The Egyptian Journal of Remote Sensing and Space Sciences xxx (2017) xxx–xxx

Contents lists available at ScienceDirect

The Egyptian Journal of Remote Sensing and Space Sciences

journal homepage: www.sciencedirect.com

Research Paper

Evaluation of automatic drainage extraction thresholds using ASTER

GDEM and Cartosat-1 DEM: A case study from basaltic terrain of Central
India
G.P. Obi Reddy ⇑, Nirmal Kumar, Nisha Sahu, S.K. Singh
ICAR-National Bureau of Soil Survey & Land Use Planning, Amravati Road, Nagpur 440 033, India

a r t i c l e i n f o a b s t r a c t

Article history: In the study, the digital elevation models from ASTER GDEM and Cartosat-1 of 30 m resolution were used
Received 10 July 2016 to evaluate the thresholds in automatic extraction of stream network at six thresholds of 0.05 km2,
Revised 31 March 2017 0.125 km2, 0.25 km2, 0.50 km2, 1.0 km2 and 2.0 km2 in Neri watershed in basaltic terrain of Central
Accepted 3 April 2017
India. At stream thresholds of 0.05 km2, Cartosat-1 DEM shows higher channel initiation points (150)
Available online xxxx
than ASTER (127). As we increase the stream thresholds from 0.05 km2 to 2.0 km2, the channel initiation
points are consistently decreased in two input DEMs. The correlation coefficient R2 between the drainage
Keywords:
area thresholds and stream length for ASTER and Cartosat-1 DEM shows 0.94 and 0.92, respectively. The
ASTER GDEM
Cartosat-1 DEM
comparative evaluation of morphometric indices shows that threshold of 0.05 km2 is found to be opti-
GIS mum to obtain the finer drainage networks particularly from Cartosat-1 DEM, when compare to the
Drainage network ASTER GDEM of same resolution.
Thresholds Ó 2017 National Authority for Remote Sensing and Space Sciences. Production and hosting by Elsevier B.
V. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-
nc-nd/4.0/).

1. Introduction
Accurate delineation of drainage networks is a prerequisite for
many natural resource management issues (Paik, 2008; Liu and
Zhang, 2011). Digital Elevation Model (DEM) refers to a quantita-
tive model of a part of the earth’s surface in digital form
(Burrough and McDonnell, 1998). Several radar satellite based
DEMs like Advanced Spaceborne Thermal Emission and Reflection
Radiometer (ASTER) 30 m and Cartosat-1 DEM 30 m are available
for the public. The availability of satellite based new topographic
datasets have opened a new venues for hydrologic and geomor-
phologic studies including analysis of surface morphology (Staley
et al., 2006; Frankel and Dolan, 2007; Patel et al., 2016) and chan-
nel network structure (Lashermes et al., 2007). The important
application that has been widely used in surface hydrology model-
ing is the automatic extraction of the channel drainage network.
Due to the increasing availability of grid DEMs, numerous research
studies have been carried out to automate the extraction of drai-
nage networks (Montgomerym and Dietrichm, 1989, 1992;
Peer review under responsibility of National Authority for Remote Sensing and
Space Sciences.
⇑ Corresponding author at: Division of Remote Sensing Applications, National
Bureau of Soil Survey & Land Use Planning, Amravati Road, Nagpur 440 033, India.
E-mail address: obireddygp@gmail.com (G.P.O. Reddy).
Laurel and Souza, 2002; Maathuis and Wang, 2006; Maathuis,
2006; Moore et al., 1991; Tribe, 1992; Bertolo, 2000; Sameena
et al., 2009; Pareta and Pareta, 2011; Singh et al., 2014; Reddy
et al., 2004a, b; Sahu et al., 2016). A variety of methods have been
developed to process raster DEMs automatically to extract drai-
nage networks and measure their properties (O’Callaghan and
Mark, 1984; Tarboton et al., 1991; Martz and Garbrecht, 1999;
Al-Fugara et al., 2016). In any method of extracting drainage net-
works from DEMs is to specify a critical support area that defines
the minimum drainage area required to initiate a channel. (Band,
1989; Tarboton et al., 1988; Gardener et al., 1991). The most com-
monly used procedures for extracting drainage networks from ras-
ter DEMs are based on O’Callaghan and Mark’s (1984) algorithm for
flow direction determination, coupled with an arbitrary constant
value for the minimum contributing area needed to form and
maintain a channel. The common problem with drainage network
delineation using DEM is the presence of sinks (Nikolakopoulos
et al., 2006). The most widespread method for handling depres-
sions is sink filling, up to the level of the sink spill point, combined
with routing through the resulting flat area (Fairfield and Leymarie,
1991; Planchon and Darboux, 2001; Wang and Liu, 2006).
Improved sink filling methods (Grimaldi et al., 2007; Santini
et al., 2009) first fill sinks, and then introduce a gradient to all flat
areas to provide non-zero gradients for flow routing. Therefore, the

http://dx.doi.org/10.1016/j.ejrs.2017.04.001
1110-9823/Ó 2017 National Authority for Remote Sensing and Space Sciences. Production and hosting by Elsevier B.V.
This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).

