Imp o emen
po en ial
Cope nicus [6,21,52]
Sola adia ion
(SR)
Imp o emen
po en ial
Digi al Su ace Model.
Spanish Na ional
Cen e o Geog aphic
In o ma ion
[51,61]
Fig. 3. alues o he co ela ion analysis conduc ed. The p- alue is p <0.005
o e e y a iable pai .
Fig. 4. Op imal k- alue esul ing om he gap analysis.
A. Villa e de e al.
Building and En i onmen 263 (2024) 111861
7
ee plan ing o cool pa emen s, as he e ec i eness o hese solu-
ions would be hinde ed by he p opo ions o he s ee canyons [18,
31,51]. On he con a y, he s ee s in Ca ego y 1 a e sui able o
implemen ing ex ile sola p o ec ion in he s ee o p o ide shade,
simila o hose used in sou he n Spain [27]. An addi ional limi a ion
o his block ca ego y is ha a signi ican pa o he ins ances lay in
his o ic u ban a eas, which could hinde d ama ic al e a ions o he
exis ing u ban a eas, limi ing he adap i e capaci y. Rega ding
imp o emen po en ial, he high oad densi y a io indica es a g ea
imp o emen po en ial h ough pedes ianisa ion wo ks. The TCC
Fig. 5. Sca e plo s o he wo i s p incipal componen s o clus e analysis o k =5, k =6, k =7, k =8 and k =9.
Fig. 6. Compa ison o he e e ence neighbou hoods a each k clus e analysis. The colou s ep esen he clus e alue o he u ban block on each k =n clus e
analysis. (Fo in e p e a ion o he e e ences o colou in his igu e legend, he eade is e e ed o he Web e sion o his a icle.)
A. Villa e de e al.
Building and En i onmen 263 (2024) 111861
8
a io is low, and he impe iousness is high. The e o e, he e is a high
po en ial o imp o emen in hese wo a eas. The isual inspec ion
o he a che ype block shows he p esence o block cou ya ds, which
a e po en ial a eas o accommoda ing adap a ion solu ions.
Ca ego y 2, labelled “new high densi y”, is mainly dis ibu ed wi hin
he ci y cen e and su ounding neighbou hoods. I ep esen s 12.64
% o he u ban su ace included in he analysis. These u ban blocks
p esen simila alues o Ca ego y 1 bu ha e a lowe RSR, lowe
albedo alues, and sligh ly highe OSR alues. In he ci y cen e, he
blocks o hese clus e s a e concen a ed a ound he s ee “G an Vía
de Don Diego L´
opez de Ha o”, a b oad a enue co e ed wi h ees. In
he pe iphe al neighbou hoods, he u ban blocks o his ca ego y a e
loca ed in highly dense neighbou hoods, gene ally om newe de-
elopmen s han Ca ego y 1. The lowe albedo alues a e indica i e
o he use o da ke oo ma e ials in hese new buildings. Thei low
OSR and OSA hinde he adap i e capaci y o he blocks in his
ca ego y. Howe e , ega ding he imp o emen po en ial, he lowe
albedo alues indica e an imp o emen po en ial by inc easing su -
ace albedo. The isual inspec ion o he a che ype block shows
limi ed open spaces wi h na ow p opo ions and la da k oo s.
These oo s, e en i hey a e no pa o he open space ne wo k,
ep esen an oppo uni y o implemen ing adap a ion solu ions ha
may ha e an impac in he mic oclima e.
Ca ego y 3, labelled “medium densi y”, is p ima ily ound in he
pe iphe al neighbou hoods o he ci y, wi h some u ban blocks in he
ci y cen e nea squa es o wide a enues. I ep esen s 18.76 % o he
u ban su ace included in he analysis. I is cha ac e ised by ha ing
medium OSR alues and low OSA alues. Rega ding he adap i e
capaci y, he medium OSR alues indica ed a medium adap i e ca-
paci y ha he na owness o he open spaces could hinde .
Rega ding he imp o emen po en ial, he high impe iousness and
low TCC a io indica e a po en ial o imp o emen by inc easing
pe ious su aces and ege a ion. The isual inspec ion o he
a che ype block shows na ow open spaces wi h some b oade e-
sidual spaces. These spaces a e gene ally impe ious and p esen a
low ee densi y, indica ing ha hey a e a eas o po en ial
imp o emen .
