Permodelan Spasial Suhu Permukaan Tanah Dan Hubungannya Dengan Faktor Ekologi Di Kota Bogor

SPATIAL MODELING OF LAND SURFACE TEMPERATURE
AND ITS RELATIONSHIP WITH ECOLOGICAL FACTORS
IN BOGOR MUNICIPALITY

EMA KURNIA

GRADUATE SCHOOL
BOGOR AGRICULTURAL UNIVERSITY
BOGOR
2016

STATEMENT
I, Ema Kurnia, hereby declare that this thesis entitled
Spatial Modeling of Land Surface Temperature and its Relationship with
Ecological Factors in Bogor Municipality
Is a result of my work under the supervision of advisory board and that it has
not been published before. The content of this thesis has been examined by the
advisory board and external examiner.

Bogor, May 2016
Ema Kurnia

G051120011

RINGKASAN
EMA KURNIA. Permodelan Spasial Suhu Permukaan Tanah dan Hubungannya
dengan Faktor Ekologi di Kota Bogor. Dibimbing oleh I NENGAH SURATI JAYA
dan WIDIATMAKA.
Salah satu parameter kunci permukaan bumi melalui proses pertukaran energi
dan air dengan udara disebut sebagai Suhu Permukaan Tanah (SPT), SPT juga
memainkan peranan penting dalam berbagai penelitian ilmiah, seperti ekologi,
hidrologi, dan studi perubahan global. Struktur buatan manusia seperti jalan dan
bangunan biasanya memiliki nilai albedo yang lebih rendah dari permukaan bumi
dan menyerap lebih banyak radiasi sinar tampak. Hal ini menunjukkan bahwa
penurunan dan peningkatan jumlah vegetasi pada permukaan bumi wilayah
perkotaan akan menyebabkan efek pulau panas. Tujuan penelitian ini adalah untuk
membangun permodelan spasial suhu permukaan tanah di Kota Bogor
menggunakan citra satelit Landsat-8, dan untuk mengidentifikasi hubungannya
dengan faktor-faktor ekologi. Algoritma Split-Window (SW) digunakan untuk
membuat SPT dengan menggunakan input berupa suhu kecerahan, emisivitas
permukaan tanah (EPT) dan uap air. Validasi citra dilakukan untuk mengevaluasi
nilai SPT dihitung dari citra satelit tahun 2014 Kota Bogor dengan pengukuran suhu

lapangan yang dilakukan selama bulan Maret 2015 pada tanggal 2, 4, 7, 9, 11, 14,
16, 18, 21, 23, 25, 28 dan 30.
Validasi citra menunjukkan bahwa suhu udara memiliki hubungan dengan
SPT pada waktu siang hari tanggal 2 dan 28 Maret 2015 dengan R² = 0,6119. Jenis
tutupan lahan memiliki peranan penting terhadap pola termal panas perkotaan. Pada
lokasi penelitian, nilai-nilai SPT meningkat dari pedesaan menuju perkotaan, yang
berkisar dari 270C sampai dengan 490C, dengan rata-rata 370C. Pola SPT yang
ditemukan tidak simetris melainkan konsentris, dengan zona suhu tinggi terpusat
menuju daerah penelitian. Daerah dengan tingkat vegetasi terendah berupa jenis
tutupan lahan pemukiman / daerah terbangun dengan nilai SPT berkisar antara 400C
sampai dengan 49.50C. Sebaliknya, daerah dengan tingkat vegetasi tinggi yang
mengindikasikan keberadaan vegetasi hijau, terutama di bagian selatan daerah
penelitian, merupakan jenis tutupan lahan dari lahan pertanian dan lahan rumput
dengan nilai SPT berkisar antara 270C sampai dengan 390C. Beberapa wilayah
dengan tingkat vegetasi tinggi juga terlihat pada bagian tengah daerah penelitian
yang merupakan daerah hutan perkotaan.
Variabel-variabel yang diduga sebagai faktor ekologi yang berkaitan dengan
suhu permukaan diwakili oleh empat faktor yaitu kelembaban, ketinggian,
kepadatan jalan dan tutupan lahan. Kelembaban memiliki hubungan negatif dengan
SPT pada waktu siang hari tanggal 2 dan 25 Maret 2015 dengan R² = 0,5309.

Elevasi memiliki hubungan negatif dengan SPT dengan R2 = 0,9256. Faktor
kepadatan jalan menunjukan nilai tertinggi pada suhu permukaan sebesar 380C
berada pada jarak 100 meter pertama dari jalan. Nilai suhu permukaan mengalami
penurunan seiring dengan bertambahnya jarak dari jalan, hal ini ditunjukkan dengan
signifikan hingga jarak 300 m dari jalan dengan suhu permukaan sebesar 34.50C.
Pembuatan peta tutupan lahan dilakukan dengan menggunakan teknik klasifikasi
terbimbing, akurasi keseluruhan untuk validasi klasifikasi tutupan lahan tahun 2014
menunjukkan nilai 90,9% dan kappa koefisien sebesar 88,6%. SPT tertinggi

ditemukan pada wilayah pemukiman / terbangun (400C), diikuti oleh lahan rumput
(35.260C), lahan pertanian dan badan air (35.10C), dan suhu terendah terdeteksi
berada di hutan kota (34.10C). Penelitian ini menunjukkan bahwa SPT dan faktorfaktor ekologi di Kota Bogor memiliki hubungan yang signifikan, menyiratkan
bahwa peningkatan jumlah vegetasi umumnya akan mengurangi suhu permukaan
dan intensitas pulau panas perkotaan. Permodelan spasial menunjukan nilai
Y = 0.33X1 + 0.33X2 + 0.34X3
dimana Y adalah suhu permukaan tanah yang diduga dari variabel X1 sebagai
tutupan lahan, X2 sebagai jarak dari jalan, dan X3 sebagai ketinggian wilayah.
Penelitian ini menunjukkan fungsi yang disediakan oleh SPT sebagai sarana dalam
memperkirakan intensitas pulau panas perkotaan dan suhu musim panas yang
diperkirakan terjadi karena perubahan iklim. Diharapkan bahwa temuan yang

