Algorithm for Automated Mapping of Land Surface Temperature Using LANDSAT 8 Satellite Data

Abstract

Land surface temperature is an important factor in many areas, such as global climate change, hydrological, geo-/biophysical, and urban land use/land cover. As the latest launched satellite from the LANDSAT family, LANDSAT 8 has opened new possibilities for understanding the events on the Earth with remote sensing. This study presents an algorithm for the automatic mapping of land surface temperature from LANDSAT 8 data. The tool was developed using the LANDSAT 8 thermal infrared sensor Band 10 data. Different methods and formulas were used in the algorithm that successfully retrieves the land surface temperature to help us study the thermal environment of the ground surface. To verify the algorithm, the land surface temperature and the near-air temperature were compared. The results showed that, for the first case, the standard deviation was 2.4°C, and for the second case, it was 2.7°C. For future studies, the tool should be refined with
in situ
measurements of land surface temperature.

Full text

Research Article
Algorithm for Automated Mapping of Land Surface Temperature
Using LANDSAT 8 Satellite Data
Ugur Avdan and Gordana Jovanovska
Research Institute of Earth and Space Sciences, Anadolu University, Iki Eylul Campus, 26555 Eskisehir, Turkey
Correspondence should be addressed to Ugur Avdan; uavdan@anadolu.edu.tr
Received  November ; Revised  January ; Accepted  February 
Academic Editor: Guiyun Tian
Copyright ©  U. Avdan and G. Jovanovska. is is an open access article distributed under the Creative Commons Attribution
License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly
cited.
Land surface temperature is an important factor in many areas, such as global climate change, hydrological, geo-/biophysical, and
urban land use/land cover. As the latest launched satellite from the LANDSAT family, LANDSAT  has opened new possibilities
for understanding the events on the Earth with remote sensing. is study presents an algorithm for the automatic mapping of land
surface temperature from LANDSAT  data. e tool was developed using the LANDSAT  thermal infrared sensor Band  data.
Dierent methods and formulas were used in the algorithm that successfully retrieves the land surface temperature to help us study
the thermal environment of the ground surface. To verify the algorithm, the land surface temperature and the near-air temperature
were compared. e results showed that, for the rst case, the standard deviation was .∘C, and for the second case, it was .∘C.
For future studies, the tool should be rened with in situ measurements of land surface temperature.
1. Introduction
Land surface temperature (LST) is dened as the temperature
feltwhenthelandsurfaceistouchedwiththehandsor
the skin temperature of the ground []. As one of the most
important aspects of the land surface, LST has been a main
topic for developing methodologies to be measured from
space. LST is an important factor in many areas of studies,
such as global climate change, hydrological and agricultural
processes, and urban land use/land cover. Calculating LST
from remote sensed images is needed since it is an important
factor controlling most physical, chemical, and biological
processes of the Earth []. ere is a growing awareness
among environmental scientists that remote sensing can and
must play a role in providing the data needed to assess
ecosystems conditions and to monitor change at all special
scales []. e tool developed in this paper is simple and does
not require any background knowledge so scientists can use
it very easy in their researches.
ealgorithmintroducedinthispaperhasbeendevel-
oped using ERDAS IMAGINE , with the Model Maker
allowing us to create a model that will repeat the process
automatically,anditiseasytodevelopasimpletooluseful
for doing pixel calculations. Without the tool, the process
of retrieving LST is very long, and it is prone to many
mistakes.etoolalsocanbedevelopedinanysoware
supporting pixel calculations from a given image, following
step by step this paper. Although an LST retrieval method for
LANDSAT  has been developed [, ], a tool is needed for
the complicated process of obtaining the LST. A similar study
for retrieving LST in ERDAS IMAGINE has been conducted
forLANDSATdata[]butnotforLANDSAT.etool
presented in this paper is used for calculating the LST of
a given LANDSAT  image with the input of the fourth
(red wavelength/micrometres, .–.), h (near infrared
(NIR) wavelength/micrometres, .–.), and tenth (ther-
mal infrared sensor (TIRS) wavelength/micrometres, .–
.) bands. Following January , , recommendations of
USGS of not using TIRS Band  due to its larger calibration
uncertainty, only Band  was included in the algorithm.
2. Data and Methods
e algorithm was created in ERDAS IMAGINE , and
itcanonlybeusedtoprocessLANDSATdatabecause
of the data complexity. e LST of any Landsat  satellite
Hindawi Publishing Corporation
Journal of Sensors
Volume 2016, Article ID 1480307, 8 pages
http://dx.doi.org/10.1155/2016/1480307

