The Egyptian Journal of Remote Sensing and Space Sciences (2013) 16, 133–139

National Authority for Remote Sensing and Space Sciences

The Egyptian Journal of Remote Sensing and Space
Sciences
www.elsevier.com/locate/ejrs
www.sciencedirect.com

REVIEW ARTICLE

A comparative analysis of different DEM

interpolation methods
P.V. Arun *

NIT, M.P., India

Received 28 June 2013; revised 1 September 2013; accepted 4 September 2013

Available online 8 October 2013

KEYWORDS Abstract Visualization of geospatial entities generally entails Digital Elevation Models (DEMs)
DEM; that are interpolated to establish three dimensional co-ordinates for the entire terrain. The accuracy
Interpolation methods; of generated terrain model depends on the interpolation mechanism adopted and hence it is needed
Kriging; to investigate the comparative performance of different approaches in this context. General inter-
IDW polation techniques namely Inverse Distance Weighted, kriging, ANUDEM, Nearest Neighbor,
and Spline approaches have been compared. Differential ground field survey has been conducted
to generate reference DEM as well as specific set of test points for comparative evaluation. We have
also investigated the suitability of Shuttle Radar Topographic Mapper Digital Elevation Mapper
for Indian terrain by comparing it with the Survey of India (SOI) Digital Elevation Model
(DEM). Contours were generated at different intervals for comparative analysis and found SRTM
as more suitable. The terrain sensitivity of various methods has also been analyzed with reference to
the study area.
Ó 2013 Production and hosting by Elsevier B.V. on behalf of National Authority for Remote Sensing and
Space Sciences.

Contents

1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 134
2. Data resources . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135
3. Methodology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 136
3.1. Comparative analysis of interpolation methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 136
3.2. Comparative analysis of SRTM and SOI DEM. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137

* Tel.: +91 4828249999.

E-mail address: [email protected].
Peer review under responsibility of National Authority for Remote
Sensing and Space Sciences.

Production and hosting by Elsevier

1110-9823 Ó 2013 Production and hosting by Elsevier B.V. on behalf of National Authority for Remote Sensing and Space Sciences.
http://dx.doi.org/10.1016/j.ejrs.2013.09.001
134 P.V. Arun

4. Results and discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 138

4.1. Comparative analysis of interpolation methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 138
4.2. Comparative analysis of SRTM and SOI DEM. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 138
5. Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 138
References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 138

1. Introduction

Remote sensing techniques are being effectively used as a tool Inverse Distance Weighting (IDW), local polynomial, Nearest
for decision making in various fields because of their spatial Neighbor (NN), and Radial Basis Functions (RBFs). On the
analysis and display capabilities. The utility of decision making other hand, global interpolation methods such as polynomial
processes is significantly improved using 3D geographical interpolation functions use all the available sample points to
models as they facilitate effective visualization. Digital Eleva- generate predictions for a particular point. These methods
tion Models (DEMs) are the generally adopted data structures facilitate to evaluate and remove global variations caused by
for storing topographic information and are usually interpo- physical trends in the data (Burrough and McDonnell, 1998).
lated to establish the values for entire terrain points. DEM is Kriging is a geo statistical interpolation method that utilizes
an array representation of squared cells (pixels) with an eleva- variogram which depends on the spatial distribution of data
tion value associated to each pixel (Manuel, 2004). DEMs can rather than on actual values. Kriging weights are derived using
be obtained from contour lines, topographic maps, field sur- a data-driven weighting function to reduce the bias toward in-
veys, photogrammetry techniques, radar interferometry, and put values, and it provides the best interpolation when good
laser altimetry (Manuel, 2004). Different interpolation meth- variogram models are available. The IDW approach is a local
ods applied over the same data sources may result in different deterministic interpolation technique that calculates the value
results and hence it is required to evaluate the comparative as a distance-weighted average of sampled points in a defined
suitability of these techniques. neighborhood (Burrough and McDonnell, 1998). It considers
Interpolation techniques are based on the principles of spa- that points closer to the query location will have more influ-
tial autocorrelation, which assumes that closer points are more ence, and weights the sample points with inverse of their dis-
similar compared to farther ones. The literature reveals a great tance from the required point.
deal of interpolation methods which are generally classified as Nearest Neighbor interpolation finds the closest subset of
local and global approaches. Local methods predict the value input samples to a query point and applies weights to them
of an unknown point based on the values of neighborhood based on proportionate areas (Sibson, 1981). It is a local deter-
pixels. Prominent local methods found in the literature include ministic method and interpolated heights are guaranteed to be

Table 1 Data resource description.