disajikan dalam penelitian ini dapat berguna dalam perencanaan kota dan ekologi
di Kota Bogor. Pengenalan dan kesadaran fungsi ini sangat penting karena dapat
berguna sebagai alat untuk beradaptasi terhadap perubahan iklim dan mengetahui
dampak negatif dari pembangunan perkotaan untuk meningkatkan perencanaan dan
manajemen strategi perkotaan.
Kata kunci: Suhu Permukaan Tanah, Emisivitas Permukaan Tanah, Split-Window,
faktor ekologi, tutupan lahan.

SUMMARY
EMA KURNIA. Spatial Modeling Land Surface Temperature and its Relationship
with Ecological Factors in Bogor Municipality. Supervised by I NENGAH
SURATI JAYA and WIDIATMAKA.
One of the key parameter in the physics of the earth surface through the
process of energy and water exchange with the atmosphere called as the Land
Surface Temperature (LST), it also plays an important role in a wide variety of
scientific studies, such as ecology, hydrology, and global change studies. Manmade structures such as roads and buildings typically have a lower albedo than the
natural surface and absorb more visible radiation. This shows that the reduced of
the vegetation and the increased in urban surface will cause heat island effect. The
objective of this study are to develop spatial modeling the land surface temperature
distribution in Bogor Municipality using landsat-8 satellite imagery, and to identify

its relationship with the ecological factors. Split-Window (SW) Algorithm was used
to derive the LST by using the inputs of the brightness temperature, land surface
emissivity (LSE) and water vapor. The image validation was done to evaluate the
relationships among LST calculated from a 2014 satellite image of the Bogor
Municipality and ground-based temperature measurements monitored during the
month of March 2015 at a specific date of 2, 4, 7, 9, 11, 14, 16, 18, 21, 23, 25, 28
and 30.
The image validation shows that the air temperature has a relationship with
the LST at a day time of date 2 and 28 of March 2015 by R² = 0.6119. The land
cover types has an important role in the thermal pattern of the urban heat island.
Across the entire study area, LST values increased from the outskirts towards the
inner urban areas, which ranged from 270C to 490C, with a mean of 370C. The LST
pattern were found to be non-symmetrical but rather concentric, with high
temperature zones clustered towards the center of study area. The areas with the
lowest vegetation levels were corresponded to the land cover types of
settlement/built-up area with the value of LST ranged from 400C to 49.50C.
Conversely, high levels of NDVI indicating the presence of green vegetation were
observed, primarily at the south part of the study area, and corresponded to the land
cover types of farmland and grass area with the value of LST ranged from 27 0C to
390C. Some pockets of high NDVI were also noticeable within the central portion

of the study area and corresponded to urban forest area.
The variables which are suspected as the ecological factors that related to the
surface temperature and represented by four factors which are humidity, elevation,
road density and land cover. The humidity has a negative relationship with the LST
at a day time of date 2 and 25 of March 2015 by R² = 0.5309. The elevation has a
negative relationship with the LST by R 2 = 0.9256. The road density exhibit the
highest value of surface temperature 380C was on the first 100 m of buffer distance,
the temperature had degradation along with the decrease of the distance, it has been
described significantly into 300 m with temperature of 34.50C. The land cover was
done by using the supervised classification technique, the overall accuracy for the
validation of land cover classification year 2014 showed the value of 90.9% and
kappa coefficient 88.6%. The highest LST was found in settlement/built-up (390C),

followed by grass (36.50C), farmland and waterbody (360C), and the lowest
temperature detected in urban forest (34.5 0C). This study found that within the
Bogor Municipality, LST and the ecological factors shared a significant relationship,
implying that increasing vegetation abundance will generally reduce surface
temperatures and thus urban heat island intensity. The spatial modeling exhibit
Y = 0.33X1 + 0.33X2 + 0.34X3
where Y was the LST suspected from the variables X1 as the land cover, X2

as road distance, and X3 as the elevation. This study pointed on the functionality
provided by LST as a means of estimating the urban heat island intensity and the
heightened summer temperatures within urban areas which are predicted to occur
due to climate change. It is hoped that the findings presented in this study may prove
useful to those involved in urban and ecological planning in Bogor Municipality.
Recognition and awareness of this functionality is critical as it can serve as a tool for
adapting cities to climate change and knowing some of the negative impacts of urban
development to improve urban planning and management strategies.
Keywords: Land Surface Temperature, Land Surface Emissivity, Split-Window,
ecological factors, land cover.

Copyright © 2016, Bogor Agricultural University
Copyright are protected by law
It is prohibited to cite all or part of this thesis without referring to and mentioning
the source. Citation is only permitted for the sake of education, research, scientific
writing, critical writing or reviewing scientific problems.
Citation does not inflict the name and honor of Bogor Agricultural University.
It is prohibited to republish and reproduce all part of this thesis without any written
permission from Bogor Agricultural University.