Journal of Sensors
Input Band 10 Input Band 4 Input Band 5
Top of atmo sp heri c
spectral radiance
(see Section 2.1)
Calculating NDVI
(see Section 2.3.1)
Determination of
ground emissivity
(see Section 2.3.3)
Calculating LST
(see (7))
LST result
Conversions of
radians to at-sensor
temperature
(see Section 2.2)
Calculating
proportion of
vegetation P

(see Section 2.3.2)
F : Flowchart for LST retrieval.
image can be retrieved following the steps of Figure . e
data of Landsat  is available at the Earth Explorer website
free of charge. In this study, the TIR band  was used to
estimate brightness temperature and bands  and  were used
for calculating the NDVI. e metadata of the satellite images
used in the algorithm is presented in Table .
2.1. Top of Atmospheric Spectral Radiance. e rst step of the
algorithm is the input of Band . Aer inputting band , in
the background, the tool uses formulas taken from the USGS
webpageforretrievingthetopofatmospheric(TOA)spectral
radiance ():
= 𝐿∗cal +𝐿−𝑖,()
where 𝐿represents the band-specic multiplicative rescal-
ing factor, cal is the Band  image, 𝐿is the band-specic
additive rescaling factor, and 𝑖is the correction for Band 
[].
2.2. Conversion of Radiance to At-Sensor Temperature. Aer
thedigitalnumbers(DNs)areconvertedtoreection,the
TIRS band data should be converted from spectral radiance
to brightness temperature (BT) using the thermal constants
provided in the metadata le. e following equation is
T : Metadata of the satellite images.
ermal constant, Band 
1.
2.
Rescaling factor, Band 
𝐿.
𝐿.
Correction, Band 
𝑖.
used in the tool’s algorithm to convert reectance to BT
[]:
BT =2
ln 1/+1 −273.15, ()
where 1and 2stand for the band-specic thermal conver-
sion constants from the metadata.
For obtaining the results in Celsius, the radiant tem-
peratureisrevisedbyaddingtheabsolutezero(approx.
−.∘C) [].

Journal of Sensors 
2.3. NDVI Method for Emissivity Correction
2.3.1. Calculating NDVI. Landsat visible and near-infrared
bandswereusedforcalculatingtheNormalDierence
Vegetation Index (NDVI). e importance of estimating the
NDVI is essential since the amount of vegetation present is
an important factor and NDVI can be used to infer general
vegetation condition []. e calculation of the NDVI is
important because, aerward, the proportion of the vegeta-
tion (V)shouldbecalculated,andtheyarehighlyrelatedwith
theNDVI,andemissivity() should be calculated, which is
related to the V:
NDVI =NIR (band 5)−(band 4)
NIR (band 5)+(band 4),()
where NIR represents the near-infrared band (Band ) and 
represents the red band (Band ).
2.3.2. Calculating the Proportion of Vegetation. Vis cal-
culated according to (). A method for calculating V[]
suggests using the NDVI values for vegetation and soil
(NDVIV=0.5and NDVI𝑠=0.2) to apply in global conditions
[]:
V=NDVI −NDVI𝑠
NDVIV−NDVI𝑠2.()
However, since the NDVI values dier for every area, the
valueforvegetatedsurfaces,.,maybetoolow.Globalvalues
fromNDVIcanbecalculatedfromat-surfacereectivities,
butitwouldnotbepossibletoestablishglobalvaluesin
thecaseofanNDVIcomputedfromTOAreectivities,
since NDVIVand NDVI𝑠will depend on the atmospheric
conditions [].
2.3.3. Calculating Land Surface Emissivity. e land surface
emissivity (LSE ()) must be known in order to estimate LST,
since the LSE is a proportionality factor that scales blackbody
radiance (Planck’s law) to predict emitted radiance, and it
is the eciency of transmitting thermal energy across the
surface into the atmosphere []. e determination of the
ground emissivity is calculated conditionally as suggested in
[]:
𝜆=
V𝜆V+𝑠𝜆 1−
V+𝜆,()
where Vand 𝑠are the vegetation and soil emissivities,
respectively, and represents the surface roughness (=
for homogenous and at surfaces) taken as a constant value
of . []. e condition can be represented with the
following formula and the emissivity constant values shown
in Table  []:
𝜆
=








