Tempo al Analysis o Land Su ace
Tempe a u e o U ban Hea Island
De ec ion in Thimphu, Bhu an
Yeshey Samd up
I
TEMPORAL ANALYSIS OF LAND SURFACE
TEMPERATURE FOR URBAN HEAT ISLAND
DETECTION
IN THIMPHU, BHUTAN
Disse a ion Supe ised by
Ma co Oc á io T indade Painho, P o esso , NOVA
In o ma ion Managemen School, Uni e sidade No a de
Lisboa
Disse a ion Co-supe ised by
Vicen e de Aze edo Tang, Resea ch Assis an P o esso ,
NOVA In o ma ion Managemen School,
Uni e sidade No a de Lisboa
. Michael Gould Ca lson, P o esso , Depa men o
Compu e Languages and Sys ems, Uni e si a Jaume I,
Spain
Feb ua y 2024
II
ACKNOWLEDGEMENT
I am ex emely pleased o con ey my hea el app ecia ion o he gene ous indi iduals
who ha e p o ided hei assis ance and suppo . My jou ney h ough he mas e ’s
p og am has been a ema kable lea ning expe ience, and he comple ion o his
disse a ion would no ha e been possible wi hou hei con ibu ions.
Fo emos , I am deeply g a e ul o my supe iso and co-supe iso s: P o . D . Ma co
Painho, P o esso , Vicen e de Aze edo Tang, and Michael Gould Ca lson, P o esso
o hei in aluable sugges ions and eedback.
I am p o oundly hank ul o he mul idisciplina y knowledge and skills bes owed upon
me by all my p o esso s a NOVA In o ma ion Managemen School, Uni e sidade
No a de Lisboa and i gi, Müns e , Ge many, con ibu ing o my p o essional
de elopmen . I would also like o exp ess my deepes g a i ude o my dea iend M .
Nyi Nyi, M . Kamal, M . Geb ial, M Sunil and all my cou se ma es o Geo Spa ial
Technologies. Th oughou his jou ney, hei unwa e ing suppo and encou agemen
ha e been a cons an sou ce o s eng h and inspi a ion.
I exp ess my deep g a i ude o he Eu opean Commission o g an ing me he
schola ship ha enabled me o pu sue my s udies in Eu ope.
Special hanks o he adminis a i e s a a NOVA In o ma ion Managemen School,
Uni e sidade No a de Lisboa o hei suppo ende ed h oughou my p og am. I am
g a e ul o all my iends o he wonde ul memo ies we sha ed.
A hea el app ecia ion goes o my pa en s, my dea wi e and my daugh e s o hei
inspi a ion, uncondi ional lo e, and unwa e ing suppo .
III
TEMPORAL ANALYSIS OF LST
FOR URBAN HEAT ISLAND DETECTION
IN THIMPHU, BHUTAN
ABSTRACT
Clima e change is a ma e o conside able global impo ance, as e idenced by he
inc eased u ban su ace empe a u es in de eloped and unde eloped a eas. Hence, his
s udy aims o analyze he h eshold and index o he u ban hea island (UHI)
phenomenon wi hin he u ban egion o Thimphu, Bhu an. This s udy conduc s a
empo al analysis o he No malized Di e ence Vege a ion Index (NDVI), No malized
Di e en Buildup Index (NDBI) and Land Su ace Tempe a u e (LST) o de ec U ban
Hea Islands (UHIs) in Thimphu, Bhu an. The esea ch ocuses on assessing changes in
ege a ion co e and su ace empe a u es o e ime o iden i y and unde s and he
pa e ns o UHI de elopmen in u ban a eas. This has po en ial bene i s o u ban
planne s in ol ed in u ban planning and managemen , as well as o inc easing public
awa eness ega ding he u ban hea island e ec .
Ad ancemen s in he mal emo e sensing, GIS, and s a is ical echniques ha e
acili a ed he moni o ing o LST and i s associa ion wi h Land Use and Land Co e
(LULC). To in es iga e his connec ion, supe ised classi ica ion, and change de ec ion
we e conduc ed o asce ain he spa ial ends in LULC changes. Subsequen ly, he
spa ial dis ibu ion o LST was acqui ed using he he mal band o Landsa image y.
Reg ession analysis was hen applied o in es iga e he link be ween su ace
empe a u e and a ious land su ace cha ac e is ics, including bo h ypes o land use
and land co e , as well as associa ed indices.
In summa y, his analysis con ibu es o he b oade unde s anding o u baniza ion
impac s on he local clima e and p o ides aluable insigh s o sus ainable u ban
planning and clima e mi iga ion s a egies in Thimphu, Bhu an.