Please cite this article in press as: Reddy, G.P.O., et al. Evaluation of automatic drainage extraction thresholds using ASTER GDEM and Cartosat-1 DEM: A
case study from basaltic terrain of Central India. Egypt. J. Remote Sensing Space Sci. (2017), http://dx.doi.org/10.1016/j.ejrs.2017.04.001
2 G.P.O. Reddy et al. / Egypt. J. Remote Sensing Space Sci. xxx (2017) xxx–xxx
sinks in the input datasets have to be removed prior to processing
of DEM for drainage identification (Simon et al., 2008).
The quantitative analysis of drainage pattern is one of the
important aspects in characterization of watersheds (Strahler,
1964). GIS based methods are being used increasingly to delineate
channels automatically for use in hydrologic models. However,
care needs to be exercised to ensure that networks are extracted
from DEMs at an appropriate scale. Many authors have attempted
to analyze the drainage characteristics of the basins using GIS tech-
niques in determining the quantitative description of the basin
geometry (Biswas et al., 1999; Reddy et al., 2002, 2004a; Vijith
and Satheesh, 2006; Kattimani and Prasad, 2016). In the present
study, in order to evaluate the performance of automatic drainage
extraction thresholds, drainage network extraction was performed
in TNTmips GIS software using ASTER GDEM (30 m) and Cartosat-1
DEM (30 m) at different threshold levels. The extracted drainage
from two input DEMs was compared with hydrological and mor-
phometric indices (Wen et al., 2006), which includes the relation-
ship between thresholds, drainage characteristics and stream
networking statistics (Strahler, 1964). We also examined the qual-
itative and quantitative comparison of the automatically extracted
drainage from ASTER GDEM and Cartosat-1 DEMs to assess the
influence of different thresholds.
2. Study area
The Neri watershed on basaltic terrain located in Nagpur dis-
trict of Maharashtra, Central India and lies between 21°030 0500
and 21°070 2400 north latitudes and 78°480 3300 and 78°530 0200 east
longitudes covering an area of about 3196.22 ha (Fig. 1). The eleva-
tion is ranging from 250 to 408 meters above mean sea level (MSL).
The climate is warm sub-humid with mean annual rainfall of
1120 mm of which about 89% is received during the south-west
monsoon (June to October). Geologically the area is covered by
basaltic lava flow of Lower Eocene to Upper Cretaceous, commonly
known as ‘‘Traps”. The dentritic drainage pattern has been
observed in the upper reaches, whereas in the lower reaches paral-
lel to sub parallel drainage pattern has been observed. The study
area has an undulating topography with landforms typical of a
basaltic terrain. The slope of the study area ranges from level to
nearly level (0–1%) in the flood plains to steep slopes (>30%)
mainly along the scarp slopes. The soils in the watershed area
ranges from very shallow (10-25 cm) in the upper reaches to deep
(100–150 cm) mainly in flood plains and broad valley floor. Clay
loam and clayey textured soils have been noticed in the upper
and lower reaches, respectively. Study area is predominantly under
well drained soils, however, excessively drained soils are asso-
ciated with the scarp slopes (Anon., 1990). The study area is mainly
under rainfed condition with cotton-fallow and cotton/pigeonpea-
fallow as main land use systems.
3. Material and methods
3.1. Processing of ASTER 30 m GDEM
The freely available ASTER 30 m GDEM (version 2) of one arc-
second (arcsec) grid was downloaded for the study area in GeoTIFF
format with geographic lat/long coordinates (USGS, 2012). The
ASTER sensor on board of the Terra spacecraft (launched in 1999)
has an along-track stereoscopic capability due to a nadir-and
27.6° backward-looking telescope in the Near Infrared (NIR) spec-
tral band. This instrumental setup allows for photogrammetric
DEM generation with vertical accuracies of ±15–30 m (Toutin,
2008). The most important specifications of the ASTER stereo sub-
system that govern the DEM generation capabilities include: stereo
Please cite this article in press as: Reddy, G.P.O., et al. Evaluation of automatic drainage extraction thresholds using ASTER GDEM and Cartosat-1 DEM: A
case study from basaltic terrain of Central India. Egypt. J. Remote Sensing Space Sci. (2017), http://dx.doi.org/10.1016/j.ejrs.2017.04.001
geometry; platform altitude of 705 km and base to height ratio of
0.6 (Abrams, 2000). ASTER DEM standard data products are pro-
duced with 30 m postings, and have Z accuracies generally
between 10 m and 25 m root mean square error (RMSE). The GDEM
is provided at a one arc sec resolution (30 m) and referenced to
World Geodetic System (WGS) 1984. With reference to GDEM1,
the improvements in GDEM2 include increased horizontal and ver-
tical accuracy, as compared to both GPS benchmarks and standard
DEMs, and improved horizontal accuracy and resolution (Dave,
2011). Downloaded ASTER GDEM 30 m data was reprojected to
Universal Transverse Mercator (UTM) projection with zone 44
and clipped to the extent of the study area to extract various drai-
nage properties at given thresholds.
3.2. Processing of Cartosat-1 (30m) DEM
The freely available resampled Cartosat-1 DEM of 1 arc-sec. i.e.
30 m resolution at equator, generated by sub sampling the original
1/3 arc-sec data was downloaded in .tiff format (http://bhuvan-
noeda.nrsc.gov.in/download/download/download.php, 2016). The
Cartosat-1 has a pair of Panchromatic cameras having an along
track stereoscopic capability using its near-nadir viewing and for-
ward viewing telescopes to acquire stereo image data with a base-
to-height ratio of about 0.63. The spatial resolution is 2.5 m in the
horizontal plane giving a swath of about 27 km. The methodology
adopted to produce the Cartosat-1 DEM involved stereo-strip tri-
angulation of 500 km strip stereo pairs using high precise ground
control points, interactive cloud masking, automatic dense conju-
gate pair generation using matching approach (Kääb, 2002). Seam-
less homogeneous Cartosat-1 DEM is produced by TIN modeling of
irregular DEM, interpolation for regular DEM generation and
automatic strip to strip mosaicing. The primary output unit is a tile
of 7.50  7.50 extents with DEM spacing of 1/3 arc-sec, and
co-registered ortho-image of resolution 1/12 arc-sec. Statistically,
Cartosat-1 DEM is meeting the specification of vertical accuracy
i.e. 8 m at 90% confidence (Radhika et al., 2007; Van Zyl, 2001).
The processed Cartosat-1 DEM of 30 m resolution was clipped to
the study area in order to extract various drainage properties at
given thresholds. Before processing, it was ensured that the two
input DEMs are errors free like removal of data voids, which could
normally found in DEMs. Besides, the remotely sensed DEMs, the
topographic data on 1:50,000 scale of the study area has been used
as reference data.
3.3. Automated detection of drainage network at different thresholds
With increasing computer capability and availability of DEMs in
public domain enable the researchers to extract drainage networks
from DEMs through computer programs (Lashermes et al., 2007).
But, the suitable drainage network extraction depends primarily
on the accuracy of the DEM and the channel initiation identifica-
tion. The more accurate the channel initiation points provided,
the more suitable the drainage network generated (Wen et al.,
2006). We used ASTER GDEM 30 m and Cartosat-1 DEM 30 m data-
sets and followed different drainage area thresholds ie., 0.05 km2,
0.125 km2, 0.25 km2, 0.50 km2 1.0 km2 and 2.0 km2 to determine
the number of upstream elements. Watershed process tools in
TNTmips software (ver. 7.7) (TNTmips, 2012) were used to extract
drainage channels at different thresholds. In the study, the
Tarboton et al. (1989) method was used to remove all sinks in
the input datasets. Then, the flow direction was calculated for
every central pixel of input blocks of 3 by 3 pixels, each time com-
paring the value of the central pixel with the value of its 8 neigh-
bours. The Watershed process tools in TNTmips create a series of
temporary vector and raster objects that present different aspects
of the results. In order to generate flow paths, the Watershed pro-
 G.P.O. Reddy et al. / Egypt. J. Remote Sensing Space Sci. xxx (2017) xxx–xxx 3

