Land Surface Temperature Estimation Using Split-Window Algorithm from Landsat 8 Thermal Infrared Sensor Data of Beas River Basin, Himachal Pradesh, India

Abstract

Melting of snow/glaciers and ice sheets are affected by the changes in Land Surface Temperature (LST). Moreover LST also affects the vegetation of the area considerably. Therefore any mechanism to infer the LST indirectly can be very useful for many applications. With the advancement in satellite remote sensing, LST can be precisely estimated with the help of satellite imageries. The LST can be estimated using algorithms, namely, Mono-Window (MW), Split-Window (SW), Dual-Angle (DA), and Single-Channel (SC). In this study, LST for the Beas River Basin, Himachal Pradesh, India, has been estimated using SW algorithm with the use of Landsat 8 Optical Land Imager (OLI) of 30 m resolution and Thermal Infrared Sensor (TIR) data of 100 m resolution. For deriving LST, the algorithm SW need spectral radiance and emissivity of two TIR bands as input. Using TIR bands 10 and 11, the spectral radiance was estimated. With the help of Normalized Difference Vegetation Index (NDVI), threshold technique OLI bands 2, 3, 4 and 5 were used to derive emissivity. The output has shown that LST was high in the barren/rocky regions, whereas it was low in the snow/glacier regions because of vegetative cover. As the SW algorithm use both the TIR bands (10 and 11) and OLI bands 2, 3, 4 and 5, the LST generated using them are expected to be more reliable and accurate.

Full text

Land Surface Temperature Estimation Using Split-Window Algorithm from Landsat 8
Thermal Infrared Sensor Data of Beas River Basin, Himachal Pradesh, India
Gopinadh Rongali, Ashok. K. Keshari, A. K. Gosain, R. Khosa
Department of Civil Engineering, Indian Institute of Technology Delhi, New Delhi, India.
Corresponding author email: gopinadh01@gmail.com
Abstract
Melting of snow/glaciers and ice sheets are affected by the changes in Land Surface
Temperature (LST). Moreover, LST also affects the vegetation of the area considerably.
Therefore any mechanism to infer the LST indirectly can be very useful for many
applications. With the advancement in satellite remote sensing, LST can be precisely
estimated with the help of satellite imageries. The LST can be estimated using algorithms,
namely, Mono-Window (MW), Split-Window (SW), Dual-Angle (DA), and Single-Channel
(SC). In this study, LST for the Beas river basin, Himachal Pradesh, India, has been estimated
using SW algorithm with the use of Landsat 8 Optical Land Imager (OLI) of 30 m resolution
and Thermal Infrared Sensor (TIRS) data of 100 m resolution. For deriving LST, the
algorithm SW need spectral radiance and emissivity of two TIRS bands as input. Using TIRS
bands 10 and 11, the spectral radiance was estimated. With the help of Normalized Difference
Vegetation Index (NDVI), threshold technique OLI bands 2, 3, 4 and 5 were used to derive
emissivity. The output has shown that LST was high in the barren/rocky regions, whereas it
was low in the snow/glacier regions because of vegetative cover. As the SW algorithm uses
both the TIRS bands (10 and 11) and OLI bands 2, 3, 4 and 5, the LST generated using them
are expected to be more reliable and accurate.
Keywords: Land Surface Temperature (LST), Split-Window (SW), Thermal Infrared Sensor
(TIRS), NDVI
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28

1. Introduction
LST estimation was difficult to estimate of an area before the Earth Observation Satellites
(EOS). To identify the point data onto area data was estimated at a certain set of sample
points and interpolated into isotherms. For the estimate of the LST spatially advantage using
remote sensing satellites and high spatial resolution sensors. LST can be estimated at a time
for an area using the thermal infrared bands from remote sensing satellites. In recent,
launched on the Landsat 8 satellite (also referrer as Landsat Data Continuity Mission,
LDCM) make sure the steadiness of remote sensing data onto high spatial resolution
assimilated by instruments on board previous Landsat satellites such as the Multispectral
Scanner System (MSS), the Thematic Mapper (TM), and the Enhanced Thematic Mapper
Plus (ETM+).
Landsat 8 carries the two sensors, OLI is the first one and TIRS is the second one. In the
electromagnetic spectrum, OLI sensor acquires data onto a 30 m spatial resolution of eight
bands situated in the visible and near-infrared and in the shortwave infrared regions, plus an
additional panchromatic band of 15 m spatial resolution. Using the two bands situated in the
atmospheric window between 10 and 12 μm (Barsi et al. 2014; Jiménez-muñoz et al. 2014;
Sobrino et al. 2015) the TIRS estimates the TIR radiance at 100 m spatial resolution. OLI
sensor was included since the beginning of the LDCM design and thermal imaging was
initially excluded from the LDCM requirements. In recent years, particularly in the
application of water resources management over the agricultural fields using Landsat 5 TM or
Landsat 7 ETM+ thermal data, was a key factor to finally include a TIR sensor as a part of
LDCM (Irons et al. 2012). These applications depend on the solution to the energy balance
equation for involved heat fluxes and finally retrieval the evapotranspiration, which is the
only way to define if TIR data is available. In certain, LST is a parameter to be estimated
from TIRS data, which also play a key role in the study of geo-biophysical (Quattrochi and
Luvall 1999; Kalma et al. 2008; Kustas and Anderson 2009; Weng 2009; Feizizadeh and
Blaschke 2012; Mallick 2014; Song et al. 2015) LST defines the surface of temperature at
which we estimate contact directly or touch with it. It also defines the surface of skin
temperature. When LST rises it leads to unbalance in the environmental condition like melt in
snow and glacier, vegetation, climate conditions of monsoon countries extends to sudden
rainfall. It is estimated in kelvin. As LST increases, it melts the snow, glaciers and ice sheets
in the polar region. As a result, it leads to floods and rise in the sea level. Increases in LST
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61