IV
KEYWORDS
1. Land use land co e .
2. Land Su ace Tempe a u e
3. Landsa
4. Reg ession analysis
5. The mal Remo e Sensing
6. U ban G ow h
7. U ban Hea Island

V
ACRONYMS
1. GIS Geog aphic In o ma ion Sys em
2. LST Land Su ace Tempe a u e
3. LULC Land use land co e
4. NDBI No malized Di e ence Buil -up Index
5. NDVI No malized Di e ence Vege a ion Index
6. NDWI No malized Di e ence Wa e Index
7. UHI U ban Hea Island
8. WGS 84 Wo ld Geode ic Sys em 84
9. GEO SAM Geo Segmen Any hing Model
VI
INDEX OF THE TEXT
ABSTRACT .................................................................................. III
KEYWORDS ................................................................................ IV
ACRONYMS ................................................................................. V
INDEX OF FIGURES ................................................................... IX
INTRODUCTION.......................................................................... 1
1.1. BACKGROUND AND MOTIVATION ............................................................... 1
1.2. AIM & OBJECTIVES ...................................................................................... 5
1.3 RESEARCH QUESTIONS ................................................................................ 5
1.4. RESEARCH WORKFLOW .............................................................................. 6
2. LITERATURE REVIEW ........................................................... 7
2.1 UHI .................................................................................................................... 7
2.2 SURFACE UHI ................................................................................................ 10
3 DATA & STUDY AREA ........................................................... 11
3.1 STUDY AREA ................................................................................................. 11
3.2 DATA .............................................................................................................. 12
3.3 SOFTWARE & TOOLS ................................................................................... 13
3.4 DATA PREPARATION ................................................................................... 14
4. RESEARCH METHODS .......................................................... 14
4.1 SUPERVISED MAXIMUM LIKELIHOOD CLASSIFICATION ......................... 14
4.2 ACCURACY ASSESSMENT .......................................................................... 15
4.3 LAND SURFACE TEMPERATURE RETRIEVAL ........................................... 16
4.4 LAND USE LAND COVER INDICES .............................................................. 18
4.5 URBAN HEAT ISLAND ANALYSIS ............................................................... 19
5. RESULT AND DISCUSSION ................................................... 20
5.1 SPATIAL PATTERN OF LST AND LULC INDICES ....................................... 20
5.2 RELATIONSHIP BETWEEN LST, NDVI AND NDBI ....................................... 22
5.3 LAND COVER CLASSIFICATION (2000, 2013, AND 2020) .......................... 25
5.4 LAND SURFACE TEMPERATURE DISTRIBUTION AND LAND COVER
CLASS .................................................................................................................. 27
6. CONCLUSIONS .................................................................. 34
VII
BIBLIOGRAPHIC REFERENCES .............................................. 36
VIII
INDEX OF TABLES
Table 1: De ails o Images used. .......................................................................................................... 12
Table 2: Tools used o he analysis. ................................................................................................... 13
Table 3: LULC Accu acy Assessmen (Con usion Ma ix) ................................................................... 15
Table 4: Pa ame e s in LST Re ie al ................................................................................................... 17
Table 5: LST A ea changes in Thimphu Ci y o 2000, 2013, and 2020 .......................................... 20
Table 6: The changes in a ea o NDVI densi y ...................................................................................... 24
Table 7: The changes in a ea o NDBI ................................................................................................ 25
Table 8: The LULC Classi ica ion 2000, 2013, and 2020 .................................................................. 26
Table 9:LST dis ibu ion o e LULC and A ea in KM2 .................................................................. 27
Table 10: UHI h eshold o 2000,2013 and 2020 .................................................................................. 31
Table 11: UHI phenomenon o 2000 based on LST ............................................................................... 32
Table 12: UHI phenomenon o 2013 based on LST ............................................................................... 33
Table 13: UHI phenomenon o 2020 based on LST ........................................................................... 33
6
1.4. RESEARCH WORK FLOW
A
UHI Index
Backg ound, Mo i a ion,
AIMS & OBJECTIVES,
& Resea ch Ques ions
Iden i ying S udy A ea &
Da a Collec ion
Landsa
2000, 2013
2020
Band
a ios
Image
Classi ica ion
LST Re ie al
Accu acy
Assessme
n
LULC Maps
Change De ec ion
LULC indices
Reg ession Analysis
LST Maps
UHI Map
UHI
Th eshold
S ep 1
S ep 2
S ep 3
S ep 4
B
C
N
Y
RESULT, DISCUSSION & CONCLUSION
Figu e 1: Resea ch wo k low

7
Figu e 1 illus a es he o e all amewo k o he hesis, ou lining he in e connec edness
among i s a ious chap e s. S ep 1 in oduces he esea ch backg ound, mo i a ion,
objec i es, and esea ch ques ions. S ep 2 p o ides a concise o e iew o he da a,
so wa e, and he s udy a ea. In S ep 3, a comp ehensi e discussion is p esen ed on he
de ailed me hodology employed in he esea ch. The indings and co esponding
discussions a e ou lined and inally conclusion summa izing he esea ch
accomplishmen s, acknowledging limi a ions, and sugges ing a enues o u u e wo k
in S ep 4.
2. LETERATURE REVIEW
2.1 UHI
Resea ch on u ban clima e has been conduc ed globally, ocusing on a ious scales such
as egional, mic o, local, and s ee le els. This explo a ion has become c ucial due o
he escala ing isk and ulne abili y o u ban a eas o clima e change. Among he
a ious ace s o u ban clima e, he UHI phenomenon s ands ou as he mos ex ensi ely
s udied ield (Rasul e . al. 2017).
The apid pace o u baniza ion and he absence o clima e-sensi i e u ban planning ha e
exposed people o he ad e se impac s o clima e change. The ecen 6 h Assessmen
Repo om he IPCC emphasizes ha he pe sis en h ea s o UHI con ibu e o
inc easing ulne abili y among u ban popula ions (IPCC, 2023). As u ban a eas expand
h ough he cons uc ion o buildings, oads, and socio-economic s uc u es, he ene gy
balance wi hin hese a eas unde goes signi ican al e a ions (Hong, Je-Woo, Hong,
Jinkyu, Kwon, Eilhann E, Yoon, DongKeun, 2019). The UHI e ec is in luenced by
he he mal en i onmen in u ban a eas, leading o empe a u e di e ences o up o 10
deg ees Celsius be ween u ban and u al a eas (San amou is, M, 2001).
The inaugu al in es iga ion in o UHI occu ed in 1818 when Luke Howa d conduc ed
a s udy in London ci y. Howa d's daily empe a u e measu emen s ac oss he s udy a ea
e ealed ha he empe a u e di e ence be ween he ci y and i s su ounding u al a eas
was mo e p onounced in win e . This di e ence was a ibu ed o he hea gene a ed
8
om buildings and he lack o g een co e (Heisle , Go don M, B azel, An hony J,
2010). Subsequen ly, nume ous schola s ha e ex ensi ely explo ed he UHI
phenomenon, es ablishing i s close associa ion wi h an h opogenic hea , su ace
cha ac e is ics and s uc u e, ege a ion co e , popula ion densi y, and me eo ological
condi ions (Feng, Rundong, Wang, Fuyuan, Wang, Kaiyong, Li, Li, 2021).
A UHI s udy in Vancou e , Canada, conduc ed by Ho, Hung Chak, Knudby, Ande s,
Walke , Blake By on, and Hende son, Sa ah B (2017), e ealed a su ace empe a u e
di e ence o 9-12 K be ween u ban and u al a eas. This dispa i y was p ima ily
a ibu ed o he highe impe ious su ace p esence in he u ban a ea.