Fig. 1. Location of study area shown with ASTER DEM (30 m).

cess tools compute for each DEM cell the number of upstream cells
that contribute flow to that cell. These flow accumulation values
are used in part to trace flow paths upstream, beginning with the
highest accumulation values, where streams reach the boundaries
of the area. The outlet parameter sets the flow accumulation
threshold for initiating a flow path at the edge of the raster. Only
boundary cells with flow accumulation values greater than the
outlet threshold are used to initiate a flow path. The flow direction
and flow accumulation rasters became an important inputs for the
final drainage extraction processes in TNTmips software. Other
intermediary steps such as catchment network extraction and
catchment network ordering were also performed. The watershed
process tools in TNTmips computes stream order for each line ele-
ment in the standard flow path vector object using four different
ordering systems (Strahler, 1964; Horton, 1932; Shreve, 1966;
Scheidegger, 1965). In the study, individual stream lengths were
computed and arranged according to the magnitude of orders as
proposed by Strahler (1952) and numbers of streams of different
orders were calculated. The drainage extracted from Topographic
maps on 1:50,000 scale was used as reference.
3.4. Determination of morphometric indices
The aim of determination of morphometric indices in the study
is to evaluate stream properties from the measurement of various
stream attributes. Evaluation of morphometric indices necessitates

Please cite this article in press as: Reddy, G.P.O., et al. Evaluation of automatic drainage extraction thresholds using ASTER GDEM and Cartosat-1 DEM: A
case study from basaltic terrain of Central India. Egypt. J. Remote Sensing Space Sci. (2017), http://dx.doi.org/10.1016/j.ejrs.2017.04.001
4 G.P.O. Reddy et al. / Egypt. J. Remote Sensing Space Sci. xxx (2017) xxx–xxx

analysis of drainage parameters i.e., ordering of the various
streams, measurement of area and perimeter of basins, length of
drainage channels, drainage density, drainage frequency,
bifurcation ratio and texture ratio (Kumar et al., 2000). Analysis
of morphometric indices is of immense help in understanding
the drainage morphometry and its interactive impacts on basin
characteristics (Reddy et al., 2004a). From the extracted drainage
network for two input DEMs at various thresholds, the morphome-
tric indices like No. of basins, basin relief (Bh), bifurcation ratio
(Rb), drainage density (Dd), stream frequency (Fu), form factor
(Rf), elongation ratio (Re), ruggedness number (Rn) and constant
channel maintenance (C) were determined using standard mathe-
matical expressions (Table 1). From comparing morphometric
indices of the drainage network extracted at different thresholds,
while keeping the spatial resolution constant at 30 m, the influence
of thresholds in drainage network extraction could be understood
clearly.
3.6. Optimization of thresholds in extraction of drainage networks
In order to optimize the threshold in extraction of drainage net-
works, the drainage network extracted from input DEMs at differ-
ent thresholds was analyzed by calculating length of stream and
stream orders, while keeping the input DEMs resolution constant
at 30 m. The upstream flow properties were computed and ana-
lyzed for the two inputs DEMs. In order to understand, the rela-
tionship between the stream lengths extracted for the two input
DEMs and influence of thresholds, the correlation graphs were
computed between stream length and thresholds. Further, mor-
phometric indices as shown in Table 1 were determined across
the thresholds for the input DEMs to find out the optimum drai-
nage area threshold for input DEMs at spatial resolution of 30 m.
4. Results and discussion