𝑠𝜆,NDVI <NDVI𝑠,
V𝜆V+𝑠𝜆 1−V+, NDVI𝑠≤NDVI ≤NDVIV,
𝑠𝜆 +, NDVI >NDVIV.
()
When the NDVI is less than , it is classied as water, and
the emissivity value of . is assigned. For NDVI values
betweenand.,itisconsideredthatthelandiscovered
with soil, and the emissivity value of . is assigned. Values
between . and . are considered mixtures of soil and
vegetation cover and () is applied to retrieve the emissivity.
In the last case, when the NDVI value is greater than ., it
is considered to be covered with vegetation, and the value of
. is assigned.
e last step of retrieving the LST or the emissivity-
correctedlandsurfacetemperature𝑠is computed as follows
[]:
𝑠=BT
1+BT/ln 𝜆,()
where 𝑠is the LST in Celsius (∘C, ()), BT is at-sensor BT
(∘C), is the wavelength of emitted radiance (for which the
peak response and the average of the limiting wavelength (=
10.895)[]willbeused),𝜆istheemissivitycalculatedin(),
and
=
=1.438×10−2 mK,()
where is the Boltzmann constant (. ×−23 J/K), is
Planck’s constant (. ×−34 J s), and is the velocity of
light (. ×8m/s) [].
3. LST Validation
e two major LST validation models are through ground
measurements or near-surface air temperature [, ]. e
LST results comparing with the ground measurements results
mayhaveanerrorupto
∘C;inthecaseofSrivastavaetal.,the
accuracy of the results in some area showed dierence of ±∘C
with actual ground temperature measurements. According to
Liu and Zhang, another method using the mean near-surface
air temperature to verify the retrieved LST results showed that
the LST retrieving error is about .∘C. For the validation, six
representative points have been used.
For the validation of the nal retrieved LST results in the
presented tool, the mean near-surface air temperature was
used [] but with bigger amount of data and taking not only
themeantemperaturebutalsotheactualtemperatureinthe
given pixel at the moment of the satellite passing over the area
for  representative points.
e comparison was made with air temperature, which
is dierent and can sometimes result in big dierences
since the resolution of LANDSAT  for the used bands
is  m for the thermal band and  m for the red and
NIR bands. e LST was calculated and taken for the
pixel in which the meteorological station fell. Sometimes,
the dierences can be very big depending on the weather
condition and other factors []. It should also be taken
into consideration that there is . to  meters’ dier-
ence between the LST and the air temperature, which
means that dierences in the temperatures are normal and
expected.