SPATIAL MODELING OF LAND SURFACE TEMPERATURE
AND ITS RELATIONSHIP WITH ECOLOGICAL FACTORS
IN BOGOR MUNICIPALITY

EMA KURNIA

A Thesis submitted for the Degree of Master of Science in
Information Technology for Natural Resources
Management Program Study

GRADUATE SCHOOL
BOGOR AGRICULTURAL UNIVERSITY
BOGOR
2016

External Examiner : Dr Ir M. Buce Saleh, MS

Thesis Title : Spatial Modeling of Land Surface Temperature and its Relationship
with Ecological Factors in Bogor Municipality
Name

: Ema Kurnia
ID
: G051120011

Approved by,
Advisory Board

Prof Dr Ir I Nengah Surati Jaya, MAgr
Supervisor

Dr Ir Widiatmaka, DAA
Co-Supervisor

Endorsed by

Program Coordinator of
MSc in IT for Natural Resources
Management

Dean of Graduate School

Dr Ir Hartrisari Hardjomidjojo, DEA

Dr Ir Dahrul Syah, MScAgr

Date of Examination:
March, 29th 2016

Date of Graduation:

ACKNOWLEDGEMENT
In the name of Allah SWT, the Most Gracious and the Most Merciful.
Alhamdulillah, all praises to Allah for the strength and His blessing in completing
this thesis, I would like to express my highly appreciation to the following:
1. Prof Dr Ir I Nengah Surati Jaya, MAgr as my supervisor for his supervision
support. His invaluable help in the forms of constructive comments and
suggestions throughout the experimental and thesis works have contributed to
the success of this research and Dr Ir Widiatmaka, DAA as the co-supervisor
for his ideas and knowledge regarding this topic during my research.
2. Dr Ir M. Buce Saleh, MS as the external examiner for his positive inputs and

ideas.
3. All my lecturer for giving knowledge and experience.
4. All my friends at MIT and at Laboratory of Remote Sensing and GIS (Alos lab)
in Faculty of Forestry IPB, for their help and support in finishing my study.
5. All of my family. Thank you for always supporting me.

Bogor, May 2016
Ema Kurnia

LIST OF CONTENTS
LIST OF TABLES

xiv

LIST OF FIGURES

xiv

LIST OF APPENDICES

xiv

1 INTRODUCTION
Background
Problem Formulation
Research Objectives

1
1
3
3

2 METHODOLOGY
Time and Location
Data and Tools
Research Conceptual
Data Analysis Procedure
Derivation of Land Surface Temperature
Ecological Factors
Spatial Modeling of LST

4
4
5
5
6
7
13
15

3 RESULT AND DISCUSSION
Land Surface Temperature Distribution
Relationship between LST and Air Temperature
Ecological Factors Relationship Analysis
LST and Humidity Relationship Analysis
LST and Elevation Relationship Analysis
LST and Road Density Relationship Analysis
LST and Land Cover Types Relationship Analysis
Spatial Modeling of LST
Model Validation

16
16
18
19
19
20
20
23
24
27

4 CONCLUSION

27

REFERENCES

28

APPENDICES

32

BIOGRAPHY

38

LIST OF TABLES
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18

Descriptions of the research conceptual
SW coefficient values
Rescaling Factor
K1 and K2 Values
Emissivity values
Sample site description
Ecological factors
LST proportion area
Mean LST of Land Cover
Separation matrix of Landsat-8 supervised classification
Confusion matrix of landsat-8 supervised classication
Ground truth of land cover classification
Land cover scoring
Road distance scoring
Elevation scoring
Weight for each variable
Spatial distribution of LST
Confusion matrix validation

6
7
8
8
9
12
13
17
17
21
22
22
24
25
25
25
26
27

LIST OF FIGURES
1
2
3
4
5
6
7
8
9
10
11

Study area of Bogor Municipality
Research operational flowchart
Design sampling
LST map
Relationship between LST and air temperature
Relationship between LST and humidity
Relationship between LST and elevation
Mean values of LST associated with road buffer
Land cover classification result
Mean values of LST associated with each LC type
Score map of LST and its spatial distribution

4
6
11
18
19
19
20
21
23
23
26

LIST OF APPENDICES
1 Wet and dry bulb thermometer
2 Ratio water vapor content to the total in different atmospheric profiles
3 Ground Truth location of land cover

32
32
33

1 INTRODUCTION
Background
Development in big cities causing global temperature changes that result in
a change of climate elements, especially the increase in temperature. This has
inspired many countries to resolve the issue, include Indonesia. Bogor Municipality
with the rapid development of economics, more and more people enter into the city,
which causes the urban area expansion. Now, the global warming has been an
international issue that attracting the international attention. At the site level, the
land surface temperatures (LST) has been used by many researchers as an indicator
of energy balance. Specifically, the LST has been used as a key parameter that
describes the land surface processes. Within the urban areas, the LST might be
closely related to the urban heat island (UHI) which is mainly affected by the human
activities. The causes of UHI mainly comes from the human activities that include
lifestyle, that use fossil fuel in main human activities, e.g., liquid petroleum gasses
for cooking, electricity for room heating and cooling, gasoline or diesel for
transportation, industry etc. The waste from the energy usage by every household
was also a secondary contributor to the heat. As the population in the city also grows,
this tends to expand the area and increase its average temperature. Development of
built-up area with the less green open area may cause global temperature changes
that result in a change of climate elements, especially the increase in temperature.
The UHI extent may vary across a city, depending on the spatial distribution
of the GHG emitter and absorber. Taha et al. (1990) reported that heat islands can
develop in ‘pockets’ around single buildings and temperature differences of 4 °C
have been reported along a single street. It is also mentioned that the UHI is affected
by the height and spacing of buildings and their orientation relative to the prevailing
wind that restrict airflow and limit cooling. Urban with very high ratio between
building height and street width, as well as the very dense settlement may have high
waste heat from human activities. In Bogor, where only a few tall buildings
available, the dominant heat contributors might become from human (metabolic
and non-metabolic) activities and road traffic. Fan and Sailor (2005) pointed out
that road traffic contributed about 32% of heat emission while the human metabolic
heat emission is only 8%.
Frequently, cities have warmer land’s average temperature than its
surrounding suburban and rural areas. The rapid development of built-up areas in
Bogor, such as settlements, tall buildings for hotel, commerce, and/or offices,
mainly alters the physical characteristics of the land surface, from vegetation to
non-vegetated areas. The replacement of green vegetated surfaces with nonvegetated and non-porous urban materials with high heat capacity and low solar
reflectivities, such as concrete masses, asphalt roads and metal surfaces exhibit a
high degree of thermal inertia (Arrau and Pena, 2010). These areas are characterized
by a high level of absorption of solar radiation, with a greater capacity for thermal
conductivity as compared to natural surfaces (Rose and Devadas, 2009).
Within the built-up area, the glass-walled buildings may reflect the
incoming short-wave solar radiation (sunlight), and consequently the surface, as
well as the air temperature in the surrounding building may increase. In the cities,