The e a e ou dis inc ypes o U ban Hea Island (UHI): Su ace UHI (SUHI/UHIsu ),
Canopy-le el UHI (CUHI/UHIucl), Bounda y Laye UHI (UHIubl), and Subs a e UHI
(UHIsub). Among hese, SUHI and CUHI a e he mos ex ensi ely s udied UHI ypes
(Wang, Jing and Zhou, Weiqi and Zhao, Wenhui 2023). UHIubl co esponds o he ai
empe a u e abo e building heigh s, while UHIsub co esponds o he soil empe a u e
below he g ound su ace. SUHI is based on he empe a u e o u ban su aces such as
he g ound, walls, and oo ops. On he o he hand, CUHI is based on nea -su ace ai
empe a u e below he building oo heigh (Ghosh, Sukanya and Kuma , Deepak and
Kuma i, Rina 2022). The ex en o CUHI is closely ied o u ban su ace p ope ies,
including he he mal cha ac e is ics o he u ban ab ic, he p esence o ege a ion, sky
iew ac o , and hea gene a ed om buildings(Aze edo, Juliana An unes and
Chapman, Lee and Mulle , Ca he ine L 2016). Figu e 1 p o ides a isual ep esen a ion
o he a ious ypes o UHI.
9
Figu e 2: UHI ypes (Sou ce: Au ho s, wi h modi ica ions based on Oke, 2017 (Oke, Mills,
Ch is en, Voog , 2017))
The ypical way o iden i y Su ace U ban Hea Island (SUHI) is h ough emo e
sensing, while Canopy-le el U ban Hea Island (CUHI) is usually de e mined using
di ec measu emen s om ield wea he s a ions o a e se measu emen s. S udies on
CUHI, conduc ed wi h bo h ixed and mobile s a ions, ha e indica ed ha i s in ensi y
is highe in u al a eas compa ed o u ban a eas, pa icula ly du ing calm nigh s wi h
clea skies(Ramak eshnan al- 2018 n.d.). SUHI is p ima ily examined using sa elli e
pla o ms o measu e Land Su ace Tempe a u e (LST), which exhibi s signi ican
spa ial a ia ion wi h highe magni udes du ing day ime(Sismanidis, Panagio is and
Bech el 2022). Remo e sensing o e s spa ial ex en s o SUHI a ci y o egional scales
(Zhou, Decheng and Xiao Al 2018) using image y om sa elli es like Landsa , Sen inel,
MODIS, e c. Landsa 's he mal image y senso s, wi h a spa ial esolu ion o 30m pixels,
a e pa icula ly well-sui ed o s udying he day ime u ban he mal en i onmen (Chen,
Yang and Yang Al 2022 n.d.).
10
2.2 SURFACE UHI
In he ealm o u ban clima e s udies, he mal in a ed (TIR) emo e sensing echniques
ha e been p edominan ly u ilized o analyzing Land Su ace Tempe a u e (LST) and
i s luc ua ions ac oss di e en su ace ypes, e alua ing U ban Hea Island (UHI), and
unde sco ing LST as a c ucial indica o o Su ace U ban Hea Island (SUHI) (Bech el
e al., 2019; Fe ei a and Dua e, 2019; Weng, 2009). Voog and Oke (2003)
concen a ed on he undamen al p inciples o he mal emo e sensing and concluded
ha LST holds he po en ial o be a signi ican pa ame e o u ban clima e s udies.
Beyond u ban clima e esea ch, LST inds applica ions in ag icul u e, disas e isk
s udies, and UHI in es iga ions (U.S. Geological Su ey, 2023).
LST is de ined as he adia i e skin empe a u e o he land su ace, measu ed in Kel in
by senso s, and is ypically associa ed wi h LULC cha ac e is ics (Weng, 2009; Bokaie
e al., 2016), including he composi ion o ege a ion, wa e , and buil -up pa e ns (Tian
e al., 2019; Zeng e al., 2015). Chen e al. (2006) explo ed he UHI in ensi y wi h
empo al and spa ial a ia ions in LST in he Pea l Ri e Del a in Guangdong P o ince,
sou he n China. Thei indings indica ed a educed UHI a ibu ed o ege a ion
in ensi y, mois u e con en , and ege a ion co e , in con as o highe UHI obse ed in
buil -up a eas.
11
3 DATA & STUDY AREA
3.1 STUDY AREA
Thimphu, he capi al o Bhu an is loca ed in he No hwes a an al i ude o 2320 me e s
abo e sea le el. The u ban a ea o Thimphu lies in he ci y su ounded by o es s. Ri e
Wang Chhu lows h ough he Thimphu ci y. The i e o igina es in he no h om
snow and glacie s, and i lows sou h-eas e ly h ough wes -cen al Bhu an. The s udy
a ea is shown in Figu e 1. Be i ing i s ole as he capi al o a small Himalayan coun y,
Thimphu s e ches along he banks o he Wangchu i e (“chu” means “ i e ” in he
na ional language, Dzongkha) a an a e age ele a ion o 7700 o 8000 ee wi h a e age
empe a u e in peak -1.1 and maximum o 28 C and in summe minimum is 16 C o
maximum o 30 C. The ci y is gene ally loca ed a 27 29N la i ude and 89 36E
longi ude. Mos o he ci y occupies he le side o he Wangchu, wi h he main ma ke
and some small indus ies such as woodwo king shops on he igh side.
Figu e 3: S udy A ea

12
3.2 DATA
The p ima y da a u ilized in his s udy consis s o Landsa sa elli e image y, speci ically
Landsa 7 Thema ic Mappe (TM) and Landsa 8 Ope a ional Land Image (OLI),
Table 1: De ails o Images used.
ob ained on 2000-12-19 o 2000; 2013-12-31 o 2013; and 2020-11-16 o 2020.
These Landsa da ase s a e accessible a no cos h ough he USGS po al and ha e been
p ocessed by NASA o c ea e adiome ic calib a ion and a mosphe ic co ec ion
algo i hms, esul ing in Le el-1 p oduc s (h p://ea hexplo e .usgs.go /). In o de o
ensu e a mo e accu a e compa ison o su ace empe a u e and U ban Hea Island (UHI)
e ec s wi hou cloud co e , sa elli e images om he mon h o No embe and
Decembe we e selec ed o all h ee yea s. Addi ional de ails ega ding he Landsa
image y a e p esen ed in he able below, and hei espec i e band designa ions can be
ound in he appendices sec ion.
