LAND SURFACE TEMPERATURE ESTIMATION USING

LANDSAT 8 IMAGE AND SINGLE-CHANNEL

ALGORITHM (SCA) METHOD IN SURABAYA CITY

Azhali, F.M. and Jaelani, L.M.

Departemen Teknik Geomatika, Fakultas Teknik Sipil, Lingkungan, dan Kebumian
Institut Teknologi Sepuluh Nopember
Kampus Sukolilo, Surabaya, 60111, Indonesia

e-mail: [email protected]

Abstract. In green open space development planning, the estimation of Land Surface
Temperature (LST) serves to find out which areas have a high LST distribution so that later it
can be used as a study for urban forest and city park development planning. LST can be obtained
by conducting research using meteorological data from the weather station. But the
meteorological data obtained is only data that is at the weather station so that the utilization of
satellite imagery becomes a solution to get the LST in wide area coverage. One image that can
be used is Landsat 8 with the Single-channel Algorithm (SCA) method. The SCA method was
chosen because at the start of the landsat 8 satellite launch, the sensor of band 11 was exposed
to light stray which affected the absolute calibration of the sensor. The results of the estimation
of SPT obtained were then validated with temperatures recorded on the BMKG Juanda and PWS
ITS stations. The data used in this study are meteorological data from BMKG Juanda Station
and PWS ITS and Landsat 8 level 1 and 2 images. This research was conducted in 2 periods
namely August 11, 2018 and September 12, 2018. The maximum temperature obtained from the
city of Surabaya was in the period August 5, 2018 to September 12, 2018 is 36oC, the minimum
temperature is 20oC and the average temperature is 27oC. The difference in temperature values
generated from the image with BMKG Juanda Station in the period 11 August 2018 is ± 0,11oC
and the value difference with PWS ITS is ± 0,88oC. Whereas in the September 12, 2018 period,
the difference in temperature values generated from the image with BMKG Juanda Station was
± 0,34oC and the different values with PWS ITS were ± 3,22oC. The test results of the correlation
between the LST Map of the SCA Method and the meteorological data of BMKG Juanda Station
and PWS ITS are 87% which means that both datas have very strong correlations.

1. Introduction

In the planning of green open space development, soil surface temperature (LST) functions to find out
which areas have high LST distribution so that later it can be used as a study for urban forest and city
park development planning. Since 2010, the Mayor of Surabaya Tri Rismaharini has been very active
in the construction of city parks and urban forests. The construction of the park caused a decrease in the
temperature of the city of Surabaya by 2oC [1].
SPT is one of the key parameters of energy balance on the surface and is the main climatological
variable. SPT controls long wave energy flux through the atmosphere. The amount of SPT depends on
the conditions of other surface parameters, such as albedo, surface moisture and land cover and
vegetation conditions [2]. Therefore, knowledge of the spatial distribution of SPT and its temporal
diversity is important for modeling accurate flow between surface and atmosphere [3].
In SPT processing using Landsat 8 imagery, channel 11 is no longer effective because there is a light
stray that affects the absolute calibration of the Landsat 8 thermal sensor (TIRS) since its launch. The
impact of stray light varies on each channel, one of which is on channel sensor 11. An algorithm is
currently being developed to calibrate channel 11 sensors [4]. Because channel 11 is not effectively used
for processing SPT, a Single-Channel Algorithm (SCA) algorithm is developed as an alternative to
estimating SPT [5].
Initially, SPT research used meteorological data from weather stations. This causes difficulty in
obtaining abundant meteorological data and it is not possible to obtain meteorological data on the surface
other than the location of the weather station. After the launch of the Sputnik satellite in 1957, many
satellite images were applied to various needs. Satellite images with thermal infrared channels such as
Landsat TM, Landsat OLI, ASTER and TERRA-MODIS can be converted to obtain SPT data [6].
Automatic Weather Station (AWS) is a station that takes measurements that send or record observations
of weather elements automatically and directly make converted codes. AWS is generally equipped with
several sensors, RTU (Remote Terminal Unit), a set of computers with LED screens, and other parts.
Sensors used include temperature sensors, wind speed and direction sensors, humidity sensors, air
pressure sensors, precipitation, pyranometers, and net radio meters. The RTU consists of a data logger
and backup power that functions as a terminal for collecting weather data from the sensor [7].
The AWS observation system has been developed for a long time in developed countries, but the price
is quite expensive so it is still very limited to be used in Indonesia [8].In this study, researchers used
Landsat 8 and AWS satellite imagery to estimate the LST in the city of Surabaya, which subsequently
produced LST maps with a scale of 1: 200.000. The purpose of LST estimation using the SCA method
is to determine the level of accuracy of the algorithm with AWS. The resulting LST data is expected to
be used to reference the development of urban forests and city parks in the city of Surabaya.

