E

SRa

b

c

d

a

ARA

KDCSAT

1

ordioTiWmmeoe1Tae

0h

International Journal of Applied Earth Observation and Geoinformation 21 (2013) 205–217

Contents lists available at SciVerse ScienceDirect

International Journal of Applied Earth Observation andGeoinformation

jo u r n al hom epage: www.elsev ier .com/ locate / jag

valuation of vertical accuracy of open source Digital Elevation Model (DEM)

andip Mukherjeea,b, P.K. Joshib,∗, Samadrita Mukherjeea, Aniruddha Ghoshb,.D. Gargc, Anirban Mukhopadhyayd

National Technical Research Organization (NTRO), Govt. of India, New Delhi, IndiaDepartment of Natural Resources, TERI University, New Delhi, IndiaIndian Institute of Technology (IIT), Roorkee, IndiaSchool of Oceanographic Studies, Jadavpur University, Kolkata, India

r t i c l e i n f o

rticle history:eceived 14 August 2012ccepted 13 September 2012

eywords:EM accuracyartosatRTM

a b s t r a c t

Digital Elevation Model (DEM) is a quantitative representation of terrain and is important for Earthscience and hydrological applications. DEM can be generated using photogrammetry, interferometry,ground and laser surveying and other techniques. Some of the DEMs such as ASTER, SRTM, and GTOPO30 are freely available open source products. Each DEM contains intrinsic errors due to primary dataacquisition technology and processing methodology in relation with a particular terrain and land covertype. The accuracy of these datasets is often unknown and is non-uniform within each dataset. In thisstudy we evaluate open source DEMs (ASTER and SRTM) and their derived attributes using high postings

STERerrain morphology

Cartosat DEM and Survey of India (SOI) height information. It was found that representation of terraincharacteristics is affected in the coarse postings DEM. The overall vertical accuracy shows RMS errorof 12.62 m and 17.76 m for ASTER and SRTM DEM respectively, when compared with Cartosat DEM. Theslope and drainage network delineation are also violated. The terrain morphology strongly influences theDEM accuracy. These results can be highly useful for researchers using such products in various modeling

exercises.

. Introduction

Digital Elevation Model (DEM) is a quantitative representationf the Earth’s surface providing basic information about the terrainelief (Guth, 2006). DEM and its derived attributes (slope, aspect,rainage area and network, curvature, topographic index, etc.) are

mportant parameters for information extraction or assessmentf any process using terrain analysis (Wolock and Price, 1994).hese are prerequisite in different applications such as model-ng water flow (Jain and Singh, 2003), estimating runoff (Cai and

ang, 2006; Chappell et al., 2006), flood simulation and manage-ent (Honghai and Altinakar, 2011; Ramlal and Baban, 2008), routeodeling (Romanowicz et al., 2008), mass movement (Iwahashi

t al., 2003), landform analysis (Weibel and Heller, 1990), creationf relief maps (Fraser et al., 2002), volcanic hazards (Vassilopoulouat al., 2002), terrain visualization and mapping (Spark and Williams,996), climate and meteorological studies (Thornton et al., 1997).he outcomes of the models depend on the accuracy of DEM (Zhang

nd Montgomery, 1994; Januchowski et al., 2010; Gómez-Gutiérrezt al., 2011).

∗ Corresponding author. Tel.: +91 11 2612 2222; fax: +91 11 2612 2874.E-mail addresses: pkjoshi27@hotmail.com, pkjoshi@teri.res.in (P.K. Joshi).

303-2434/$ – see front matter © 2012 Elsevier B.V. All rights reserved.ttp://dx.doi.org/10.1016/j.jag.2012.09.004

© 2012 Elsevier B.V. All rights reserved.

DEM is generated using different techniques such as pho-togrammetric method using stereo data (San and Suzen, 2005;Hohle, 2009), interferometry (Kervyn, 2001), airborne laser scan-ning (Favey et al., 2003), aerial stereo photograph (Schenk, 1996)and topographic surveys (Wilson and Gallant, 2000) using inter-polation of contours maps (Taud et al., 1999). Like any spatialdataset, DEM is subject to different type of errors such as gross errorduring data collection (Rodgriguez et al., 2006), deficient orienta-tion of stereo images (systematic error) with photogrammetricallydetermined elevation values (Mukherjee et al., 2011) and unknowncombinations of errors (random error) which cannot be avoided.These errors vary geographically depending on terrain conditions(Holmes et al., 2000). The other issues related to DEM accuracyare grid spacing and interpolation techniques (Mukherjee et al.,2011).

Acquisition of quality DEM data over large area is a challengingtask because of the complicated generation process. The availableopen source DEMs such as SRTM (1′′ for USA and 3′′ for otherareas, Di Luzio et al., 2005), ASTER GDEM (30 m, Frey and Paul,2012), GTOPO 30 (30′′ which is ∼1000 m, Kiamehr and Sjöberg,2005) and many others have coarser resolution. Small scale DEM

representation is required for global and regional scale simulationstudies, but the feasibility of application depends on vertical accu-racy (Brasington and Richards, 1998; Dragut and Eisank, 2011).Attempts have been made to examine the vertical accuracy of DEM

2 Earth Observation and Geoinformation 21 (2013) 205–217

(2eewrs

dietG(Cwhrn

2

RhttsDtds

RdIAoatthpr18tlrrf

bcflSaAf2parTi

Table 1Horizontal and vertical datum of elevation data sets.

Data Horizontal datum Vertical datum

Cartosat DEM WGS 84 WGS 84ASTER DEM WGS 84 EGM 96SRTM DEM WGS 84 EGM 96GCPs WGS 84 WGS 84

77◦43′2′′E to 78◦2′52′′E (Fig. 2). The study area encompasses partof Dehradun, the capital city of Uttarakhand state, India. A stretchof 27 km × 27 km, equivalent to Cartosat-1 single swath is taken.The study area is highly rugged and significant variation of relief

06 S. Mukherjee et al. / International Journal of Applied

Wu et al., 2008; Vaze et al., 2010; Zhou et al., 2012; Hirano et al.,003; Bourgine and Baghdadi, 2005; Kornus et al., 2006; Tarekegnt al., 2010; Frey and Paul, 2012). Still, there is enough scope tovaluate the open source DEMs because ASTER GDEM Version 2as released on October 17, 2011. Also, analysis of DEM accu-

acy in Indian landscape especially in Himalayan terrain is verycarce.