2
it is also quite common that emission from public traffic as well as private traffic
may increase the GHG released to the atmosphere, then eventually increase the
surface temperature. Some gasses that are emitted into the atmosphere will act as
a greenhouse gas that is transparent to short-wave solar radiation and absorb longwave radiation of the earth thus increasing global warming. Urban development
can raise the local temperature of the city where the rate of temperature rise is
proportional to the rate of urban development (Fukui, 2003). The increase of
population also caused an increase of heat waste, mainly emitted from nonmetabolic activities such as vehicles, personal waste, energy consumption and
manufactures etc. This may affect local climate change especially air temperature
directly and indirectly (Mas'at, 2008). The temperature of the urban areas might be
effectively managed and slightly modified by increasing the extent of GHG
absorber or by reducing the source of heat gasses. The common strategy applied is
by increasing the amounts of heat energy absorbed and stored in the vegetation.
Vegetation would be a very effective way as it delivers several mechanisms of
cooling simultaneously and it had been recognized as a very cheap way to
implement the cooling strategy.
To spatially measure the spatial distribution of the heat emitter and heat
absorber, it needs to develop a technique to derive land surface temperature quickly,
consistently, accurately, comprehensively and with a reasonable cost. One
technique that frequently applied is by using the remote sensing approaches.
Traditionally, the land surface temperature is mapped by using interpolation
technique using the data recorded by each national weather station. Interpolation
method for sparsely stationed network had been a focus of many researchers,
including the examination of geostatistics and deterministic approaches. This
method will provide very rough map since the distances between climatology
stations are very low. The different interpolation method can provide different
accuracy and deviation.
Now, the availability of remotely sensed data recorded using thermal bands,
coupling the available middle-infrared, near-infrared, as well as the visible bands,
has given a very good prospect. There is no interpolation method required in this
approach. The temperatures were derived from every grid of the data. The algorithm
used by combining the thermal, near-infrared and red-band of Landsat data, the
pixel-based temperature might be derived. On the LST estimation, the accuracy of
LST estimation is mainly affected by the surface capability of emitting radiation.
In many algorithms, the LST estimations are based on the assumption that the
ground surface acts as a blackbody (emissivity equals one).
The Split-Window (SW) algorithm is the most commonly used, given that
this algorithm removes the atmospheric effect and obtains the LST from the linear
or nonlinear combination of the brightness temperatures of two adjacent channels
centered at 11 and 12 m. Wan (2014) proposed a new refinement of the generalized
SW algorithm added with a quadratic term of the difference amongst the brightness
temperature of the adjacent thermal infrared channels. Remote sensing is a practical
way to accomplish the monitoring and assessing the LST because it represents a
relatively low-cost and rapid method to acquire up-to-date information over a large
geographical area. The studies related to the land surface temperatures can also be
found in Dash (2005) and Akhoondzadeh and Saradjian (2008). The LST has been
used as a key parameter in the physics of land surface processes. Other related

3
studies on land surface temperature can be found in Liang et al. (2012); Zhang and
He (2013). The main objective of the study was to develop spatial modeling of land
surface temperature (LST) based on the Landsat-8 OLI and TIRS imageries using
Split-Window (SW) algorithm and its relationship with the ecological factors in
Bogor Municipality, Indonesia.
Problem Formulation
The most basic needs of the population are housing and vacancies. This has
led to the land conversion of green area into built area. Green area has always been
a victim because of the assumption that green area is not included in the mechanism
of the market economy, and has a market value that is lost by the price of land
(Irwan, 2008). The population and land use that are not controlled can lead to
reduced quality of the environment. Good environmental quality conditions can be
comfortable, healthy, and aesthetically pleasing for residents. According to Oliver
(1981) in Irwan (2008), human comfort can be determined quantitatively by
temperature and relative humidity indicators. Air temperature is determined to a
great extent by surface properties that vary in both space and time (Oke, 1987) and,
therefore, directly responds to local changes in land surface (Arribas et al., 2003).
Despite the fact that ground meteorological stations provide important local point
data, such as air temperature, which is needed to monitor the ecosystem on a macro
scale, their spatial density is highly variable and their distribution is usually not
optimal for regional and local applications. However, remote sensing data, due to
their high sampling rate and repetitive basis over large and heterogeneous regions,
offer us a continuous surface that provides additional information between ground
meteorological stations where there is a lack of meteorological data (Vogt et al.,
1997).
In order to control the urban development especially as a prevent of urban
heat island in Bogor Municipality and also as a prediction of the land surface
temperature due to relationship with the ecological factors, so it is important to
make problem definition. There is some research question that should be known:
1. How is the land surface temperature distribution in Bogor Municipality?
2. How the ecological factors can give a contribution in affecting the land surface
temperature in the research location?
Research Objectives
The main objective of this research is to develop spatial modeling of land
surface temperature (LST) in Bogor Municipality. As for the specific purposes of
this study are:
1. To estimate the land surface temperature distribution in Bogor Municipality
using landsat-8 satellite imagery.
2. To identify the ecological factors affecting the land surface temperature and its
relationships.