Journal of Sensors
N
EW
S
(a)
(b)
(c)
(d)
F : Application of algorithm in Ontario and Quebec, Canada. (a) Geographic location of Ontario in Canada; (b) frames of satellite
images of study areas; (c) rst case located between Toronto and Huntsville; (d) second case located in surroundingarea of the city of Moncton.
3.1. Application of the Algorithm to Ontario and Quebec,
Canada. Hourly data were collected from the Canadian
Weather and Meteorology website (http://climate.weather.gc
.ca/) and used for comparison with the retrieved LST for
which, according to the available data, satellite images were
downloaded for // (Toronto area) and //
(Moncton area) for the areas shown in Figure .
e study area was the Canadian provinces of Ontario
and Quebec (Figure ). One satellite image was downloaded
from each of the two provinces. ese areas were chosen
because of their specications. at is, both study areas
included water, urban areas, and green areas.
3.2. Comparison of LST Validation Results. To c o m p a r e t h e
results, two dierent satellite images from two dierent
dates in two dierent areas were chosen according to the
available data. Aer downloading the satellite images from
http://earthexplorer.usgs.gov/, LSTs were retrieved in ERDAS
using the algorithm presented in this paper. In the rst case,
the satellite image was located between Toronto and the city
of Huntsville near Lake Simcoe in Ontario, Canada. For this
area,  meteorological stations were found, but only  of
them were used for the accuracy assessment because of the
presence of clouds or other unwanted events. e dierences
between the retrieved LSTs and the air temperatures and
details on the stations are presented in Table  and Figure .
Inthesecondcase,thestudyareawaslocatedinthe
area surrounding the city of Moncton and included part of
New Brunswick, Prince Edward Island, and Nova Scotia in
Canada. For this area, we found  meteorological stations,
and all of them were used for the accuracy assessment. e
details are presented in Table  and Figure .
T : Emissivity of representative ter restrial materials for LAND-
SAT  TIRS Band .
Terrestrial material Water Building Soil Vegetation
Emissivity . . . .
4. Conclusion
is paper presented a new LST soware tool and its
algorithm created in ERDAS for calculating the LST from
LANDSAT  TIRS. e algorithm was derived using the
observed thermal radiance of the TIRS Band  of LANDSAT
 TIRS. To verify the nal retrieved LST results, the near-
surface air temperature method was used. From the anal-
ysis of the two areas in Canada from two dierent dates,
the standard deviation calculated for the rst case based
on  meteorological stations was .∘C, and that for the
second case based on  stations was .∘C. It should be
mentioned that sometimes, the dierence between the near-
surface temperature and the LST can be drastic since we
are comparing two dierent temperatures in dierent places
(ground temperature and . to . m o the ground). It
shouldalsobetakenintoconsiderationthattheresolution
of the LANDSAT  TIRS data is  m for the thermal band
and  m for the red and NIR bands. Values smaller than
−∘C in the two cases were considered to be clouds or other
unwanted events on the satellite images since the data were
from springtime; it was not expected. From Tables  and 
and Figures  and , it can be concluded that, for the rst
case, the smallest dierence between the LST retrieved from
thepresentedtoolandthenear-airtemperaturewas.
∘Cand

Journal of Sensors 
Collingwood
Lagoon City
Pa Udora Strong
Barrie Oro
Mono Centre
Pa Atmos Erin
Borden Awos
Uxbridge West
Pa Uxbridge Taris
Pa Claremont Silo Farm
Pa Atmos Claremont
Toronto Button Ville A
Egbert CS
Pa Tornto
Inter.
Pa Atmos Vaughan
Pa Hardwood
Mountain Bike Park
Pa Caledon E Park
Pa Atmos Bran.
Pa Atmos Vaughan
Pa Markham North
N
E
S
WS
Meteorological stations
<−5 ∘C
−5–3∘C
3–6∘C
6–9∘C
9-10∘C
24-25∘C
20–24∘C
17–20∘C
10–17∘C
>−25∘C
Toy o t a
0 5 10 20
(km)
LST (∘C)
F : Retrieved LST image and meteorological stations from rst study area used in accuracy assessment.

Journal of Sensors
N
EW
S
0 5 10 20
(km)
Doaktown Auto RCS
Buctouche CDA Cs
Parsboro
Mechan ic
Settlement
Gagetown Awos
Pa Atmos Erin
Gagetown A
Napan Auto
Fundy Park (Alma) Cs
Saint Jhon A
Moncton Intl A
Meteorological stations
<−5 ∘C
−5–3∘C
3–6∘C
6–9∘C
9-10∘C
10-11∘C
13–17∘C
11–13∘C
17–20∘C
>−20∘C
LST (∘C)
F : Retrieved LST image and meteorological stations from second study area used in accuracy assessment.