The present study is to assess the accuracy of DEM in fourifferent ways. First, visualizing the effect of DEM grid size (post-

ng) for representation of surface and analysing the smootheningffect on surface representation. Second, examining the magni-ude of vertical error of ASTER GDEM version 2 and SRTM usingCPs collected from field, surveyed elevation from Survey of India

SOI) topographic sheets and regular grid surface of high resolutionartosat-1 DEM. Third, examining the association of DEMs errorith terrain morphology and identifying the variation of error ineight with respect to relief and slope. Fourth, analysing the accu-acy of DEM derived terrain attributes such as slope and drainageetwork.

. Cartosat, ASTER and SRTM Digital Elevation Model

Cartosat-1 is an Indian satellite launched by Indian Spaceesearch Organisation (ISRO) on May 5, 2005. Cartosat-1 satelliteas forward (F) and aft (A) panchromatic camera which gives alongrack stereo, with a tilt in flight direction of ±26◦ and ±5◦, respec-ively (Baltsavias et al., 2007). It is the first Indian remote sensingatellite which provides in-orbit stereo images which is useful forEM generation. LPSz® software (9.3 version) was used for orien-

ation of the stereo block. High resolution (2.5 m) Cartosat-1 stereoata of 2nd October 2005 (path/row 0526/0258) was used in thistudy for DEM generation.

The Advanced Space borne Thermal Emission and Reflectionadiometer (ASTER) Global Digital Elevation Model (GDEM) waseveloped jointly by the METI (Ministry of Economy, Trade, and

ndustry) of Japan and the NASA (National Aeronautics and Spacedministration) of United States. The ASTER sensor was launchednboard NASA’s Terra spacecraft in December 1999. It has anlong-track stereoscopic capability using its near infrared spec-ral band (3N) and its nadir-viewing and backward-viewing (3B)elescopes (27.7◦ angle) to acquire stereo image with a base-to-eight ratio of 0.6. The GDEM product is generated from automaticrocessing of 1.5 million stereo pairs by applying the stereo cor-elation methodology. The spatial resolution of band 3N and 3B is5 m and generated DEM is 30 m. It covers land surfaces between3◦N and 83◦S (ASTER GDEM Readme Handbook). The absolute ver-ical accuracy of ASTER GDEM version-1 is 20 m at 95% confidenceevel. The improved vertical accuracy of ASTER GDEM version-2,eleased on October 17, 2011, is 8.86 m (ASTER GDEM V2 validationeport). For this study, ASTER GDEM version-2 was downloadedrom http://demex.cr.usgs.gov/DEMEX/.

Shuttle Radar Topography Mission (SRTM) was a joint missiony National Imagery and Mapping Agency (NIMA) and NASA toollect global elevation data set. Specially modified radar systemew for an 11-day mission in February 2000 with onboard Spacehuttle Endeavour (Sun et al., 2003). The SRTM elevation datare derived from X-band and C-band Interferometric Syntheticperture Radar (InSAR) sensor (5.6 cm wavelength and 5.3 GHz

requency) (Gorokhovich and Voustianiouk, 2006; Van Niel et al.,008). SRTM data are available at 30 m posting for USA and 90 mosting for entire globe (60◦N to 56◦N). The absolute vertical height

ccuracy is 16 m (90% linear error) and the absolute horizontal accu-acy is 20 m (90% circular error) (http://www.jpl.nasa.gov/srtm/).he seamless voids filled (SRTM upgraded version-4) data setss available at Consultative Group for International Agriculture

SOI Everest MSL

Source: ASTER and SRTM data user Handbook.

Research Consortium for Spatial Information (CGIAR-CSI) websitehttp://srtm.csi.cgiar.org, from which tile No. 52 06 was down-loaded.

3. Reference datum of various DEMs

The vertical datum of Cartosat, ASTER GDEM, SRTM and SOItopographic sheet heights are different. The vertical and horizontaldatum of all these data sets is given in Table 1. The global posi-tioning system (GPS) uses WGS84 vertical datum as default andcomputes the height relative to this (Kaplan and Hegarty, 2006).The elevation of a point on Earth surface computed from Mean SeaLevel (MSL) can vary from GPS derived elevation because of the vari-ation between WGS84 ellipsoid and Geoid (local MSL). The Geoidsurface is an equipotential or constant geopotential surface whichcorresponds to MSL. The geoid height/geoid undulation (N) is thedifference in height between geoid and ellipsoid at a point. Fig. 1represents ellipsoid height (h) and height above geoid surface (H)which is orthometric height. However, it can be derived that:

h = H + N (1)

The magnitude of geoid height/undulation ranges from a lowof about −105 m at southern tip of India to a height above+85 m at New Geinea (Kaplan and Hegarty, 2006). Error in DEMis usually estimated by taking the difference with referencedata elevation in which two additional factors, vertical datumand horizontal mismatch introduce the apparent vertical error(Shortridge and Messina, 2011). Hence, for the comparison of GPSelevation/WGS84 reference elevation data with SRTM, conversionof the vertical datum (datum matching) should be considered(http://hydrosheds.cr.usgs.gov/hydro.php).

4. Study area

The study is carried out in western part of Shiwalik Himalaya,geographically situated between 30◦8′30′′N to 30◦27′3′′N and

Fig. 1. Relation between ellipsoid height, orthometric height and geoid undulation.

S. Mukherjee et al. / International Journal of Applied Earth Observation and Geoinformation 21 (2013) 205–217 207

Fig. 2. DEM of study area is shown using 10 m grid spacing. 3D perspective view of the terrain shows the lower middle part, highly dissected rugged terrain by the rivers.The middle part is the Dun valley and foot hill zone.

iaddwtp

s present. Geomorphologically, the area is characterized by hillsnd valleys. The lower middle part of the study area is mainlyominated by hills. Due to the pressure of steep slope, the area is

issected by number of small rivers. The type of forests is mixed; inhich Sal forest is dominating and its density varies from medium

o scattered. On the side of the river valley scattered vegetation isresent.

5. Methodology

5.1. Generation of Cartosat DEM for reference data

To generate the Cartosat DEM, orientation of stereo pair wascarried out to solve the basic problem of determining object spacecoordinate (x, y and z) of a point in the image space. For this,

208 S. Mukherjee et al. / International Journal of Applied Earth Observation and Geoinformation 21 (2013) 205–217

n usi

trimtTbSagpvfiM

5

auAtesvlgsr

mut2c

5

sdet[r

Fig. 3. Part of the study area is show

wo types of orientation are needed, interior orientation and exte-ior orientation (Grodecki and Gene, 2003). Initially stereo blocks oriented with RPCs, and then GCPs are added in the stereo

odel to avoid systematic errors. 31 well-distributed GCPs overhe scene were collected using Lieca single frequency GPS receiver.he received signals are differentially corrected with the help ofase station receiver data during the post processing of GCPs usingki Pro software. 15 GCPs were used as control points and 5 GCPss check points in order to orient the stereo model. DEM at 10 mrid size was generated using oriented stereo block. The DEM isrojected to UTM projection with zone 43 and the horizontal andertical datum was considered as WGS 84. No land cover basedltering or objects height removal was performed. Digital Surfaceodel (DSM) was used for the further processing and usage.

