Remote Sensing and Geoinformation Lena Halounová, Editor

not only for Scientific Cooperation EARSeL, 2011

587

Comparison of the Accuracies of DTM‟s Obtained from

SRTM, Topographic Maps and ALOS/PRISM

Cihan Uysal, Derya Maktav

Istanbul Technical Univ., Geomatics Engineering Dept., Istanbul, Turkey;

uysalc@yahoo.com, maktavd@itu.edu.tr

Abstract. Recent improvements in satellite technologies and information systems have

caused the frequently usage of remote sensing technology and its integration with

geographic information systems (GIS). Especially, due to the increases of the spatial

resolution of satellite sensors, remote sensing technology has been also used for

archaeological applications. To analyse this type of archaeological applications 3D models

are needed. These models are an important tool to demonstrate the terrain topography. In

this study, digital terrain models (DTM) obtained from 1:25000 topographic maps, Shuttle

Radar Topography Mission (SRTM) and Advanced Land Observing Satellite (ALOS)/ (The

Panchromatic Remote Sensing Instrument for Stereo Mapping (PRISM) stereo satellite

images were compared. The two DTMs from 1:25000 topographic maps and stereo

PRISM images show approximately same accuracy values, whereas the DTM generated

from SRTM has lower accuracy. Kurşunlugerme aqueduct in Istanbul,Turkey, which has

been an important part of the Roman and Byzantine water supply system, and its

environment have been chosen as the study area.

Keywords. Remote Sensing, SRTM, ALOS/PRISM, DTM.

1. Introduction

Turkey has a remarkable strategic and geographical location that has been influenced by many

empires in the past. Therefore, Turkey has a rich cultural inheritance. Today, in Istanbul, there

are many monuments remained from Roman, Byzantine and Ottoman Empire. Aqueducts,

main monuments of water supply systems, which carry water to towns can be shown as an

example of that kind of monuments. A system that provided water for Istanbul in ancient

time, lies 250-300 km west of the city [1]. Unfortunately, most of the aqueducts are damaged

or destroyed. Moreover, difficulties were encountered while revealing these aqueducts in

detail because of the difficult terrain conditions and the very wide area. There were

uncertainty about the route of the aqueducts in some locations. Classical terrestrial and

archaeological methods were used for this study. In addition, advanced satellite imageries

having high spatial resolution encourage users to manage archaeological investigations by

using remote sensing technology. Moreover, satellite remote sensing with synoptic view

ability provides working easily in large areas and gives satisfactory results.

The Kurşunlugerme aqueduct and its vicinity were chosen as the study area. First, the

route of the aqueduct was determined by intensive GPS measurements. Satellite images were

also used during determination and analyses. Correct analyses could be made by using 3D

models which were obtained by draping satellite images over DTMs. To compare the

accuracies of the DTMs, they were obtained using elevation data from 1/25000 topographic

maps, SRTM data and stereo PRISM data. Thus, GIS was established by using 3D models

with DTMs, satellite images and GPS measurements.

Uysal, C. and Maktav, D.:Comparison of the accuracies of DTM‟s obtained from SRTM, topographic maps and

ALOS/PRISM

588

2. Study area

Kurşunlugerme aqueduct and its vicinity were chosen as the study area. This area is located

near the Gümüşpınar village in Çatalca district of Istanbul (Fig. 1). It covers an area of 15

km². This aqueduct is one of the magnificent aqueducts of the late Roman and Byzantine

water supply system of Istanbul.

Figure 1: Study area.

3. Data and method

In this study, IKONOS pansharpened image (2005) and orthophotos (2005) were used for

visualation, location and draping over process. To generate DTM, three types of images were

considered: SRTM (3 arcsecond), 1/25000 topographic maps and stereo PRISM image. All

these data have WGS84 datum and UTM 35N zone projection. In this study, Trimble GeoXT

handheld GPS was used which has a horizontal accuracy of <1 meter [4]. ArcGIS 9.2

(ArcView), ERDAS 9.1 and PCI Geomatica 10.1 (OrthoEngine module) software were used.

3.1 DTM generation from topographic maps

1/25000 scaled topographic maps were digitized by using 10 m contour line intervals (Fig. 2).

First the Triangle irregular Network (TIN) (Fig. 3) was generated from the contour line curves

and altitude points using the 3D Analyst module. All points in this data structure are generally

interconnected via Delaunay triangulation. The TIN model is then converted into raster format

Uysal, C. and Maktav, D.:Comparison of the Accuracies of DTM‟s Obtained from SRTM, Topographic Maps

and ALOS/PRISM

589

and DTM is generated accordingly (Fig. 4). The grid size of this conversion was 10 m. Also,

DTM of study area and overlapping of IKONOS image on this DTM are shown Fig. 5.

Figure 2: Contour line. Figure 3: Triangle Irregular Network (TIN).

Figure 4: DTM generation from topographic maps.