will also affect the climatic condition of the monsoon countries leading to unpredictable
rainfall. LST has affected the vegetation in the whole earth surface. The amount of LST is
estimated by using the Land use/ Land cover (LU/LC) of an area. The LU/LC of an area is
changed by the natural and anthropogenic activity. As its value changes the local climate of
the area also changes. It is a very important phenomenon to be studied. Hence, most of the
researchers had estimated LST using various techniques and algorithms. For studying the
thermal characteristic of land surface the thermal infrared remote sensing technology has
become important. Various algorithms are designed by many researchers to estimate land
surface temperature like MW algorithm, SW algorithm, DA algorithm and SC algorithm etc.
In Landsat 8 TIRS contains two thermal bands (10 and 11). We choose SW algorithm to
estimate LST using moderate resolution Landsat 8 bands (30m). For estimation of LST, we
need OLI sensor Bands (2-5) for estimation of Land Surface Emissivity (LSE) done with
Fractional Vegetation Cover (FVC). SW Algorithm combined Brightness Temperature (TB)
band 10 and 11 with LSE to estimate LST for each ground pixels vector. In the present study,
we estimate LST for entire Beas river basin.
1.1 Literature Reviewed
LST was estimated using Moderate-resolution Imaging Spectroradiometer (MODIS) bands of
31 and 32 and Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER)
data of same date and time for one of the semi-arid regions in Iran. The output of LST of
MODIS exactly matches the output of ASTER (Akhoondzadeh and Saradjian 2008;
Khandelwal et al. 2016). To estimate LST, MW and SC algorithms were used in Alastair city,
Iran. Using NDVI the emissivity was estimated and LU/LC by supervised classification
method. The LST results were compared with in situ and the results show a positive
relationship with NDVI and LU/LC method (Alipour and Esmaeily 2005). In Kunsan city
Chollabuk do, Korea the NDVI was fined in four different years. LST was derived by using
threshold technique of NDVI. The output of LST was compared with the NDVI output and a
positive correlation was found in between them. Changes over a time period were also
measured and found that increase in LST (Anbazhagan and Paramasivam 2016; Nikam et al.
2016). Using the Landsat 7 image the LST and NDVI were estimated. The FVC for each
pixel was calculated and it was used in LST analysis. A correlation study was conducted
between LST and NDVI data. The output shown that there was a strong positive correlation
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94

between them and the methodology is more reasonable to estimate NDVI and surface
emissivity (Javed Mallick and B.D.Bharath 2008). SW algorithm for estimating LST was
adjusted for Landsat 8 data for better accuracy. The main inputs for SW algorithm were TB
and LSE. Using the MODTRAN 4.0 software the adjusted SW algorithm accuracy was
estimated. The study has been carried out in the northern Negev Desert, Israel (Du et al.
2014; Rozenstein et al. 2014). In Hebei and Shanxi, North China Plain, Sobrino and Mao
were used individual methods to retrieve LST using MODIS data. The maximum, minimum
and mean of Sobrino and Mao LST methods were compared with the standard LST values
and found that Sobrino output range greater while Mao method had less value than standard
LST. Hence, a combined method of Sobrino and Mao was evolved as Sobmao method. It is
as accurate as Sobrino and simple to use (Mao et al. 2005; Galve et al. 2008; Zhao et al.
2009). Along-Track Scanning Radiometer-2 (ATSR-2) data was handled in MODTRAN 3.5
simulations. In a part of New South Wales the SW, DA, and mixed structures algorithms were
used to generate LST and Sea Surface Temperature (SST).
So many researchers had estimated LST using satellite image. To estimate LST many
numbers of algorithms were developed and adopted. Some of the commonly used algorithms
are SW, Sobrino, Mao, DA, and Sobmao. A maximum number of the studies were done for
urban areas and arid and semi-arid regions and in many of the studies, single thermal band
was used. In the present study, LST was estimated for the entire Beas river basin using two
TIRS bands and four OLI bands. Many studies have been conducted to retrieve LST using
TIRS radiation emitted from surfaces using a SW algorithm (Becker and Li 1990; Ulivieri et
al. 1994; Coll and Caselles 1997; Sobrino and Raissouni 2000; Li et al. 2013; Xia et al.
2014). The SW algorithm uses two adjacent TIRS channels, centred at 11 mm and 12 mm, for
the Advanced Very High Resolution Radiometer (AVHRR) to retrieve surface temperatures
because of their different atmospheric transmittance. The SW LST method corrects the
atmospheric effects based on the differential absorption in the infrared bands. The accuracy of
the SW algorithm depends on the magnitude of the difference between the emissivities of the
surface in the two bands (Becker 1987).
The major objectives of the study are to find the TB using band 10 and band 11 of TIRS,
calculate the LSE using NDVI threshold technique and estimate the LST of Beas river basin,
Himachal Pradesh, India using SW algorithm.
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127