.2. Vertical datum matching

In order to match the same vertical datum, Cartosat DEMnd GPS height were transferred from WGS84 to EGM96 datumsing mathematical geoid model. National Imagery and Mappinggency (NIMA) and National Aeronautics and Space Administra-

ion (NASA) provide well-known global geoid model by whichlevation derived relative to WGS84 can be transferred to EGM96urface, formally referred as EGM96 geo-potential model. It pro-ides correction coefficient and computes geoid height overand area (http://cddis.gsfc.nasa.gov/926/egm96/egm96.html). Theeoid height (N) grid is available at 15′ spacing. The Cartosat DEMurface and GPS heights have been converted into EGM96 geoideference surface using the geo-potential model.

Bench mark (BM) and spot height (surveyed elevation pointarked from the SOI Toposheet) No. 53J/3, 4, 53F/15, 16 were also

sed. The vertical datum of SOI Toposheet is considered as MSL. Ashe EGM96 surface is very close approximation of MSL (Sun et al.,003), the height taken from SOI topographic sheets was directlyompared with other DEM derived elevation values.

.3. Computation of terrain attributes

The two necessary terrain parameters derived from DEMs arelope and drainage network. Accuracy of slope and the variation inrainage network delineation derived from open source DEMs were

valuated. Slope represents the magnitude of the terrain inclina-ion. Slope at a given point on a surface is a function of height valuesslope(x,y) = f(Z)]. It is the first derivative of elevation describingate of change of elevation. Together, the slope in the x direction

ng Cartosat, ASTER and SRTM DEM.

and slope in the y direction (partial derivatives of Z with respectto the x and y directions) define gradient vector of the surface (Eq.(2)). The maximum slope can be determined by taking the normof this vector. On a grid DEM, slope calculation is performed using3 × 3 moving window to derive finite differential. In this study, sec-ond order finite difference was used. Four Closest Neighbors (FCN)algorithm (Guth, 1995; Raaflaub and Collins, 2006) was used forcomputing slope. It takes into account two orthogonal componentsof slope: in x and y directions. In other words, the algorithm usedfour cardinal neighbors, that is North, South, East and West repre-senting a second order finite difference relationship (Eq. (2)). Thisdefines steepness and downhill direction.

slope =√(

dz

dx

)2

+(

dz

dy

)2

(2)

dz

dx= z8 − z2

2g,

dz

dy= z6 − z4

2g(3)

Drainage network are the channels along which fluvial pro-cesses work to transport water and mineral material of a localregion, allowing gravity process on slope to continue to lowerlandscape. Identification of catchment/watershed is carried outbased on drainage network which is important for geomorpho-logical and hydrological studies. It is basically a channelized flowby which water drains towards down slope (O’Callaghan andMark, 1984). Drainage network was derived from Cartosat, ASTERand SRTM DEM. Flow direction and flow accumulation rasterwas generated using blocks of 3 × 3 pixels based on the steepestslope. Based on flow direction and accumulation raster, drainagenetwork was extracted (Tarboton, 1997). Drainage lines havingspecific catchment area greater than 0.03 km2 were taken intoconsideration. Drainage network ordering was done based on themethod given by Strahler (Tarboton et al., 1991).

5.4. Validation

Accuracy is the closeness of observation to a true value (Mauneet al., 2001). Accuracy assessment of Cartosat DEM was performedusing 11 GCPs and elevation values (BM and Spot heights) of SOItopographic sheet. The accuracy of ASTER and SRTM DEMs surfacesis estimated by comparing with GCPs, BM and spot heights of topo-

graphic sheet (30 point locations) and 10 m grid spacing CartosatDEM. The ASTER and SRTM DEMs were found to be shifted approx-imately 60 m in x-direction and 40 m y-direction as well as 250 min x-direction and 200 m in y-direction with respect to Cartosat

S. Mukherjee et al. / International Journal of Applied Earth Observation and Geoinformation 21 (2013) 205–217 209

Fig. 4. Representation of surface in various grid sizes, (a) Cartosat DEM (vertical datum is transformed to EGM96), (b) ASTER and (c) SRTM, Simulated (d) ASTER and (e) SRTMelevation from Cartosat are also shown.

Fig. 5. Terrain profile derived from three DEMs along the section line. Profiles showing the smoothening affect in representation of surface by ASTER and SRTM elevationdata.

210 S. Mukherjee et al. / International Journal of Applied Earth Observation and Geoinformation 21 (2013) 205–217

0

5

10

15

20

25

30

35a

b

c d

1 3 5 7 9 11 13 15 17 19 21 23 25 27 29

ASTE

R -T

opos

heet

hei

ght (

in m

)

Sample points

0

10

20

30

40

50

1 3 5 7 9 11 13 15 17 19 21 23 25 27 29

SRTM

-Top

oshe

et H

eigh

t (in

m)

Sample point

y = 1.0005x + 1.8205

R² = 0.9828

200

400

600

800

1000

300 500 700 900

ASTE

R he

ight

(in

m)

Toposheet height (in m)

y = 0.9906x + 2.4595

R² = 0.9762

200

400

600

800

1000

300 500 700 900

SRTM

hei

ght (

in m

)

Toposheet height (in m)

oshee

DpiRacte(

R

Fig. 6. (a) Absolute difference in ASTER – Toposheet height, (b) SRTM – SOI Top

EM. To account for this spatial misalignment, a co-registrationrocess with Cartosat was applied. The validation was performed

n terms of root mean square error (RMSE) and mean error (ME).MSE exhibits on average how far observed values differ from thessumed true value. The ME tells us whether set of measurementsonsistently underestimate (negative ME) or overestimate (posi-ive ME) the true value. The RMSE is a single quantity characterizingrror surface, and mean error reflects the bias of the error surfaceMukherjee et al., 2011). The equations as follows:

MSE =

√√√√[n−1

n∑i=−1

(DEMref10 − DEM)2

](4)

t height, (c) level of agreement among Toposheet height with ASTER and SRTM.

ME =[

n−1n∑

i=1

(DEM − DEMref10)

](5)

The level of agreement between ASTER and SRTM derived eleva-tion values and reference data are also evaluated in terms of linearregression and correlation.