S. no. Image used Resolution (m) Satellite Area Date of procurement
1 PAN 2.5 IRS-P5 (Cartosat-1) Bhopal November 2012
2 LISS-IV 5.8 IRS P6 Bhopal September 2012
3 Google Earth 0.15 (Highest) MANIT –
4 SOI DEM As per 1:50,000 scale topo sheet – Bhopal November 2012
5 SRTM DEM 3-ARC Shuttle Radar Bhopal August 2012

Arc map/Arc catalog

GNSS solution

DGPS survey and

Import the GCPs to ESRI 20 control points taken out as check points to
preprocessing of GCPs
shape using Arc GIS analyze the accuracy of DEM
software.

Magellan Promark-3 Analyze, Compare and DEM preperation using IDW, Kriging,
single frequency Display the Results Nearest Neighbor, spline and ANUDEM.
DGPS system.

Raster 3D analyst/Spatial
Interpolation Analyst

Figure 1 Methodology for comparative analysis of interpolation methods.

A comparative analysis of different DEM interpolation methods 135

Table 2 Ellipsoidal heights at test GCPs from interpolated and DGPS observed values.
Control points ID Ellipsoidal heights in meter
IDW Value ANUDEM Nearest neighbor Spline Kriging DGPS observed value
FID-24 474.97 474.89 476.93 476.33 476.02 476.72
FID-81 476.64 476.78 476.72 477.54 476.68 478.54
FID-39 479.42 478.63 479.22 479.37 478.90 477.60
FID-7 476.78 477.30 475.59 475.66 475.83 478.40
FID-11 477.75 477.30 477.78 477.30 477.29 478.24
FID-14 478.06 477.48 478.59 480.76 479.63 479.58
FID-17 479.27 477.85 479.32 480.36 479.44 479.37
FID-71 477.38 476.26 477.18 477.65 477.64 476.68
FID-64 477.39 478.41 478.05 478.18 477.93 477.28
FID-61 479.30 480.06 479.59 477.79 479.03 479.69
FID-56 477.72 478.11 478.40 477.96 478.20 475.72
FID-45 477.93 478.93 477.57 477.36 477.43 477.97
FID-39 479.450 478.32 479.22 479.123 478.75 477.60
FID-91 473.11 471.95 473.94 474.50 474.18 475.83
FID-87 473.02 471.95 473.56 473.25 473.39 476.42
FID-30 474.32 474.83 473.67 471.40 472.58 473.19
FID-28 473.80 475.57 473.41 473.32 473.47 471.82
FID-89 473.14 471.95 472.31 472.33 472.22 469.89
FID-95 471.08 471.95 471.27 471.15 471.11 472.14
FID-34 477.07 478.01 477.90 477.14 477.43 477.82

Table 3 RMSE values with reference to terrain variation.

Type of test GCPs used RMSE values
IDW ANUDEM NN Spline Kriging
Mild slope areas 0.93 0.87 0.72 0.91 0.70
Steep slope area 1.45 1.82 1.34 1.37 1.31
Combined slope area 1.73 2.02 1.53 1.62 1.49

within the range of the samples used. It does not produce neighborhood approaches such as IDW or RBFs were found
peaks, pits, ridges or valleys that are not already present in to be as accurate as kriging or even better. The ANUDEM
the input samples and adapts locally to the structure of the in- interpolation method is specifically designed for the creation
put data. It does not require input from the user and works of hydrologically correct terrain surfaces.
equally well for regularly as well as irregularly distributed data In this paper, we evaluate the comparative suitability of dif-
(Watson, 1992). The Spline interpolation approach uses math- ferent interpolation techniques based on their accuracy and
ematical function to minimize the surface curvature and pro- sensitivity to terrain variations. Performance of different inter-
duces a smooth surface that exactly fits the input points. The polation methods namely IDW, ordinary kriging (KRG),
ANUDEM method uses an interpolation technique specifically ANUDEM, NN and spline has been evaluated with reference
designed to create a surface that more closely represents a nat- to the study area. Generally available DEMs for Indian terrain
ural drainage surface and preserves both ridgelines as well as namely Shuttle Radar Topographic Mapper (SRTM) and Sur-
stream networks (Hutchinson, 1989). vey of India (SOI) Digital Elevation Models (DEMs) are also
Zimmerman et al. (1999) showed that kriging yielded better evaluated based on the contours generated at different
estimations of altitude than inverse distance weighting (IDW) intervals.
irrespective of the landform type and sampling pattern. This
result is attributed to the ability of kriging to adjust itself to 2. Data resources
the spatial structure of the data. However, in other studies
(Weber and Englund, 1992; Gallichand and Marcotte, 1993; Investigations have been conducted over MANIT campus and
Brus et al., 1996; Declercq, 1996; Aguilar et al., 2005), surrounding areas of Bhopal city in India; variation of the ter-

Table 4 Comparison of contours.