2. LST Retrieval Algorithms
2.1 Landsat 8 TIRS Bands
Landsat 8 was successfully launched on 11 February 2013 and deployed into orbit with two
instruments onboard: (1) the Operational Land Imager (OLI) with nine spectral bands in the
Visible (V), Near-Infrared (NIR), and the Shortwave Infrared (SWIR) spectral regions; and
(2) the Thermal Infrared Sensor (TIRS) with two spectral bands in the Longwave Infrared
(LWIR). The relative spectral response of the TIRS bands is presented in Fig. 1. The two
TIRS bands were selected to enable the atmospheric correction of the thermal data using a
SW algorithm (Cuenca et al. 2013). The use of two separate, relatively narrow, thermal
bands has been shown to minimize the error in the retrieval of LST (Caselles et al. 1998). The
spatial resolution of TIRS data is 100 m with a revisit time of 16 days, and as a result,
applications are different than those of other sensors with coarser spatial resolutions and
shorter revisit times. While Landsat 8 images are already freely distributed through the U.S.
Geological Survey (USGS), to the best of our knowledge, no SW algorithm for LST retrieval
in the Beas river basin from TIRS has been published. Although several SW algorithms have
been developed for use with other sensors (Sobrino and Romaguera 2004; Tang et al. 2008)
some adaptations are required in order to implement them for the TIRS spectral bands.
Therefore, the objective of this study is to develop an SW algorithm adapted for use with
Landsat 8 TIRS data.
The difference between the new TIRS and earlier TM/ETM sensors (apart from differences
related to sensor design) is the existence of two TIR bands in the atmospheric window
between 10 and 12 μm, which represents an advancement over the single thermal band
present on TM and ETM sensors ( Fig. 1). Since the previous single band has been split into
two TIR bands, the bandwidths of TIR bands are narrower than the previous TM/ETM TIR
band. The presence of two TIR bands opens the possibility to apply SW algorithm instead of
other algorithms for LST retrieval, as will be presented in subsequent sections.
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154

Fig.1. Landsat 8 TIRS bands relative spectral response functions
2.2 MW Algorithm
Qin developed the following MW algorithm for obtaining LST from
TS=
[
a(1−C−D)+(b(1−C−D)+C+D)Tsensor −DT a
]
/C
(1)
with
C=ετ , D=(1−τ)
[
1+( 1−ε)τ
]
, a=−67 . 355351, b=0 . 458606 ,
and where
ε
is the
land surface emissivity,
τ
is the total atmospheric transmissivity,
Tsensor
is the at-sensor
brightness temperature and
Ta
represents the mean atmospheric temperature given by
Ta=16. 01110 +0 . 92621 T0
(2)
T0
being the near-surface air temperature. Qin also estimate the atmospheric transmissivity
from
w
, the atmospheric water vapor content, for the range 0.4–1.6
g/cm2,
according to
τ=0 .974290 −0 . 08007 w(highT 0)
(3a)
τ=0 . 982007 −0 . 09611 w(lowT 0)
(3b)
More details about this algorithm and its sensitivity can be found in the work of (Qin et al.
2001).
2.3 SW Algorithm
The SW technique uses two TIR bands typically located in the atmospheric window between
10 and 12 μm. The basis of the technique is that the radiance attenuation for atmospheric
absorption is proportional to the radiance difference of simultaneous measurements at two
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175

different wavelengths, each of them being subject to different amounts of atmospheric
absorption (McMillin 1975). The SW algorithm proposed in this letter is based on the
mathematical structure proposed by (Sobrino et al. 1996) and applied to different EOS in
(Jiménez-Muñoz and Sobrino 2008) i.e.,
Ts=Ti+c1(Ti−Tj)+c2(Ti−Tj)2+c0+(c3+c4w)(1−ε)+(c5+c6w)Δε
(4)
where
Ti
and
Tj
are the at-sensor TB at the SW bands
i
and
j
(in kelvins),
ε
is
the mean emissivity,
ε=0 .5 (εi−εj)
,
Δε
is the emissivity difference,
Δε=( εi−εj)
,
w
is the total atmospheric water vapour content (in
g.cm−2
), and
c0
to
c6
are the
SW coefficients to be determined from simulated data.
3. Description of the Study Area
For carrying out this work, the Beas river basin up to Pandoh dam has been selected as the
study location (Fig. 2). The Beas river basin lies in between 31º N to 32º N latitude and 77º E
to 78º E longitude. The Beas river is an important river of the Indus river system. It starts
from the Rothang pass at an elevation of 3900 m and flows near north-south direction up to
Larji. At Larji, it takes a right angle and turns towards the southwest and flows in the same
direction up to Pandoh dam. The length of the Beas river up to Pandoh is 116 km. Among its
tributaries, Parbati and Sainj Khad rivers are glacier fed. The Beas river catchment up to
Pandoh dam is 5383 km2. The catchment area mostly contains impulsive slopes and the rocks
are commonly bare. There are high peaks in the east as well as in the north of the river valley.
The altitude varies from 846 m near Pandoh dam up to 6500 m on the northeast border of the
Parbati sub-catchment. A substantial portion of the river becomes snow-covered during the
winter season. During the summer, the Beas river is mainly fed by snowmelt. Some of the
major tributaries which join upstream of Pandoh dam are Parbati river near Bhuntar, Tirthan
and Sainj rivers near Larji, Sabari Nala near Kulu and Bakhli Khad near Pandoh dam. Pandoh
dam is primarily a dam to divert water from the Beas river to the Satluj river for power
generation at Dehar powerhouse located on the right bank of the river Satluj.
The Beas river basin obtains heavy rainfall during the monsoon season, which generally
ranges from July to late September. It collects the moisture-bearing winds from both the
Arabian Sea and the Bay of Bengal. The upper portion of the basin receives snowfall during
the winter season. A study conducted on the rainfall data indicated that the heaviest rainfall
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206