2. Method

2.1. Research Sites

This research was conducted in Surabaya City, which is located at 07o12’ - 07o21’ LS and 112o36’ -
112o 54’ BT. The following is an image of the research location shown as shown below:

Figure 1. Surabaya City location

2.2. Data and Equiptment
The data used for this study are level 1 and 2 Landsat 8 satellite images covering the city of Surabaya
acquired on August 11, 2018 and September 12, 2018, Surabaya City administration boundary vectors
obtained from BIG scale 1: 25,000, temperature data from PWS ITS and BMKG Juanda. The software
used in this study is ArcMAP 10.6.1 ITS education license for image processing, IDW processing and
map making.

2.3. Land Surface Temperature

The processing of LST begins with radiometric calibration. Radiometric calibration is done to convert
pixel values from digital number (DN) to brightness temperature (BT). In addition to the conversion of
DN to BT, DN is also converted into radians and reflectants. Following are the equations for radiometric
calibration calculations [9]:

𝐿𝐿𝜆𝜆 = 𝑀𝑀𝐿𝐿 𝑄𝑄𝑐𝑐𝑐𝑐𝑐𝑐 + 𝐴𝐴𝐿𝐿 (1)

Where Lλ = Top of atmosphere spectral radiance (w� 2 ), ρλ = Top of atmosphere spectral

m sradμm
reflectance, Q cal = Band specific digital number, ML = Band specific multiplicative rescaling factor
dan AL = Band specific addiitive rescaling factor.
𝐾𝐾2
𝐵𝐵𝐵𝐵 = 𝐾𝐾 (2)
ln( 𝐿𝐿 1 +1)
𝜆𝜆

Where BT = Brightness temperature (oK), Lλ is computed in (2), K1 and K2 = Thermal conversion bands
10. Atmospheric correction is done in bands 4 and 5 to convert radians to satellite sensors (LTOA) and
reflectances on satellite sensor (ρTOA) becomes a radians value below the atmospheric layer (LBOA)
and the reflectance value below the atmospheric layer (ρBOA). In this study atmospheric correction uses
Landsat 8 level 2 images. 8 level 2 Landsat imagery was chosen because the level 2 algorithm is better
than the 6S (Second Simulation of Satellite Signal in Solar Spectrum) algorithm that requires many
parameters, so that if there is a parameter selection error it will have an impact on subsequent processing.
NDVI is carried out by image processing by combining 4 and 5 channels that have been corrected by
the atmosphere into a new image that shows the greenness value of the image. The equation used to
obtain the NDVI value is as follows [10]:
(𝜌𝜌𝐵𝐵𝐵𝐵𝐵𝐵 𝑁𝑁𝑁𝑁𝑁𝑁−𝜌𝜌𝐵𝐵𝐵𝐵𝐵𝐵 𝑅𝑅𝑅𝑅𝑅𝑅)
𝑁𝑁𝑁𝑁𝑁𝑁𝑁𝑁 = (3)
(𝜌𝜌𝐵𝐵𝐵𝐵𝐵𝐵 𝑁𝑁𝑁𝑁𝑁𝑁+𝜌𝜌𝐵𝐵𝐵𝐵𝐵𝐵 𝑅𝑅𝑅𝑅𝑅𝑅)

