Paleochannel Delineation Using Landsat 8 OLI and Envisat ASAR Image Fusion Techniques

Paleochannel delineation using
Landsat 8 OLI and Envisat ASAR

image fusion techniques in Cholistan
desert, Pakistan
Zaheer Ul Islam
Javed Iqbal
Junaid Aziz Khan
Waqas A. Qazi
Zaheer Ul Islam, Javed Iqbal, Junaid Aziz Khan, Waqas A. Qazi, Paleochannel delineation using Landsat
8 OLI and Envisat ASAR image fusion techniques in Cholistan desert, Pakistan, J. Appl.
Remote Sens. 10(4), 046001 (2016), doi: 10.1117/1.JRS.10.046001.
Downloaded From: http://remotesensing.spiedigitallibrary.org/ on 10/16/2016 Terms of Use: http://spiedigitallibrary.org/ss/termsofuse.aspx
Paleochannel delineation using Landsat 8 OLI and

Envisat ASAR image fusion techniques in
Cholistan desert, Pakistan
Zaheer Ul Islam,a Javed Iqbal,a,* Junaid Aziz Khan,a and Waqas A. Qazib
a
National University of Sciences and Technology, Institute of Geographical Information

Systems, Sector H-12, Islamabad 44000, Pakistan
b
Institute of Space Technology, Geospatial Research and Education Lab, Department of Space
Science, Islamabad Highway, Islamabad 44000, Pakistan
Abstract. Sustainability of desert ecosystem is highly dependent upon water availability from
different sources. Paleochannels are important sources of groundwater, and exploiting such
resources involves their identification/mapping and subsequent investigation for fresh groundwater. A study in which multisensor (optical/infrared Landsat 8 OLI and active microwave
Envisat ASAR) images of the Cholistan desert of Pakistan were processed and analyzed to identify and map Hakra River paleochannels is presented. Radiometrically corrected optical and
synthetic aperture radar datasets were fused using principal components image fusion method.
The paleochannels were extracted from the analysis of this fused output, and normalized difference vegetative index analysis of Landsat 8 OLI atmospheric corrected images was used as supporting information. Identification and alignment of an identified paleochannel was validated
with geophysical ground measurements (electrical resistivity and conductivity surveys) and
historical records. The presence of high apparent electrical resistivity with corresponding low
soil water conductivity values intersects well with the paleochannels identified from the remote
sensing data. The results were also confirmed with historical evidence such as old wells beside
forts and proposed ground water harvesting sites. The proposed methodology in this study could
be adopted in other parts of the world for mapping of paleochannels. 2016 Society of PhotoOptical Instrumentation Engineers (SPIE) [DOI: 10.1117/1.JRS.10.046001]
Keywords: paleochannel; aquifer; geophysical; Landsat 8; Envisat ASAR.

Paper 16003 received Jan. 8, 2016; accepted for publication Sep. 20, 2016; published online Oct.
13, 2016.
1 Introduction
Arid and semiarid regions like Cholistan are mostly water-deprived areas, with sole dependence
on rainfall. Rain freshwater is collected in natural depressions/manmade ponds locally called
tobas. Out of 1500 tobas, only 500 are operational in the area due to absence of a sustainable
system. Though Pakistan Council of Research in Water Resources (PCR WR) has constructed
92 rainwater harvesting systems as pilot projects in the region, far flung areas are not in a good
state with respect to water availability.1 Freshwater demand of such arid regions could be best met
by subsurface massive storage capacity, which should prevent seepage and evaporation and could
be recharged through rainwater/diverted floodwater. Recharging of paleochannels is one of the
suitable measures for drought proofing;2 so, such ideal storage capacity could naturally exist in
the form of the Hakra river paleochannel in the Cholistan desert.3 Traces of ancient civilizations
provide evidence that this desert used to be the site of plenty of water and agricultural activity.
Paleochannels are dried up old river beds, which either change their courses due to past
seismic or flooding activities or cease to exist due to various climatological factors.4,5 In
arid regions, these relic waterways get filled with highly porous, permeable fluvial, and aeolian
*Address all correspondence to: Javed Iqbal, E-mail: javed@igis.nust.edu.pk