occurs at Kothi near Manali, while the lowest occurs near Pulga (Singh et al. 1995). The
distribution of rain gauge stations in the basin shown in (Fig. 2) demonstrates that no rain
gauge is located in the eastern part of the basin, where two important tributaries, namely the
Parvati and Sainj rivers, originate. Overall, the climate of the study area is cold.
The discharge of the Beas river has two constituents one resulting from melting of the snow
and glaciers, and the other causing from the rainfall in the catchment. The accessibility of
flow from snowmelt, rainfall or baseflow creates this river a perennial one. The discharge due
to snowmelt differs from year to year due to differences in temperature and the extent of
snowfall in the basin. The discharge at various sites in the basin is measured by reading the
stages. The average river discharge increases to a considerable extent with the onset of the
monsoon season. Sharp-peaked high floods occur during this period. The pattern of discharge
and flood magnitude in the river and its tributaries depend mainly upon the intensity and
extent of rainfall and its relative occurrence in their catchments. The contribution to the
annual flow for the pre-monsoon season (April–June) is mainly due to the snow and glacier
melt runoff. There may also be a little contribution from rains. The contribution for the
monsoon season (July–September) is primarily due to rains along with some contribution
from snow and glacier melt runoff.
Fig. 2. Area of study illustrating Beas river basin in Himachal Pradesh, India
102°0'0"E99°0'0"E
96°0'0"E
96°0'0"E
93°0'0"E
93°0'0"E
90°0'0"E
90°0'0"E
87°0'0"E
87°0'0"E
84°0'0"E
83°0'0"E
81°0'0"E
79°0'0"E
78°0'0"E75°0'0"E
75°0'0"E
72°0'0"E
71°0'0"E
69°0'0"E
67°0'0"E
37°0'0"N
38°0'0"N
39°0'0"N
36°0'0"N
36°0'0"N
35°0'0"N
35°0'0"N
34°0'0"N
34°0'0"N
33°0'0"N
33°0'0"N
32°0'0"N
32°0'0"N
31°0'0"N
31°0'0"N
30°0'0"N
30°0'0"N
29°0'0"N
29°0'0"N
28°0'0"N
28°0'0"N
27°0'0"N
27°0'0"N
26°0'0"N
26°0'0"N
25°0'0"N
25°0'0"N
24°0'0"N
24°0'0"N
23°0'0"N
23°0'0"N
22°0'0"N
22°0'0"N
21°0'0"N
21°0'0"N
20°0'0"N
20°0'0"N
19°0'0"N
19°0'0"N
18°0'0"N
18°0'0"N
17°0'0"N
17°0'0"N
16°0'0"N
16°0'0"N
15°0'0"N
15°0'0"N
14°0'0"N
14°0'0"N
13°0'0"N
13°0'0"N
12°0'0"N
12°0'0"N
11°0'0"N
11°0'0"N
10°0'0"N
10°0'0"N
9°0'0"N
9°0'0"N
8°0'0"N
8°0'0"N
7°0'0"N
7°0'0"N
6°0'0"N
6°0'0"N
INDIA
Himachal
Pradesh
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234

4. Data and Software Used
Landsat 8 is one of the Landsat series of National Aeronautics and Space Administration
(NASA). The data of Landsat 8 is available in EarthExplorer website at free of cost. In the
present study, the TIRS bands 10 and 11 were used to estimate TB and OLI spectral bands 2,
3, 4 and 5 were used to generate NDVI of the study area. The Beas river basin was covered in
one tile (Table 1). Landsat 8 provides metadata of the bands such as thermal constant,
rescaling factor value etc., which can be used for calculating various algorithms like LST.
Landsat 8 TIRS Band 10 and 11 and OLI sensor Band (1-9) of Beas river basin of date 24
April 2015. Thermal constant K1 and K2 and other image statistics are obtained from
metadata of the image file. ERDAS IMAGINE 9.2, ARC MAP 10.3, ENVI 4.8.
Table 1. Metadata of the Satellite Images
Sensor No. of Bands Resolution (m) Path/Row Date of Acquisition
OLI 9 30 147/38 24 April 2015
TIR 2 100
Table 2. K1 and K2 Values
Thermal Constant Band10 Band11
K1 1321.08 1201.14
K2 777.89 480.89
Table 3. Rescaling Factor
Rescaling Factor Band10 Band11
ML0.0003342 0.0003342
AL0.1 0.1
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250