Da a
Resolu ion
(M)
Bands
Bands
Name
Sou ce
Landsa 8 OLI_TIRS
30
1
Ul a-blue
USGS
30
2
Blue
30
3
G een
30
4
Red
30
5
NIR
30
6
SWIR 1
30
7
SWIR 2
30
10
TIRS
Landsa 7 TM
30
1
Blue
30
2
G een
30
3
Red
30
4
NIR
30
5
SWIR
30
6
The mal
30
7
SWIR
Da a on
30
Open s ee
map
(Roads, buildings & i e s)
13
Landsa images a e ex ensi ely employed in sa elli e emo e sensing due o hei b oad
applica ion in mapping and planning p ojec s, acili a ed by hei ad an ageous spa ial,
spec al, and empo al esolu ions (Wulde e al. 2008). Landsa images we e used o
classi y land use land co e classes, e ie e LST and calcula e NDVI, NDBI and
NDWI indices. Besides Landsa images, he seconda y da a used in his esea ch
di e en laye s o Thimphu ci y such as oad ne wo ks, wa e bodies and building
oo p in s which a e p epa ed by he Na ional Land Commission Sec e a ia o Bhu an.
Road ne wo k da a includes in o ma ion abou he layou , ype, and connec i i y o
oads wi hin Thimphu ci y and i can in luence UHI by a ec ing su ace albedo, hea
abso p ion and c ea ing u ban hea island along he pa ed su aces. S udying he oad
ne wo ks helps us o unde s and he ole o impe ious su aces in UHI o ma ion and
i s impac on local empe a u e pa e ns. Wa e bodies ep esen he spa ial dis ibu ion
o wa e bodies such as ponds, i e s, and lakes a ound he ci y. Wa e bodies ac as
cooling elemen s in u ban a eas. S udying wa e pa e ns can help us access hei
mode a ing e ec on empe a u e as he a ea wi h wa e bodies expe iences lowe
empe a u e by i s na u e o e apo a ion and causing cooling e ec in he a ea. Building
oo p in s ep esen s spa ial ex en s o indi idual buildings wi hin Thimphu ci y.
Building densi y and heigh can impac UHI by in luencing by abso bing and e-
emission o hea . A eas wi h high building densi y and all buildings can con ibu e o
inc ease o UHI and s udying he building oo p in s can help us iden i y po en ial a eas
wi h UHI ho spo s in he ci y.
3.3 SOFTWARE & TOOLS
Se e al so wa e ools we e employed o image p ocessing, spa ial analysis, and map
gene a ion. These ools include as show in he Table 2 below.
Table 2: Tools used o he analysis.
Sl
Tools/ So wa e
Rema ks
1
Excel and wo d
Documen a ion and epo ing
2
A cGIS P o/ QGIS
Analysis, mapping and isualiza ion
3
SankeyMATIC
S a is ical analysis and da a isualiza ion
4
GEO SAM
Segmen a ion ool o label Land o ms
14
Mos o he spa ial analyses such as change de ec ion, U ban Hea Island, de e mina ion
o LST, de i ing indices we e conduc ed using A cGIS P o and QGIS, while Geo SAM
was speci ically used o he da a p epa a ion – anno a ion o aining samples and
alida ion samples o image classi ica ion o gene a e LULC o he s udy a ea. The
Landsa images, and digi al image classi ica ion. Linea eg ession was pe o med wi h
ARCGIS P o, MS O ice packages (Wo d, Excel) we e used o documen a ion,
abula ion and g aphical ep esen a ion o he esul s.
3.4 DATA PREPARATION
The p ocess en ails ca e ully examining and unde s anding da a, ollowed by he
applica ion o emo e sensing echniques o analysis. In his speci ic s udy, Landsa 7
TM and Landsa 8 OLI/TIR images om 2000, 2013, and 2020, sou ced om he
Uni ed S a es Geological Su ey (USGS), a e u ilized. The esea ch ocuses on
Thimphu Ci y, Bhu an, and he da a p ocessing inco po a es Geog aphic In o ma ion
Sys em (GIS) me hods in conjunc ion wi h emo e sensing me hodologies. All da ase s
use he WGS 1984, UTM zone 45 N spa ial e e ence sys em. Consequen ly, any da a
no o iginally in his sys em, especially ec o laye s and o he Thimphu ci y laye s,
we e ans o med o align wi h his sys em and inally, Landsa images we e clipped o
ob ain he a ea o in e es .
4. RESEARCH METHODS
This sec ion add esses he s a egies employed o achie e he p e iously men ioned aim
and objec i es. These s a egies demons a e he p ac ical applica ions o GIS and
Remo e Sensing, speci ically in u ilizing spa ial- empo al da ase s o ackle eal-wo ld
issues, wi h a ocus on he U ban Hea Island (UHI) phenomenon in ou s udy. The
p ima y echniques employed in ou esea ch encompass supe ised maximum
likelihood classi ica ion, change de ec ion analysis, u ban hea island assessmen and
eg ession analysis.
4.1 SUPERVISED MAXIMUM LIKELIHOOD CLASSIFICATION
Supe ised Maximum Likelihood Classi ica ion was employed o ca ego izing he
s udy a ea based on land use and land co e classes. This me hod in ol ed de ining he
spec al cha ac e is ics o he classes by iden i ying aining samples, wi h c ucial inpu
15
om knowledge abou he a ea o in e es . Following he collec ion o aining samples,
he applica ion o he Maximum Likelihood Classi ica ion algo i hm acili a ed image
classi ica ion. In his algo i hm, each cell is assigned o he class wi h he highes
p obabili y, whe e he p obabili y alue is de e mined by he s a is ical dis ance using
mean alues and co a iance ma ix in o ma ion o he clus e s(O ukei and Blaschke
2010).
The classi ica ion esul included h ee land use land co e classes: Build Up,
Vege a ion and Ba e soil. In his way he inal land use land co e maps we e p oduced
o all h ee yea s 2000, 2013 and 2020 espec i ely. These maps enabled spa ial-
empo al change analysis.