6. Results and discussion

6.1. Representation of surface

The oriented stereo block of Cartosat data was used to gener-ate10 m grid spacing DEM. The elevation range is 275–973 m withmean and standard deviation of 552 m and 111 m, respectively. Dueto presence of cloud in the data, some bright patches were seen in

S. Mukherjee et al. / International Journal of Applied Earth Observation and Geoinformation 21 (2013) 205–217 211

0

5

10

15

20

25

30

a

b

c d

1 3 5 7 9 11 13 15 17 19 21 23 25 27 29

AS

TE

R

-C

art

osa

t h

eigh

t (i

n m

)

Sample points

0

10

20

30

40

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30

SR

TM

- C

art

osa

t H

eigh

t (i

n m

)

Sample points

y = 1.0036x + 0.9937

R² = 0.9929

200

400

600

800

1000

300 400 500 600 700 800 900

AS

TE

R

Cartosat

y = 0.9947x + 1.0192

R² = 0.9884

300

400

500

600

700

800

900

300 400 500 600 700 800 900

SR

TM

Cartosat

F Carto

tStaw6

eoidovm((rStwAsm

ig. 7. (a) Absolute difference in ASTER height – Cartosat height, (b) SRTM height –

he DEM which were masked out. The overlap region of ASTER andRTM with Cartosat DEM (27 km × 27 km) was considered. Eleva-ion range of ASTER is 324–1026 m. The mean elevation is 603 mnd standard deviation is 114 m. In SRTM data, the elevation variesithin 333 m and 1050 m with mean and standard deviation of

04 m and 113 m.The surface is represented using Cartosat, ASTER and SRTM

levation model. Representation of terrain surface depends on res-lutionof DEM. In the finer resolution, surface can be representedn detail and small undulation characteristics of the terrain areepicted (Fig. 3). In the coarse postings, terrain heights are averagedut and representation becomes smooth. To analyse this, elevationalues of 90 m × 90 m area, equivalent to a pixel of SRTM elevationodel were visualized by Cartosat (10 m), ASTER (30 m) and SRTM

90 m) DEM. It covers 81 Cartosat pixels (Fig. 4a), 9 ASTER pixelsFig. 4b) and 1 SRTM pixel (Fig. 4c). It shows that details of the ter-ain are better represented in Cartosat and ASTER comparison withRTM due to the finer spacing. If we simulate the coarse resolu-ion surface by aggregating the Cartosat DEM using 3 × 3 and 9 × 9

indow (Fig. 4d and e), the elevation values are not matching withSTER and SRTM elevation. It signifies that due to the systematicampling in finer spacing, terrain characteristics representation isore detailed and close to real ground. The terrain cross profile

sat height, level of agreement among Cartosat height with (c) ASTER and (d) SRTM.

along a section line is shown in Fig. 5. It shows that the slope gradi-ent is steep in finer spaced DEM. The detail of the surface is averagedout in coarser grid spacing and terrain undulation characteristicsare also generalized.

6.2. Evaluation of Cartosat DEM

Overall accuracy for Cartosat DEM (RMSE) was 1.06 pixels andvertical check point accuracy was 1.11 m (RMSE) derived from LPS.The DEM was also evaluated using 11 GCPs collected by DGPS(Table 2). As the Cartosat DEM height and GCPs height are refer-enced with WGS 84 datum, both can be compared directly. It hasbeen found that the average vertical error is 3.2 m (calculated fromabsolute difference), RMSE is 3.64 m and mean bias −0.42 m.

The Cartosat DEM surface was also validated using 30 surveyedpoint elevations (BM/Spot height) taken from SOI Toposheet. Theelevation values given in the SOI Toposheet and Cartosat DEM arewith respect to MSL and WGS84, respectively. Hence, the elevationvalues taken from Cartosat DEM (for the same 30 points) were con-

verted from WGS84 into EGM96 by calculating the geoid heightsthrough the EGM96 geoid model available in the open source(http://earth-info.nga.mil/GandG/wgs84/gravitymod/egm96/intpt.html) which is a close approximation of MSL. The difference in

212 S. Mukherjee et al. / International Journal of Applied Earth Observation and Geoinformation 21 (2013) 205–217

Table 2Error in Cartosat DEM with respect to GCPs.

GPS point UTM coordinates ZGCPs (m)

ZDEM (m)

Error in DEM (m)

X Y

Point 1 769,986.27 3,367,905.01 454.65 456.28 1.63Point 2 777,381.85 3,369,567.47 648.50 643.32 −5.18Point 3 771,532.48 3,370,524.05 492.87 494.12 1.25Point 4 777,340.91 3,358,636.80 490.26 484.76 −5.5Point 5 770,981.69 3,360,624.53 480.39 482.69 2.3Point 6 772,976.63 3,360,482.88 467.71 464.66 −3.05Point 7 768,903.00 3,345,906.56 371.58 373.54 1.96Point 8 765,861.25 3,347,501.29 371.40 374.57 3.17Point 9 783,589.54 3,368,403.41 775.17 768.77 −6.4Point 10 772,341.35 3,343,955.79 364.73 366.36 1.63

43

bD

6

dWfefT9btbo

TE

Point 11 770,017.87 3,364,620.82

oth the elevation was shown in Table 3. The RMSE of CartosatEM calculated is 4.83 m and the mean error is 0.19 m.

.3. Comparison of ASTER and SRTM with GCPs

The ASTER GDEM and SRTM data were compared with 11 DGPSerived GCPs. The GPS system provides elevation reference toGS84 surface but ASTER and SRTM are referenced to EGM96 sur-

ace. These two heights cannot be compared directly. The GPS basedllipsoid height was converted into EGM96 geoid reference sur-ace and compared with ASTER and SRTM height shown in Table 4.he RMS error for the ASTER and SRTM calculated is 6.08 m and.2 m with mean error of −2.58 m and −2.94 m, respectively. In

oth cases, the error is less than the error specification given byhe nodal agency (8.86 m for ASTER and 16 m for SRTM). This coulde because the GCPs are taken in the flat terrain due to accessibilityf the field work.

able 3rror in Cartosat DEM with respect to SOI Toposheet derived height.

