Land Surface Temperature Retrieval of Landsat-8 Data using Split Window Algorithm-A Case Study of Mosul District

Abstract

The main purpose of this paper is to present an operational algorithm to retrieve the Land Surface Temperature (LST) and Land Surface Emissivity (LSE) in Mosul District, Ninawa Province, Iraq from Landsat-8 data of September 16, 2015. The proposed algorithm is Split-Window (SW) with brightness temperature value of both band 10 (10.60 - 11.19µm) and band 11 (11.50 - 12.51 µm) of Landsat-8 in thermal infrared range. Normalized Difference Vegetation (NDVI) threshold values have been determined to separate the bare soil, and vegetated areas from each other. Fractional Vegetation Cover (FVC) was derived with the help of NDVI threshold technique. Emissivity values of bands 10 and 11 are calculated through FVC. The results showed that the spatial variation of land surface temperature was more reliable and accurate in entire Mosul District.

Full text

Journal of American Science 2017;13(12) http://www.jofamericanscience.org
62
Land Surface Temperature Retrieval of Landsat-8 Data Using Split Window Algorithm-A Case Study of
Mosul District
Amal Muhammad Saleh
College of Agriculture /University of Baghdad
geetakh@gmail.com
Abstract: The main purpose of this paper is to present an operational algorithm to retrieve the Land Surface
Temperature (LST) and Land Surface Emissivity (LSE) in Mosul District, Ninawa Province, Iraq from Landsat-8
data of September 16, 2015. The proposed algorithm is Split-Window (SW) with brightness temperature value of
both band 10 (10.60 - 11.19µm) and band 11 (11.50 - 12.51 µ m) of Landsat-8 in thermal infrared range. Normalized
Difference Vegetation (NDVI) threshold values have been determined to separate the bare soil, and vegetated areas
from each other. Fractional Vegetation Cover (FVC) was derived with the help of NDVI threshold technique.
Emissivity values of bands 10 and 11 are calculated through FVC. The results showed that the spatial variation of
land surface temperature was more reliable and accurate in entire Mosul District.
[Amal Muhammad Saleh. Land Surface Temperature Retrieval of Landsat-8 Data Using Split Window
Algorithm-A Case Study of Mosul District. J Am Sci 2017;13(12):62-75]. ISSN 1545-1003 (print); ISSN 2375-
7264 (online). http://www.jofamericanscience.org. 8. doi:10.7537/marsjas131217.08.
Keywords: Land Surface Temperature; Land Surface Emissivity; Fractional Vegetation Cover; Split-Window;
NDVI.
1. Introduction
Land surface temperature (LST) is one of the
most important parameters in the physical processes
of surface energy and water balance at local through
global scales (Karnieli et al., 2010). Knowledge of the
LST provides information on the temporal and spatial
variations of the surface equilibrium state and is of
fundamental importance in many applications (Kerr et
al., 2000). As such, the LST is widely used in a
variety of fields including evapotranspiration, climate
change, hydrological cycle, vegetation monitoring,
urban climate and environmental studies, among
others (Hansen et al., 2010). and has been recognized
as one of the high-priority parameters of the
International Geosphere and Biosphere Program
(IGBP) (Townshend et al., 1994). Due to the strong
heterogeneity of land surface characteristics such as
vegetation, topography, and soil, LST changes rapidly
in space as well as in time and an adequate
characterization of LST distribution and its temporal
evolution, therefore, requires measurements with
detailed spatial and temporal sampling (Neteler,
2010).
Given the complexity of surface temperature
over land, ground measurements cannot practically
provide values over wide areas. With the development
of remote sensing from space, satellite data offer the
only possibility for measuring LST over the entire
globe with sufficiently high temporal resolution and
with complete spatially averaged rather than point
values. (Li et al., 2013).
The major objectives of the study are to find the
brightness temperature using band 10 and band 11 of
TIR, calculate the LSE using NDVI threshold
technique and estimate the LST of Mosul District
using Split-Window (SW) algorithm.
2. Material and Methods
Mosul is a city in northern Iraq Located some
400 km (250 mi) north of Baghdad, the original city
stands on the west bank of the Tigris River, opposite
the ancient Assyrian city of Nineveh on the east bank
(Figure 1a). It is geographically situated on latitude
35º 33 54.67-36º 32 03.79 North and longitude 42º
43 12.46 - 43º 03 21.78 East. Mosul experiences a
hot semi-arid climate (Köppen climate classification
BSh) (Peel et al., 2007). The winter is mild, but
certainly it's not tropical; the January average is 7 °C.
The summer in Mosul is very hot, with a relentless
sun, and with daytime temperatures of 43°C in July
and August, but with peaks of 47/48°C; however, the
relative humidity is between 25% - 79% (Brugge,
2014). Mosul is at an altitude of about 228 metres
(748ft) above sea level. Throughout the year, in
Mosul, 365 mm of rain fall: they are not many, but
they are concentrated between November and April,
with very few rains in May and October, while
between June and September it almost never rains
(Dessler and Parson, 2006). The average annual wind
speed of Mosul is 1.26 m/s, (Hassoon, 2013).
Landsat 8 is one of the Landsat series of NASA.
The data of Landsat 8 is available in Earth Explorer
website at free of cost. In the present study, the TIR
bands 10 and 11 were used to estimate brightness
temperature and OLI spectral bands 2, 3, 4, and 5
were used to generate NDVI of the study area (Figure