1931-3195/2016/$25.00 2016 SPIE
Journal of Applied Remote Sensing
046001-1
OctDec 2016
Vol. 10(4)
Islam et al.: Paleochannel delineation using Landsat 8 OLI and Envisat ASAR image fusion techniques. . .
sediments, which act as preferred pathways for groundwater flow. Identification of paleochannels can be a good indicator for exploring new sites of groundwater.6 Several studies have been
performed all over the world to extract these dried up river courses, such as published works.2,5,712
These paleochannels are considered very economical and an appropriate option to address
freshwater scarcity problems in arid regions. Though geophysical techniques are resource extensive, they are the most precise and accurate methods for paleochannel studies. Spatial mapping
using electromagnetic (EM) conductivity can quickly define past sedimentary environments within
meandering river floodplains.13 Similarly, electrical resistivity measurements can help classify
groundwater conditions for past river course identification. Low electrical resistivity values
represent either soils having fine grained rock type or coarse grained formation containing
saline water. High resistivity values, on the other hand, represent freshwater in sandy soil
formations.14
To explore changes of drainage pattern and paleochannels through traditional methods is
resource extensive in terms of time, money, labor, and materials, whereas use of remote sensing
methods circumvents these issues. Ancient rivers abandoned channels can be identified by combining and enhancing multiple bands of Landsat 7 ETM+ data to bring out slight differences in
the spectral reflection of soil moisture and vegetation.10 Image-enhancement techniques, such as
linear contrast enhancement of optical data, and fusion of microwave and optical image data
bring out more information about the paleochannels.15 Imaging radar datasets of L, C, and
X band frequencies have been used to identify shallow depth subsurface structures up to
few meters.16 Lancaster et al.17 identified paleodrainages of former courses of the Tsondab
and Kuiseb rivers in the Central Namib Desert using orbital radar images of C-band frequency;
they stated that 0.50 m of sand appears to be sufficient to prevent penetration of the radar
energy of X-, C-, and L-band frequencies to the underlying calcrete-cemented fine gravel.
Sukumar and Sankar18 note that radar imaging depths for X-, C-, L- and P-band frequencies
can range from 0.1 to 2.9 m for lightly compacted, fine-grained, sandy alluvium, and smallgravel alluvium. Similarly, Schaber et al.19 measured radar imaging depth for the X-band to
be 0.10 to 0.3 m, C-band to be 0.2 to 0.5 m, and the L-band to be 0.8 to 2.0 m deep into
blow sand and sandy alluvium in Bir Safsaf, within the hyper arid region of the Sahara desert,
Egypt. They ranked the utility of SIR-C/X-SAR frequency bands and polarizations in order of
decreasing priority, to be L-HV, L-HH (VV), C-HV, C-HH (VV), and X-VV. Zhang et al.20 fused
SAR images and pixel level optical imagery for feature enhancement. They proposed a new
fusion technique called block-based synthetic variable ratio (block-SVR) based on multiple linear regression of block regions to fuse optical and SAR imagery data. The fusion results indicate
enhancement in certain features, i.e., spatial and textural content. Various techniques like compressed sensing-based21 pan sharpening method and adjustable model-based22 fusion method
have also been used for fusion of multispectral and panchromatic imageries. Meng et al.23 proposed an algorithm for spatiotemporal-spectral fusion of multispectral and panchromatic images.
They used very a coarse image of MODIS having a spatial resolution of 1 km to be fused with
high resolution SPOT 5 and Landsat ETM+ with spatial resolutions of 5 and 30 m, respectively.
2 Paleochannel Extraction Techniques

Passive multispectral satellite images record information of EM waves reflected from surface
features. Due to difference in ground objects structure, mineral compositions, and physical and
chemical properties, their reflective response would differ. As paleochannels contain alluvial
soils, they have different structures and mineral compositions than their surrounding soils.
A number of places would be covered by aeolian surrounding sand, but its uncovered segments
would have a different spectral response from surroundings in multispectral as well as radar
images. Second, paleochannels are dried water bodies, but still may often contain sufficient
surface and subsurface soil moisture, especially in rainy seasons. Due to the richness in soil
moisture, such locations have vegetation growth and have high infrared (IR) reflectance, and
thus would appear brighter in IR bands, whereas spots where there is no vegetation will appear
dark in IR bands due to IR spectra absorption. The X-, C-, L-, and P-band microwave frequencies
can penetrate from 0.1 to 2.9 m for lightly compacted, fine-grained, sandy alluvium, and
046001-2
OctDec 2016
Vol. 10(4)
small-gravel alluvium soils and could bring out subsurface geology.18 At longer wavelengths,
like L- and C-bands, the radar signal would penetrate into dry aeolian sands and would bring out
subsurface paleochannels information. For plainer areas, paleochannels would have more fine
grained soils than the surroundings and, therefore, would appear dark in radar image, whereas
paleochannels with gravel beds would appear bright due to more volumetric scattering of the
radar backscatter.
2.1 Multisensor Data Fusion

Data fusion deals with data and information from multiple sources to achieve refined/improved
information. With the availability of multisensor, multitemporal, multiresolution, and multifrequency satellite data, extraction of different information by dataset fusion is a valuable tool in
remote sensing image analysis.16 The term image fusion covers multiple techniques which are
used to combine the geometric detail of a high-resolution panchromatic image and the color
information of a low resolution multispectral image to produce a final image with the highest
possible spatial information content while still preserving good spectral information quality.24
A good example of data fusion in remote sensing is the fusion of images acquired by sensors
sensitive to visible/infrared (VIR) wavelengths with data from active high-resolution synthetic
aperture radar (SAR). The information contained in VIR imagery depends on the multispectral
reflectivity of the target illuminated by sunlight, while SAR image intensities depend on the
characteristics of the illuminated surface target as well as on the signal itself. The fusion of
these disparate data contributes to an even more detailed understanding of the features being
observed.16 Spectral differences in multispectral and radar images can further be enhanced
by image fusion techniques and resultant fused image could be used to distinguish paleochannel
information from its surroundings.
2.2 Segregating Spectral Differences using Normalized Difference

Vegetative Index
Active or intermittent freshwater aquifers would exist under paleochannels or at least would
have more surface moisture due to recharging during rainy seasons. Therefore, multispectral
images during rainy season in arid regions would have some vegetation response along expected
alignments of a paleochannel. Such vegetation can be identified using normalized difference
vegetation index (NDVI) technique:
NDVI
EQ-TARGET;temp:intralink-;e001;116;325
NIR IR
:
NIR IR
(1)
The NDVI ranged25 from 1 to 1. Thus, highest values of the NDVI represent dense biomass
and the lowest values represent bare earth, snow, or ice.26
The objectives of this study were (1) to use image fusion technique on multisensor satellite
imagery (optical/IR Landsat 8 OLI & active microwave Envisat ASAR) for identification and
mapping of paleochannels of the Cholistan desert, Pakistan, and (2) to validate the results with
ground survey data, i.e., geophysical ground measurements (electrical resistivity and conductivity surveys) and historical records.
3 Study Area and Data Sources