4.2 ACCURACY ASSESSMENT
Accu acy assessmen is a undamen al s ep in e alua ing he eliabili y o land co e
classi ica ion me hods, such as Maximum Likelihood Classi ica ion (MLC), in emo e
sensing s udies. The p ocess in ol es compa ing he esul s ob ained om he MLC
algo i hm wi h g ound u h da a, collec ed h ough ield su eys o o he eliable
sou ces. To conduc accu acy assessmen o MLC, a ep esen a i e se o e e ence
da a is collec ed, co e ing a ious land co e ypes in he s udy a ea. Random sampling
ensu es a s a is ically signi ican selec ion o pixels o assessmen . The cons uc ion o
a con usion ma ix acili a es he calcula ion o accu acy me ics, including o e all
accu acy, p oduce 's accu acy, use 's accu acy, and he Kappa coe icien . Visual
in e p e a ion and e o analysis a e c ucial componen s o accu acy assessmen ,
p o iding quali a i e insigh s in o misclassi ica ions and guiding imp o emen s o he
classi ica ion model o pa ame e s. Accu acy assessmen ensu es ha emo e sensing-
based land co e maps accu a ely ep esen he ue condi ions on he g ound,
enhancing he eliabili y o in o ma ion o decision-making in di e se applica ions.
Table 3: LULC Accu acy Assessmen (Con usion Ma ix)
2000
2013
2020
LULC
Use Ac.
P o Ac,
Use Ac.
P o Ac,
Use Ac.
P o Ac,
Buil Up
0.93
0.78
0.73
0.81
0.88
0.83
Vege a ion
0.82
0.98
0.87
0.96
0.89
0.9
Ba e Land
0.65
0.87
0.84
0.64
0.82
0.86
O e All
0.76
0.81
0.87
Kappa
0.64
0.71
0.8
22
5.2 RELATIONSHIP BETWEEN LST, NDVI AND NDBI
Co ela ion coe icien be ween LST and NDVI was calcula ed and o 2000 is -0.41
0
10
20
30
-0.2 0 0.2 0.4 0.6
LST
NDVI
LST Vs NDVI 2000
0
10
20
30
-0.1 0 0.1 0.2 0.3 0.4
LST
NDVI
LST Vs NDVI 2013
0
10
20
30
40
-0.1 0 0.1 0.2 0.3 0.4 0.5
LST
NDVI
LST Vs NDVI 2020
0
10
20
30
-0.4 -0.2 0 0.2 0.4
LST
NDBI
LST Vs NDBI 2000
0
5
10
15
20
25
-0.3 -0.2 -0.1 0 0.1 0.2
LST
NDBI
LST Vs NDBI 2013
0
10
20
30
40
-0.3 -0.2 -0.1 0 0.1 0.2
LST
NDBI
LST Vs NDBI 2020
Figu e 4: Co ela ion LST wi h Indices

23
in 2013 is 0.05 and in 2020 is -0.10. Fo 2000, he nega i e co ela ion coe icien o -
0.41 be ween LST and NDVI in 2000 sugges s a mode a e nega i e co ela ion. This
indica es ha as ege a ion (NDVI) inc eases, he e is a mode a e endency o land
su ace empe a u e o dec ease. Nega i e co ela ions be ween LST and NDVI a e
common, as ege a ion ends o ha e a cooling e ec on he en i onmen h ough
e apo anspi a ion and shading. While o 2013, he co ela ion coe icien o 0.05
be ween LST and NDVI in 2013 indica es a e y weak posi i e co ela ion. This
sugges s a minimal associa ion be ween ege a ion and land su ace empe a u e in
2013. The posi i e co ela ion implies ha as ege a ion inc eases, he e is a weak
endency o land su ace empe a u e o inc ease. The weak posi i e co ela ion may
indica e ha o he ac o s play a mo e dominan ole in in luencing LST du ing his
yea . Finally o 2020, he nega i e co ela ion coe icien o -0.10 be ween LST and
NDVI in 2020 sugges s a e y weak nega i e co ela ion. This indica es a e y weak
endency o land su ace empe a u e o dec ease as ege a ion inc eases in 2020. The
weak nega i e co ela ion implies ha he in luence o ege a ion on land su ace
empe a u e is minimal du ing his yea , and o he ac o s may ha e a mo e signi ican
impac . The o e all inding is he a ying co ela ion coe icien s sugges changes in
he ela ionship be ween land su ace empe a u e and ege a ion o e he yea s. The
nega i e co ela ions in 2000 and 2020 indica e a gene al end whe e inc eased
ege a ion is associa ed wi h lowe land su ace empe a u es. The weak o e y weak
co ela ions in 2013 sugges ha o he ac o s, such as u ban de elopmen o speci ic
local condi ions, may be in luencing he ela ionship du ing ha yea .
Rega ding LST and NDBI, he co ela ion coe icien be ween LST and NDBI in 2000
is 0.033, o 2013 is 0.49 and o 2020 is 0.54. The co ela ion coe icien o 0.033
be ween LST and NDBI in 2000 indica es an ex emely weak posi i e co ela ion. The
weak posi i e co ela ion sugges s ha he e is a minimal associa ion be ween land
su ace empe a u e and buil -up index in he yea 2000. The p ac ical signi icance o
his co ela ion may be limi ed, as changes in one a iable a e ba ely associa ed wi h
changes in he o he .
Whe eas o 2013, he co ela ion coe icien o 0.498 be ween LST and NDBI in 2013
indica es a mode a e posi i e co ela ion. This sugges s a s onge associa ion compa ed
24
o he co ela ion in 2000. Changes in land su ace empe a u e a e mode a ely
associa ed wi h changes in he buil -up index in 2013. The posi i e co ela ion implies
ha as buil -up a eas inc ease, he e is a mode a e endency o land su ace empe a u e
o inc ease. Finally o 2020, he co ela ion coe icien o 0.541 be ween LST and
NDBI in 2020 indica es a mode a e posi i e co ela ion simila o 2013. This sugges s
a consis en ela ionship be ween land su ace empe a u e and buil -up index o e ime.
The posi i e co ela ion implies ha as buil -up a eas inc ease, he e is a mode a e
endency o land su ace empe a u e o inc ease.