4.1. Determination of channel initiation points at different thresholds

3.5. Comparative analysis of drainage parameters
The threshold value represents the minimum upstream area
required to form a channel segment in which water starts to flow
The comparative analysis of obtained drainage parameters was
as channel runoff (Rieger, 1993). The stream network is normally
achieved through analysis of extracted drainage in conjunction
displayed as inter-connected linear features and these features
with shaded relief and flow accumulation raster’s obtained from
begin where the threshold value is exceeded. In the study, the drai-
the input DEM. The total length of streams, stream orders and total
nage streams delineated using stream thresholds of 0.05 km2,
length of streams at each stream order for the input datasets at dif-
0.125 km2, 0.25 km2, 0.50 km2 1.0 km2 and 2.0 km2 from input
ferent thresholds has been analyzed. The channel initiation points
DEMs. Drainage channel network patterns, lengths and number
at different thresholds were also examined. Further, inventory
of channel initiation points have been obtained as a result of differ-
parameters compiled for ASTER GDEM 30 m and Cartosat-1 DEM
ent threshold areas. The channel initiation points from the input
30 m were compared in three ways: first, the drainage features
DEMs at different thresholds have been computed using TNTmips
as extracted from input DEMs at different thresholds were com-
software. The analysis shows that Cartosat-1 DEM yields consis-
pared through visual inspection and examination. Secondly,
tently higher channel initiation points across the thresholds. At
stream length extracted for different orders at different thresholds
stream thresholds of 0.05 km2, Cartosat-1 DEM shows the higher
for the input DEMs were analyzed. Thirdly, automatically extracted
channel initiation points of 150, than the ASTER GDEM (127). As
drainage features at different thresholds for the inputs DEMs were
we increase the stream thresholds from 0.05 km2 to 2.0 km2, the
compared in conjunction with the drainage extracted from topo-
channel initiation points are consistently decreased in both
graphical sheet at 1:50,000 scale to investigate the influence of dif-
the input DEMs. Interestingly, across the stream area thresholds,
ferent thresholds on the output.
the ASTER GDEM shows the lower channel initiation points, when
compare to the Cartosat-1 DEM. From the stream thresholds of
0.25 km2 to 2.0 km2, the difference between the channel initiation
Table 1 points is progressively decreased in two input DEMs.
Formulae adopted for computation of morphometric parameters.

S. Morphometric Formula/Relationship Reference 4.2. Determination of drainage parameters at different thresholds

No. Parameters
1. Stream order (u) Hierarchical order Strahler The drainage networks delineated at stream thresholds of
(1964) 0.05 km2, 0.125 km2, 0.25 km2, 0.50 km2 1.0 km2 and 2.0 km2 for
2. Stream length Length of stream Horton
ASTER GDEM and Cartosat-1 DEM are shown in Fig. 2. As expected,
(Lu) (1945)
3. Basin Relief (Bh) (Elevation of basin mouth)-(Elevation Schumn with the increase in stream thresholds, the number of streams
of highest point on the basin) (1956) gradually decreases. The influence of threshold shows the signifi-
4. Bifurcation ratio Rb = Nu/Nu + 1, Nu = total number of Schumn cant difference in extraction of drainage networks. The analysis
(Rb) stream segments of order ‘‘u”, Nu (1956) indicates that at lowest stream threshold area of 0.05 km2, the total
+ 1 = number pf stream segments of the
next higher order.
stream length of 1st order was observed 45.2 km and 45.7 km in
5. Drainage D = Lu/A, where A is the total area of the Horton ASTER GDEM and Cartosat-1 DEM, respectively. At stream thresh-
density (Dd) basin (km)2, Lu is the total stream (1932) old of 0.05 km2, the per cent drop of stream length from 1st to 2nd
length of all orders. order was observed as 57.6% and 61.8% in ASTER GDEM and
6. Stream Fs = Nu/A, Where Nu is the total Horton
Cartosat-1 DEM, respectively. At stream threshold of 0.125 km2,
frequency (Fu) number of streams of all order, A is (1932)
basin area in km2 the highest total stream length of 1st order (23.2 km) was noticed
7. Form factor (Rf) Ff = A/Lb2 is the square of the basin Horton in ASTER GDEM. Compare to the Cartosat-1 DEM, the per cent drop
length (km), A is the basin area in km2 (1932) of streams length from 1st to 2nd order is higher (47.5%) in ASTER
8. Elongation ratio Re = 2sqrt (A/p)/Lb, where A is the area Schumn GDEM. At stream threshold of 0.25 km2, the highest total stream
(Re) (km)2 and Lb = basin length. (1956)
9. Ruggedness Rn = Bh  Dd Where, Bh = Basin relief; Schumn
length of 1st order (13.1 km) was noticed in ASTER GDEM and low-
number (Rn) Dd = Drainage density (1956) est values (10.7 km) was observed in Cartosat-1 DEM. The highest
10. Constant Inverse of drainage density Schumn (50.7%) increase per cent of streams length from 1st to 2nd order
channel (1956) was observed in Cartosat-1 DEM. Significantly, the sudden drop
maintenance (C)
of stream length of 4th order was noticed in ASTER GDEM. At high-

Please cite this article in press as: Reddy, G.P.O., et al. Evaluation of automatic drainage extraction thresholds using ASTER GDEM and Cartosat-1 DEM: A
case study from basaltic terrain of Central India. Egypt. J. Remote Sensing Space Sci. (2017), http://dx.doi.org/10.1016/j.ejrs.2017.04.001
 G.P.O. Reddy et al. / Egypt. J. Remote Sensing Space Sci. xxx (2017) xxx–xxx 5

Fig. 2. Extracted drainage from ASTER GDEM and Cartosat-1 DEM at 0.05, 0.125, 0.25, 0.50, 1.0 and 2.0 km2 thresholds.

est stream threshold of 2.0 km2, the total stream length of 1st order
is higher in Cartosat-1 DEM (31.1 km) when compared to the
ASTER GDEM (24.8 km). In two cases, at stream threshold of
2.0 km2, the increase of 2nd order stream length was observed.
However, in both cases, 4th order streams could not be obtained
at stream threshold of 1.0 km2 and 2.0 km2 (Fig 3a and b). Across
the stream thresholds, the Cartosat-1 DEM shows consistently
higher stream length (Fig 3c).
4.3. Comparative analysis of drainage parameters with reference
topographic data
The visual inspection and examination of drainage network
extracted at different thresholds from input DEMs of same resolu-
tion (30 m) shows that at stream threshold of 0.05 km2 the total
number of streams are even more when compare to the reference
topographic datasets, more particularly at 1st order level (Fig. 4).

Please cite this article in press as: Reddy, G.P.O., et al. Evaluation of automatic drainage extraction thresholds using ASTER GDEM and Cartosat-1 DEM: A
case study from basaltic terrain of Central India. Egypt. J. Remote Sensing Space Sci. (2017), http://dx.doi.org/10.1016/j.ejrs.2017.04.001
6 G.P.O. Reddy et al. / Egypt. J. Remote Sensing Space Sci. xxx (2017) xxx–xxx

Fig. 3. (a) Stream lengths extracted from ASTER GDEM at different thresholds and stream orders, (b) Stream lengths extracted from Cartosat-1 DEM at different thresholds
and stream orders and (c) Total stream lengths extracted from ASTER GDEM and Cartosat-1 DEM at different thresholds.