3.1 Study Area
Cholistan, the largest desert in the Punjab province of Pakistan, lies along eastern border of
Pakistan-India between longitudes 6952 to 7324 E and latitudes 2842 to 2925 N having
an area of 26;300 km2 at an elevation of 89 m amsl, (Fig. 1). It extends from east to south of an
irrigated track of Bahawalpur district and is bounded by districts Bahawalnagar in the north and
Rahimyar Khan district in the south. Its climate is extremely hot and dry with mean maximum/
minimum temperatures of 55 and 27.5C, respectively, with low and sporadic average rainfall
046001-3
OctDec 2016
Vol. 10(4)
Fig. 1 Map of study area showing the extent of the Cholistan desert, major builtup areas of
the region, and network of Punjab irrigation main canals.
about 160.5 mm annually.1 High temperatures, low humidity, high rate of evaporation, and
strong dust storms are the main characteristics of the summer climate. The winter season is
short starting from December to February with mean maximum/minimum temperature values
of 22 and 6C, respectively.
3.2 Datasets Used

Remote sensing datasets used for mapping the Hakra paleochannel were multiband images of
Landsat 8 with 30-m spatial resolution acquired on 7 and 14 September 2013, and Envisat ASAR
images27 with 12.5-m spatial resolution, swath coverage of 100 km with VV-polarization
acquired on 2, 5, 22, and 25 August 2010. To capture some signs of paleochannels in arid regions
like limited scattered vegetation, variation in soil moisture content, small beaded lakes, swamps,
and some perennial surface flows and so on, Landsat 8 data of the last period of the summer
monsoon were used. Landsat 8 OLI band 3 (Green), 5 (NIR), and (SWIR) were selected to
capture slight variations in soil moisture content and limited scattered vegetation growth.
Similarly, to capture maximum subsurface geology, Envisat ASAR images for the hottest/driest
period of the year were acquired from European Space Agency.
For validation of the Hakra paleochannel extracted from remote sensing datasets, electrical
resistivity and conductivity survey data were acquired from Water and Power Development
Authority (WAPDA), Government of Pakistan. Electrical resistivity method measures the resistance of the subsurface layers to the flow of electric current transmitted into the ground through a
pair of electrodes. The potential difference developed is measured by another pair of electrodes.
Resistivity is calculated using the relationship between resistivity, electric field, and current density (Ohms law), and the geometry of the current and potential electrodes.28 Thickness and true
resistivity of the subsurface layers occurring within the probing depth can be determined either
by matching the observed data with master curves or through computer software. The geological
interpretations could then be made from the obtained results. The WAPDA carried out geoelectrical investigations of an 170 35 km2 area in the Cholistan desert during 1986 in collaboration with Gesellschaft fr Technische Zusammenarbeit, Germany.14 Resistivity readings were
taken along 16 horizontal profiles almost at every 2 km, as shown in Fig. 2. These resistivity
046001-4
OctDec 2016
Vol. 10(4)
Fig. 2 Map of resistivity profiles taken from WAPDA on extreme left.14 The same map is transformed in the GIS environment with locations of electrical resistivity sounding and EC sounding.
Fig. 3 Soil resistivity sectional drawing along resistivity profile G showing locations of resistivity
sounding points (G2, G3,. . . , G10), and location of borehole (CTN19), representing variations in
soil resistivity values in m up to 150-m depth along profile G.14
readings were compiled in the form of sectional drawings (Fig. 3), which represent resistivity
distribution in a vertical section. 14 Electrical conductivity (EC) data of more than 300 existing
water sources (tube wells, dug wells either lined or unlined, canals, and hand pumps) were
acquired from WAPDA.29 EC is a measure of the capacity of water to conduct electrical current;
it is directly related to the concentration of salts dissolved in water or the total dissolved solids
(TDS). It is quickly measured using portable meters in the field. It is used to determine quality of
ground water.30 The EC was converted to TDS using the following calculation:
TDSppm 0.64 X EC Scm 640 X EC dSm:
(2)
EQ-TARGET;temp:intralink-;e002;116;178
4 Methods
This study was divided into two parts: mapping of paleochannels using satellite datasets, and
validation of the mapped Hakra paleochannel by comparison with geophysical-based field
survey data and historical evidence. Figure 4 shows the complete methodology flow chart of
the study.
046001-5
OctDec 2016
Vol. 10(4)
Fig. 4 Data processing methodology flow chart.
4.1 Mapping of Paleochannel