4

2 METHODOLOGY
Time and Location
The research was conducted from November 2014 to November 2015 in
Bogor Municipality. Geographically, site location is in coordinate of: 106o48'40''106o46'22'' E and 6o30'53'' - 6o40'08'' S (Figure 1) with the total area of 11.694 Ha.
Bogor Municipality is located in a hinterland area of the Capital City Jakarta that
consisted of 6 districts and 68 villages (BPS, 2013). The city has been the main
destination of many domestic tourist within the Jabodetabek, and the residential
area of many people who are working in the Capital City Jakarta. Now, the city has
been frequently selected by His Excellency President Joko Widodo as his second
working office. Thus, Bogor Municipality which is now becoming the center of
various activities such as commerce, tourism, residential and president palace has
been chosen as a study site. Bogor Municipality has a high rate of population growth
and development.

Figure 1 Study area of Bogor Municipality

5
Data and Tools
The Supporting Data
The main data used are Digital Satellite Landsat-8 OLI (Operational Land
Imager) imageries path 112-row 65; acquired on 13th September 2014. The bands
used particularly red band (band 4), near infrared band (band 5), and Thermal
Infrared Sensor (TIRS), namely band 10 and 11. The data was captured at
approximately 10:00:37 a.m. local time. Landsat-8 provides metadata of the bands
such as thermal constant and rescaling factor value that used for calculating the LST.
Although the Landsat-8 OLI recorded on 13th September 2014, while the ground
measurement was done in March 2015, the difference between the temperature in
March 2015 and in September 2014 in the day time is not significantly different (at
measurement time). The land cover and land use within at the measurement points
didn’t change drastically. Data satellite Aster GDEM has the spatial resolution of
30 m was obtain from the United States Geological Survey (USGS). The humidity
data was obtain from the field measurement and from BMKG. While the vector
data of basic geospatial information of Bogor Municipality was obtain from the
BAPPEDA (Spatial Planning Agency) year 2011.
Software, Hardware, and Tools Used
Software
: Erdas Imagine 9.1 and ArcGIS 9.3 for the Remote Sensing (RS)
and Geographic Information System (GIS) spatial analysis.
Hardware
: Personal computer and printer
Tools
: - Thermometer (dry and wet bulb) was obtain from laboratory
of Geophysics and Meteorology of Bogor Agricultural
University (Appendix 1).
- GPS
- Camera.
Research Conceptual
This research is pointed in combining the information of Remote Sensing to
retrieve the land surface temperature on a micro-scale (Bogor Municipality) and
Geographic Information System by means of spatial modeling. The research will
be focusing on the land surface temperature (LST) and its relationship with
ecological factors. The increasing of greenhouse gases in the atmosphere was
greatly affected the LST. The land use/land cover of an area can be used for
estimating the amount of LST influenced by the natural and anthropogenic activities
change of an area. The climate of the city was determined by the characteristic of
the spatial elements that will be interacting ecologically with the area. The natural
and anthropogenic activities will give effect to the LST. Estimating of land surface
temperature will be done by using the split-window algorithm to derive the LST
and validate by the actual data of air temperature. The regression analysis will be
done to see the relationship of the suspected ecological factors with the LST.
Scoring and weighting will be done to develop spatial modeling and its spatial
distribution in Bogor Municipality. The research conceptual will be shown in Table
1 as follow.

6
Table 1 Descriptions of the research conceptual
No
1.

2.

3.

Objective
Data
To estimate the - Landsat-8 OLI
land
surface
Bands 4, 5, and
temperature map
TIRs Bands 10,
in
Bogor
11.
Municipality
- Actual air
temperature
To
see
the - Humidity
ecological factors - Elevation
that affecting the - Road density
land
surface - Land cover
temperature and its
relationships
To develop spatial - LST
modeling of land - Elevation
- Road density
surface
- Land cover
temperature

Process
- Split-Window
Algorithm

Result
- Land
surface
temperature (LST)
map

- Regression
analysis
- Regression
analysis

- LST validation
accuracy
- Relationship
analysis

- Rescaling

- Score and weight

- Spatial
modeling
analysis

- Spatial
distribution of
LST

- Validation

- Model Validation

Data Analysis Procedures
The methodology of this study described the four main procedures which
are lined from data inventory until implementation (Figure 2). The procedures are
consisted of: (1) image analysis: derived brightness temperature, LSE, water vapor
and LST; (2) field measurement; (3) relationship analysis, (4) spatial modeling.
Bogor

Subset Area

Landsat-8
path 122 row 65

Data Inventory
Band 10,11

Band 4,5,6,8
Data Processing

Radiance

NDVI

Brightness
T emperature (K)

LSE

LST ( 0 C)
(Split-Window Algorithm)

Water Vapour

Land Cover
Classification
Elevation,
Road Density,
Land Cover T ypes
Field Measurement
- Air T emperature
- Humidity

Relationship
Analysis

Spatial Distribution of LST

Spatial Modeling
(Scoring and Weighting)