The analysis of extracted stream network demonstrates that the
input DEMs have an enough resolution for extracting of first and
second-order streams in upper reaches at 0.50 km2 and
0.125 km2 thresholds values. However, at the outlet of the water-
shed, across the thresholds, the deviation of drainage was
observed, when compared to the reference topographic data. When
compare to the drainage networks extracted from ASTER GDEM,
the drainage extracted from Cartosat-1 DEM found to be more
accurate particularly in flat areas.
4.4. Comparative analysis of upstream flow distance
Maximum upstream flow distance maps indicate the longest
flow distance from any part of the upstream watershed boundary
to the current cell. Whereas, downstream flow distance maps
shows the flow distance downstream from each cell to the outlet
of its watershed. The mean upstream flow distance and standard
deviation have been determined from the drainage channels
extracted from ASTER GDEM and Cartosat-1 DEMs in the study
area. A higher mean upstream length of 239.5 m was observed in
ASTER GDEM, when compare the Cartosat-1 DEM. Whereas, higher
mean downstream length of 7438.4 m was observed in Cartosat-1
DEM, when compare the ASTER GDEM. The standard deviation of
upstream flow distance varies from 997.1 in Cartosat-1 DEM to
961.2 in ASTER GDEM.
4.5. Influence of area threshold in extraction of drainage network
In order to analyze the relationship between the flow accu-
mulation and extracted drainage network across the thresholds,
the visual inspection between the flow accumulation and
extracted drainage channels has been carried out at the selected
window. The analysis shows that across the drainage thresholds,
the Cartosat-1 DEM shows the finer drainage channels more par-
ticularly at the lower thresholds. In order to investigate, the
relationship between the stream lengths extracted for the two
input DEMs and influence of thresholds, the correlation graphs
were plotted between stream length and thresholds. The

Please cite this article in press as: Reddy, G.P.O., et al. Evaluation of automatic drainage extraction thresholds using ASTER GDEM and Cartosat-1 DEM: A
case study from basaltic terrain of Central India. Egypt. J. Remote Sensing Space Sci. (2017), http://dx.doi.org/10.1016/j.ejrs.2017.04.001
 G.P.O. Reddy et al. / Egypt. J. Remote Sensing Space Sci. xxx (2017) xxx–xxx 7

Fig. 4. Drainage networks derived at 0.05, 0.125, 0.25, 0.50, 1.0 and 2.0 km2 thresholds from ASTER and Cartosat-1 DEMs, respectively. The drainage extracted from
Topographic sheet on 1:50,000 scale is also shown as background.

correlation coefficient R2 between the stream length and thresh-
olds for ASTER GDEM and Cartosat-1 DEM shows 0.94 and 0.92,
respectively (Fig 5a and b). It indicates that the thresholds
applied on input DEMs in automatic extraction of drainage are
strongly and negatively related to the stream length. The
comparative analysis of stream length and thresholds between
the two input DEMs, Cartosat-1 DEM shows consistently higher
stream length across the thresholds more particularly at lower
thresholds.
Stream threshold is assumed as a constant value, based on per-
sonal judgment or visual comparison of the networks generated
with the streamlines identified or digitized from a topographical
map (Gardener et al., 1991). However, to arrive an optimum
threshold value in extraction of drainage channels, keeping the
spatial resolution constant needs the evaluation of various mor-
phometric indices. The determined morphometric indices in the
study are shown in Table 2. The analysis shows that at drainage
area threshold of 0.05 and 0.125 km2, ASTER and Cartosat-1 DEMs
show 69 and 73 basins, respectively. When compare to the ASTER
GDEM, Cartosat-1 DEM shows better performance in delineation of
basins, this might be attributed that Cartosat-1 DEM was resam-
pled from its original high resolution of 2.5 m. Basin relief (Bh)
shows the maximum vertical distance between the lowest and
the highest points of a basin and at drainage area threshold of
0.05 and 0.125 km2, ASTER and Cartosat-1 DEMs show basin relief
as 45.38 and 43.95, respectively. At drainage area threshold of
0.05 km2, Cartosat-1 DEM shows the higher bifurcation ratio (Rb)
of 1.29, when compare to the ASTER GDEM (1.23). When compare
to the ASTER GDEM, Cartosat-1 DEM consistently shows higher
drainage density (Dd) across the thresholds. Stream frequency
(Fu) computed as the ratio between the total number of streams
and area of the basin, the Fu values also consistently higher in
Cartosat-1 DEM, when compare to the ASTER DEM across the
thresholds. The form factor (Rf), elongation ratio (Re) and rugged-
ness number (Rn) also indicates the higher values across the
thresholds in Cartosat-1 DEM, when compare to the ASTER DEMs.
At highest area thresholds of 2.0 km2 the constant channel mainte-
nance (C), which expressed as the inverse of drainage density

Please cite this article in press as: Reddy, G.P.O., et al. Evaluation of automatic drainage extraction thresholds using ASTER GDEM and Cartosat-1 DEM: A
case study from basaltic terrain of Central India. Egypt. J. Remote Sensing Space Sci. (2017), http://dx.doi.org/10.1016/j.ejrs.2017.04.001
8 G.P.O. Reddy et al. / Egypt. J. Remote Sensing Space Sci. xxx (2017) xxx–xxx

Fig. 5. (a) Correlation between drainage area thresholds and total stream lengths derived from ASTER GDEM and (b) Correlation between drainage area thresholds and total
stream lengths derived from Cartosat-1 DEM.

shows the lower values in Cartosat-1 DEM (10.61), when compare
to the ASTER GDEM (14.27). The morphometric indices as shown
Table 1 illustrates that higher drainage area threshold values
extract less numbers of basins and streams orders; thus, have less
network order, which ultimately reflects less values of morphome-
tric indices. The results of this study demonstrate that the
Cartosat-1 DEM shows the better performance more particularly
at lower threshold of 0.05 km2 in extraction of drainage channels.
As we increase the thresholds, the difference between the
extracted drainage parameters is getting reduced in both the input
DEMs. Therefore, caution must be taken while applying higher
thresholds on 30 m DEMs in extraction of drainage parameters.
The analysis of morphometric indices of the drainage network
extraction demonstrates that at threshold of 0.05 km2, Cartosat-1
DEM provides finer drainage networks as compare to the ASTER
GDEM of same resolution.