Journal of American Science 2017;13(12) http://www.jofamericanscience.org
63
1b), (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.
Figure 1. (a) Location map of the study area, (b) The Satellite Landsat OLI and TIRS scene.
The value of Top of Atmospheric (TOA) spectral
radiance (L) was determined by multiplying
multiplicative rescaling factor (0.0003342) of TIR
bands with its corresponding TIR bands and adding
additive rescaling factor (0.1) with it (Table 2).
Lλ =      (1)
Where:
Lλ - Top of Atmospheric Radiance in watts/(m2 *
srad * m).
ML - Band specific multiplicative rescaling factor
(radiance_mult_band_10/11).
Qcal - band 10/ 11 image.
AL - Band specific additive rescaling factor
(radiance_add_band_10/11).
Estimation of Brightness Temperature (TB) of
Band 10 and 11 is the electromagnetic radiation
traveling upward from the top of the Earth’s
atmosphere. Thermal calibration process has been
done by converting thermal DN values of raw thermal
bands of TIR sensor into Top of Atmospheric (TOA)
Spectral Radiance.
Using ERDAS IMAGINE 9.2 Modeler we
implement algorithm of equation-2.
TB = 
󰇡
󰇢 (2)
Where:
TB - Brightness Temperature.
Lλ - Top of Atmospheric spectral radiance in
watts/
(m2 * srad * m).
K1 and K2 - Band-specific thermal conversion
constant from the metadata image file and it varies for
both TIR bands (Table 3).