Figure 2 Research operational flowcharts

7
Derivation of Land Surface Temperature
Land surface temperature calculated by transforming two bands (10, 11) of
TIR DN into a radiance value, and then convert the radiance into the brightness
temperature (BT). It also uses the mean and difference in land surface emissivity
(LSE) for estimating the LST of an area. LSE was derived from OLI bands of 4 and
5 by using the NDVI threshold method.
Subset area
Subset area was done to define the area location of the research. Landsat-8
scene covers an area of 170 km by 185 km, subset area was to cut out of this larger
image to simplify the analysis and focus on the scene portion of primary interest.
Land Surface Temperature
Landsat-8 TIR sensor is capable of recording the radiance heat data on the
earth's surface in thermal infrared spectrum. The radiance heat information in the
thermal spectrum is strongly influenced by the surface temperature and the object
emissivity. The higher temperature of an object is the higher intensity of the
radiance. Split-window (SW) algorithm was used to determine the LST of an area.
The basis of the SW technique is that the radiance attenuation for
atmospheric absorption is proportional to the radiance difference of simultaneous
measurements at two different wavelengths (Sobrino et al., 1996). The basic inputs
for SW algorithm were the brightness temperature and land surface emissivity
(LSE). The Split-Window technique uses two TIR bands typically located in the
atmospheric window between 10 and 12 m. The SW algorithm only requires the
knowledge of water vapor (w).
The brightness temperature of two bands of TIR, mean and difference in
land surface emissivity was used to calculate the LST. The formula is in equation1 (Jimenez-Munoz, 2014):
LST = BT10 + C1 (BT10-BT11) + C2 (BT10-BT11)2 + C0 + (C3+C4w) (1- ε) + (C5+C6w)
Δ ε …... (1)
where: LST is Land surface temperature (0Kelvin); C0 to C6 are Split-window
Coefficient values (as shown in Table 2); BT10 and BT11 are Brightness
temperatures of band 10 and band 11 (in 0K); ε is mean LSE of TIR bands
(see equation-7); w is atmospheric water vapor content; and Δ ε is difference
in LSE (see equation-8)
Table 2 SW coefficient values
Constant
C0
C1
Value
-0.268
1.378
(Source: Skokovic et al., 2014)

C2
0.183

C3
54.300

C4
-2.238

C5
C6
-29.200 16.400

Radiance conversion to Brightness Temperature (BT)
Landsat 8 translates into 65536 potential grey levels, compared with only
256 grey levels in previous 8-bit instruments. Thus, the obtaining of brightness
temperature from Landsat 8 data includes the estimation of radiance from its DN

8
value and the conversion of the radiance into brightness temperature. Once the
spectral radiance L ( ) is obtained by using of equation-2, the brightness
temperature at the satellite level can be directly calculated by either inverting
Planck’s radiance function for temperature (Sospedra et al., 1998). Brightness
Temperature (BT) of Band 10 and 11 is the electromagnetic radiation traveling
upward from the top of the Earth’s atmosphere. Thermal calibration process done
by converting thermal DN values of raw thermal bands of TIR sensor into TOA
Spectral Radiance. TIRS band 10 and 11 was taken to estimate Brightness
Temperature (BT) in 0Kelvin using the algorithm of equation-3 by converting the
spectral radiance to brightness temperature (BT) for both the TIRs bands.
L = ML * Qcal + AL ….…………………………….(2)
whereμ L is Top of atmospheric radiance in watts/ (m 2 srad m); ML is Band
specific multiplicative rescaling factor; Qcal is Band 10/11 image and AL is
Band specific additive rescaling factor (Table 3).
Table 3 Rescaling factor
Rescaling Factor
ML
AL

Band 10
0.0003342
0.1

Band 11
0.0003342
0.1

TIRS bands 10 and 11 were taken to estimate Brightness Temperature for
thermal band ith (BTi) in 0Kelvin using the algorithm of equation-3. Brightness
temperature for both the TIRs bands:
�

=

(

1
�

+ )

…. (3)

where: K1 and K2 are thermal conversion constant and they vary for both TIR bands
(Table 4); L is Top of atmospheric radiance.
Table 4 K1 and K2 values
Thermal Constant
K1
K2

Band 10
774.89
1321.08

Band 11
480.89
1201.14

In the IR-range, brightness temperature is also known as equivalent
blackbody temperature. The blackbody obtained from the top of atmospheric
(TOA) radiance is the brightness temperature of TOA and the blackbody obtained
from surface radiance (which contains effect of emissivity and reflected
downwelling radiance) is the brightness temperature of the surface. Brightness
temperature of the TOA is generally lower than the LST and the brightness
temperature of the surface (except for an atmosphere that is warmer than the
surface) (Dash, 2005).

9
NDVI
The Normalize Difference Vegetation Index (NDVI) is an index that
describes the level of greenness of vegetation. The NDVI value was calculated by
dividing the difference of the near infrared from red band and sum of the near
infrared and red band obtained from satellite imagery (Lillesand and Kiefer, 1997).
Considering pixels as a mixture of soil and vegetation, NDVI was used to calculate
the fractional vegetation cover in order to obtain the emissivity variation. The
method classifies each pixel as a combination of either bare soil or vegetation and
then combines their separate effects to derive LST. The NDVI formula is shown in
equation-4, this equation result in an image with pixel value ranging between -1 to
1.

� �� =

�� −
�� +

……………………(4)

Land Surface Emissivity (LSE)
Emissivity describes the object’s ability to emit the radiance energy, as the
emissivity of the land surface varied dynamically (Nerry et al., 1988), its effect
must also be compensated for. Unlike the oceans, the emissivity of land, can differ
significantly from unity and vary with vegetation types, surface moisture,
roughness, and viewing angles (Salibusbury and D’Aria, 1992). The mean and the
difference LSE map of band 10 and 11 in Bogor Municipality is an important
parameter when deriving LST as the emissivity of a surface will influence the
amount of thermal radiation that it emits. LSE is determined by several factors
including the chemical composition, roughness and moisture content of a surface.
The emissivity of a surface can have values between 0 and 1, however, for most
objects spectral emissivity is very close to 1 (Stathopoulou et al., 2007).
Since LSE can change substantially over short distance, it is important to
estimate its value for every pixel prior to applying the split window algorithm.
NDVI map was used to retrieving the proportional of soil and vegetation in order
to estimate the LSE (Sobrino et al., 2000; Qin et al., 2006). LSE is estimated from
information collected by OLI in VNIR bands (reflectance or vegetation indices)
depending on the Fractional Vegetation Cover (FVC) for a given pixel (Sobrino et
al., 2008). The formula of LSE as shown in equation-5.
LSE = εs(1-FVC) + εv*FVC ………………………………(5)
whereμ εs and εv are soil and vegetation emissivity values, respectively (Table 5)
and FVC is Fractional Vegetation Cover
Table 5 Emissivity values
Emissivity
εs
εv

Band 10
0.971
0.987

Band 11
0.977
0.989

(Skokovic et al., 2014)
The FVC for an image was derived from the NDVI values that had been
calculated previously. In this study, the pixel was considered as bare soil when the

10
NDVI values are < 0.2; while it would be considered as full vegetation when the
NDVI values are > 0.5 (Stathopoulou et al., 2007). The calculation as shown in
equation-6.