Please cite this article in press as: Reddy, G.P.O., et al. Evaluation of automatic drainage extraction thresholds using ASTER GDEM and Cartosat-1 DEM: A
case study from basaltic terrain of Central India. Egypt. J. Remote Sensing Space Sci. (2017), http://dx.doi.org/10.1016/j.ejrs.2017.04.001
 G.P.O. Reddy et al. / Egypt. J. Remote Sensing Space Sci. xxx (2017) xxx–xxx 9

Table 2
Morphometric indices of the drainage channels extracted from ASTER GDEM and Cartosat-1 DEM at different thresholds.

Input DEMs and No. of Stream Basin relief Bifurcation Drainage Stream Form Elongation Ruggedness Constant of channel
thresholds (sq.km) basins Length (Bh) (mts) ratio (Rb) density frequency ratio ratio (Re) number (Rn) maintenance (C)
(mts) (Dd) (Fu) (Rf)
ASTER
GDEM
0.05 69 85992.87 45.38 1.23 2.70 7.54 0.08 0.27 0.13 0.41
0.125 69 60191.72 45.38 1.12 1.45 6.41 0.08 0.27 0.08 1.10
0.25 66 44295.93 46.45 1.07 0.69 5.57 0.09 0.28 0.04 3.24
0.50 58 35640.44 48.02 1.09 0.47 4.74 0.11 0.30 0.03 4.82
1.0 49 25559.94 50.96 1.00 0.29 4.32 0.13 0.34 0.02 7.91
2.0 40 18941.72 55.30 1.02 0.20 3.78 0.14 0.36 0.02 14.27
Cartosat-1 DEM
0.05 73 94755.36 43.95 1.29 2.96 7.87 0.10 0.30 0.13 0.36
0.125 73 63494.35 43.95 1.14 1.54 7.19 0.10 0.30 0.08 1.09
0.25 65 46908.22 46.48 1.12 0.74 6.07 0.11 0.32 0.05 3.10
0.50 57 36877.76 48.26 1.13 0.47 5.13 0.13 0.34 0.03 5.03
1.0 47 26960.79 50.17 1.20 0.30 4.75 0.14 0.35 0.02 7.43
2.0 41 20987.15 53.07 1.00 0.23 4.36 0.15 0.36 0.02 10.61