The o e all indings on LST and NDBI e eals ha he co ela ion coe icien s show a
a ia ion in he s eng h o he ela ionship be ween LST and NDBI o e he yea s. The
consis en posi i e co ela ion in 2013 and 2020 indica es a mode a e associa ion
be ween u ban de elopmen (as indica ed by NDBI) and land su ace empe a u e. The
ac o s con ibu ing o his ela ionship could include hea e en ion in buil -up a eas,
changes in land use pa e ns, and u ban hea island e ec s.
Table 6: The changes in a ea o NDVI densi y
Tables 5, 6, and 7 p o ide insigh s in o he a ia ions obse ed in LST, NDVI, and
NDBI o he yea s 2000, 2013, and 2020. The s udy e eals a s a is ically signi ican
mode a e posi i e co ela ion be ween LST and NDVI. The NDBI indica es ha he
expansion o buil -up a eas is posi i ely co ela ed wi h an inc ease in LST alues,
sugges ing a ise in LST o e ime. Speci ically, he LST alues o buil -up a eas
expanded om 1.95 o 28.06 °C in 2000 o 3.1 o 29.06 °C in 2013, co e ing 5 KM2
(14.29% o he o al a ea). In 2020, he LST ange will be ex ended om 9.81 o 32.99
°C, co e ing 8 KM2 (22.9% o he o al a ea).
Vege a ion
Densi y
A ea (KM2)
NDVI A ea Change
2000-2013
2000-2020
2000
%
2013
%
2020
%
∆
∆%
∆
∆%
No
Vege a ed
6
18
5
14
6
17
-1
-4
0
-1
Low
G eenness
6
17
10
29
7
20
4
11
1
3
Low G een
8
22
7
20
8
23
-1
-2
0
1
Medium
5
15
6
17
9
26
1
3
4
11
High
10
28.6
7
20
5.0
14
-3
-9
-5
-14
25
Table 7: The changes in a ea o NDBI
5.3 LAND COVER CLASSIFICATION (2000, 2013, AND 2020)
The land co e classi ica ion o Thimphu Ci y encompasses h ee dis inc ca ego ies,
namely ege a ion, buil -up, and Ba e land. A ea in squa e kilome e s, and pe cen ages
o each class a e summa ized in Table 7 and Figu e 4. Table 3 displays he esul o
LULC accu acy es o he s udy. The indings poin ou ha he Buil -Up a ea has
expe ienced signi ican g ow h o e he yea s, inc easing om 10 KM2 in 2000 o 19
KM2 in 2020. This subs an ial ise indica es u ban expansion and de elopmen ,
encompassing an expansi e a ea o 19 KM2, accoun ing o app oxima ely 54% o he
o al land a ea. While Vege a ion emains he dominan class, i s pe cen age has
dec eased om 44% in 2000 o 27% in 2020, sugges ing po en ial changes in land use,
de o es a ion, o u baniza ion a ec ing g een spaces. The pe cen age o Ba e Land has
dec eased om 27% in 2000 o 18% in 2020, indica ing a educ ion in open, non-
ege a ed a eas. The da a sugges s a dynamic shi in land co e , wi h u baniza ion
leading o inc eased Buil -Up a eas a he expense o Vege a ion and Ba e Land. This
ans o ma ion has implica ions o ecological balance, biodi e si y, and u ban
planning. Figu e 4 depic s he g aphical ep esen a ion o he each LULC class changes
om 2000 o 2020.
As he Figu e 5 depic s ha he e was educe by -3 KM2 and -6 Km2 om 2000 o
2020 om he o al a ea. Con e sely, he e is a signi ican inc ease o 9 KM2 in Build
up om 2000 o 2020 o he o al a ea.
Building
Densi y
A ea (KM2)
NDBI A ea Change
2000-2013
2013-2020
2000
%
2013
%
2020
%
∆
∆%
∆
∆%
Non Building
5
15
8
23
6.0
17
3
8
1
3
Ve y Low
5
15
7
20
7.0
20
2
5
2
5
Low
8
23
7
20
8.0
23
-1
-3
0
0
Medium
11
31
6
17
5.0
14
-5
-14
-6
-17
High
5
14
7
20.0
8.0
22.9
2
6
3
9
26
Table 8: The LULC Classi ica ion 2000, 2013, and 2020
Figu e 6 isualizes he LULC co e changes o e he yea 2000 o 2020 and he changes
is shown in KM2. As he s udy ocuses on Build up and Vege a ion o ind ou he
ela ion wi h LST o de e mine which and classes do con ibu e o UHI by inc easing
empe a u e. The diag am show inc ease in Buil up a ea om 0.1224 o 0.05 KM2 and
simila ly ege a ion oo om 5.571 o 11.24 KM2 om ba e land buil up.
Yea
LULC
%
A ea KM2
2000
Buil Up
28
10
Vege a ion
45
16
Ba e Land
27
10
2013
Buil Up
40
14
Vege a ion
36
12
Ba e Land
24
8
2020
Buil Up
54
19
Vege a ion
27
10
Ba e Land
19
6
9
-6
-3
-8 -6 -4 -2 0 2 4 6 8 10
Buil Up
Vege a ion
Ba e Land
KM2
Change in LULC 2000-2020
Buil Up
Vege a ion
Ba e Land
Figu e 5: LULC changes om 2000 o 2020 in g aph
27
5.4 LAND SURFACE TEMPERATURE DISTRIBUTION AND LAND COVER
CLASS
The empo al a ia ions in Thimphu Ci y's land su ace empe a u e (LST) ac oss he
yea s 2000, 2013, and 2020 a e e iden in he spa ial dis ibu ion. As pe sa elli e image
analysis, he LST displayed a spec um o empe a u es, anging om a ound 1.95 °C
o 28.06 °C in 2000, app oxima ely 3.10 °C o 29.0 °C in 2013, and abou 9.81 o 32.99
°C in 2020, as illus a ed in able 5. This s udy in es iga es he dis ibu ion o Land
Su ace Tempe a u e (LST) and analyzes he spa ial ex en and loca ion o empe a u e
changes wi hin each land co e class in Thimphu Ci y o he yea s 2000, 2013, and
2020. Table 9 displays he changes in LST o Thimphu Ci y in 2000, 2013 and 2020
and igu e 10 displays he LST o e h ee LULC classes.