4.1.1 Preprocessing
Before analyzing information from multispectral images obtained from different sensor systems,
sensor calibration is mandatory to convert multisensor image pixel values to physical values of
radiance, which is the at-sensor radiance in Wm2 :sr:m.25 Radiometric calibration and atmospheric correction of Landsat 8 data was performed using the FLAASH model in ENVI 5.1 with
parameter values including radiance scale factor 1, ground elevation 1.37 km, tropical atmospheric and aerosol rural models with 2-Band (K-T) retrieval, initial visibility set up to 40 km,
and multispectral setting as the default with over-land retrieval standard (660:2100 nm).
Multispectral radiance image was prepared for multisensor data fusion and an atmosphericcorrected reflectance image was created for calculation of NDVI. The mosaicked preprocessed
Landsat 8 OLI radiance image is shown in Fig. 5(a).
Radiometric calibration of SAR imagery requires recovery of normalized radar cross-section
termed sigma-nought ( ). Six tiles of ASAR data intensity images were radiometrically
calibrated in ESAs NEST software using each products auxiliary file. SAR images contain
a characteristic grainy speckle noise, which is suppressed by multilook processing on
the raw SAR data.31 Although the Envisat ASAR images were already multilooked, some
speckle noise was still seen in the calibrated images, which was suppressed using the cascade
spatial filtering technique,32 in which 3 3 kernel size Lee and Median filters (twice) were
applied successively on the SAR images. The cascade filtering method attempts to preserve
edges while smoothing the data. The different Envisat ASAR image tiles were then mosaicked;
the preprocessed mosaicked Envisat ASAR data are shown in Fig. 5(b).
046001-6
OctDec 2016
Vol. 10(4)
Fig. 5 Images from (a) 3-band composite of bands 3, 5, and 7 of Landsat 8 OLI sensor and
(b) Envisat ASAR datasets. In both images, a visible portion of Hakra channel entering
Pakistan just north of Fort Abbas area can be seen. On the Envisat ASAR image, two other
linear features extending southwest direction from visible portion of Hakra channel can also
be seen.
4.1.2 Multisensor data fusion

After preprocessing of Landsat 8 OLI dataset, a false-color 3-band composite was extracted,
utilizing the following three bands:
i. Band 3 (green): 0.525 to 0.6 m
ii. Band 5 (near IR): 0.845 to 0.885 m
iii. Band 7 (shortwave IR): 2.1 to 2.3 m
046001-7
OctDec 2016
Vol. 10(4)
Selection of bands 3, 5, and 7 is based on sensitivity to capture the scattered vegetation

growth and sensitivity to soil water. The combination of these three bands serves as one of
the indicators of groundwater at the probable location of paleochannels in the arid region.
The low spatial resolution (30 30 m2 ) Landsat 8 radiance image composite (bands 3, 5, and
7) was fused with Envisat ASAR preprocessed radiometrically calibrated high resolution
(12.5 12.5 m2 ) image. This image fusion was performed using principal components (PCs)
spectral sharpening technique, supplemented with nearest neighbor resampling. This technique
is used by many24 because the PC spectral sharpening collects the maximum information from
all the bands used as input data in the PCA, i.e., the spatial information, whereas the spectral
information is segregated in the other PCs.
The PC transformation was applied on the selected bands (i.e., RGB composite band 7, 5, and
3, respectively) in order to remove redundancy from the data. Most of the information was
compressed in PC 1. This component was replaced with the high resolution Envisat ASAR
band. The Envisat ASAR band was then scaled to match PC band 1. This process reduced
the spectral information distortion.33 Then inverse PC transformation was applied on the selected
composite. Finally, the nearest neighborhood algorithm34 was used to perform the resampling
of multispectral data to match the pixel size of the high-resolution image (Envisat ASAR).
To visually improve the resultant image, a nonlinear histogram equalization technique was
applied, which enhanced the brightness values such that the peaks and valleys of the histogram
were preserved.26
4.1.3 National difference vegetation index calculation

All the Landsat 8 OLI images were atmospherically corrected and mosaicked. The NDVI was
calculated before fusion. False colors were assigned to the output image by color slicing for
display and interpretation. The positive values in the range of 0.2 to 0.3 indicate the presence
of shrubs.
4.2 Paleochannel Validation

Paleochannels can be mapped using geophysical survey; depending on how obvious the physical
property contrast is between their fill and the surrounding rocks. Many studies have identified
and verified paleochannels using geophysical methods2,35 and validated identified paleochannels
by making comparisons with field geophysical data and historical evidence. Three classical geophysical methods that are used to investigate paleochannels are EM profiling supplemented by
electrical soundings, electrical resistivity tomography, and ground penetrating radar.4 In this
study, the available datasets of soil resistivity and EC were used to validate the existence of
the paleochannel.
4.2.1 Validation by resistivity method

Resistivity data were acquired from WAPDA in the form of soil resistivity sectional drawings
(Fig. 3), showing soil resistivity variability in the vertical downward direction up to 150-m depth.
These measurements were performed at resistivity sounding points along 16 resistivity profiles
(Fig. 2). Resistivity values at each sounding point were recorded at 50-, 100-, and 150-m depths.
Reclassified raster surfaces were generated from these point measurements using inverse distance weighting (IDW) interpolation technique, according to the resistivity and groundwater
quality ranges specified in Table 1. These raster surfaces represent the aquifer condition in
terms of water quality at 50-, 100-, and 150-m depths.
4.2.2 Validation by conductivity method

Data of soil water EC in Scm of more than 300 wells (including tube wells, dug wells, boreholes and hand pumps, and so on) located in Cholistan were attained from WAPDA. A raster
map of water quality was generated from these point measurements using IDW interpolation as
per EC range and its relation with water quality (Table 2).
046001-8
OctDec 2016
Vol. 10(4)
Table 1 Soil resistivity ranges and their relationship with groundwater quality
and water retaining soil lithology.14
Resistivity range (Ohm-m)
Groundwater quality
Soil lithology
Fresh
Sand
20 to 35
Low quality fresh
Clayey sand, silty sand
10 to 20
Brackish
Silt, sandy clay, clay
Saline
Clay
>35
<10
Table 2 EC ranges and corresponding groundwater quality.36