Where NDVI = Normalized Difference Vegetation Index, ρBOA (λ) = Bottom of atmosphere spectral
reflectance. Proportion of vegetation (Pv) is needed to find out the emissivity parameters by looking at
the NDVI values that have been previously processed. In this study, the NDVIs coefficient and NDVIv
coefficients are determined with a value of 0.2 for NDVIs and 0.5 for NDVIv. This value can be applied
in all conditions globally [11]. The equation for obtaining a Pv value is as follows:

𝑁𝑁𝑁𝑁𝑁𝑁𝑁𝑁− 𝑁𝑁𝑁𝑁𝑁𝑁𝑁𝑁𝑠𝑠 2
𝑃𝑃𝑃𝑃 = � � (4)
𝑁𝑁𝑁𝑁𝑁𝑁𝑁𝑁𝑣𝑣 − 𝑁𝑁𝑁𝑁𝑁𝑁𝑁𝑁𝑠𝑠

Land Surface Emissivity (LSE) is needed to obtain the emissivity value (ε) which will be included in
the SPT calculation formula. The emissivity of soil pixels (NDVI <NDVIs) is 0.96 (Pv = 0 and 𝐶𝐶𝜆𝜆 = 0,
so 𝜀𝜀𝑠𝑠𝑠𝑠 = 𝜀𝜀𝑠𝑠𝑠𝑠 = 0.96) and the emissivity value of vegetation pixels (NDVI> NDVIv) is 0.99 (Pv = 1 and
𝐶𝐶𝜆𝜆 = 0.005, so 𝜀𝜀𝜆𝜆 = 𝜀𝜀𝑣𝑣𝑣𝑣 + 𝐶𝐶𝜆𝜆 = 0.985 + 0.005 = 0.99). Mixed emissivity values can be calculated using
the following equation [12]:
 𝜀𝜀𝜆𝜆 = 𝜀𝜀𝑣𝑣𝑣𝑣 𝑃𝑃𝑃𝑃 + 𝜀𝜀𝑠𝑠𝑠𝑠 (1 − 𝑃𝑃𝑃𝑃) + 𝐶𝐶𝜆𝜆 (5)

Where εvλ = 0,985 εsλ = 0,960 and Cλ = 0,005 (constant). Atmospheric Correction Parameter Calculator
is a web-based tool used to obtain atmospheric correction parameters. Furthermore, the atmospheric
correction parameters are processed using the following equation to obtain atmospheric function values
[13]:

1 𝐿𝐿↑
𝜓𝜓1 = 𝜓𝜓2 = −𝐿𝐿↓ − 𝜓𝜓3 = 𝐿𝐿↓ (6)
𝜏𝜏 𝜏𝜏

Where τ is atmospheric transmissivity, L↑ is up-welling radiance and L↓ is down-welling radiance.

Correction of atmospheric influences requires specific atmospheric parameters (𝜏𝜏, 𝐿𝐿↑ , 𝐿𝐿↓ ) [14]. LST is
the equation used to obtain LST distribution from satellite imagery. The LST estimation of the SCA
method can be done using the following equation [15]:

1
𝑆𝑆𝑆𝑆𝑆𝑆 = 𝛾𝛾 � (𝜓𝜓1 𝐿𝐿𝑠𝑠 + 𝜓𝜓2 ) + 𝜓𝜓3 � + 𝛿𝛿 (7)
𝜀𝜀

Where ε is computed in (5), γ dan δ is computed in (8) and (9), dan ψ1, ψ2 dan ψ3 is computed in (6). In
an effort to obtain the results of LST in celcius units, each stage of processing using kelvin units must
be converted in celsius first. The stages intended are BT and γ processing. To obtain the values γ and δ
can use the following equation:

𝐶𝐶2 𝐿𝐿𝑠𝑠 𝜆𝜆4 𝐿𝐿𝑠𝑠 1

𝛾𝛾 = { � + �}−1 (8)
𝑇𝑇𝑠𝑠2 𝐶𝐶1 𝜆𝜆

𝛿𝛿 = −γ. 𝐿𝐿𝑠𝑠 + 𝑇𝑇𝑠𝑠 (9)

𝐶𝐶2 = λ × 𝐾𝐾2 (10)

C1 and C2 is planck radiation constant (C1 =1,19104 108 W μm4 m-2 sr-1; C2 = 1,43877 104 μmoK); λ is
effective wavelength (λ = 10,895 μm for Landsat 8 OLI/TIRS band 10) and K2 thermal convertion
constant band 10. K2 values for Landsat-8 OLI and band 10 TIRS are 1321,0789oK atau 1047,9289oC.
After all variables have Celsius units, they are then included in the SCA equation.

2.4. Final Analysis

The final analysis phase consists of correlation tests, data validation and extrapolation of AWS
meteorological data. Pearson correlation test is used to see the existence of the relationship of the value
of surface temperature generated from the SPT processing method SCA with the value of surface
temperature generated by AWS. The correlation coefficient is obtained using the following equation
[17]:

𝑛𝑛 ∑ 𝑥𝑥𝑖𝑖 𝑦𝑦𝑖𝑖 −∑ 𝑥𝑥𝑖𝑖 ∑ 𝑦𝑦𝑖𝑖

𝑅𝑅𝑥𝑥𝑥𝑥 = (11)
�𝑛𝑛 ∑ 𝑥𝑥𝑖𝑖 −(∑ 𝑥𝑥𝑖𝑖 )2 .�𝑛𝑛 ∑ 𝑦𝑦𝑖𝑖 2 −(∑ 𝑦𝑦𝑖𝑖 )2
2

The results of the calculation of the correlation coefficient are then used to determine the correlation
strength. The strength of the correlation can be seen in Table 1 about the correlation strength interval
[18]:
 Table 1. Correlation strength intervals
Correlation
No. Correlation Strength
Coeficient
1 0 No
2 0,00 - 0,25 Very weak
3 0,25 - 0.50 Average
4 0,50 - 0,75 Strong
5 0,75 - 0,99 Very strong
6 1 Perfect

Validation is used to check the value of land surface temperature resulting from image processing with
surface temperature data obtained from AWS. Interpolation is a step to determine the value between two
known values based on an equation function.

3. Results
The results of LST processing are obtained by the SCA method. In facilitating the analysis process, the
upper and lower limits of the LST processing results are set at 40oC and 25oC. The following are the
results of SPT processing:

Figure 2. Results of processing LST using SCA method on Landsat 8 acquisition images August 11, 2018 and
September 12, 2018

The results of the LST processing using SCA method can be seen in Figure 2. The color changes in the
processing results have different meanings, red means that the LST is high and the blue color means a
low LST. In the middle of Surabaya City, LST is increasingly high due to population activities and
vehicles. The results of comparison of LST obtained from BMKG Juanda with AWS processing of the
SCA method on August 11, 2018 are ± 0.11oC. Whereas the LST comparison obtained from PWS ITS
is ± 0.88oC. The results of the SPT comparison obtained from the PWS ITS with the AWS processing
of the SCA method on September 12, 2018 are ± 0.34oC. While the LST comparison obtained from
PWS ITS is ± 3.22oC. Furthermore, the temperature data is entered as a variable in the calculation of the
correlation test. This can be seen in Table 3 about the comparison of LST obtained from AWS by
processing the LST of the SCA method. The LST obtained by AWS and the results of the LST
processing of the SCA method were tested for correlation to get the correlation coefficient value.
Correlation test used to obtain the correlation coefficient value of LST obtained from AWS with the
results of the LST processing of the SCA method is the pearson correlation test [19]:
 Table 2. Comparison of SPT obtained from AWS by processing the SPT method of the SCA
Temperature (⁰C)
Station
Time SPT using SCA ∆t (⁰C)
Name AWS
methods
Juanda 27,98 28,09 0,11
11/08/18
ITS 33 32,12 -0,88
Juanda 31 31,34 0,34
12/09/18
ITS 33 36,22 3,22