Journal of Sensors 
T : Details and dierences of station from rst study case.
ST name Data pm LST Dierence Latitude Longitude 
Barrie-Oro . . −1.0 ∘󸀠.󸀠󸀠 ∘󸀠.󸀠󸀠 . m
Pa Hardwood Mountain Bike Park . . −5.7 ∘󸀠.󸀠󸀠 ∘󸀠.󸀠󸀠 . m
Borden Awos . . −0.7 ∘󸀠.󸀠󸀠 ∘󸀠.󸀠󸀠 . m
Pa Udora Strong . . −2.5 ∘󸀠.󸀠󸀠 ∘󸀠.󸀠󸀠 . m
Lagoon City . . 1.5 ∘󸀠.󸀠󸀠 ∘󸀠.󸀠󸀠 . m
Collingwood . . −4.9 ∘󸀠.󸀠󸀠 ∘󸀠.󸀠󸀠 . m
Mono Centre . . 0.7 ∘󸀠.󸀠󸀠 ∘󸀠.󸀠󸀠 . m
Uxbridge West . . −2.9 ∘󸀠.󸀠󸀠 ∘󸀠.󸀠󸀠 . m
Pa Uxbridge Taris . . −3.5 ∘󸀠.󸀠󸀠 ∘󸀠.󸀠󸀠 . m
Pa Atmos Vaughan . . 3.1 ∘󸀠.󸀠󸀠 ∘󸀠.󸀠󸀠 . m
Pa Angus Glen Golf Club . . −2.9 ∘󸀠.󸀠󸀠 ∘󸀠.󸀠󸀠 . m
Toronto Buttonville A . . −2.1 ∘󸀠.󸀠󸀠 ∘󸀠.󸀠󸀠 . m
Pa Claremont Silo Farm . . −2.0 ∘󸀠.󸀠󸀠 ∘󸀠.󸀠󸀠 . m
Pa Markham North Toyota . . −5.8 ∘󸀠.󸀠󸀠 ∘󸀠.󸀠󸀠 . m
Pa Atmos Claremont . . −2.0 ∘󸀠.󸀠󸀠 ∘󸀠.󸀠󸀠 . m
Pa Atmos Erin . . −1.8 ∘󸀠.󸀠󸀠 ∘󸀠.󸀠󸀠 . m
T : Details and dierences of station from second study case.
ST name Data pm LST Dierence Latitude Longitude 
Mechanic Settlement . . −0.2 ∘󸀠.󸀠󸀠 ∘󸀠.󸀠󸀠 . m
Moncton Intl A . . 1.0 ∘󸀠.󸀠󸀠 ∘󸀠.󸀠󸀠 . m
Fundy Park (Alma) Cs . . 2.2 ∘󸀠.󸀠󸀠 ∘󸀠.󸀠󸀠 . m
Buctouche Cda Cs . . 0.2 ∘󸀠.󸀠󸀠 ∘󸀠.󸀠󸀠 . m
Nappan Auto . . 2.1 ∘󸀠.󸀠󸀠 ∘󸀠.󸀠󸀠 . m
Doaktown Auto Rcs . . 4.4 ∘󸀠.󸀠󸀠 ∘󸀠.󸀠󸀠 . m
Saint John A . . 2.6 ∘󸀠.󸀠󸀠 ∘󸀠.󸀠󸀠 . m
Parrsboro . . 7. 8 ∘󸀠.󸀠󸀠 ∘󸀠.󸀠󸀠 . m
Gagetown A . . 4.8 ∘󸀠.󸀠󸀠 ∘󸀠.󸀠󸀠 . m
Gagetown Awos A . . −2.3 ∘󸀠.󸀠󸀠 ∘󸀠.󸀠󸀠 /
Summerside . . 2.2 ∘󸀠.󸀠󸀠 ∘󸀠.󸀠󸀠 . m
the biggest was .∘C. In the second study case, the smallest
dierence was .∘Candthebiggestwas.
∘C.
e tool can be signicant since it is opening an opportu-
nity to many researchers to get to the LST values easy and they
canapplytheminanumberofresearchs.etoolpresented
in this paper produced quite good results considering that the
accuracy assessment was made with near-air temperatures
and it can be used for dierent types of research. For future
studies,thetoolshouldberenedwithin situ measurements
of LST.
Conflict of Interests
e authors declare that there is no conict of interests
regarding the publication of this paper.
References
[] A. Rajeshwari and N. Mani, “Estimation of land surface tem-
perature of dindigul district using landsat  data,” International
Journal of Research in Engineering and Technology,vol.,no.,
pp. –, .
[] F. Becker and Z.-L. Li, “Towards a local split window method
over land surfaces,” International Journal of Remote Sensing,vol.
, no. , pp. –, .
[] S. Ustin, Manual of Remote Sensing: Remote Sensing for Natural
Resource Management and Environmental Monitoring,John
Wiley&Sons,Hoboken,NJ,USA,.
[] F.Wang,Z.Qin,C.Song,L.Tu,A.Karnieli,andS.Zhao,“An
improved mono-window algorithm for land surface temper-
ature retrieval from landsat  thermal infrared sensor data,”
Remote Sensing,vol.,no.,pp.–,.
[] Q. Q. Sun, J. J. Tan, and Y. H. Xu, “An ERDAS image processing
method for retrieving LST and describing urban heat evolution:
a case study in the Pearl River Delta Region in South China,”
Environmental Earth Sciences,vol.,no.,pp.–,.
[]J.A.Barsi,J.R.Schott,S.J.Hook,N.G.Raqueno,B.L.
Markham, and R. G. Radocinski, “Landsat- thermal infrared
sensor (TIRS) vicarious radiometric calibration,” Remote Sens-
ing,vol.,no.,pp.–,.
[] USGS, , http://landsat.usgs.gov/Landsat Using Product
.php.