BM/spot height Toposheet height Cartosat height Geoid u

1 489.1 447.99 −44

2 544 495.5 −44.44

3 618 575.47 −44.15

4 601.2 551.84 −44.63

5 638.8 597.22 −45.02

6 453.8 405.68 −46.84

7 468 419.54 −44.57

8 482 441.24 −44.32

9 561.4 522.22 −44.29

10 675 631.77 −45.55

11 713 664.37 −46.06

12 774 723.02 −46.55

13 515 468.58 −46.68

14 418.6 375.23 −46.15

15 410 359.94 −45.21

16 543 493.77 −45.09

17 581 534.79 −45.22

18 579 538.18 −45.18

19 848 789.89 −46.6

20 672 640.6 −46.12

21 818 774.08 −45.98

22 468 419.54 −44.56

23 478 439.47 −44.14

24 695 653.34 −43.27

25 631 582.97 −44.04

26 705 658.05 −45.77

27 415 373.37 −47.03

28 410 358.94 −47.21

29 624 579.75 −45.31

30 491 450.55 −44.72

3.73 437.28 3.55

6.4. ASTER and SRTM height comparison with SOI Toposheet

Accuracy of ASTER and SRTM DEMs was also evaluated using 30point location taken from SOI Toposheet height (BM/Spot height).The vertical datum of SOI Toposheet is MSL which is closely com-parable with EGM96. Therefore, matching datum is not necessary.The absolute differences in heights are shown in Fig. 6. The RMSerror calculated is 16.06 m and 18.91 m with mean bias of −2.14 mand 2.10 m for ASTER and SRTM, respectively. ASTER is showingnegative bias which suggests the under estimation of elevationvalues but SRTM gives the positive bias which leads to over estima-tion. The RMS error is much higher compared to error specified inthe validation report of the nodal agency. The level of agreements

(R2) between Toposheet height and ASTER and SRTM are 0.982 and0.976 (Fig. 6c). Correlation between ASTER and Toposheet height ishigher which signifies that accuracy of ASTER is better comparedto SRTM.

ndulation Transform Cartosat height Error in Cartosat height

491.99 2.89539.94 −4.06619.62 1.62596.47 −4.73642.24 3.44452.52 −1.28464.11 −3.89485.56 3.56566.51 5.11677.32 2.32710.43 −2.57769.57 −4.43515.26 0.26421.38 2.78405.15 −4.85538.86 −4.14580.01 −0.99583.36 4.36836.49 −11.51686.72 14.72820.06 2.06464.1 −3.9483.61 5.61696.61 1.61627.01 −3.99703.82 −1.18420.4 5.4406.15 −3.85625.06 1.06495.27 4.27

S. Mukherjee et al. / International Journal of Applied Earth Observation and Geoinformation 21 (2013) 205–217 213

Table 4Error in ASTER and SRTM DEM with respect to GPS derived height.

GPS point ZGCPs (m)

Geoid undulation Transform GPS height Z SRTMDEM (m)

ErrorSRTM DEM (m)

Z ASTERDEM (m)

ErrorASTER DEM (m)

Point 1 454.65 −42.89 497.54 488 −9.54 490 −7.54Point 2 648.50 −42.10 690.60 695 4.40 692 1.40Point 3 492.87 −42.40 535.27 526 −10.27 531 −4.27Point 4 490.26 −43.86 534.12 521 −9.12 530 −4.12Point 5 480.39 −43.93 524.32 514 −2.32 521 −3.32Point 6 467.71 −43.81 511.52 504 −8.52 503 −8.52Point 7 371.58 −46.08 417.66 420 −17.66 412 −5.66Point 8 371.40 −46.03 417.43 422 4.57 410 −7.43Point 9 775.17 −41.74 816.91 804 −12.91 828 11.09Point 10 364.73 −46.18 410.91 424 −10.91 414 3.09Point 11 433.73 −43.39 477.12 483 3.88 474 −3.12

EGM9

6

ttacbb

pcCDHvefmCea5a

1eiGsaTvi

Terrain morphology is one of the major influencing factors forvertical accuracy of DEM. In order to evaluate this, Cartosat DEM isdivided into 5 altitudinal zones (<400 m, 401–500 m, 501–600 m,

27

25

19

12

18

31

22

16

13

18

5

10

15

20

25

30

35

Are

a (

in %

)

ASTER

SRTM

Fig. 8. Geoid undulation calculated from (a)

.5. ASTER and SRTM height comparison with Cartosat DEM

ASTER and SRTM height value was also compared with Car-osat height (reference vertical datum converted into EGM96) onhe same 30-point location mentioned in the previous section. Thebsolute differences in heights are shown in Fig. 7. The RMS erroralculated for ASTER and SRTM are10.68 m and 13.18 m with meanias of 3.09 and 2.37 m, respectively. The level of agreements (R2)etween Cartosat height with ASTER and SRTM are 0.992 and 0.988.

The assessment of a DEM surface with respect to few sampleoints is not good enough. Due to this reason, surface-to-surfaceomparison of ASTER and SRTM with Cartosat DEM was carried out.artosat does not provide a perfect elevation model. But CartosatEM is considerably more accurate compared to ASTER and SRTM.ence, Cartosat DEM is considered as reference data. To match theertical datum, Cartosat surface was transferred to EGM96 refer-nce datum. The height differences between WGS84 and EGM96or the entire study area (Fig. 8) have a mean value of −44.71 with

inimum and maximum of −46.71 and −41.42. Using this offset,artosat DEM ellipsoid height was converted into EGM96 geoid ref-rence surface (refer Eq. (1)) and the minimum, maximum, meannd standard deviation of the surface were changed to 321, 1015,96 and 110. ASTER and SRTM were then compared with Cartosatnd different surfaces were generated.

The ASTER elevation difference surface ranges from 138 m to99 m, but major area comes under −30 m to 15 m. The overall RMSrror for the surface is 12.62 with mean bias of −5.53 m. Mean biasndicates that due to smoothening effect in coarser posting ASTERDEM, the elevation is underestimated. The differenced surfacehows the positive bias in western face of the mountainous slope

nd agricultural and bare land of north western part of study area.he negative bias is associated with mountain ridge, plain land andegetated area. The aerial distribution height error shown in Fig. 9ndicates 27% area under less than 5 m error and 52% area under

6 model and (b) transformed Cartosat DEM.

less than 10 m elevation error. 30% areas have greater than 15 melevation error where variation of relief is higher.

The elevation difference surface of SRTM ranges from −142 mto 206 m, though maximum pixels fall under −25 m to 12 m. TheRMS error for the surface is 17.76 m with mean bias of −5.42 m. Itwas found from the difference surface that positive bias is associ-ated with vegetation covered mountainous area and negative biasassociated with open land and other land cover types. Fig. 8 showsin 69% of the study area; the elevation error is less than 15 m whichsatisfy the error specification of the SRTM mission. 31% area witherror greater than 15 m is mainly due to rough terrain. The his-togram of the error surface of ASTER and SRTM also shows a normaldistribution (Fig. 10) which is statistically significant.

6.6. Effect of terrain morphology in DEM accuracy

0

< 5m 5-10 m 10-15 m 15-20 m >20m

Err or in height (i n m)

Fig. 9. Aerial distribution of vertical accuracy of ASTER and SRTM.

214 S. Mukherjee et al. / International Journal of Applied Earth Observation and Geoinformation 21 (2013) 205–217

Fig. 10. Statistical distribution of ASTER – Cartosat DEM (a) and SRTM – Cartosat DEM surface (b).