Table 9:LST dis ibu ion o e LULC and A ea in KM2
The in o ma ion p o ided in Table 9 indica es ha in 2000, Buil -Up A eas expe ienced
an LST o 14°C, co e ing app oxima ely 29% o he o al a ea (KM2). By 2013, he
LST dec eased o 8°C, encompassing 42% (KM2) o he o al a ea. Howe e , in 2020,
LULC Class
2000
2013
2020
LST °C
% KM2
LST °C
% KM2
LST °C
% KM2
Buil Up
14
29
8
42
16
55
Vege a ion
9
45
5
35
13
27
Ba e Land
15
28
11
24
18
19
Figu e 6: LULC changes om 2000 o 2020 in de ail wi h Sankey diag am

28
he e was a no able inc ease in LST o 16°C, expanding he co e age o 55% (KM2) o
he o al a ea. The Buil -Up A eas showed a subs an ial dec ease in LST om 2000 o
2013, sugges ing a po en ial cooling e ec . Howe e , a signi ican inc ease in LST was
obse ed in 2020, indica ing possible u ban hea island e ec s o changes in land use
pa e ns. The pe cen age o he o al a ea co e ed by buil -up a eas consis en ly
inc eased, indica ing u ban expansion.
In 2000, he LST o Vege a ion A eas shows 9°C, co e ing 45% (KM2) and
ep esen ing 35% o he o al a ea and, in 2013, he LST dec eased o 5°C, co e ing
35% (KM2) o he o al a ea. Howe e , in 2020, LST signi ican ly inc eased o 13°C,
co e ing 27% (KM2) o he o al a ea. O e all, Vege a ion A eas consis en ly show a
dec ease in LST, indica i e o a cooling e ec . The pe cen age o he o al a ea co e ed
by ege a ion emains ela i ely s able, wi h a sligh dec ease.
Acco ding o (Sene i a hne e al. 2021) which use low albedo ma e ials leading o high
hea abso p ion in u ban cen e s. In addi ion, emo al o ege a ion co e and emissions
o was e hea om a ious sou ces con ibu e o he accumula ion o hea ene gy,
leading o o ma ion o u ban hea islands. O e all, Buil -Up A eas show a no able
luc ua ion in LST, possibly in luenced by u baniza ion ends. Con e sely, Vege a ion
A eas consis en ly exhibi a cooling e ec , con ibu ing o empe a u e educ ion.
Figu e 7: The esul s o LST, NDVI, NDBI, and land co e o 2000
29
Simila ly, Ba e Land A eas ollow a cooling end, e lec ing changes in land co e and
use in he s udy a ea.
Figu e 8: The esul s o LST, NDVI, NDBI, and land co e o 2020
Figu e 9: The esul s o LST, NDVI, NDBI, and land co e o 2013
30
5.5 URBAN HEAT ISLAND DETECTION
De ec ing he exis ence o he u ban hea island (UHI) in Thimphu Ci y in ol es
employing equa ions (11), (12), and (13) o de e mine he a mosphe ic condi ions.
empe a u e h eshold o UHI occu ence. Addi ionally, UHI h eshold calcula ions
acili a e he spa ial mapping o UHI phenomena dis ibu ion wi hin Thimphu Ci y.
Fu he mo e, equa ion (14) is u ilized o UHI Index analysis, enabling an assessmen
Figu e 11: UHI maps o 2000, 2013 and 2020
12
16
913
15
18
0
5
10
15
20
2000 2020
LST C°
Yea
A e age LST o LULC
Buil Up Vege a ion Ba e Land
Figu e 10: LST dis ibu ion on LULC classes
31
o he ex en o UHI phenomenon po en ial. The UHI index is ca ego ized in o h ee
dis inc classes: non-UHI, po en ial UHI, and UHI.Thimphu Ci y unc ions as a cen al
hub o se ices, playing a pi o al ole as he p ima y economic cen e o he egions.
I suppo s di e se communi y ac i i ies, encompassing ade hospi al and educa ion.
The con inuous de elopmen and g ow h o Thimphu Ci y o e ime unde sco e he
impe a i e acknowledgmen o he ine i able p esence o he u ban hea island
phenomenon in he a ea.
The examina ion o he empe a u e h eshold igge ing he U ban Hea Island (UHI)
phenomenon in Thimphu Ci y e eals ha he egion is indeed unde going he e ec s
o UHI. The iden i ied h eshold empe a u e o he UHI phenomenon in 2000 was
documen ed as 14.21%. When he land su ace empe a u e su passes his h eshold, i
indica es he mani es a ion o he u ban hea island phenomenon in he co esponding
a ea. The ou comes o he analysis pe aining o he u ban hea island phenomenon in
Thimphu Ci y o he yea 2000, 2013 and 2020 a e de ailed in Table 10. As illus a ed
in Figu e 11, he indings align wi h he de e mined h eshold empe a u e, he UHI
phenomenon exhibi s a no able concen a ion wi hin he ci y’s cen al egion,
subsequen ly ex ending o he su ounding a eas.
Table 10: UHI h eshold o 2000,2013 and 2020
The u ban hea island phenomenon’s p e alence in Thimphu Ci y is 11.43% o i s o al
a ea. Addi ionally, 45.71% o he ci y’s a ea exhibi s po en ial o expe iencing he UHI
phenomenon, while he emaining 42.86% o Thimphu Ci y’s o al a ea emains
una ec ed by he UHI phenomenon. The p oduc ion o a map illus a ing he
dis ibu ion o UHI phenomena in Thimphu Ci y in 2000 can be acili a ed by u ilizing
he UHI index, as depic ed in Figu es 11 and Figu e 12.
Yea
Tmax
Tmin
Median (μ)
S d
UHI Th eshold
2000
28.06
1.95
12.12
4.19
14.55
2013
30
3.10
7.75
3.89
14.38
2020
33
9.81
15.44
3.65
20.79
38
Moye , Ashley N., and Timo hy W. Hawkins. 2017. “Ri e E ec s on he
Hea Island o a Small U ban A ea.” U ban Clima e 21:262–77.
doi: h ps://doi.o g/10.1016/j.uclim.2017.07.004.