Electrical conductivity (Scm)
Groundwater quality
<1500
Fresh
1500 to 2500
Low quality fresh
2500 to 8500
Brackish
>8500
Saline
4.2.3 Validation by physical and historical evidences

Locations of six forts and around 300 existing/abandoned water sources were overlaid on the
detected paleochannels to ascertain their location. In addition, the alignment of the middle and
lower portions of identified paleochannels was also be checked against potential ground water
harvesting sites reported in a study.37
5 Results
Signs of ancient river courses like old channel scars, meandering loops, ox-bow lakes, water
filled abandoned channels, dissected fluvial deposit, and different types and texture of vegetation
may provide hints about the existence and nature of a paleochannel.4 The running Hakra river
channel has a very noticeable and visibly elongated portion of highly dense vegetation belt
stretching from Pakistan-Indian border up to Marot area that is clearly visible on both optical
and microware imagery (Fig. 5).
The surface and subsurface paleochannel information was contained in both the optical and
active radar datasets, respectively, which was enhanced after PC fusion of both datasets. The
output of the Landsat 8 OLI and Envisat ASAR image fusion is shown in the top part of
Fig. 6. The linear green segments extending toward the southwestren direction from the visible
portion of Hakra were taken as patches representing possible paleochannels and were converted
to vector format created through digitization for further analysis and validation; this vector
output is shown in the bottom part of Fig. 6. The paleochannel mapped using remote sensing
data is an extension of the visible portion along with a meandering nature and wide spans, which
are very typical of riverine behavior in plain areas.
The NDVI output created from Landsat 8 OLI atmospheric corrected reflectance image is
shown in Fig. 7. False colors were assigned to this image by color slicing for easy visual interpretation. This image shows the presence of shrubs growth (positive value in the range of 0.2 to
0.3) along the paleochannel alignment detected in the PC fused image generated from SAR and
optical data (Fig. 6). These linear patterns of shrubs growth are indicated by oval shapes, which
show a clear spatial correlation with the Hakra paleochannel.
The raster surfaces were generated from soil conductivity readings, representing the aquifer
condition in terms of water quality at 50-, 100-, and 150-m depths, overlaid with the detected
046001-9
OctDec 2016
Vol. 10(4)
Fig. 6 (a) Signature of Hakra paleochannel visible in a multisensor fused image as linear green
segments extending toward southwestren direction from the visible portion of Hakra and (b) vector
format created using on-screen digitizing by keeping a fused image as a base.
paleochannels, as shown in Fig. 8. Existence of fresh to brackish water is evident along the
mapped paleochannels up to Mojgarh area of Cholistan at the 50-m depth. Similar spatial patterns exist at 100- and 150-m depths, however, the amount of fresh water decreases at higher
depths, as expected.
Raster surface generated from soil resistivity data is overlaid with the detected paleochannels,
as shown in Fig. 9. The spatial distribution of water quality derived from soil resistivity measurements is reasonably well aligned with the detected paleochannels; however, it is not as strong
as the results from soil conductivity measurements, especially at the 50-m depth. It is important
046001-10
OctDec 2016
Vol. 10(4)
Fig. 7 NDVI image created from Landsat 8 atmospheric corrected data confirming presence of
shrubs growth (positive value in the range of 0.2 to 0.3) along paleochannel alignment. Linear
pattern shrubs growth is indicated by oval shapes. This has linkage with visible portion of the
Hakra paleochannel.
Fig. 8 Groundwater quality maps at (a) 50-m depth, (b) 100-m depth, and (c) 150-m depth, generated using reference soil resistivity measurements, overlaying the detected paleochannel from
this study. Presence of fresh groundwater confirms the northern portion of Hakra paleochannel.
to highlight here that in desert areas, slightly brackish water with EC values of 1500 to
2500 Scm is still regarded as drinking water.38
The supplementary validation of extracted paleochannels with physical and historical evidences also supports the validation with geophysical datasets. Figure 10 shows the location of
046001-11
OctDec 2016
Vol. 10(4)
Fig. 9 Groundwater quality map created from water conductivity readings of more than 300 water
sources, overlaying the detected paleochannel from this study.
Fig. 10 Location map of forts and water sources. The location of water sources is mapped by
plotting the spatial attribute of 300 existing/old water sources (at which water conductivity readings
were taken).
six forts and around 300 existing/abandoned water sources in the study region. The presence of
historical forts at Mojgarh, Dingarh, Derawar, and Islamgarh is historical evidence of the
paleochannel because forts in this region were often constructed around water channels, i.e.,
Hakra River in this case. Concentration of existing/abandoned water sources along identified
paleochannels is yet more supporting evidence.
046001-12
OctDec 2016
Vol. 10(4)
6 Discussion
In this study, Hakra paleochannels in the Cholistan region were mapped using the SAR and
optical images fusion technique. The results of the study were validated using different methods,