Type DEM used Number of contours generated
1 m Interval 2 m Interval 5 m Interval 10 m Interval
SOI 3303 1637 663 342
SRTM 12,182 11,212 4274 2153
136 P.V. Arun

rain, spread over more than 1000 acres made it optimal for the test points in WGS-84 datum. Collected raw data have been
analyses. Satellite images of Bhopal along with SOI and pre-processed using GNSS software (Trimble Survey Division,
SRTM DEMs have been used for comparative analysis of var- 2013) to remove various errors and to calibrate the readings at
ious methodologies. Details of the satellite data used for these centimeter level accuracy. The processed data (GCPs) have
investigations are summarized in Table 1. The ground truthing been imported in the ArcGIS environment (ArcGIS, 2013)
information has been collected using Differential Global Posi- and plotted to a shape file. About 680 GCPs were used as sam-
tioning System (DGPS) survey conducted over Bhopal during ple points to generate the DEM and rest were used as test
October 2012. points to estimate accuracy of interpolation. Raster surface
has been generated from reference DEM using different inter-
polation methods namely IDW, Kriging, NN, ANUDEM and
3. Methodology Spline. Accuracies of generated surfaces have been evaluated
using 320 reference GCPs as test points. Visual analyses as well
3.1. Comparative analysis of interpolation methods as statistical parameters have been adopted for comparative
evaluation of the interpolated surfaces. In the visual analysis,
Commonly used interpolation approaches have been evaluated DEM generated heights were verified in the ground by field
with reference to the study area and adopted methodology is visit using GPS. Mathematical analysis has been done by cal-
summarized in Fig. 1. Differential Global Positioning System culating the deviations of interpolated height values from cor-
(DGPS) survey has been conducted over the study area to col- responding observed values in terms of root mean square error
lect three-dimensional coordinates of around 1000 samples and (RMSE).

(a) IDW (b) ANUDEM

(c) Nearest Neighbor (d) Spline

(e) Kriging (f) Index

Figure 2 Slope maps generated using different interpolation methods.
A comparative analysis of different DEM interpolation methods 137

3.2. Comparative analysis of SRTM and SOI DEM corresponding contours using the Kriging interpolation tech-
nique in the ArcGIS environment. Contours with interval
Comparative suitability of SRTM and SOI DEMs has been 10 m, 5 m, 2 m and 1 m were generated from SRTM as well
analyzed with reference to the generation of contours. Con- as SOI DEM using ArcGIS 3D analyst extension. Compara-
tours of the study areas have been digitized from SOI Topo tive analysis has been done with reference to the nature and
sheet No. 55E7 and 55E8 and contour heights were recorded number of contours generated from DEMs. Further, visual
in the attribute table. SOI DEM has been generated from analysis has been conducted based on the 3D view generated

(a) SOI-55E8 DEM draped with exaggeration 10 ( b) SRTM DEM draped with exaggeration 10

( c) SOI-55E8 DEM draped with exaggeration 15 (d) SRTM DEM draped with exaggeration 15

( e) SOI-55E8 DEM draped with exaggeration 20. ( f ) SRTM DEM draped with exaggeration 20.

Figure 3 3D surfaces generated from SOI & SRTM DEM at different exaggeration levels.
138 P.V. Arun