Table 3. Pearson correlation test results

Station SPT using SCA
Time AWS (X) X2 Y2 XY
Name Methods (Y)
Juanda 27,98 28,09 782,88 789,05 785,96
11/08/2018
ITS 33,00 32,12 1089,00 1031,69 1059,96
Juanda 31,00 31,34 961,00 982,20 971,54
12/09/2018
ITS 33,00 36,12 1089,00 1304,65 1191,96
Total (∑) : 124,98 127,67 3921,88 4107,59 4009,42
Rata-rata : 31,25 31,92 980,47 1026,90 1002,35
Rxy : 0,87

Pearson correlation test results obtained are 0.87 or 87% which has a very strong correlation meaning
this can be seen in Table 3 about pearson correlation test results. Correlation coefficient is obtained
because of the pattern of temperature changes indicated by the AWS and SPT temperature data. The
SCA method is relatively the same, which is high at PWS ITS and low at BMKG Juanda Station. The
location of the PWS ITS is surrounded by buildings and settlements while the BMKG Juanda Station is
surrounded by rice fields and airport buildings.

Figure 3. Graph of correlation test results

Figure 3 describes the graph of linear equations from the results of the correlation test that has been
done. In the graph it can be seen that the LST of the AWS and LST methods acquired by AWS have
an adjacent linear relationship. In the sample there is 1 point that has the most distant relationship, that
is LST acquired by PWS ITS and LST using SCA Method on September 12, 2018 which has a
temperature difference of 3.22oC.
4. Discussion and Conclussion

In the SPT processing the SCA method needs to be converted into units from kelvin to celcius on each
parameter. When processing is carried out using kelvin units and the results are converted to Celsius,
the results of the temperature obtained will not match the temperature acquired at AWS. So it is
important to change the constant thermal conversion (K2) for channel 10 to Celsius first. From the
results of the SPA processing of the SCA method it can be concluded that the SCA method is well used
in the SPT estimation because the level of confidence in the SPT processing results of the SCA method
with AWS temperature data is 87%, which means the SPT estimation method can be trusted