i.e., geophysical (resistivity and conductivity methods), remote sensing based, and historical
evidence using old maps.
The Landsat 8 OLI data for the year 2013 were used to get vegetation response, as the presence of vegetation in arid/semiarid regions would be an indication of some underground water
existence if no obvious water source is found nearby. Similarly, Envisat ASAR image data of dry
and hot weather were selected as this radar signal has maximum soil penetration and subsurface
mapping capability in arid/dry regions during such climatic conditions. Visually, no obvious
signatures could be seen in Landsat 8 OLI image except elongated vegetation segment protruding from Pakistan-India border up to Marot. However, beside the visible channel segment, a few
additional linear features were visible on the Envisat ASAR image (Fig. 5). Fusion of these
datasets were performed because microwave images have a better ability to retain soil moisture
information, whereas optical and IR bands preserve better spatial and spectral information about
vegetation. In the PC fused image generated from the fusion of optical/IR and SAR data (Fig. 6),
segmented green patterns extending toward the southwestern direction from an elongated segment of vegetation were visible. Further south, another continuous green pattern was also visible
along the Pakistan-Indian border and entering Pakistan north of Winjor (Fig. 6). These green
patterns followed the same exact alignment, which could be seen in the Envisat ASAR image
(Fig. 5); however, these seem more obvious on the fused image. Furthermore, a number of studies
have been conducted to map paleochannels that drain the Sahara desert,6,11,15,39 and similar work
has been done in the neighboring regions of India2,5,79,40 and China.12 However, to the authors
knowledge, so far there has been no specific remote sensing based research conducted to investigate paleochannels in the Pakistan region, and this study may be a first step in this regard.
For validation by geophysical techniques, both kinds of groundwater quality maps generated
either from soil resistivity data (Fig. 8) or water conductivity data (Fig. 9) represent the availability of fresh water along the mapped Hakra paleochannel up to the Mojgarh area in Cholistan.
The presence of freshwater along the identified alignment of paleochannel confirms our methodology. Geophysical data of only northern Cholistan was provided by WAPDA, which was
used for validation of the upper portion of the mapped paleochannel. Existence of freshwater
in the northern portion of the paleochannel is probably due to its proximity to a canal system in
Punjab and underground recharging from a perennial channel of the old Indian Saraswati river.
A color sliced NDVI image (Fig. 7) also shows the same trend of linear features extending
from the visible elongated portion of vegetation (visible portion of Hakra paleochannel), which
is visible on the multisensor fused image (Fig. 6) as well as in the Envisat ASAR preprocessed
image [Fig. 5(b)]. The NDVI value range of these linear features corresponds to around 0.2 to
0.3, which indicates sparse vegetation or shrubs, indicating the existence of subsurface water.
Other supporting evidence for extracted alignment of the Hakra paleochannel are the proximity of old water sources and historical forts, which traditionally existed along rivers (Fig. 10).
The possible existence of many paleochannels in the study area has been reported by Sinha.2
To validate the alignment of the extracted paleochannels, the map2 shown in Fig. 2 was overlaid
with paleochannels identified in this study. While analyzing the map in Fig. 11, it should be kept
in mind that the reported map2 may not be of as high geolocation accuracy as that of the remote
sensing datasets used in our work.
In addition, the middle and lower portion of the identified channel also aligns with potential
ground water harvesting sites identified in a study Runoff farming in reducing rural poverty in
Cholistan desert,37 which further validates the identified channels middle and lower segment.
This study has used Envisat ASAR C-band imagery to identify a buried Hakra paleochannel
in a dry arid region of Cholistan, which has relatively less subsurface penetration (may be less
than 1 m) as compared to L-band imagery. The use of L-band SAR imagery (such as from
ALOS-1 or ALOS-2 satellites) may help in better delineation of a subsurface paleochannel
in dry arid regions up to 2-m in depth. Furthermore, resistivity data of only the northern portion
of the Cholistan desert were available from WAPDA. The lower reach of the identified paleochannel should also be geophysically analyzed to ascertain ground water conditions in the
046001-13
OctDec 2016
Vol. 10(4)
Fig. 11 Paleochannels extracted from this study were overlaid on paleochannels reported in
the other study.2 The alignments of paleochannels extracted in both studies serve as another
validation of existence of paleochannels in the study area.
southern part of the desert. However, this study has been limited by the availability of survey data
for validation only in the northern part of the study area.
7 Conclusions
Satellite imagery (Landsat 8 OLI Visible/IR imagery and Envisat ASAR C-band SAR imagery)
and geophysical ground survey data (soil resistivity and conductivity measurements) have been
used to map paleochannels in the Cholistan desert. The visible/IR and SAR datasets were fused
to extract paleochannels, which were then validated against geophysical field measurements and