Table 5Statistical characteristics of altitudinal zone (Cartosat DEM).

Altitudinal zone Elevation Slope

Mean Variance Mean Variance

<400 377.80 14.67 3.83 4.81401–500 457.12 30.72 7.17 8.58501–600 555.79 29.19 8.58 10.36601–700 646.56 28.27 12.26 12.16

6vbvahvecncuso

i

F

9.44 9.46 10.67

13.01

17.59

9.22 9.99

12.52

17.25

22.25

0

5

10

15

20

25

<2 2-5 5-10 10-20 > 20

RM

S E

rror

(in

m)

Slope (i n degree )

ASTER

SRTM

700–800 741.64 27.93 17.59 12.27>800 836.67 36.65 19.16 12.99

01–700 m and >701 m). The statistical characteristics of each ele-ation zones are derived (Table 5). The variance of elevation cane an indicator of terrain roughness. A high variance indicates highariability in local terrain surface (Holmes et al., 2000). The vari-nce of elevation and slope indicates roughness of the terrain beingigher in high altitudinal zone (Table 5). The RMS error of elevationalue of ASTER and SRTM within each zone is calculated from differ-nced surfaces (Fig. 11) showing the effect of terrain morphologicalharacteristics on DEM accuracy. The DEM surface is more erro-eous in high altitudinal zone where terrain is rugged. The RMSEurve shows an increasing trend indicating that relief increases thencertainty of height measurement. The ASTER surface provideslightly more accuracy compared to SRTM which could be the effect

f grid spacing.

The effect of slope on the height accuracy is shown in Fig. 12. Itndicates that in relatively flat area (slope less than 5◦), the accuracy

7.11

12.15 12.18

15.16

19.17

22.40

4.25

11.39

13.63

17.34

22.86

26.05

0

5

10

15

20

25

30

<400 40 1-50 0 50 1-60 0 60 1-70 0 70 1-80 0 >800

RM

S E

rror

(in

m)

Height (i n m)

ASTER

SRTM

ig. 11. Vertical accuracy of ASTER and SRTM relative to terrain morphology.

Fig. 12. Vertical accuracy of ASTER and SRTM relative to slope of the terrain.

of DEMs is less than 10 m. The error increases rapidly when theslope value is greater than 10◦ which is also found in the study ofGorokhovich and Voustianiouk (2006). It also signifies the effect ofrelief on DEM accuracy.

The areal distribution of Cartosat, ASTER and SRTM within eachaltitudinal zone is compared. The % of area to total area failingwithin the altitudinal zone is computed. Fig. 13 shows variationin areal distribution among Cartosat, ASTER and SRTM which again

signifies the error in estimating the height value in ASTER and SRTMin rugged terrain.

0

50

100

150

200

250

300

<400 401 -500 501 -600 601 -700 701 -800 >800

Area

(in

Sq K

m)

Al�tud inal Zone (in m)

Cartosat

ASTER

SRTM

Fig. 13. Areal distribution of DEMs within each altitudinal zone.

S. Mukherjee et al. / International Journal of Applied Earth Observation and Geoinformation 21 (2013) 205–217 215

Table 6Statistical characteristics slope derived from ASTER and SRTM elevation data.

Altitudinal zone Maximum Mean Std

Statistics of ASTER DEM derived slope<400 29.34 4.96 2.92401–500 45.98 6.60 5.19501–600 56.14 8.31 7.25601–700 57.11 10.13 8.04701–800 56.71 13.08 8.02>800 53.34 14.61 8.29

Statistics of SRTM DEM derived slope<400 6.16 1.07 0.56401–500 28.75 3.26 3.85501–600 37.18 4.71 5.83601–700 40.93 5.84 5.94

6

pRaifts

AMca

F

701–800 39.35 8.65 5.79>800 36.05 10.62 6.54

.7. Evaluation of slope and drainage network

The slope of the surface derived from ASTER and SRTM is com-ared with Cartosat DEM derived slope. The validation result showsMS error of 7.27◦ and 8.18◦ as well as mean error of −1.21◦

nd −2.54◦ for ASTER and SRTM, respectively. Due to smoothen-ng effect, slope of the surface is underestimated when calculatedrom ASTER and SRTM that is indicated by the negative bias. Dueo coarser resolution of ASTER and SRTM, the representation of theurface slope decline.

The statistical characteristics of the slope maps calculated fromSTER and SRTM for each altitudinal zone are given in Table 6.

aximum, mean and standard deviation of slope reduce when

alculated from SRTM, which again signifies the effect of gener-lization due to coarse posting. Fig. 14 shows RMS error and mean

5.41

8.08 8.56

10.28

11.81

12.85

4.86

7.58 8.20

10.13

11.41

12.54

0

2

4

6

8

10

12

14

<40 0 40 1-50 0 50 1-60 0 60 1-70 0 70 1-80 0 >800

Slo

pe

(in

deg

ree)

Height (i n m)

ASTER

SRTM

-11

-9

-7

-5

-3

-1

1 <40 0 40 1-50 0 50 1-60 0 60 1-70 0 70 1-80 0 >800

Mea

n e

rror

(in

m)

Height (i n m)

ASTER

SRTM

a

b

ig. 14. (a) RMS error and (b) mean error of slope relatives to altitudinal zone.

Fig. 15. Drainage network extracted from Cartosat, ASTER and SRTM elevationmodel.

error of slope within each altitudinal zone. The accuracy curvesof the slope suggest that the decrease of slope accuracy is influ-enced by terrain roughness. With coarser spacing, the relative reliefof an area with its neighborhood decreases and slope calculationbecomes more erroneous.

The drainage network generated from Cartosat, ASTER andSRTM elevation models are shown in Fig. 15. Due to higher reliefvariation in the central and northern part of the study area, numbersof small channels originate. Hence, drainage line having specificcatchment area >0.03 km2 is considered in the drainage network.Cartosat DEM delineated streams are seen as very smooth in charac-ter compared to ASTER and SRTM derived stream lines. The higherorder stream (main stream having 4th or 5th order) delineatedfrom ASTER shows more similarity with Cartosat derived streamcompared to SRTM. The major variation is found in the first andsecond order stream. The small channels are found to be hiddenwhen derived from coarse posting DEM.

7. Conclusion

The vertical accuracy of ASTER GDEM Version 2 and SRTM ele-vation model was tested in the present study. The validation wasperformed based on three reference data; GCPs, height taken fromSOI Toposheet and high posting Cartosat DEM. Matching of ver-tical datum was carried out by comparing the height values. The

slope and drainage network derived from these DEMs were eval-uated. The effect of terrain morphological characteristic was alsoanalysed.