5. Methodology
5.1 Flow Chart for the SW Algorithm
Fig. 3. Flow chart of the SW algorithm
6. Process for the Estimation of LST
6.1 SW Algorithm
LST =TB10 +C1(TB10−TB11)+C2(TB10−TB 11 )2+C0+(C3+C4W)(1−m)+( C5+C6W)Δm
(5)
Where,
LST
= Land surface temperature (
K
)
LANDSAT-8 (OLI+TIRS)
TIRS BAND 10 AND 11 OLI BAND 1-9
BRIGHTNESS
TEMPERATURE
(TB) OF BAND 10
BRIGHTNESS
TEMPERATURE
(TB) OF BAND 11
LAYER STACK
BAND 2,3,4,5
NDVI USING
BAND 4 AND 5
FRACTIONAL
VEGETATION
COVER (FVC)
MEAN OF LSE (m)
LST = +-) +-) 2 +++
NDVI THRESHOLDING
FOR SOIL AND
VEGETATION
LAND SURFACE
EMISSIVITY (LSE)
OF BAND 10
LAND SURFACE
EMISSIVITY (LSE)
OF BAND 11
DIFFERENCE OF LSE (�m)
251
252
253
254
255
256
257
258
259
260

C0
to
C6
= SW coefficient values (Table 4) (Sobrino et al. 2003; Zhao et al. 2009;
Skokovic et al. 2014)
TB10
and
TB11
= TB of band10 and band11 (
K
)
m
= mean
LSE
of TIR bands
W
= Atmospheric water vapor content
Δm
= Difference in
LSE
Table 4. SW Coefficient Values
Step-1: Estimation of Top of Atmospheric (TOA) spectral radiance of TIRS Band 10 and 11
and OLI sensor of Band 2-5 individually using the algorithm given below. This algorithm
transform raw image into spectral radiance image. Using ARC MAP 10.3 raster calculator
implement algorithm of equation 6 to perform task,
Lλ=
(
Lmax−Lmin
DN max
)
¿Band +Lmin
(6)
Where,
Lλ
– Top of atmospheric spectral radiance in watts/ (
m2×srad×μm
)
Lmax
– Maximum spectral radiance of respective band
Lmin
– Minimum spectral radiance of respective band
DN max
=
Qcalmax −Qcalmin
= Difference of maximum and minimum calibration of
sensor
Constant Value
C0-0.268
C11.378
C20.183
C354.300
C4-2.238
C5-129.200
C616.400
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288

Lλ
= 0.000342
¿
Band 10 + 0.1 (Band 10_radiance) and 0.000342
¿
Band 11 + 0.1
(Band 11_radiance)
Step-2: Estimation of TB of Band 10 and 11. TB is the electromagnetic radiation traveling
upward from the top of the Earth’s atmosphere. Thermal calibration process done by
converting thermal Digital Number (DN) values of raw thermal bands of TIR sensor into
TOA spectral radiance and after using TB equation shown in equation 7.
TB
=
K2/(ln(K1/Lλ+1) )−273 . 15
(7)
Where,
K1
and
K2
– Thermal constant of bands from metadata image file
L
– Top of atmospheric spectral radiance layer
TB
= 1321.01/
ln
(774.89/Band 10_radiance+1) – 273.15 for Band 10_
TB
and
1201.14/
ln
(480.89/Band 11_radiance+1) – 273.15 for Band 11_
TB
Fig. 4. Flowchart of thermal calibration process
Step-3: Estimation of NDVI using OLI sensor optical Band after layer stacking of Band
2,3,4,5 using algorithm shown in equation 8.
NDVI=Band 5−Band 4
Band5+Band4
=NIR−RED
NIR+RED
(8)
Range: -1<
NDVI
<+1
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310

Step-4: Estimation of FVC for an image using NDVI image obtain from Step-3 using the
equation 8. FVC estimate the fraction of area under vegetation. Fig. 6 show the flowchart to
perform FVC. SW algorithm utilize FVC to estimate LSE. Using ARC MAP 10.3 we
reclassify the NDVI layer into soil and vegetation and calculate NDVI for soil and vegetation
and implement the algorithm of FVC of equation 9.
FVC=NDVI −NDVI (Soil)
NDVI (Vegetation)−NDVI (Soil)
(9)
FVC=NDVI −0 .15
0. 48−0 . 15
Fig. 5. Flowchart for FVC
Step-5: Estimation of LSE from FVC layer obtain from step-4 using algorithm in equation
10. LSE measure the inherent characteristic of earth surface. It measure its ability to convert
thermal or heat energy into radiant energy. LSE estimation required emissivity of soil and
vegetation of both Band 10 and 11 are given in Table 5. LSE of Band 10 and 11 are
individually calculated.
LSE=εs×(1−FVC )+εv×FVC
(10)
Where,
εs
= Emissivity for soil
εv
= Emissivity for vegetation
FVC
= Fractional vegetation cover
311
312
313
314
315
316
317
319
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335