�� =

� ��−� ��

� �� −� ��

……………………(6)

where: NDVIs is the NDVI of soil with value of 0.2; while NDVIv is the NDVI of
vegetation with the maximum value of vegetation cover (more than 0.5).
The mean of emissivity (ɛ) and difference of emissivity (Δε) are calculated
using equation-7 and equation-8.
ε = (ε10+ε11)/2 …………………….(7)
whereμ ε is mean emissivity; ε10 and ε11 is the emissivity of band 10 and 11
Δε = ε10-ε11 ……………………..(8)
where: Δε is difference of emissivity; ε10 and ε11 is the emissivity of band 10 and 11
Water Vapor
Water vapor (w) had been recognized as one factor that affecting the land
surface temperature. Most atmospheric water vapor is concentrated in the lower
atmosphere (Sobrino et al., 1991) especially in the first 3 km of the profile
(appendix 2). Though the total water vapor content is different, the distributions of
the ratio of water vapor content to the total in the atmospheric profiles are very
similar. The standard distributions of the atmospheres or just their average for
simplification can be used to develop a simple method to generate the required
distribution of water vapor content at each layer of the atmosphere, from the
measurement of total water vapor content in the atmosphere as shown in equation9 (Qin et al., 2001):
w(z) = w Rw(z) ………………………(9)
where: w(z) is water vapor content at altitude z; Rw(z) is the ratio of water vapor
content to the total in the standard atmospheric profiles.
If total atmospheric water vapor content is not available, it can be
approximately estimated as w=w(0)/Rw(0) in which w(0) is water vapor content
near the surface. Usually, the w(0) can be obtain from local meteorological data.
The average of humidity for Bogor area in September 2014 (w=0.73) was
multiplied by the ratio of water vapor content to the total in the standard
atmospheric profiles for the tropical area (Rw(z)=0.425043). The water vapor
content at altitude z was 0.31.

11
Ground Measurement
Land Surface Temperature (LST) is defined as the temperature of the
interface between the earth’s surface and its atmosphere and thus it is a critical
variable to understand land-atmosphere interactions and a key parameter in
meteorological and hydrological studies, which involve energy fluxes (Niclòs et al.,
2009). Land surface air temperature (LSAT) is the temperature of the air near the
earth´s surface which is routinely measured at 1.5 to 2 m by common
meteorological stations distributed across a territory. For validating the land surface
temperature derived from the model in equation 1, the authors made ground
measurements. Ground-based temperature measurements were taken in 13 days
during March 2015 starting from 2nd March to 28th March. The measurement days
were selected randomly, at the following specific date: 2nd, 4th, 7th, 9th, 11th, 14th,
16th, 18th, 21st, 23rd, 25th, 28th, and 30th. The temperature measurements were done
between 12.00 – 14.00 pm local time, which is the closest time to the Landsat-8
image captured. A total of five thermometers at measurement points were mounted
at 1.5 m height in an open space that protected from solar radiation. Each of the
stations was also measured relative humidity at the same height and at the same
time. Field visits was undertaken to collect the data actual of air temperature in the
area under study based on the land use/land cover (LU/LC) data year 2011. The
LU/LC was divided into three classes to make easier in defining the design samples.
The design sampling was done by using a purposive sampling method (Nasir, 2005)
where the location of the samples was determined at a specific location with
consideration to represent the characteristics of the region such as built-up, urban
forest, rice field, housing and manufactured area (Table 6). Camera and GPS are
also used to precisely locate the ground sites investigated and location pictures.
Design sampling was shown in Figure 3.

Figure 3 Design sampling

12
Table 6 Sample site description
No

Sample Site

Description of location

1.

-06.57574 S, 106.79682 E
Goodyear
manufacture
located in Pemuda street,
North Bogor. It is a
manufacture of tire and
the location is in the
middle of built-up area.

2

-06.56745 S, 106.77704 E
Yasmin housing is located
in Cemara boulevard
street, West Bogor.

3.

-06.5988 S, 106.7991 E
Urban Forest Botanical
Garden is located in
Central Bogor. Its location
also surrounding by the
dense settlement and
roads.

4.

-06.6701 S, 106.8251 E
Pamoyanan village is
located in South Bogor. It
is a rice field area
surrounding
by
the
settlement.

5.

-06.55174 S, 106.82093 E
Cibuluh street is located in
North Bogor.

To know the consistency and the relationship between these ground-based
temperatures and the average of LSTs estimation, then the Pearson’s correlation
coefficients were derived. Besides, the deviation between the land’s surface
temperature estimate (LST) and the actual temperature was calculated by using
mean deviation (MD) of equation-10:

13
MD = {

̂ −y
y
|
y

|∑n
i=1

}×

% ………. (10)

The mean deviation is the percentage amount absolute value of the
difference quotient between the estimated value and the actual to the predicted
values of the samples (n). A good model has a value of less than 10%.