Journal of American Science 2017;13(12) http://www.jofamericanscience.org
64
OLI bands 2, 3, 4 and 5 were layer stacked and
NDVI was calculated using algorithm shown in
equation-3 (Rouse et al., 1974).
(3)
Range: -1 < NDVI < + 1
Where:
NDVI - Normalized Difference Vegetation
Index.
BAND4 - TOA planetary reflectance values from
Red band.
BAND5 - TOA planetary reflectance values from
Near Infrared band.
The relationship between LST and NDVI takes
into account that vegetation and soils are the main
surface cover for the terrestrial component (Snyder et
al., 1998).
Table 1. Metadata of Satellite Image.
Sensor
No. of Bands
Resolution (m)
Path/ Row
Date of Acquisition
OLI
Panchromatic (1)
15.00
170 / 035
16th September
2015
Reflective (8)
30.00
170 / 035
TIRS
Thermal (2)
30.00
170 / 035
Table 2. Radiometric rescaling Factor.
Rescaling Factor
Band 10
Band 11
ML
0.0003342
0.0003342
AL
0.10000
0.10000
Table 3. Thermal constant values.
Band
K1
K2
10
774.8853
1321.0789
11
480.8883
1201.1442
Estimation of Fractional Vegetation Cover
(FVC) was implemented using NDVI image.
Fractional Vegetation Cover estimate the fraction of
an area under vegetation. Split-Window algorithm
utilize FVC to estimate Land Surface Emissivity
(LSE). Using ARC MAP 10.3 we reclassify the NDVI
layer into soil and vegetation and calculate NDVI for
Soil and Vegetation (Table 4).
Table 4. NDVI for Soil and Vegetation.
NDVI for Soil
0.060304167
NDVI for Vegetation
0.557708025
Using ERDAS IMAGINE 9.2 Modeler we
calculated the algorithm of FVC of equation-4
(Carlson and Ripley, 1997; Gutman and Ignatov,
1998).
Where:
FVC - Fractional Vegetation Cover.
NDVIs - NDVI reclassified for soil.
NDVIv - NDVI reclassified for vegetation.
The correct determination of the surface
temperature is constrained to an accurate knowledge
of surface emissivity. The emissivity of a surface is
controlled by such factors as water content, chemical
composition, structure and roughness. It can be
determined as the contribution of the different
components that belong to the pixels according to
their proportions (Snyder et al., 1998).
Estimation of Land Surface Emissivity (LSE)
was obtained from FVC layer. Land Surface
Emissivity measure the inherent characteristic of earth
surface. LSE estimation required emissivity of soil
and vegetation of both Band 10 and 11 (Table 5).
Table 5. Emissivity values.
Emissivity
Band 10
Band 11
Ɛs
0.971
0.977
Ɛv
0.987
0.989
[Source: Skoković et al., 2014]
LSE of Band 10 and 11 were individually
calculated as shown in equation-5.
LSE = Ɛs * (1 - FVC) + Ɛv * FVC (5)
Where:
LSE - Land Surface Emissivity.
Ɛs - Emissivity for soil.
Ɛv - Emissivity for vegetation.
FVC - Fractional Vegetation Cover.
Combination of LSE of Band 10 and LSE of
Band 11 was obtained through Mean and Difference