Mukhe jee, Am i endu, A jun Anil Kuma , and Pa hasa a hy
Ramachand an. 2021. “De elopmen o New Index-Based
Me hodology o Ex ac ion o Buil -Up A ea F om Landsa 7
Image y: Compa ison o Pe o mance Wi h SVM, ANN, and
Exis ing Indices.” IEEE T ansac ions on Geoscience and Remo e
Sensing 59(2):1592–1603. doi: 10.1109/TGRS.2020.2996777.
Na ional S a is ics Bu eau o Bhu an. 2018. “Na ional S a is ics Bu eau o
Bhu an (2018) ’2017 Popula ion & Housing Census o Bhu an. ’
Thimphu: Na ional S a is ics Bu eau, Royal Go e nmen o
Bhu an.”
Nicholas E. Young. 2017. “A Su i al Guide o Landsa P ep ocessing.”
O ukei, J. R., and T. Blaschke. 2010. “Land Co e Change Assessmen
Using Decision T ees, Suppo Vec o Machines and Maximum
Likelihood Classi ica ion Algo i hms.” In e na ional Jou nal o
Applied Ea h Obse a ion and Geoin o ma ion 12:S27–31. doi:
h ps://doi.o g/10.1016/j.jag.2009.11.002.
P. Dash, F. S. Olesen, F. M. Gö sche, and H. Fische . 2002. “Land
Su ace Tempe a u e and Emissi i y Es ima ion om Passi e
Senso Da a: Theo y and P ac ice-Cu en T ends.” In e na ional
Jou nal o Remo e Sensing 23(13):2563–94. doi:
10.1080/01431160110115041.
Pe i colin, F ançois, and E ic Ve mo e. 2002. “Land Su ace Re lec ance,
Emissi i y and Tempe a u e om MODIS Middle and The mal
In a ed Da a.” Remo e Sensing o En i onmen 83(1):112–34. doi:
h ps://doi.o g/10.1016/S0034-4257(02)00094-9.
Phelan, Pa ick E., Kamil Kaloush, Ma k Mine , Jay Golden, Be nade e
Phelan, Humbe o Sil a, and Robe A. Taylo . 2015. “U ban Hea
Island: Mechanisms, Implica ions, and Possible Remedies.” Annual
Re iew o En i onmen and Resou ces 40(1):285–307. doi:
10.1146/annu e -en i on-102014-021155.
Pu a, Bayu Ta una Widjaja, and Peeyush Soni. 2017. “E alua ing NIR-
Red and NIR-Red Edge Ex e nal Fil e s wi h Digi al Came as o
Assessing Vege a ion Indices unde Di e en Illumina ion.”

39
In a ed Physics & Technology 81:148–56. doi:
h ps://doi.o g/10.1016/j.in a ed.2017.01.007.
Radwan, Tahe M., G. Alan Blackbu n, J. Duncan Whya , and Pe e M.
A kinson. 2019. “D ama ic Loss o Ag icul u al Land Due o
U ban Expansion Th ea ens Food Secu i y in he Nile Del a,
Egyp .” Remo e Sensing 11(3). doi: 10.3390/ s11030332.
Rai, C.M., Do ji, Y. and Zangmo, S. 2022. “Use Sa is ac ion and he
Social and En i onmen al Bene i s o U ban G een Spaces: A Case
S udy o Thimphu Ci y.”
Ramak eshnan al- 2018. n.d. “A C i ical Re iew o U ban Hea Island
Phenomenon in he Con ex o G ea e Kuala Lumpu , Malaysia.”
Rich e , R and Schl. 2011. “A mosphe ic/Topog aphic Co ec ion o
Ai bo ne Image y.”
Royal Go e nmen o Bhu an. 2022. “~Thimphu Pa o S a egic
P ospec us.”
Sene i a hne, D. M., V. M. Jayasoo iya, S. M. Dassanayake, and S.
Mu hukuma an. 2021. “E ec s o Pa emen Tex u e and Colou on
U ban Hea Islands: An Expe imen al S udy in T opical Clima e.”
U ban Clima e 40:101024. doi:
h ps://doi.o g/10.1016/j.uclim.2021.101024.
Sismanidis, Panagio is and Bech el. 2022. “The Seasonali y o Su ace
U ban Hea Islands ac oss Clima es.”
Taylo , Lucy, and Die e F. Hochuli. 2015. “C ea ing Be e Ci ies: How
Biodi e si y and Ecosys em Func ioning Enhance U ban
Residen s’ Wellbeing.” U ban Ecosys ems 18(3):747–62. doi:
10.1007/s11252-014-0427-3.
Wang, Jing and Zhou, Weiqi and Zhao, Wenhui. 2023. “Highe UHI
In ensi y, Highe U ban Tempe a u e? A Syn he ical Analysis o
U ban Hea En i onmen in U ban Mega egion.”
Wang, Lamchin, and Lee. 2022. “The E olu ion o he Capi al Ci y.
Thimphu.”
Wulde , Michael A., Joanne C. Whi e, Samuel N. Gowa d, Je ey G.
Masek, James R. I ons, Ma in He old, Wa en B. Cohen, Thomas
R. Lo eland, and Cu is E. Woodcock. 2008. “Landsa Con inui y:
Issues and Oppo uni ies o Land Co e Moni o ing.” Remo e
40
Sensing o En i onmen 112(3):955–69. doi:
h ps://doi.o g/10.1016/j. se.2007.07.004.
Yang, Xinyan, Yuguo Li, Zhiwen Luo, and Pak Wai Chan. 2017. “The
U ban Cool Island Phenomenon in a High-Rise High-Densi y Ci y
and I s Mechanisms.” In e na ional Jou nal o Clima ology
37(2):890–904. doi: h ps://doi.o g/10.1002/joc.4747.
Yangka, Do ji and Newman, Pe e and Rauland, Vanessa and De e eux,
Pe e . 2018. “Sus ainabili y in an Eme ging Na ion: The Bhu an
Case S udy.”
Zhou, Decheng and Xiao Al. 2018. “Sa elli e Remo e Sensing o Su ace
U ban Hea Islands: P og ess, Challenges, and Pe spec i es.”