other secondary sources of validation. The presence of existing freshwater aquifer zones can
provide a big step forward in identification of a suitable recharge mechanism, which if implemented, keeping site specifications in view, would provide the optimal yield in terms of freshwater storage. The paleochannels are considered as possible and credible reservoirs of ground
water with a vast storage capacity. Paleochannel features have not only supported human civilization in the past, but still have potential to help present day water-deprived societies especially
in the arid to semiarid regions all over the world if these features are properly identified, delineated, understood, and managed. Remote sensing along with surface and subsurface geophysical
methods are highly appropriate tools to map such channels in desert regions. There is a need to
identify and assess the storage potentialities of these landforms, particularly in regions where
groundwater emergency situations are anticipated in addition to developing their suitable recharging mechanisms. Flooding of Indus plains is an almost regular phenomenon during the summer
monsoon season in Pakistan. The subsurface aquifers in desert areas astride the Indus valley have
to be analyzed for developing into subsurface freshwater storage facilities and should be evaluated
for suitable recharge mechanism through storm water or flood water in subsequent studies.
Acknowledgments
The authors would like to thank the European Space Agency (ESA) for provision of Envisat
ASAR data under Cat-1 project 20851, United States Geographical Survey (USGS) for free
046001-14
OctDec 2016
Vol. 10(4)
distribution of Landsat 8 data, Water and Power Development Authority (WAPDA) Pakistan for
providing soil resistivity and water conductivity data, IGIS, National University of Sciences and
Technology (NUST), Islamabad, Pakistan for research funding and literary guidance. We are
grateful to Dr. A. D. Khan at PCRWR for valuable suggestions and guidance on interpretation
of resistivity data and its relation with groundwater condition. We also thank engineer Dr. Ejaz
Hussain, associate dean, IGIS, NUST, Islamabad, for valuable guidance on atmospheric
correction of Landsat 8 data. Dr. Farrukh Chishtie and Dr. Saleem Ullah at the Geospatial
Research & Education Lab (GREL), Institute of Space Technology, Islamabad, Pakistan, are
thanked for advice on manuscript preparation. Mr. Muhammad Usman at Korea Advanced
Institute of Science & Technology (KAIST) is thanked for ensuring timely availability of
required library material.
References
1. M. A. Kahlown, Rainwater Harvesting in Cholistan Desert: A Case Study of Pakistan,
PCRWR, Islamabad, Pakistan (2007).
2. A. Sinha, Palaeochannels as groundwater storage-a promising option to cope up with
emergency situations in Rajasthan, Western India, in Proc. of the Int. Workshop,
pp. 2931 (2006).
3. A. Stein, A survey of ancient sites along the lost Sarasvati River, Geogr. J. 99, 173182
(1942).
4. M. Mohammed-Aslam and A. Balasubramanian, History of river channel modifications-a
review, J. Ecol. Nat. Environ. 2(10), 207212 (2010).
5. C. V. Nandini and S. Sanjeevi, Identification and mapping of palaeochannels in a part of
south India using remote sensing and field studies, Int. J. Earth Sci. Eng. 4(8), 3945
(2011).
6. S. I. Elmahdy and M. Mostafa Mohamed, Remote sensing and GIS applications of surface
and near-surface hydromorphological features in Darfur region, Sudan, Int. J. Remote Sens.
34(13), 47154735 (2013).
7. R. Chopra, C. M. Bhatt, and P.K. Sharma, Elucidation of palaeochannels in parts of Punjab
(India) using remote sensing and geographical information system, ISPRS Archives,
Symposium of ISPRS Commission IV, Vol. XXXVI(4), pp. 16, Goa, India (2006).
8. C. V. Nandini, S. Sanjeevi, and A. S. Bhaskar, An integrated approach to map certain
palaeochannels of South India using remote sensing, geophysics, and sedimentological
techniques, Int. J. Remote Sens. 34(19), 65076528 (2013).
9. V. S. Rathore, M. S. Nathawat, and P. K. Champatiray, Palaeochannel detection and aquifer
performance assessment in Mendha River catchment, Western India, J. Hydrol. 395(34),
216225 (2010).
10. R. A. Wray, Palaeochannels of the Namoi River Floodplain, New South Wales, Australia:
the use of multispectral Landsat imagery to highlight a Late Quaternary change in fluvial
regime, Aust. Geogr. 40(1), 2949 (2009).
11. S. I. Elmahdy and M. M. Mohamed, Remote sensing and geophysical survey applications
for delineating near-surface palaeochannels and shallow aquifer in the United Arab
Emirates, Geocarto International, 30(7), 114 (2015).
12. X. Wang et al., Extraction of palaeochannel information from remote sensing imagery in
the east of Chaohu Lake, China, Front. Earth Sci. 6(1), 7582 (2012).
13. B. Waske and S. van der Linden, Classifying multilevel imagery from SAR and optical
sensors by decision fusion, IEEE Trans. Geosci. Remote Sens. 46(5), 14571466 (2008).
14. K. Fielitz, Groundwater Investigations in Desert Areas of Pakistan, pp. 125, Federal
Republic of Germany, Federal Ministry of Economic Cooperation, Germany (1986).
15. C. Robinson et al., Use of radar data to delineate palaeodrainage leading to the Kufra Oasis
in the Eastern Sahara, J. Afr. Earth. Sci. 44(2), 229240 (2006).
16. C. Pohl and J. L. Van Genderen, Review article: multisensor image fusion in remote sensing: concepts, methods and applications, Int. J. Remote Sens. 19(5), 823854 (1998).
17. N. Lancaster, G. G. Schaber, and J. T. Teller, Orbital radar studies of paleodrainages in