Journal of American Science 2017;13(12) http://www.jofamericanscience.org
65
between them as shown in equation-6 & 7 (Sobrino et
al., 1996).
∆m = LSE10 – LSE11 (7)
Where:
m - Mean of LSE.
LSE10 - LSE of band 10.
LSE11 - LSE of band 11.
∆m - Difference of LSE
Land Surface Temperature (LST) was calculated
by applying a structured mathematical algorithm viz.,
Split-Window (SW) algorithm. It uses brightness
temperature of two bands of TIR, mean and difference
in land surface emissivity for estimating LST of an
area. Using ERDAS IMAGINE 9.2 Modeler we
implement the algorithm of equation-8.
LST = TB10 + C1 (TB10 -TB11) +C2 (TB10 -TB11)2 +
C0 + (C3+C4W) (1-m) + (C5+C6W)∆m (8)
Where:
LST - Land Surface Temperature (Kº).
TB10 and TB11 - Brightness Temperature of band
10 and band 11 (Kº).
C0 To C6 - Split-Window Coefficient values
(Table 6).
m - mean LSE of TIR bands.
∆m - Difference of LSE.
W- Atmospheric water-vapour content =14.215
g/cm2 [Source: Iraqi Meteorological Organization And
Seismology, 2015].
Table 6. Split-Window coefficient values.
Constant
Value
C0
-0.268
C1
1.378
C2
0.183
C3
54.300
C4
-2.238
C5
-129.200
C6
16.400
[Source: Skoković et al., 2014]
3. Results
NDVI map revealed that the NDVI value varies
between -0.0923236 to 0.557708. Healthy and green
vegetation cover area had highest NDVI value
whereas area under water body had negative value
(Figure 2). The NDVI value of area under vegetation
was more than 0.060304167 and for built-up and
barren land was 0 - 0.060304167. We calculate
Fractional vegetation cover (FVC) using equation-4 as
shown in Figure (3) and LSE using equation-5.
We implement algorithm in ERDAS IMAGINE
9.2 Modeler to calculate difference and mean LSE
Layer shown in Figure (4) and Figure (5). Mean LSE
Layer of Mosul Province ranged between 0.969 -
0.988. Highly elevated regions in the Province had
more vegetative cover, hence LSE was high in these
regions.
We take TIRS band 10 and 11 to estimate
Brightness Temperature (TB) in Kelvin using the
algorithm of equation-2 shown in Figure (6) and
Figure (7). From Figure (8) and Figure (9) of
statistical graph, we observe that class 5 exhibit
28.90% of the total area at a temperature between
302.28 -315.60 Kº from TB of Band 10 and also class
5 of TB of Band 11 exhibit maximum of 27.68% of
total area at a temperature between 298.52-311.11Kº.
Figure (10) represent the final LST layer of
Mosul Province on 16th September 2015. Area
statistics graph of LST layer in Figure (11) proof that,
it divide into two major land surface temperature class
of 1 and 4 with statistics of 28.97% of an area under
foothill regions at a temperature less than 269.20 Kº
by class 1 and 30.66% of an area under healthy
vegetation land cover at a temperature in between
311.54 - 334.64 Kº by class 4. Water bodies exhibit
20.83% of the total area at a temperature in between
292.29 - 311.54 Kº by class 3. Cultivable land exhibit
17.52% of the total area at a temperature in between
269.20 - 292.29 Kº by class 2. Remain 2.02% of an
area under barren land and wasteland exhibit an LST
more than 334.64 Kº by class 5.
Validation of LST is of crucial importance for
estimating the accuracy of the products and
understanding the potential and limitations of satellite
observations of LST. Validation of LST was carried
out using a direct comparison of satellite-derived LST
with collocated and simultaneously acquired LST
from ground-based on 16th September 2015 (Iraqi
Meteorological Organization And Seismology, 2015).
Figure (12) shows a Linear regression of estimated vs.
actual LST, the small bias & RMSE and the high
coefficient of determination R2 demonstrate the
excellent quality of the Satellite Application Facility.
The cold outliers are assumed to be caused by
undetected cloud contamination or instrument -related
problems.
4. Discussions
LST and LSE are two significant parameters in
global change studies, in estimating radiation budgets
in heat balance studies and as a control index for
climate models. Emissivity, is an indicator of land-
cover type and resources, and also a necessary
element in the calculation of Land surface temperature
(LST) from remotely sensed data. For LSE mapping,
Fractional Vegetation Cover (FVC) of optical bands
of OLI sensor of Landsat 8 had been derived. Split-
Window algorithm (SW) is a dynamic mathematical

Journal of American Science 2017;13(12) http://www.jofamericanscience.org
66
tool, provide the (LST) information using brightness
temperature of thermal bands of TIRS sensor and
Land surface emissivity (LSE). The study clearly
revealed that 30.66% of the total area are under
vegetation land cover at a temperature of 311.54 -
334.64 Kº and other 28.97% of the total area are under
hilly regions at a temperature less than 269.20 Kº.
Therefore the results indicate that the proposed (SW)
algorithm can be a suitable and robust method to
retrieve the LST map from Landsat-8 satellite data.
Figure 2. NDVI Layer of Mosul District on 16th September 2015.

Journal of American Science 2017;13(12) http://www.jofamericanscience.org
67
Figure 3. FVC Layer of Mosul District on 16th September 2015.

Journal of American Science 2017;13(12) http://www.jofamericanscience.org
68
Figure 4. Difference LSE layer between Band 10 and 11.

Journal of American Science 2017;13(12) http://www.jofamericanscience.org
69
Figure 5. Mean of LSE layer between band 10 and 11.

Journal of American Science 2017;13(12) http://www.jofamericanscience.org
70
Figure 6. TB of Band 10 with label of Temperature intervals.

Journal of American Science 2017;13(12) http://www.jofamericanscience.org
71
Figure 7. TB of Band 11 with label of Temperature intervals.