5. Conclusions References

The comparative evaluation of thresholds in automatic extrac-
tion of drainage parameters from ASTER GDEM and Cartosat-1
30 m DEM at thresholds of 0.05 km2, 0.125 km2, 0.25 km2,
0.50 km2 1.0 km2 and 2.0 km2 shows that the smaller the thresh-
old, higher the drainage channels. The study demonstrates that
Cartosat-1 DEM yield consistently higher channel initiation points
across the thresholds. As we increase the stream thresholds from
0.05 km2 to 2.0 km2, the channel initiation points are consistently
decreased in both the input DEMs. The analysis also indicates that
at lowest stream threshold area of 0.05 km2, the total stream
length of 1st order was observed 45.7 km and 50.9 km in ASTER
GDEM and Cartosat-1 DEM, respectively. The maximum upstream
length of 14053.1 m was noticed in Cartosat-1 DEM, as compare to
the ASTER GDEM. The comparative analysis of stream length and
thresholds in the input DEMs shows that Cartosat-1 DEM shows
consistently higher stream length across the thresholds when com-
pare to the ASTER DEMS. Analysis of various morphometric indices
demonstrate that Cartosat-1 DEM performs better more particu-
larly at lower threshold of 0.05 km2 in extraction of drainage chan-
nels. As we increase the thresholds, the difference between the
extracted drainage parameters is getting reduced in both the
DEMs. The analysis of morphometric indices demonstrates that
thresholds have immense influence on extraction of drainage
channels, the threshold of 0.05 km2 is found to be optimum in
order to obtain the finer drainage network particularly from
Cartosat-1 DEM. The similar results were also observed in delin-
eation of drainage length from Cartosat-1 DEM (NRSC, 2011). The
study also demonstrates the efficiency of GIS based techniques in
automated extraction of drainage network using satellite based
DEMs, which have immense potential in geological and hydrolog-
ical applications.
Conflict of interest
None.
Acknowledgement
The authors would like to thank Director, National Bureau of
Soil Survey & Land Use Planning, Nagpur for providing the neces-
sary facilities to execute the work. This study was supported under
institute funded project.
Abrams, M., 2000. ASTER. Data products for the high spatial resolution imager on
NASA’s EOS-AM1 platform. Int. J. Remote Sens. 21, 847–861.
Al-Fugara, A., Al-Adamat, R., Al-Kouri, O., Taher, S., 2016. DSM derived stereo pair
photogrammetry: Multitemporal morphometric analysis of a quarry in karst
terrain. Egypt J. Remote Sens. Space Sci. 19 (1), 61–72.
Anon., 1990. Soils of Nagpur District, Maharashtra, Report No. 514, NBSS&LUP (RC),
Nagpur, 68pp.
Band, L.E., 1989. A terrain based watershed information system. Hydrol. Process. 3
(2), 151–162.
Bertolo, F., 2000. Catchment Delineation and Characterization: A Review. Joint
Research Centre European Commission, Ispra, Italy, p. 30. Technical Report EUR
19563 EN.
Biswas, S., Sudhakar, S., Desai, V.R., 1999. Prioritization of sub watersheds based on
Morphometric analysis and drainage basin: a remote sensing and GIS approach.
J. Indian Soc. Remote Sens. 27 (3), 155–166.
Burrough, P.A., McDonnell, R.A., 1998. Principles of Geographic Information
Systems. Oxford University Press, New York, p. 333.
Dave, M., 2011. ASTER Global Digital Elevation Model Version 2 – Summary of
Validation Results. US Geological Survey, p. 28.
Fairfield, J., Leymarie, P., 1991. Drainage networks from grid elevation models.
Water Resour. Res. 27 (5), 709–717.
Frankel, K.L., Dolan, J.F., 2007. Characterizing arid region alluvial fan surface
roughness with airborne laser swath mapping digital topographic data. J.
Geophys. Res. 112, F02025.
Gardener, T.W., Sasowsky, K.C., Day, R.L., 1991. Automatic extraction of geomorphic
properties from digital elevation data. Z. Geomorphol. Suppl. 80, 57–68.
Grimaldi, S., Nardi, F., Di Benedetto, F., Istanbulluoglu, E., Bras, R.L., 2007. A
physically-based method for removing pits in digital elevation models. Adv.
Water Resour. 30 (10), 2151–2158.
Horton, R.E., 1932. Drainage basin characteristics. Trans. Am. Geophys. Union 13,
350–361.
Horton, R.E., 1945. Erosional development of streams and their drainage basins;
Hydrophysical approach to quantitative morphology. Geol. Soc. Am. Bull. 56,
275–370.
http://bhuvan-noeda.nrsc.gov.in/download/download/download.php, (accessed on
10.06.16).
Kääb, A., 2002. Monitoring high mountain terrain deformation from air and
spaceborne optical data: Examples using digital aerial imagery and ASTER data.
ISPRS J. Photogramm. Remote Sens. 57 (1–2), 39–52.
Kattimani, J.M., Prasad, T.J.R., 2016. Cartosat Digital Elevation Model (DEM) to
drainage extraction techniques of Vrushabhavati basin of Karnataka, India using
remote sensing and GIS techniques. Int. J. Adv. Res. 4 (7), 2008–2013.
Kumar, R., Kumar, S., Lohani, A.K., Nema, R.K., Singh, R.D., 2000. Evaluation of
geomorphological characteristics of a catchment using GIS. GIS India 9 (3), 13–
17.
Lashermes, B., Foufoula-Georgiou, E., Dietrich, W.E., 2007. Channel network
extraction from high resolution topography using wavelets. Geophys. Res.
Lett. 34, L23S04.
Laurel, A.F., Souza Jr., C., 2002. The role of remote sensing and GIS in enforcement of
areas of permanent preservation in the Brazilian Amazon. Geocarto Int. 17 (2),
53–58.
Liu, X., Zhang, Z., 2011. Drainage network extraction using LiDAR-derived DEM in
volcanic plains. Area 43 (1), 42–52.
Maathuis, B.H.P., 2006. Digital elevation model based hydro-processing. Geocarto
Int. 21 (1), 21–26.
Maathuis, B.H.P., Wang, L., 2006. Digital elevation model based hydro-processing.
Geocarto Int. 21 (1), 21–26.