LSE
=0.971
¿
(1−FVC )
+0.987
¿FVC
(Band10_
LSE
) and 0.977
(1−FVC )
+0.989
¿FVC
(Band11_
LSE
).
Table 5. Emissivity Values
Emissivity Band10 Band11
εs0.971 0.977
εv0.987 0.989
(Source: Sobrino et al. 2003; Zhao et al. 2009; Skokovic et al. 2014)
Step-6: Combination of LSE of Band10 and LSE of Band 11 obtain from step-5 through
mean and difference in between them as shown in equation 11 and 12.
Mean of
LSE=m=LSE10 +LSE11
2
(11)
Difference of
LSE=Δm=LSE10 −LSE11
(12)
Step-7: Estimation of LST using the algorithm in equation 13 implemented. Finally, the LST
in kelvin was determined using SW algorithm.
LST =TB10 +C1(TB10−TB11)+C2(TB10−TB 11 )2+C0+(C3+C4W)(1−m)+( C5+C6W)Δm
(13)
LST =TB10 +1 . 378 (TB10−TB11 )+0 . 183 (TB10−TB11 )2−0 . 268 +(54 .300−2 . 238(0 . 013 )
(1−MeanLSE )+(−129 .200+16 . 400 (0 . 013 ))( DifferenceofLSE )
Where,
TB10
and
TB11
– TB of band 10 and 11
C0
-
C9
– SW coefficient values
m
– Mean
LSE
Δm
– Difference of
LSE
W
– Atmospheric water-vapour content
Table 6. NDVI for Soil and Vegetation
NDVI for Soil 0.15
NDVI for Vegetation 0.48
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358

7. Analysis and Discussion
(Fig. 6) represent NDVI layer of Beas river basin of date 24 April 2015 derived from Band
5(NIR) and Band 4(RED) of OLI sensor using ARC GIS 10.3 raster calculator. The range of
NDVI is varied from -0.262529 to 0.610834 for SW algorithm (zero for negative NDVI
values using condition). The increase in NDVI range from 0 to 1 indicate for healthy and
green vegetation cover area. From NDVI image we perform reclassification of NDVI image
of (Fig. 6) for soil and vegetation separately and calculate NDVI for soil and vegetation
shown in (Table 6). Take the value of NDVI for soil and vegetation form (Table 6) we
calculate FVC using equation 9 and LSE using equation 10. We implement equation in ARC
GIS 10.3 raster calculator as shown in (Fig. 7) to calculate the difference and mean LSE layer
shown in (Fig. 8) from SWA.
We take TIRS band 10 and 11 to estimate TB in Celsius using the algorithm of equation 7
shown in (Fig. 9 and 11). From (Fig. 10 and 12) of the statistical graph, we observe that class
2 exhibit 36.29% of the total area at a temperature in between -4.98 – 3.28ºC from TB of
Band10 and class 5 of TB of Band 11 exhibit maximum of 31.73% of the total area at a
temperature in between 18.13–26.17ºC. Equation 13 estimate the LST using raster calculator.
(Fig. 13) represent the final LST image of Beas river basin on 24 April 2015. Area statistics
graph of LST layer in (Fig. 14) proof that, it divides the whole area into three major surface
temperature class of 2, 3 and 4 with statistics of 44.85 % of the area under an average surface
temperature of -22.24ºC in class 3 the statistics of 21.25% area under an average surface
temperature of -11.21 ºC and in class 4 the statistics of 30.23% area under an average surface
temperature of -0.18 ºC. Remain 2.38% of an area by class 1exhibit and an average of
-33.28ºC.
NDVI map revealed that the NDVI value ranged between -0.26 to 0.61. The south-western
part of Beas river basin had highest NDVI value whereas area under snow cover area had a
negative value (Fig. 6). LSE was created using NDVI threshold technique (Fig. 8). The LSE
of Beas river basin ranged between 1.44 and 1.46. Highly elevated regions in the basin had a
more snow cover, hence LSE was low in these regions. High LSE was found in western and
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389

south-western parts of the basin, whereas low LSE was noticed in northern and eastern parts
of the study area.
(Fig. 13) has been derived using LSE, TB and emissivity difference between LSE of band 10
and 11 of TIR. LST output portrayed that it varied from less than -38.79 ºC to more than 27.4
ºC. The highest LST of more than 27.4ºC was traced in the southern plains of the study area,
where barren lands and wastelands were mostly found. The lowest category of less than
-38.79 ºC was seen in the highly elevated regions with snow cover area.
Fig. 6. NDVI image of 24 April 2015 from SWA
Fig. 7. Difference LSE image between Band 10 and 11 of 24 April 2015
390
391
392
393
394
395
396
397
398
399
400
401
402
403

Fig. 8. Mean of LSE image between Band 10 and 11 of 24 April 2015
Fig. 9. TB of Band 10 with label of temperature intervals
404
405
406
407
408
409
410
411

Fig. 10. Area statistics of classified TB image of Band10
Fig. 11. TB of Band 11 with label of temperature intervals
Fig. 12. Area statistics of classified TB image of Band11
412
413
414
415
416
417
418
419

Fig. 13. LST layer of Beas river basin of 24 April 2015
Fig. 14. Graph of area occupied (%)
Table 6. Statistics of LST layer of Fig. 13
Temperature Class Temperature Interval Average Temperature Area (%)
1 -38.79 – -27.76 -33.27 2.382699
2 -27.76 – -16.73 -22.24 44.857007
3 -16.73 – -5.69 -11.21 21.252092
4 -5.69 – 5.33 -0.18 30.232649
5 5.33 – 16.36 10.84 1.271105
6 16.36 – 27.40 21.88 0.000903
420
421
422
423
424
425
426
427