Journal of American Science 2017;13(12) http://www.jofamericanscience.org
72
Figure 8. Area statistics of classified TB layer of Band 10.
Figure 9. Area statistics of classified TB layer of Band 11.

Journal of American Science 2017;13(12) http://www.jofamericanscience.org
73
Figure 10. Land Surface Temperature Layer of Mosul District on 16th September 2015.

Journal of American Science 2017;13(12) http://www.jofamericanscience.org
74
Figure 11. Graph of Area occupied (%).
Figure 12. Scatter plot of estimated vs. actual LST.

Journal of American Science 2017;13(12) http://www.jofamericanscience.org
75
Reference
1. Karnieli A, Agam N, Pinker RT, Anderson M,
Imhoff ML, Gutman GG, Panov N, Goldberg A.
Use of NDVI and land surface temperature for
drought assessment: Merits and limitations.
Journal of Climate 2010;23:633-618.
2. Kerr YH, Lagouarde JP, Nerry F, Ottlé C. Land
surface temperature retrieval techniques and
applications. In D. A. Quattrochi, & J. C. Luvall
(Eds.), Thermal remote sensing in land surface
processes, Boca Raton, Fla.: CRC Press,
2000:109-33.
3. Hansen J, Ruedy R, Sato M, Lo K. Global
surface temperature change. Reviews of
Geophysics 2010;48(4): RG4004.
4. Townshend JRG, Justice CO, Skole D,
Malingreau JP, Cihlar J, Teillet P, Sadowski F,
Ruttenberg S. The 1 km resolution global data
set: needs of the International Geosphere
Biosphere Programme. International Journal of
Remote Sensing 1994;15(17):3441-3417.
5. Neteler M. Estimating daily land surface
temperatures in mountainous environments by
reconstructed MODIS LST Data. Remote
Sensing 2010;2(1):351-333.
6. Li ZL, Tang B-H, Wu H, Ren H, Yan G, Wan Z,
Trigo IF, Sobrino JA. Satellite-derived land
surface temperature: Current status and
perspectives. Remote Sensing of Environment
2013;131:37-14.
7. Peel MC, Finlayson BL, McMahon TA. Updated
world map of the Köppen -Geiger climate
classification. Hydrology and Earth System
Sciences 2007;11:1644-1633.
8. Brugge R. World weather news, August 2011.
Department of Meteorology, University of
Reading, Archived from the original on 29 June
2014.
9. Hassoon AF. Determination trends and abnormal
seasonal wind speed in Iraq. International
Journal of Energy and Environment
2013;4(4):628-615.
10. Rouse JW, Haas RH, Schell JA, Deering DW.
Monitoring vegetation systems in the Great
Plains with ERTS. In Fraden S.C., Marcanti E.P.
& Becker M.A. (eds.), Paper presented at the 3rd
ERTS-1 Symposium, NASA SP-351,
Washington D.C. NASA 1974a:317-309.
11. Snyder WC, Wan Z, Zhang Y, Feng YZ.
Classification based emissivity for land surface
temperature measurement from space,
International Journal of Remote Sensing
1998;19(14):2774-2753.
12. Carlson TN, Ripley DA. On the relation between
NDVI, fractional vegetation cover, and leaf area
index. Remote Sensing of Environment
1997;62(3):252-241.
13. Gutman G, Ignatov A. The derivation of the
green vegetation fraction from NOAA/AVHRR
data for use in numerical weather prediction
models. International Journal of Remote Sensing
1998;19(8):1543-1533.
14. Skoković D, Sobrino JA, Jimenez-Munoz JC,
Soria G, Julien Y, Mattar C, Cristobal J.
Calibration and validation of land surface
temperature for Landsat 8-TIRS sensor. Land
product Validation and Evolution, ESA/ESRIN
Frascati (Italy), 2014:9-6.
15. Sobrino JA, Li ZL, Stoll MP, Becker F. Multi-
channel and multi-angle algorithms for
estimating sea and land surface temperature with
ATSR data. International Journal of Remote
Sensing 1996;17(11):2114-2089.
12/20/2017