2 Earth O

tteoaIhowm

tahwwbcSToiis

a2vmrhtf

A

DCIe

R

B

B

B

C

C

D

D

F

F

F

16 S. Mukherjee et al. / International Journal of Applied

The experimental results conclude that the surface represen-ation is highly influenced by DEM posting. The detail of theerrain variation reduces in the coarser posting. The slope gradi-nt becomes less due to smoothening effect. The vertical accuracyf the DEMs is affected by the terrain morphological characteristicsnd terrain roughness negatively influences the vertical accuracy.n the higher altitude (>600 m) where the variance of elevation isigh, the error of height is also increased. The slope characteristicf the terrain has significant impact on ASTER and SRTM accuracyhere terrain slope is above 10◦. The error values of SRTM exceedission specifications for this study site.Interestingly, the accuracy of ASTER and SRTM heights exceeds

he mission specification when compared with Toposheet heightnd Cartosat DEM. Overall both DEMs show underestimation ofeight relative to Cartosat surface. The positive bias is associatedith vegetated mountainous region and negative bias is associatedith open and agricultural land cover. Slope calculation is affected

y the DEMs posting. The mean and variance of slope reduce whenalculated from coarser posting DEM. The accuracy of ASTER andRTM derived slope is highly violated in comparison with Cartosat.he terrain morphology also has an impact on slope accuracy. Slopef the terrain is underestimated in the rugged terrain. The variations seen in the drainage network delineation due to the smoothen-ng effect in larger posting. The first order streams are hidden andhifting of higher order channels is also seen.

Findings of the study are important to understand the errorssociated with SRTM and recently released ASTER GDEM version

data. It also depicts the spatial characteristics of DEM error forarious land cover, slope and terrain morphology. The study recom-ends that the elevation models can be very useful for small scale

egional level study. Further studies can be carried out to find theorizontal accuracy of open source DEM and also to examine howhe first and second order terrain attributes vary while calculatingrom SRTM and ASTER DEM.

cknowledgements

Author is thankful to Indian Institute of Remote Sensing,ehradun for providing the dataset for this study. AG acknowledgesouncil of Scientific and Industrial Research (CSIR), Government of

ndia for support. Authors are also thankful to anonymous review-rs for their comments.

eferences

altsavias, E., Kocaman, S., Wolff, K., 2007. Geometric and radiometric investigationsof Cartosat-1 data. In: ISPRS Hannover Workshop 2007, High Resolution EarthImaging for Geospatial Information, Hannover, Germany, May 29–June 1.

ourgine, B., Baghdadi, N., 2005. Assessment of C-band SRTM DEM in a dense equa-torial forest zone. Computers and Geoscience 337 (14), 1225–1234.

rasington, J., Richards, K., 1998. Interactions between model predictions, parame-ters and DTM scales for top model. Computers & Geosciences 24 (4), 299–314.

ai, X., Wang, D., 2006. Spatial autocorrelation of topographic index in catchments.Journal of Hydrology 328, 581–591.

happell, N.A., Vongtanaboon, S., Jiang, Yi., Tangtham, N., 2006. Return-flow predic-tion and buffer designation in two rainforest headwaters. Forest Ecology andManagement 224, 131–146.

i Luzio, M., Arnold, J.G., Srinivasan, R., 2005. Effect of GIS data quality on smallwatershed stream flow and sediment simulation. Hydrological Processes 19,629–650.

ragut, L., Eisank, C., 2011. Object representations at multiple scales from digitalelevation models. Geomorphology 129, 183–189.

avey, E., Geiger, A., Gudmundsson, G.H., Wehr, A., 2003. Evaluating the potentialof an airborne laser-scanning system for measuring volume changes of glaciers.Geografiska Annaler: Series A Physical Geography 81 (4), 555–561.

raser, C.S., Baltsavias, E., Gruen, A., 2002. Processing of IKONOS imagery for subme-

ter 3D positioning and building extraction. ISPRS Journal of Photogrammetry &Remote Sensing 56, 177–194.

rey, H., Paul, F., 2012. On the suitability of the SRTM DEM and ASTER GDEM forthe compilation of topographic parameters in glacier inventories. InternationalJournal of Applied Earth Observation and Geoinformation 18, 480–490.

bservation and Geoinformation 21 (2013) 205–217

Gómez-Gutiérrez, A., Schnabel, S., Contador, F.L., GarcíaMarín, R., 2011. Testing thequality of open-access DEMs and their derived attributes in Spain: SRTM, GDEMand PNOA DEM. Geomorphometry, 53–56.

Gorokhovich, Y., Voustianiouk, A., 2006. Accuracy assessment of the processedSRTM-based elevation data by CGIAR using field data from USA and Thailandand its relation to the terrain characteristics. Remote Sensing of Environment104 (4), 409–415.

Grodecki, J., Gene, D., 2003. Block adjustment of high resolution satellite imagesdescribed by rational polynomials. Photogrammetric Engineering and Remotesensing 69 (1), 59–68.

Guth, P.L., 1995. Slope and aspect calculations on gridded digital elevation models:examples from a geomorphometric toolbox for personal computers. Zeitschriftfür Geomorphologie 101, 31–52.

Guth, P.L., 2006. Geomorphometry from SRTM – comparison to NED. Photogram-metric Engineering & Remote Sensing 72 (3), 269–277.

Hirano, A., Welch, R., Lang, H., 2003. Mapping from ASTER stereo image data:DEM validation and accuracy assessment. ISPRS Journal of Photogrammetry andRemote Sensing 57 (5–6), 356–370.

Hohle, J., 2009. DEM generation using a digital large format frame camera. Pho-togrammetric Engineering & Remote Sensing 75 (1), 87–93.

Holmes, K.W., Chadwick, O.A., Kyriankidis, P.C., 2000. Error in USGS 30-meter digitalelevation model and its impact on terrain modelling. Journal of Hydrology 233,154–173.

Honghai, Qi., Altinakar, M.S., 2011. A GIS-based decision support system for inte-grated flood management under uncertainty with two dimensional numericalsimulations. Environmental Modelling & Software 26 (6), 817–821.

Iwahashi, J., Watanabeb, S., Furuyac, T., 2003. Mean slope-angle frequency distribu-tion and size frequency distribution of landslide masses in Higashikubiki area,Japan. Geomorphology 50 (4), 349–364.

Jain, S.K., Singh, V.P., 2003. Water Resources Systems Planning and Management 51,858.

Januchowski, S.R., Pressey, R.L., VanDerWal, J., Edwards, A., 2010. Characterizingerrors in digital elevation models and estimating the financial costs of accuracy.International Journal of Geographical Information Science 24 (9), 1327–1347.