Journal of Sensors
[] H.-Q. Xu and B.-Q. Chen, “Remote sensing of the urban heat
island and its changes in Xiamen City of SE China,” Journal of
Environmental Sciences,vol.,no.,pp.–,.
[] Q. H. Weng, D. S. Lu, and J. Schubring, “Estimation of
land surface temperature-vegetation abundance relationship for
urban heat island studies,” Remote Sensing of Environment,vol.
,no.,pp.–,.
[ ] J. A. Sobrino, J. C. Jim´
enez-Mu˜
noz, and L. Paolini, “Land surface
temperature retrieval from LANDSAT TM ,” Remote Sensing of
Environment, vol. , no. , pp. –, .
[] J. C. Jim´
enez-Mu˜
noz, J. A. Sobrino, A. Plaza, L. Guanter, J.
Moreno, and P. Mart´
ınez, “Comparison between fractional
vegetation cover retrievals from vegetation indices and spectral
mixture analysis: case study of PROBA/CHRIS data over an
agricultural area,” Sensors,vol.,no.,pp.–,.
[] J. C. Jimenez-Munoz, J. A. Sobrino, A. Gillespie, D. Sabol,
and W. T. Gustafson, “Improved land surface emissivities over
agricultural areas using ASTER NDVI,” Remote Sensing of
Environment,vol.,no.,pp.–,.
[]J.A.SobrinoandN.Raissouni,“Towardremotesensing
methods for land cover dynamic monitoring: application to
Morocco,” International Journal of Remote Sensing,vol.,no.
, pp. –, .
[] M. Stathopoulou and C. Cartalis, “Daytime urban heat islands
from Landsat ETM+ and Corine land cover data: an application
to major cities in Greece,” Solar Energy,vol.,no.,pp.–
, .
[] B. L. Markham and J. L. Barker, “Spectral characterization of
the Landsat ematic Mapper sensors,” International Journal of
Remote Sensing,vol.,no.,pp.–,.
[] Z.-L. Li, B.-H. Tang, H. Wu et al., “Satellite-derived land surface
temperature: current status and perspectives,” Remote Sensing
of Environment,vol.,pp.–,.
[]P.K.Srivastava,T.J.Majumdar,andA.K.Bhattacharya,
“Surface temperature estimation in Singhbhum Shear Zone of
India using Landsat- ETM+ thermal infrared data,” Advances
in Space Research,vol.,no.,pp.–,.
[] L. Liu and Y. Z. Zhang, “Urban heat island analysis using the
landsat TM data and ASTER data: a case study in Hong Kong,”
Remote Sensing, vol. , no. , pp. –, .
[] K.Gallo,R.Hale,D.Tarpley,andY.Yu,“Evaluationoftherela-
tionship between air and land surface temperature under clear-
and cloudy-sky conditions,” Journal of Applied Meteorology and
Climatology,vol.,no.,pp.–,.