Please cite this article in press as: Reddy, G.P.O., et al. Evaluation of automatic drainage extraction thresholds using ASTER GDEM and Cartosat-1 DEM: A
case study from basaltic terrain of Central India. Egypt. J. Remote Sensing Space Sci. (2017), http://dx.doi.org/10.1016/j.ejrs.2017.04.001
10 G.P.O. Reddy et al. / Egypt. J. Remote Sensing Space Sci. xxx (2017) xxx–xxx
Martz, L.W., Garbrecht, J., 1999. An outlet breaching algorithm for the treatment of
closed depressions in a raster DEM. Comput. Geosci. 25, 835–844.
Montgomerym, D.R., Dietrichm, W.E., 1989. Source areas, drainage density, and
channel initiation. Water Resour. Res. 25 (8), 1907–1918.
Montgomerym, D.R., Dietrichm, W.E., 1992. Channel initiation and the problem of
landscape scale. Science 255 (5046), 826–830.
Moore, I.D., Grayson, R.B., Ladson, A.R., 1991. Digital terrain modelling: a review of
hydrological, geomorphological and biological applications. Hydrol. Process. 5,
3–30.
Nikolakopoulos, K.G., Karmaratakis, E.K., Chrysoulakis, N., 2006. SRTM vs ASTER
elevation products. comparison for two regions in Crete Greece. Int. J. Remote
Sens. 27 (21), 4819–4838.
NRSC, 2011. Evaluation of Indian National DEM from Cartosat-1 Data, Summary
Report (Ver. 1). National Remote Sensing Centre (Indian Space Research
Organization), Hyderabad, p. 19.
O’Callaghan, J.F., Mark, D.M., 1984. The extraction of drainage networks from digital
elevation data. Comput. Vision Graph. Image Process. 28, 323–344.
Paik, K., 2008. Global search algorithm for nondispersive flow path. J. Geograph. Res.
113 (F04001), 9. http://dx.doi.org/10.1029/2007JF000964.
Pareta, K., Pareta, U., 2011. Quantitative morphometric analysis of a watershed of
Yamuna basin, India using ASTER (DEM) data and GIS. Int. J. Geomatics Geosci. 2
(1).
Patel, A., Katiyar, S.K., Prasad, V., 2016. Performances evaluation of different open
source DEM using Differential Global Positioning System (DGPS). Egypt J.
Remote Sens. Space Sci. 19, 7–16.
Planchon, O., Darboux, F., 2001. A fast, simple and versatile algorithm to fill the
depressions of digital elevation models. Catena 46, 159–176.
Radhika, V.N., Kartikeyan, B., Krishna, B.G., Chowdhury, S., Srivastava, P.K., 2007.
Robust stereo image matching for spaceborne imagery. IEEE Trans. Geosci.
Remote Sens. 45 (8), 2993–3000.
Reddy, G.P.O., Maji, A.K., Gajbhiye, K.S., 2002. GIS for morhophometric analysis of
river basins. GIS India 11 (9), 9–14.
Reddy, G.P.O., Maji, A.K., Chary, G.R., Srinivas, C.V., Tiwary, P., Gajbhiye, K.S., 2004a.
GIS and remote sensing applications in prioritization of river sub basins using
morphometric and USLE parameters—a case study. Asian J. Geoinform. 4 (4),
35–49.
Reddy, G.P.O., Maji, A.K., Gajbhiye, K.S., 2004b. Drainage morphometry and its
influence on landform characteristics in Basaltic Terrain – a remote sensing and
GIS approach. Int. J. Appl. Earth Obs. Geoinform. 6, 1–16.
Rieger, W., 1993. Hydrological terrain features derived from a pyramid raster
structure. In: Kovar, K., Natchtnebel, H.P. (Eds.), Application of Geographical
Information Systems in Hydrology and Water Resources Management. IAHS
Publication 211, Wallington, Oxford.
Sahu, N., Reddy, G.P.O., Kumar, N., Nagaraju, M.S.S., Srivastava, R., Singh, S.K., 2016.
Morphometric analysis in basaltic Terrain of Central India using GIS techniques:
a case study. Appl. Water Sci. http://dx.doi.org/10.1007/s13201-016-0442-z.
Sameena, M., Krishnamurthy, J., Jayaraman, V., Ranganna, G., 2009. Evaluation of
drainage networks developed in hard rock terrain. Geocarto Int. 24 (5), 397–
420.
Please cite this article in press as: Reddy, G.P.O., et al. Evaluation of automatic drainage extraction thresholds using ASTER GDEM and Cartosat-1 DEM: A
case study from basaltic terrain of Central India. Egypt. J. Remote Sensing Space Sci. (2017), http://dx.doi.org/10.1016/j.ejrs.2017.04.001
Santini, M., Grimaldi, S., Rulli, M.C., Petroselli, A., Nardi, F., 2009. Pre-processing
algorithms and landslide modeling on remotely sensed DEMs. Geomorphology
113 (1–2), 110–125.
Scheidegger, A.E., 1965. The algebra of stream order numbers. U.S. Geol. Surv. Prof.
Pap. 525 B, 187–189.
Schumn, S.A., 1956. Evaluation of drainage systems and slopes in badlands at Perth
Amboy, New Jersy. Bull. Geol. Soci. Am. 67, 597–646.
Shreve, R.L., 1966. Statistical law of stream numbers. J. Geol. 74 (1), 17–37.
Simon, W., Jonathan, L., Huang, G.H., 2008. A study on DEM-derived primary
topographic attributes for hydrologic applications: sensitivity to elevation data
resolution. Appl. Geograph. 28, 210–223.
Singh, P., Gupta, A., Singh, M., 2014. Hydrological inferences from watershed
analysis for water resource management using remote sensing and GIS
techniques. Egypt J. Remote Sens. Space Sci. 17, 111–121.
Staley, D.M., Wasklewicz, T.A., Blaszczynski, J.S., 2006. Surficial patterns of debris
flow deposition on alluvial fans in Death Valley, CA using airborne laser swath
mapping. Geomorphology 74, 152–163.
Strahler, A.N., 1952. Hypsometric (area-altitude) analysis of erosional topography.
Geol. Soc. Am. Bull. 63, 1117–1142.
Strahler, A.N., 1964. Quantitative geomorphology of drainage basins and channel
networks. Handbook of Applied Hydrology. McGraw-Hill, New York, pp. 4.40–
4.74.
Tarboton, D.G., Bras, R.L., Rodridguez Iturbe, I., 1988. The fractal nature of river
networks. Water Resour. Res. 24, 1317–1322.
Tarboton, D.G., Bras, R.L., Rodriguez-Iturbe, I., 1989. The Analysis of River Basins and
Channel Networks Using Digital Terrain Data, TR 326. Department of Civil and
Environmental Engineering, Massachusetts Institute of Technology, Boston.
Tarboton, D.G., Bras, R.L., Rodriguez-Iturbe, I., 1991. On the extraction of channel
networks from digital elevation data. Hydrol. Process. 5 (1), 81–100.
TNTmips, 2012. Tutorial on modeling watershed geomorphology. Microimages Inc.,
USA, p. 27.
Toutin, T., 2008. ASTER DEMs for geomatic and geoscientific applications: a review.
Int. J. Remote Sens. 29 (7), 1855–1875.
Tribe, A., 1992. Automated recognition of valley lines and drainage networks from
grid digital elevation models: a review and a new method. J. Hydrol. 139, 263–
293.
USGS, 2012. Global data explorer, http://gdex.cr.usgs.gov/ (accessed 11.12.12).
Van Zyl, J., 2001. The shuttle radar topography mission (SRTM): a breakthrough in
remote sensing of topography. Acta Astronautica 48 (5–12), 559–565.
Vijith, H., Satheesh, R., 2006. GIS based morphometric analysis of two major upland
sub-watersheds of Meenachil river in Kerala. J. Indian Soc. Remote Sens. 34 (2),
181–185.
Wang, L., Liu, H., 2006. An efficient method for identifying and filling surface
depressions in digital elevation models for hydrologic analysis and modelling.
Int. J. Geogr. Inf. Sci. 20, 193–213.
Wen, T.L., Wen, C.C., Chao, Y.L., Pi, H.H., Jing, S.T., 2006. Automated suitable drainage
network extraction from digital elevation models in Taiwan’s upstream
watersheds. Hydrol. Process. 20 (2), 289–306. http://dx.doi.org/10.1002/
hyp.5911.