References
[1] Hakim, A. (2018, September 13). Risma : Suhu di Surabaya Turun 2 Derajat Celcius. Antaranews Jatim.
[2] Prasasti, I., Sambodo, K. A., & Carolita, I. (2007). Pengkajian Pemanfaatan Data TERRA-MODIS Untuk
Ekstraksi Data Suhu Permukaan Lahan (SPL) Berdasarkan Beberapa Algoritma. LAPAN, 4(1), 1–8.
[3] Rajeshwari, A., & Mani, N. D. (2014). Estimation of Land Surface Temperature of Dindigul District Using
Landsat 8 Data. International Journal of Research in Engineering and Technology (IJRET), 3(5), 122–126.
[4] Landsat 8 OLI And TIRS Calibration Notices. (2017). Retrieved June 13, 2019, From
Https://www.Usgs.Gov/Land-Resources/Nli/Landsat/Landsat-8-Oli-And-Tirs-Calibration-Notices
[5] Walawender, J. A. P. W., Zymanowski, M. A. S., Ajto, M. O. J. H., & Okwa, A. N. B. (2014). Land Surface
Temperature Patterns in the Urban Agglomeration of Krakow ( Poland ) Derived from Landsat-7 / ETM +
Data, 171, 913–940. https://doi.org/10.1007/s00024-013-0685-7
[6] Tsou, J., Zhuang, J., Li, Y., & Zhang, Y. (2017). Urban Heat Island Assessment Using The Landsat 8 Data:
A Case Study In Shenzhen And Hong Kong. Urban Science, 1(1), 10.
[7] Qudratullah, M. I., Kamus, Z., & Asrizal. (2017). Analisis Unsur-Unsur Cuaca Berdasarkan Hasil
Pengukuran Automated Weather System (AWS) Tipe Vaisala Maws 201. Pillar Of Phyiscs, 9, 17–24.
[8] Nugroho, G. A., Machfud, M. S., & Sanjaya, M. (2016). Rancang Bangun Automatic Weather Station (Aws)
Menggunakan Raspberry Pi. Alhazen Journal Of Physics, II(2), 48–57.
[9] Jaelani, L. M. (2013). Kalibrasi Radiometrik: Mengubah Digital Number (DN) ke Radiance dan/atau
Reflectance. Retrieved January 16, 2019, from http://lmjaelani.com/2013/12/kalibrasi-radiometrik-
mengubah-digital-number-dn-ke-radiance-danatau-reflectance/
[10] Waas, H. J. D., & Nababan, B. (2010). Pemetaan dan analisis index vegetasi mangrove di pulau Saparua,
Maluku Tengah. E-Jurnal Ilmu Dan Teknologi Kelautan Tropis, 3(2), 50–58.
[11] Sobrino, J. A., & Raissouni, N. (2000). Toward Remote Sensing methods for land cover dynamic monitoring.
Application to Morocco. International Journal of Research in Engineering and Technology (IJRET), 21, 353–
366.
[12] Sobrino, J. A., Casseles, V., & Becker, F. (1990). Significance of the remotely sensed thermal infrared
measurements obtained over a citrus orchard. ISPRS J. Photogramm, 44, 343–354.
[13] Walawender, J. P., Szymanowski, M., Hajto, M. J., & Bokwa, A. (2014). Land Surface Temperature Patterns
in the Urban Agglomeration of Krakow (Poland) Derived from Landsat-7/ETM+ Data. Pure and Applied
Geophysics, 171(6), 913–940. https://doi.org/10.1007/s00024-013-0685-7
[14] Barsi, J. A., Barker, J. L., & Schott, J. R. (2003). An Atmospheric Correction Parameter Calculator for a
Single Thermal Kanal Earth-Sensing Instrument. IEEE IGARSS. Toulouse, France.
[15] Jimenez-Munoz, J. C., Cristobal, J., Sobrino, J. A., Soria, G., Ninyerola, M., & Pons, X. (2009). Revision of
the Single- Channel Algorithm for Land Surface Temperature Retrieval From Landsat Thermal-Infrared
Data. IEEE Trans. Geosci. Remote Sensing, 47, 339–349.
[16] Sobrino, J. (2010). A Single-Channel Algorithm for Land-Surface Temperature Retrieval From ASTER Data,
(February), 3–7. https://doi.org/10.1109/LGRS.2009.2029534
[17] Hidayat, A. (2012). Uji Pearson Product Moment dan Asumsi Klasik. Retrieved June 23, 2019, from
https://www.statistikian.com/2012/07/pearson-dan-asumsi-klasik.html/amp
[18] Sarwono, J. (2006). Metode Penelitian Kuantitatif dan Kualitatif (Edisi pert). Yogyakarta: Graha Ilmu.
[19] Istantinova, D. B., Hadiwidodo, M., & Handayani, D. S. (1995). Terhadap Konsentrasi Gas Pencemar Sulfur
Dioksida ( So 2 ) Dalam Udara Ambien Di Sekitar Pt . Inti General Yaja Steel Semarang Abstrak. Tentang
Konsentrasi Gas Sulfur, 10(Gas Sulfur), 1–10.