Ecological Factors
Spatial dynamics of the population is rooted in social and economic
activities. Activities of the urban population represented by the increasing demand
for built area. Built area in this case represented by the settlement, either for
residential, industrial, and educational. Each of these activities result in spatial
changes over time into a certain extent, this activity will eventually be limited by
biophysical carrying capacity of the region. The city climate is influenced by the
development of the city which cannot be separated from the physical development
of the city, due to the population growth of the city; and increased human activity
in the city; this is caused by the industry and life sustaining needs of the people.
According to Tjasjono (1986), physical development and the increase of human
activity in the city resulted in the higher city temperatures. Industrial development
of the city will cause the layer of the atmosphere above becomes soiled by dust
particles, smoke and others caused by the increasing number of motor vehicles and
factories. Man-made structures such as roads and buildings typically have a lower
albedo than the natural surface and absorb more visible radiation. This shows that
the reduced of the vegetation and the increased in urban surface will cause heat
island effect. This is also increasing the mass of the earth's surface that can store
(and release) a lot more energy than the surface. Urban surfaces tend to heat faster
than natural surfaces that hold water; the water will be released rapidly by urban
surfaces in contrast with natural surfaces such as vegetation that can hold water.
Anthropogenic heat source comes from the heating and ventilation systems,
industrial processes, and internal combustion engines. In general, the consumption
of energy will result heat.
Based on the analysis above, then variables in this study are variables which
are suspected as the ecological factors that related to the surface temperature and
represented by five factors (Table 7):
Table 7 Ecological factors
No
Variables
Method
Data Source
1. LST
Split-window algorithm
Landsat-8 TIRs bands
2. Humidity
Direct measurement from
Actual data
certain location
3. Elevation
Spatial analysis
Aster Image
4. Road density Spatial
operation
Vector data
(buffering technique) of
the existing road
5. Land cover
Supervised classification
Landsat-8 OLI bands

Unit
0
C
%
m
km

ha

14
The underlying assumption of the ecological factor was the relationship
between the LST and the areal differentiation of the certain location. Simple linear
regression was used to see the relationship between them. The variables used in this
research were:
1. Humidity
This factor suspected with the assumption that the higher humidity then the
lower LST. Relative humidity is the percent of water vapor, in the air, compared to
what would be present if it were saturated. The data measurement of humidity was
collected at the same time and using the same method with the air temperature.
Relative humidity can be found on psychometric charts using the difference value
of dry and wet bulb temperatures. Relative humidity is always expressed as a
percentage, and a value of 100% means the air is saturated.
2. Elevation
This factor suspected with the assumption that the lower elevation, then the
higher LST. Elevation significantly affects the surface temperature and relative
humidity, temperature typically decreases with an increase in elevation. Therefore,
daytime mixing of the atmosphere near the ground will yield a uniform amount of
moisture, and the decrease in temperature will yield an increase in relative humidity.
3. Road density
The proximity factor to the road has led to the increasing of the temperature
due to road surface temperature (RST) is the highest among land use/land cover
(Chudnovsky et al., 2004). This factor suspected had a relationship that the closer
to the road then the higher probability of temperature increasing. The assumption
of coating or road pavement material of all road in studied area are the same. The
road density analyzed by using the road buffer of 100 m distance.
4. Land cover
Land cover classification of Landsat-8 imagery was performed using the
maximum likelihood classification (MLC) algorithm. The MLC has been the most
popular parametric classifier used for remote sensing data classification (Foody et
al., 1992; Jia et al., 2011). The MLC assumed that for a given unknown pixel, the
probability of membership in each class is calculated using the mean feature vectors
of the classes, the covariance matrix, and the prior probability. The unknown pixel
is consider belong to the class with the maximum probability of membership. The
classification process started with creating the training area of Landsat-8 bands
composite (654). Evaluation of the training area that has been made for land
classification is based on the value of separability. The separability measurement
was done to obtain the quality classification accuracy. The method chose was
Transformed Divergence (TD), because it is good in evaluating the separability of
each land cover type. It also give the best estimation (Jaya, 1996). The value of TD
for each land cover type was calculated by using the formula as shown in equation9 and equation-10 (Swain dan Davis, 1978; Singh, 1984).
Dij=0.5 tr[(Ci−Cj)(Cj−1− Ci−1)]+0.5 tr[(Ci−1+ Cj−1)(Mi−Mj)(Mi−Mj)T] … (11)

15
TDij = 2000 [1 − exp [−Dij / 8]] ………………………………………….(12)
where: i and j is two class to compare; Ci is covariance matrix of class i; M is the
average of vector class i; tr is the matrix algebra; T is the transpose function;
and TDij is the separation between class i and j.
The separability analysis was needed to exhibit the separability statistics of
each land cover type based on the average digital number classified so that it can
concluded whether it needs to clustered or not. The criteria of value TD in this
research were (Mather et al., 1990; Jaya and Kobayashi, 1995; Jensen, 1996):
1. Inseparable: ≤ 1600
2. Poor: 1601 – 1699
3. Fair: 1700 – 1899
4. Good: 1900 – 1999, and
5. Excellent: 2000
The accuracy analysis of land cover classification was done by calculating
the overall accuracy and kappa accuracy as shown in equation 13 and 14 (Jensen,
1996):

……………………………………………… (13)
………………………………….. (14)
where: N is the number of all pixel (training area); r is the number of rows in error
matrices; Xkk is the number of pixel in the class (diagonal matrices); X k+ is
the number of all columns at rows-i; X+k is the number of all rows at columnj.
The accuracy was calculated using the overall accuracy equation and kappa
accuracy equation. Kappa accuracy was used to takes into account all the element
in the matrix contingency. Overall accuracy (producer accuracy and user accuracy)
describe the truth, but not considering the possibility of similarity/agreement map
related to the reference data. A field survey was done in order to improve and
examine the sample unit by reviewing locations using GPS and camera for each
point which is useful for the visual interpretation.
Spatial Modeling of Land Surface Temperature
Variables used in the model are the ecological factors that affecting