Fig. 15. Graph of mid value LST
8. Conclusions
SW algorithm is a dynamic mathematical tool provide the LST information using TB of
thermal bands of TIRS and LSE factor derived from FVC of optical bands of OLI sensor.
44.857007% of the total area is under the temperature of -27.76 – -16.73ºC and 21.252092%
of the total area is under the temperature of -16.73 – -5.69ºC other 30.232649% of the total
area are under the temperature of -5.96 – 5.33ºC. Overall we say around 96.341748% of the
total area exhibit a surface the temperature of -11.21ºC in 24 April 2015. From (Fig. 15) we
observe that average LST is -5.695ºC Thus, LST can be calculated using SW algorithm on
Landsat 8 with multiband OLI and TIRS images.
LST of an area was determined based on its TB and LSE using SW algorithm. In this study,
OLI and TIRS bands of Landsat 8 had been used. The study clearly revealed that as the basin
had more snow cover in hilly regions the LST in north-eastern part was low and the south-
western plains with barren lands, uncultivable land, and urban areas experienced high LST.
Beas river basin being a snow cover area under vegetation was less and it is restricted to hilly
areas in the south-western and south-eastern parts. Thus, LST can be calculated using SW
algorithm on Landsat 8 with multiband OLI and TIRS images.
Acknowledgement
The authors would like to thank USGS EarthExplorer for online availability of Landsat 8 and
ASTER GDEM data with free of cost.
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453

References
Akhoondzadeh M, Saradjian MR (2008) Comparison of Land Surface Temperature Mapping
Using Modis and Aster Images in Semi-Arid Area. Int Arch Photogramm Remote Sens
Spat Inf Sci XXXVII:873–876.
Alipour T, Esmaeily A (2005) Land Surface Temprature Estimation From Thermal Band of
Landsat Sensor , Case Study : Alashtar City.
Anbazhagan S, Paramasivam CR (2016) Statistical Correlation between Land Surface
Temperature (LST) and Vegetation Index (NDVI) using Multi-Temporal Landsat TM
Data. Int J Adv Earth Sci Eng 5:333–346.
Barsi JA, Schott JR, Hook SJ, et al (2014) Landsat-8 thermal infrared sensor (TIRS)
vicarious radiometric calibration. Remote Sens 6:11607–11626. doi:
10.3390/rs61111607
Becker F (1987) The impact of spectral emissivity on the measurement of land surface
temperature from a satellite. INT J Remote Sens 8:1509–1522. doi:
10.1080/01431168708954793
Becker F, Li ZL (1990) Toward a local split window method over land surface. Int J Remote
Sens 11:369–393. doi: 10.1080/01431169008955028
Caselles V, Rubio E, Coll C, Valor E (1998) Thermal band selection for the PRISM
instrument 3. Optimal Band Configurations. J Geophys Res 103:17057–17067.
Coll C, Caselles V (1997) A split-window algorithm for land surface temperature from
advanced very high resolution radiometer data: Validation and algorithm comparison. J
Geophys Res 102:16697. doi: 10.1029/97JD00929
Cuenca RH, Ciotti SP, Hagimoto Y (2013) Application of landsat to evaluate effects of
irrigation forbearance. Remote Sens 5:3776–3802. doi: 10.3390/rs5083776
Du C, Ren H, Qin Q, et al (2014) Split-Window algorithm for estimating land surface
temperature from Landsat 8 TIRS data. Int Geosci Remote Sens Symp 3578–3581. doi:
10.1109/IGARSS.2014.6947256
Feizizadeh B, Blaschke T (2012) Thermal Remote Sensing for Land Surface Temperature
Monitoring : Maraqeh County , Iran 2 . Methods and Data. IEEE Int Geosci Remote
Sens Symp (IGARSS 2012) 2217–2220. doi: 10.1109/IGARSS.2012.6350808
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483

Galve JM, Coll C, Caselles V, et al (2008) Comparison of split-window and single-channel
methods for land surface temperature retrieval from MODIS and aatsr data. Int Geosci
Remote Sens Symp 3:294–297. doi: 10.1109/IGARSS.2008.4779341
Irons JR, Dwyer JL, Barsi JA (2012) The next Landsat satellite: The Landsat Data Continuity
Mission. Remote Sens Environ 122:11–21. doi: 10.1016/j.rse.2011.08.026
Javed Mallick YK, B.D.Bharath (2008) Estimation of land surface temperature over Delhi
using Landsat-7 ETM+. J Ind Geophys Union 12:131–140.
Jiménez-Muñoz JC, Sobrino JA (2008) Split-window coefficients for land surface
temperature retrieval from low-resolution thermal infrared sensors. IEEE Geosci Remote
Sens Lett 5:806–809. doi: 10.1109/LGRS.2008.2001636
Jiménez-muñoz JC, Sobrino J a, Skokovi D, et al (2014) Land Surface Temperature Retrieval
Methods From Landsat-8 Thermal Infrared Sensor Data. 11:1840–1843.
Kalma JD, McVicar TR, McCabe MF (2008) Estimating land surface evaporation: A review
of methods using remotely sensed surface temperature data. Surv Geophys 29:421–469.
doi: 10.1007/s10712-008-9037-z
Khandelwal S, Goyal R, Kaul N, Mathew A (2016) Assessment of land surface temperature
variation due to change in elevation of area surrounding Jaipur, India. Egypt J Remote
Sens Sp Sci 1–8. doi: 10.1016/j.ejrs.2017.01.005
Kustas W, Anderson M (2009) Advances in thermal infrared remote sensing for land surface
modeling. Agric For Meteorol 149:2071–2081. doi: 10.1016/j.agrformet.2009.05.016
Li ZL, Tang BH, Wu H, et al (2013) Satellite-derived land surface temperature: Current status
and perspectives. Remote Sens Environ 131:14–37. doi: 10.1016/j.rse.2012.12.008
Mallick J (2014) Land Characterization Analysis of Surface Temperature of Semi-Arid
Mountainous City Abha, Saudi Arabia Using Remote Sensing and GIS. J Geogr Inf Syst
664–676. doi: 10.4236/jgis.2014.66055
Mao K, Qin Z, Shi J, Gong P (2005) A practical split‐window algorithm for retrieving land‐
surface temperature from MODIS data. Int J Remote Sens 26:3181–3204. doi:
10.1080/01431160500044713
McMillin L (1975) Estimation of sea surface temperatures from two infrared window
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512