Kaplan, D., Hegarty, C.J., 2006. Understanding GPS: Principles and Applications.Artech House, Boston, London, pp32.

Kervyn, F., 2001. Modelling topography with SAR interferometry: illustrationsof a favourable and less favourableenvironment. Computer & Geoscience 27,1039–1050.

Kiamehr, R., Sjöberg, L.E., 2005. Effect of the SRTM global DEM on the determinationof a high-resolution geoid model: a case study in Iran. Journal of Geodesy 79 (9),540–551.

Kornus, W., Alamús, A., Ruiz, A., Talaya, J., 2006. DEM generation from SPOT-5 3-fold along track stereoscopic imagery using autocalibration. ISPRS Journal ofPhotogrammetry and Remote Sensing 60 (3), 147–159.

Maune, D.F., Maitra, J.B., McKay, E.J., 2001. In: Maune, D. (Ed.), Accuracy Standards,Digital Elevation Model Technologies and Applications: The DEM User Manual.American Society for Photogrammetry and Remote Sensing, Bethesda, MD, pp.61–82.

Mukherjee, S., Garg, R.D., Mukherjee, S., 2011. Effect of Systematic error on DEM andits derived attributes: a case study on Dehradun area using Cartosat-1 stereodata. Indian Journal of Landscape System and Ecological Studies 34 (1), 45–58.

O’Callaghan, J.F., Mark, D.M., 1984. The extraction of drainage network from digitalelevation data. Computer Vision, Graphics, and Image Processing 28, 323–344.

Raaflaub, L.D., Collins, M.J., 2006. The effect of error in gridded digital elevationmodels on the estimation of topographic parameters. Environmental Modelling& Software 21, 710–732.

Ramlal, B., Baban, S., 2008. Developing a GIS based integrated approach to floodmanagement in Trinidad, West Indies. Journal of Environmental Management88 (4), 1131–1140.

Rodgriguez, E., Morris, C., Belz, J., 2006. A global assessment of SRTM performance.Photogrammetric Engineering and Remote Sensing 72, 249–260.

Romanowicz, R.J., Young, P.C., Beven, K.J., Pappenberger, F., 2008. A data based mech-anistic approach to nonlinear flood routing and adaptive flood level forecasting.Advances in Water Resources 31 (8), 1048–1056.

San, B.T., Suzen, M.L., 2005. Digital elevation model (DEM) generation and accuracyassessment from ASTER stereo data. International Journal of Remote Sensing 26(22), 5013–5027.

Schenk, T., 1996. Digital aerial triangulation. Archives of Photogrammetry andRemote Sensing 31 (B3), 735–745.

Shortridge, A., Messina, J., 2011. Spatial structure and landscape association of SRTMerror. Remote Sensing of Environment 115, 1576–1587.

Spark, R.N., Williams, P.F., 1996. Digital Terrain Models and the visualization ofstructural geology. In: Declan, G., De Paor (Eds.), Computer Methods in theGeosciences, 15. Pergamon, pp. 421–446.

Sun, G., Ranson, K.J., Kharuk, V.I., Kovacs, K., 2003. Validation of surface height fromshuttle radar topography mission using shuttle laser altimeter. Remote Sensingof Environment 88, 401–411.

Tarboton, D.G., 1997. A new method for the determination of flow directions andcontributing areas in grid digital elevation models. Water Resource Research 33,309–319.

Tarboton, D.G., Bras, R.L., Rodriguez-Iturbe, I., 1991. On the extraction of channelnetworks from digital elevation data. Hydrological Processes 5, 81–100.

Tarekegn, T.H., Haile, A., Rientjes, T., Reggiani, P., Alkema, D., 2010. Assessment ofan ASTER-generated DEM for 2D hydrodynamic flood modeling. InternationalJournal of Applied Earth Observation and Geoinformation 12 (6), 457–465.

arth O

T

T

V

V

V

W

representation, and hydrologic simulation. Water Resources Research 30 (4),1019–1028.

S. Mukherjee et al. / International Journal of Applied E

aud, H., Parrot, J., Alvarez, R., 1999. DEM generation by contour line dilation. Com-puters & Geosciences 25 (7), 775–783.

hornton, P.E., Running, S.W., White, M.A., 1997. Generating surface of daily mete-orological variable over large regions of complex terrain. Journal of Hydrology190 (3–4), 214–250.

an Niel, T.G., McVicar, T.R., Li, L., Gallant, J.C., Yang, Q., 2008. The impact of misreg-istration on SRTM and DEM image differences. Remote Sensing of Environment112 (5), 2430–2442.

assilopouloua, S., Hurnia, L., Dietrichb, V., Baltsaviasc, E., Paterakic, M., Lagiosd, E.,Parcharidis, L., 2002. Orthophoto generation using IKONOS imagery and highresolution DEM: a case study on volcanic hazard monitoring of Nisyros Island(Greece). ISPRS Journal of Photogrammetry & Remote Sensing 57, 24–38.

aze, J., Teng, J., Spencer, G., 2010. Impact of DEM accuracy and resolution on topo-graphic indices. Environmental Modelling and Software 25 (10), 1086–1098.

eibel, R., Heller, M., 1990. A framework for digital terrain modeling. In: ForthInternational Symposium on Spatial Data Handling, Zurich, Switzerland, pp.219–229.

bservation and Geoinformation 21 (2013) 205–217 217

Wilson, J.P., Gallant, J.C., 2000. Secondary Topographic Attributes. In: Wilson, J.P.,Gallant, J.C. (Eds.), Terrain Analysis: Principles and Applications. John Wiley andSons, New York, pp. 87–131.

Wolock, D.M., Price, C.V., 1994. Effects of digital elevation model map scale and dataresolution on a topography-based watershed model. Water Resources Research30 (11), 3041–3052.

Wu, S., Li, J., Huang, G.H., 2008. A study on DEM-derived primary topographicattributes for hydrological applications: sensitivity to elevation data resolution.Applied Geography 28, 210–223.

Zhang, W., Montgomery, D.R., 1994. Digital elevation model grid size, landscape

Zhou, H., Zhang, J., Gong, L., Shang, X., 2012. Comparison and validation of dif-ferent DEM data derived from InSAR. Procedia Environmental Sciences 12,590–597.

本文献由“学霸图书馆-文献云下载”收集自网络，仅供学习交流使用。

学霸图书馆（www.xuebalib.com）是一个“整合众多图书馆数据库资源，

提供一站式文献检索和下载服务”的24 小时在线不限IP

图书馆。

图书馆致力于便利、促进学习与科研，提供最强文献下载服务。

图书馆导航：

图书馆首页 文献云下载 图书馆入口 外文数据库大全 疑难文献辅助工具