Uysal, C. and Maktav, D.:Comparison of the accuracies of DTM‟s obtained from SRTM, topographic maps and

ALOS/PRISM

590

Figure 5: IKONOS image draped over DTM generated from topographic maps.

3.2 DTM generation from SRTM

The DEMs based on the US C-band are available free of charge on the internet with a spacing

of 3 arcsec, corresponding to approximately 92 m at the equator [2]. Also, DTM of study area

and overlapping of IKONOS image on this DTM are shown below (Figs. 6 and 7).

Figure 6: DTM generation from SRTM.

Uysal, C. and Maktav, D.:Comparison of the Accuracies of DTM‟s Obtained from SRTM, Topographic Maps

and ALOS/PRISM

591

Figure 7: IKONOS image draped over DTM generated from SRTM.

3.3 DTM generation from stereo images

In this study, to generate DTM, stereo PRISM images (georeferenced, nearest neighbour,

UTM projection, level 1B2) were used (Fig. 8). Backward and nadir images were selected to

generate DTM. Their swath area consists of 35 km²[5].

Figure 8: ALOS PRISM image.

Firstly, a new project was created with using „OrthoEngine‟ module (Fig. 9). Then

„Project information‟ was selected. The next step was to define the projection. It was entered

in the appropriate projection information for the PRISM dataset. It was recommended that we

use „Set GCP Projection based on Output Projection‟ option for the GCP projection [3].

Uysal, C. and Maktav, D.:Comparison of the accuracies of DTM‟s obtained from SRTM, topographic maps and

ALOS/PRISM

592

Figure 9: Project information.

Secondly, PRISM imagery was imported into project and the imagery has been formatted

as PIX. In the processing step, in Orthoengine menu, ‟GCP/TP Collection‟ and „Open a new

or existing image‟ submenus were selected. Once both images were opened, one of them was

selected as „reference‟, and the other as „working‟. The next step was the collection of stereo

GCPs (Fig. 10). Here, 82 points were selected as GCPs but only 32 points were used and the

other points were eliminated. GCPs were selected from IKONOS images, DTM generated

from 1/25000 topographic maps and GPS measurements in the fieldwork.

After that, all GCPs have been collected, switched to „Model Calculations‟ in the

processing step of Orthoengine and the „Compute Model‟ button clicked. After model

calculation, RMS value was found as 0.6. Then the epipolar pairs have been generated,

clicking the „Extract DEM automatically‟ button in Orthoengine. After clicking „Extract

DEM‟, the result was acquired successfully (Fig. 11). Also, DTM of the study area and the

IKONOS image draped over this DTM are shown Figs. 12 and 13.

Figure 10: Selection of GCP‟s.

Uysal, C. and Maktav, D.:Comparison of the Accuracies of DTM‟s Obtained from SRTM, Topographic Maps

and ALOS/PRISM

593

Figure 11: DTM generated from PRISM image.

Figure 12: DTM generation from PRISM image.

Uysal, C. and Maktav, D.:Comparison of the accuracies of DTM‟s obtained from SRTM, topographic maps and

ALOS/PRISM

594

Figure 13: IKONOS image draped over DTM generated from PRISM image.

4. Results

Improvements in satellite technologies and information systems have caused the frequently

usage of integration of remote sensing and GIS in archaeological applications. Istanbul has a

remarkable strategic and geographical location that has been influenced by many empires in

the past. Therefore, it has a rich cultural inheritance. In this study, the Kurşunlugerme

aqueduct and its environment have been selected as the study area. Three types of DTM were

generated. Comparison of the two DTM‟s (from 1/25000 topographic maps and stereo PRISM

images), showed approximately the same accuracies. The other DTM generated from SRTM

has a lower accuracy (6-7 meters). If GCPs used in DTM generation from stereo PRISM

images had have a better accuracy, the result would be more precise and reliable. The 3D

model obtained by draping satellite images over regional DTMs provides a better

visualization of the terrain [6].

Acknowledgement

We would like to thank TUBITAK (The Scientific and Technological Research Council of

Turkey) and the British Academy for their valuable support to this study.

References

[1] Çeçen, K. (1996), The Longest Roman Water Supply Line, First Edition, Türkiye Sınai Kalkınma Bankası,

Aksoy Matbaası, İstanbul.

[2] NASA, (2007), “SRTM”, http://www.jpl.nasa.gov/srtm.

[3] PCI Geomatics, (2007), “DEM Extraction of PRISM Stereo Imagery Tutorial”.

[4] TRIMBLE, (2005), “Getting Started Guide, GeoExplorer 2005 Series”,

http://www.trimble.com/geoxt.shtml.

[5] ALOS, (2008), “ALOS PRISM”, http://www.eorc.jaxa.jp/ALOS/.

[6] Uysal, C. (2008), “Integration of Remote Sensing and Geographic Information Systems in Archaeological

Applications”, Master Thesis, Geomatics Engineering, İstanbul.