from the two DEMs. Satellite images were draped over the been generated using 3D analyst extension of Arc GIS soft-
DEMs using Virtual GIS viewer in ERDAS and were analyzed ware and outcomes of these investigations are tabulated in
at different exaggeration levels. Table 4.
From the table, it is evident that the contours generated
4. Results and discussion from SOI DEM are sparse while that from SRTM are compar-
atively denser. Therefore we can conclude that SOI DEM is
4.1. Comparative analysis of interpolation methods having very poor data quality compared to SRTM.
The suitability of DEMs has also been evaluated based on
the comparative visualization of 3D models generated
We have investigated the comparative performance of different
from these DEMs at different exaggeration levels as given in
interpolation techniques with reference to various terrain con-
Fig. 3.
texts. Visual comparisons as well as mathematical analyses
Visual comparison also reveals that SRTM is performing
have been conducted. Visual comparison of slope map gener-
better than the SOI DEMs. Reason behind the poor perfor-
ated using different interpolation techniques is presented in
mance of the SOI DEM may be attributed to its construction
Fig. 2.
from 1:50,000 scale topographic maps. Open source SRTM
DGPS survey data revealed that the kriging approach per-
data are giving more reliability and accuracy than the SOI
formed accurately in average cases when compared to others.
DEM due to the usage of radar technology.
Interpolated heights at different test points (points having
coordinates from DGPS survey) have been also compared
for the five different methods and results are summarized in 5. Conclusion
Table 2.
Table 2 reveals that different approaches produce varied re- The generated DEMs are found to be sensitive to height inter-
sults over the same points. Interpolated height values for dif- polation methods as well as the terrain nature. Investigations
ferent methods at each test point have been plotted. revealed that the Kriging method performs better when com-
Deviations of interpolated height values from the actual values pared to other contemporary methods in most contexts.
(DGPS observed) at each test point give a better understand- DEM generated from the DGPS data was found to be better
ing about the performance of each method and reveals a better than the DEM available from SOI or SRTM data. Number
performance of the kriging approach. of contours extracted from SRTM DEM was found to be bet-
In order to investigate the sensitivity of interpolation meth- ter than that from SOI DEM, which may be attributed to the
ods to the nature of terrain, the test GCPs were divided into better accuracy of SRTM data source. Kriging has been found
two zones namely mild slope and steep slope areas. Average to adapt itself to terrain variations while ANUDEM is found
RMSE values of the test points have been also calculated with preferable for streams and ridge lines.
reference to terrain variations and are summarized in Table 3.
IDW and Kriging have been found to adjust themselves to the References
terrain variations when compared to other methods. ANU-
DEM has been found to yield a better performance for ridges Aguilar, F.J., Agüera, F., Aguilar, M.A., Carvajal, F., 2005. Effects
as well as stream areas. of terrain morphology, sampling density, and interpolation meth-
The investigations have shown that interpolation results ods on grid DEM accuracy. Photogrammetr. Eng. Remote Sens.
vary with variation in a spatial structure and terrain nature 71, 805–816.
of input data. As far as our data are concerned, we have more ArcGIS, 2013. Online data available at: http://www.arcgis.com/
home/ (accessed 24.06.13).
samples at slope areas than at plane areas. Kriging and NN
Brus, D.J., Gruijter, J.J., Marsman, B.A., Visschers, R., Bregt, A.K.,
were found to perform well in these contexts and can be Breeuwsma, A., 1996. The performance of spatial interpolation
adopted for geomorphologically smooth and small areas. In methods and choropleth maps to estimate properties at points: a
stream and ridge line areas, the ANUDEM method has shown soil survey case study. Environmetrics 7, 1–16.
lowest RMSE value. The NN method has shown nearly opti- Burrough, P.A., McDonnell, R.A., 1998. Principles of Geographical
mal values over smooth surfaces, i.e. second lowest. This trend Information Systems. Oxford University Press, New York, pp.
in RMSE values of Kriging has continued even for steep slope 333–335.
areas as well as for areas covering both steep and mild slopes. Declercq, F.A.N., 1996. Interpolation methods for scattered sample
IDW and NN methods have been found to be good for inter- data: accuracy, spatial patterns, processing time. Cartogr. Geogr.
polation of geo-morphologically smooth areas. Kriging meth- Inform. Syst. 23, 128–144.
Gallichand, J., Marcotte, D., 1993. Mapping clay content for
ods take into consideration autocorrelation structures of
subsurface drainage in the Nile delta. Geoderma 58, 165–179.
elevations in order to define optimal weights. The method re- Hutchinson, M.F. 1989. Calculation of hydrologically sound
quires a skilled user with geo-statistical knowledge. Spline- digital elevation models, Third International Symposium on
based methods fit a minimum-curvature surface through the Spatial Data Handling at Sydney, Australia, vol. 3, no. 1, pp.
input points, and ensure preservation of trend in the sample 120–127.
data along with rapid changes in gradient or slope. Manuel, P., 2004. Influence of DEM interpolation methods in
drainage analysis. GIS Hydro 04. Texas, USA.
4.2. Comparative analysis of SRTM and SOI DEM Sibson, R., 1981. A Brief Description of Nearest Neighbor Interpo-
lation. Interpolating Multivariate Data. John Wiley & Sons, New
York, pp. 21–36, Chapter 2.
We have investigated the accuracy of DEMs namely SRTM Trimble survey division. Online data available at: https://www.trim-
and SOI with reference to contour extraction. Contours have ble.com/survey/, (accessed 24.06.13).
A comparative analysis of different DEM interpolation methods 139

Watson, D., 1992. Contouring: A Guide to the Analysis and Display of Zimmerman, D., Pavlik, C., Ruggles, A., Armstrong, M., 1999. An
Spatial Data. Pergamon Press, London, pp. 120–123. experimental comparison of ordinary and universal kriging and
Weber, D., Englund, E., 1992. Evaluation and comparison of spatial inverse distance weighting. Math. Geol. 31, 375–390.
interpolators. Math. Geol. 24, 381–391.