the central Namib Desert, Remote Sens. Environ. 71(2), 216225 (2000).
046001-15
OctDec 2016
Vol. 10(4)
18. S. Sukumar and K. Sankar, Delineation of potential zones for artificial recharge using GIS
in Theni district, Tamilnadu, India, Int. J. Geomatics Geosci. 1(3), 639 (2010).
19. G. G. Schaber, J. F. McCauley, and C. S. Breed, The application of SIR-C/X-SAR images
to mapping of sand-blanketed basement rocks in southern Egypt in Final Reports of SIR-C/
X-SAR Principal Investigators, Jet Propulsion Laboratory, Pasadena, California (1997).
20. J. Zhang et al., Block-regression based fusion of optical and SAR imagery for feature
enhancement, Int. J. Remote Sens. 31(9), 23252345 (2010).
21. C. Jiang et al., A practical compressed sensing-based pan-sharpening method, IEEE
Geosci. Remote Sens. Lett. 9(4), 629633 (2012).
22. L. P. Zhang et al., Adjustable model-based fusion method for multispectral and panchromatic images, IEEE Trans. Syst. Man Cybern. B 42(6), 16931704 (2012).
23. X. Meng et al., A unified framework for spatio-temporal-spectral fusion of remote sensing
images, in IEEE Int. Geoscience and Remote Sensing Symp., pp. 25842587 (2015).
24. K. G. Nikolakopoulos, Comparison of nine fusion techniques for very high resolution
data, Photogramm. Eng. Remote Sens. 74(5), 647659 (2008).
25. P. M. Mather, Computer Processing of Remotely-Sensed ImagesAn Introduction, 3rd ed.,
John Wiley and Sons Ltd., West Sussex, United Kingdom (2004).
26. J. B. Campbell and R. H. Wynne, Introduction to Remote Sensing, 5th ed., The Guilford
Press, New York (2011).
27. R. Barstow, ENVISAT-1 Products Specifications Volume 8: ASAR Products
Specifications, Technical Report 4 (2007).
28. K. Voudouris, Artificial recharge via boreholes using treated wastewater: possibilities and
prospects, Water 3(4), 964975 (2011).
29. H. S. Fielitz et al., Groundwater Investigations in Desert Areas of Pakistan, Vol II A: Results
of Hydrogeological Investigations (Annexures), pp. 112, B. A. WAPDA, Pakistan (1991).
30. M. A. Wahid et al., Stormwater treatment using porous rock matrices, in Int. Symp. on
Flood Research and Management, pp. 195209, Springer (2015).
31. D. W. Hawkins, Synthetic aperture imaging algorithms: with application to wide bandwidth sonar, PhD Thesis, University of Canterbury, Christchurch, New Zealand (1996).
32. W. A. Qazi, W. J. Emery, and B. Fox-Kemper, Computing ocean surface currents over
the coastal California current system using 30-Min-Lag Sequential SAR Images, IEEE
Trans. Geosci. Remote Sens. 52(12), 75597580 (2014).
33. P. Chavez, S. C. Sides, and J. A. Anderson, Comparison of three different methods to
merge multiresolution and multispectral data- Landsat TM and SPOT panchromatic,
Photogramm. Eng. Remote Sens. 57(3), 295303 (1991).
34. R. Welch and M. Ehlers, Merging multiresolution SPOT HRV and Landsat TM data,
Photogramm. Eng. Remote Sens. 53, 301-303 (1987).
35. M. K. Mayilvaganan, P. Mohana, and K. Naidu, Delineating groundwater potential zones
in Thurinjapuram watershed using geospatial techniques, Indian J. Sci. Technol. 4(11),
14701476 (2011).
36. K. Fielitz, Geoelectrical investigations for ground water exploration in desert areas of
Pakistan, Hannover, Germany, (unpublished report, BGR archives no. 13054/87) (1987).
37. F. Ahmad, Runoff farming in reducing rural poverty in Cholistan desert, Soc. Nat. 20(1),
177188 (2008).
38. K. Fielitz, Groundwater Investigations in Desert Areas of Pakistan: Results of the 1987/
1988 BGR Heliborne Survey Cholistan, Tharparkar, Windar Valley, p. 41, BGR, Germany
(1988).
39. J. F. McCauley et al., Subsurface valleys and geoarcheology of the eastern Sahara revealed
by shuttle radar, Science 218(4576), 10041020 (1982).
40. K. Singh, Significance of palaeochannels for hydrogeological studies-a case study from
alluvial plains of Punjab & Haryana states, India, in Subsurface-Water Hydrology,
pp. 245249, Springer, Netherlands (1996).
Zaheer Ul Islam received his BE (civil engineering) from MCE, NUST, in 2007. He has
completed his MS RS&GIS from Institute of Geographical Information System (IGIS),
National University of Science and Technology (NUST), Islamabad, Pakistan.
046001-16
OctDec 2016
Vol. 10(4)
Javed Iqbal (research supervisor) is an associate professor and HOD IGIS, School of Civil &
Environmental Engineering (SCEE). His research interests include geospatial applications in
glaciology, climate change and variability, water resources, hydrological modeling, natural
hazard and disaster management, site-specific agriculture, soil sciences, and geology.
Junaid Aziz Khan is a lecturer at IGIS, School of Civil & Environmental Engineering (SCEE).
He has a background in geology, remote sensing and GIS. He has published several publications
in the field of GIS & RS application in geology.
Waqas A. Qazi is an assistant professor, Institute of Space Technology (IST), Islamabad,
Pakistan and received his PhD from University of Colorado, Boulder, United States, in
2013. In his PhD thesis, he worked on remote sensing using SAR and oceanography.
046001-17
OctDec 2016
Vol. 10(4)

Paleochannel Delineation Using Landsat 8 OLI and Envisat ASAR Image Fusion Techniques

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Paleochannel Delineation Using Landsat 8 OLI and Envisat ASAR Image Fusion Techniques

Uploaded by

Copyright:

Available Formats

Paleochannel delineation using

Landsat 8 OLI and Envisat ASAR

Downloaded From: http://remotesensing.spiedigitallibrary.org/ on 10/16/2016 Terms of Use: http://spiedigitallibrary.org/ss/termsofuse.aspx