measurements with different absorption. J Geophys Res 80:5113–5117.
Nikam BR, Ibragimov F, Chouksey A, et al (2016) Retrieval of land surface temperature from
Landsat 8 TIRS for the command area of Mula irrigation project. Environ Earth Sci
75:1–17. doi: 10.1007/s12665-016-5952-3
Qin Z, Karnieli A, Berliner P (2001) A mono-window algorithm for retrieving land surface
temperature from Landsat TM data and its application to the Israel-Egypt border region.
Int J Remote Sens 22:3719–3746. doi: 10.1080/01431160010006971
Quattrochi D a., Luvall JC (1999) Thermal Infrared Remote Sensing for Analysis of
Landscape Ecological Processes: Methods and Applications. 577–598. doi:
10.1002/9781118801628.ch03
Rozenstein O, Qin Z, Derimian Y, Karnieli A (2014) Derivation of land surface temperature
for landsat-8 TIRS using a split window algorithm. Sensors (Switzerland) 14:5768–
5780. doi: 10.3390/s140405768
Singh P, Ramasastri KS, Kumar N (1995) Topographical Influence on Precipitation
Distribution in Different Ranges of Western Himalayas. Hydrol Res 26:259–284.
Skokovic D, Sobrino JA, Jiménez Muñoz JC, et al (2014) Calibration and Validation of land
surface temperature for Landsat8- TIRS sensor TIRS LANDSAT-8
CHARACTERISTICS. L Prod Valid Evol ESA/ESRIN 27.
Sobrino JA, El Kharraz J, Li ZL (2003) Surface temperature and water vapour retrieval from
MODIS data. Int J Remote Sens 24:5161–5182. doi: 10.1080/0143116031000102502
Sobrino JA, Li Z, Stoll MP, Becker F (1996) Multi-channel and multi-angle algorithms for
estimating sea and land surface temperature with ATSR data. Int J Remote Sens
17:2089–2114. doi: 10.1080/01431169608948760
Sobrino JA, Raissouni N (2000) Toward remote sensing methods for land cover dynamic
monitoring: Application to Morocco. Int J Remote Sens 21:353–366. doi:
10.1080/014311600210876
Sobrino JA, Romaguera M (2004) Land surface temperature retrieval from MSG1-SEVIRI
data. Remote Sens Environ 92:247–254. doi: 10.1016/j.rse.2004.06.009
Sobrino JA, Skoković D, Jiménez-Muñoz JC (2015) Spatial analysis of the homogeneity of
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541

the land surface temperature in three Spanish test sites. Int J Remote Sens 36:1–15. doi:
10.1080/01431161.2015.1055611
Song L, Liu S, Kustas WP, et al (2015) Using the Surface Temperature-Albedo Space to
Separate Regional Soil and Vegetation Temperatures from ASTER Data. Remote Sens
7:5828–5848. doi: 10.3390/rs70505828
Tang B, Bi Y, Li Z-L, Xia J (2008) Generalized Split-Window Algorithm for Estimate of
Land Surface Temperature from Chinese Geostationary FengYun Meteorological
Satellite (FY-2C) Data. Sensors 8:933–951. doi: 10.3390/s8020933
Ulivieri C, Castronuovo MM, Francioni R, Cardillo A (1994) A split window algorithm for
estimating land surface temperature from satellites. Adv Sp Res 14:59–65. doi:
10.1016/0273-1177(94)90193-7
Weng Q (2009) Thermal infrared remote sensing for urban climate and environmental
studies: Methods, applications, and trends. ISPRS J Photogramm Remote Sens 64:335–
344. doi: 10.1016/j.isprsjprs.2009.03.007
Xia L, Mao K, Ma Y, et al (2014) An algorithm for retrieving land surface temperatures using
VIIRS data in combination with multi-sensors. Sensors (Switzerland) 14:21385–21408.
doi: 10.3390/s141121385
Zhao S, Qin Q, Yang Y, et al (2009) Comparison of two split-window methods for retrieving
land surface temperature from MODIS data. J Earth Syst Sci 118:345–353. doi:
10.1007/s12040-009-0027-4
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563