Ca ego y 4, labelled “low densi y,”is mainly ound in he pe iphe al
neighbou hoods o he ci y. I ep esen s 16.35 % o he u ban su -
ace included in he analysis. I is cha ac e ised by a high OSR, me-
dium OSA and a ela i ely high TCC a io. Rega ding adap i e
capaci y, he OSR, OSA, and SVF le els indica e ha his ca ego y
can accommoda e e ec i e solu ions. E en i his ca ego y has one o
he highes TCC a ios, i is below 0.3, indica ing imp o emen po-
en ial by ee plan ing. The isual inspec ion o he a che ype block
shows he p esence o la ge pa king lo s, which a e po en ial a eas
o accommoda ing adap a ion solu ions.
Ca ego y 5, labelled “g ey open spaces”, is spa sely dis ibu ed
wi hin he ci y, wi h a sligh concen a ion a ound he i e sides. I
ep esen s 6.54 % o he u ban su ace included in he analysis. This
ca ego y is cha ac e ised by ha ing a e y high OSR, OSA and
impe iousness. The u ban blocks wi hin his ca ego y a e gene ally
esidual spaces be ween oads o la ge open spaces wi h low ege-
a ion co e and high impe iousness. The adap i e capaci y o he
blocks in his ca ego y is high due o hei OSR and OSA le els. TCC is
low, e en lowe han highly buil -up clus e s. This ai deno es an
Fig. 7. U ban block ca ego y dis ibu ion wi hin he ci y o Bilbao.
A. Villa e de e al.
Building and En i onmen 263 (2024) 111861
9
excellen imp o emen po en ial h ough inc easing pe ious su -
aces and ee co e . The high RSR also shows imp o emen po en-
ial by educing oad su ace o accommoda e adap a ion solu ions.
The isual inspec ion o he a che ype block shows a esidual open
space be ween oads ha , e en i small in size, p esen s an oppo -
uni y o accommoda ing adap a ion solu ions.
Ca ego y 6, labelled “g een open spaces”, is o med by u ban pa ks
and g een a eas o he u ban bounda y. I ep esen s 26.74 % o
u ban su ace. I p esen s he ypical alues o g een spaces, such as
high OSR, low impe iousness o low RSR. These ai s make he
ca ego y sui able o hos adap a ion solu ions, showing g ea adap-
i e capaci y. Rega ding imp o emen po en ial, e en hough he
TCC is high compa ed o o he clus e s, i is s ill low, wi h a median
alue o 0.25, deno ing ha he e is oom o imp o emen . The
isual inspec ion o he a che ype block con i ms his ac , showing
ha mos o he block’s ees a e medium o small and lack ex ensi e
clus e ed ee a eas.
Looking a he alue anges o he indica o s (Fig. 8), some collin-
ea i y is obse ed be ween OSR, OSA, and SVF and be ween
impe iousness and TCC a io. E en i his could indica e edundancy
among he indica o s used o he ca ego isa ion, he median alues and
he alue anges ha e di e en ends. This ac highligh s ha he in-
dica o s a e o e ing addi ional in o ma ion. I hese indica o s we e
excluded, hese nuances be ween he ca ego ies would ha e been los .
Fo ins ance, clus e s 4 and 6 ha e simila median alues o TCC a io,
bu clus e 6 has lowe impe iousness han clus e 4. This implies ha
clus e 6 has a highe p opo ion o g ass su ace unco e ed by ees
han 4, indica ing highe imp o emen capaci y and majo impac o
po en ial adap a ion ac ions.
An o e iew o he ca ego ies shows signi ican dissimila i ies wi hin
he ci y ega ding he adap i e capaci y o he open spaces. Ca ego ies 1
(“old high densi y”), 2 (“new high densi y”), and 3 (“medium densi y”),
mainly due o hei low OSR, e eal limi ed space o accommoda e
adap a ion solu ions in hei open spaces. On he con a y, ca ego ies 4
(“low densi y”), 5 (“g ey open spaces”), and 6 (“g een open spaces”)
display a high adap i e capaci y. Rega ding imp o emen po en ial,
mos ca ego ies, e en hose wi h a highe TCC a io, show signi ican
imp o emen po en ial. These indings a e in line wi h p e ious s udies
on he ci y o Bilbao. Fo ins ance, Fe n´
andez de Manuel e al. [67]
Fig. 8. (a) No malised a iables o each u ban block a che ype. (b) Ae ial image o he a che ype blocks (sou ce google ea h). (c) Boxplo o a iables. The boxes
ex end om he i s qua il o he hi d qua il o he da a. The whiske s ep esen he lowe and uppe limi s o he non-ou lie alues. The do s ep esen
ou lie alues.
A. Villa e de e al.
