96
DEM extraction from ALOS-PRISM data in the region of Sahel-Oualidia (Moroccan Atlantic coast) HABIB A.(1), LABBASSI K.(2), MENENTI M.(2) & KHOSHELHAM K.(2) (1)
(2)
Department of Earth Sciences, Faculty of Sciences El Jadida, Chouaïb Doukkali University, B.P. 24000 El Jadida, Morocco, E-mail:
[email protected],
[email protected] Delft Institute of Earth Observation and Space Systems, Delft University of Technology, Kluyverweg 1, 2629 HS Delft, The Netherlands, E-mail:
[email protected],
[email protected]
Abstract. With the launch of the Advanced Land Observing Satellite (ALOS) carrying the PRISM sensor (Panchromatic Remote Sensing Instrument for Stereo Mapping), a large scale solution for achieving a high resolution DEM (Digital Elevation Model) has been found. This work presents the results of extracting a high resolution DEM of 5m resolution, from two ALOS-PRISM scenes available at the Level 1B and covering the northern region of Sahel-Oualidia (Moroccan Atlantic coast), with a RMSE of ±2.5m.ALOSPRISM level 1B1 data are delivered with ephemeris data, so that the most suitable model is Orbital Pushbroom, which fully utilizes orbital ephemeris information. The quality of the resulting DEM is evaluated by a DEM developed by digitizing the contour lines from a topographic map at scale 1:50000. The mean difference between the two DEM is calculated, and it is about 10m. The accuracy of the DEM obtained remains the best in our study area and is sufficient for the production of orthoimages or for digital terrain analysis for Hydrogeology. Keywords: ALOS, PRISM, DEM, Sahel-Oualidia. 1. Introduction DEM is one basic material of the geographic information systems (GIS) and is a good source to produce information of land physical parameters, which are useful for supporting many kinds of activities such as disaster and water resources management. There are several methods to generate a DEM. One of them is parallax calculation from stereoscopic data of optical satellite sensor, as for the case of our study, the PRISM sensor which has the capacity to produce stereoscopic data. PRISM can observe the surface of the earth with a high spatial resolution (2.5m resolution) and it can acquire three images simultaneously in the along-track direction of the satellite using three independent optical systems for forward, nadir and backward views. The forward and backward telescopes are inclined by ±23.8° from nadir to realize a base to height ration of 1 and 0.5 for nadir and backward at an orbital altitude of 992km (Jaxa, 2006). In this paper, we present the results of extracting a high resolution DEM of 5m resolution from ALOS PRISM data. 2. Materials and methods b. Datasets 1. ALOS PRISM The ALOS PRISM data used are selected as the bases images in DEM generation and consist of two scenes (Nadir & Forward), the Nadir scene is centred at 8°42'50''E longitude and 32°49'30''N latitude, the Forward the scene is centred at 8°42'3''E longitude and 32°48'250''N latitude. The maximum altitude is 160m. Datasets are acquired on 2/07/2008 with good quality and there is no cloud cover. Datasets are delivered with CEOS format and they have the level 1B1.
Fig. 1 The location of the study area (red box: ALOS PRISM data used)
Colloque International des Utilisateurs de SIG, Taza GIS-Days, 23-24 Mai 2012 Recueil de Proceeding
97
2. Topographic maps Topographic maps at the scale of 1:50000 are used in this work as a basis for GCPs (Ground Control Points), and are as follows: Oualidia-Sidi Moussa, Sidi Smaïl, Zmamra and Tnine Gharbia. c. Sensor model ALOS PRISM level 1A and 1B data are delivered with ephemeris data. The most suitable model is orbital pushbroom, which fully utilizes orbital ephemeris information. Three GCPs are needed at least to correct the major error of this ephemeris data. d. Reference system Considering the available GCPs, it’s decided to use the Lambert Conformal Conic of Morocco as a reference system. e. Ground Control Points (GCPs) Good CGPs for DEM generation should satisfy two conditions. Firstly, the horizontal and vertical accuracy of the GCPs should be high enough; secondly, GCPs should be recognizable and locatable on the stereo image pairs. The available GCPs are read from topographic maps at scale 1:50000. f. Generate DEM The steps of the DEM generation are illustrated in the following Flowchart:
ALOS PRISM data (Nadir & Forward)
Setting & pyramid layer
- Sensor model - Define the reference system
Insert GCPs (XYZ) Insert Check points and Tie points
Triangulation process
Error report
DEM generation
DEM Fig. 2 Flowchart used for the DEM generation process 3. Results and discussion Level 1B1 data used the orbital pushbroom model for DEM generation. The ephemeris data included in the level 1B1 are used to build the model, so few GCPs could be sufficient to help in improving the sensor model. Therefore, these points in the image are located approximately, which can be translated by the large residuals in the report of triangulation provided by the software, which indicates the difference between the calculated and measured values of the coordinates in the image. Thus, the residual value is so great between ground coordinates calculated and original coordinates of the point. In the report of triangulation, ground coordinates and image coordinates of the GCPs or check points are evaluated as a residue. Residues checkpoints are considered an objective assessment of the overall accuracy DEM. more the residue is minimal, more the sensor model is built well. In the tables below, we have the results of the triangulation of GCPs and Check Points, so you must be careful when looking at the statistics of Check Points in Table 2. The performance of Check Point is not as good as the GCPs. We note that the residue for Check Point is big for the XY, on the other side; it is more or less acceptable for Z. Number of GCPs 6 Number of GCPs 6
Ground X
Ground Y Ground Z Meter 5.5978780 2.3085012 1.7744694 Table1. GCPs triangulation report Number of Ground X Ground Y Ground Z check points Meter 5 163.0807037 262.5836182 6.2385364 Table2. Check points triangulation report
Image X
Image Y Pixel
2.1126139
1.0859189
Image X
Image Y Pixel 34.8210602 116.8509750
International Conference of GIS-Users, Taza GIS-Days, May 23-24, 2012 Proceeding Book
98
After analysing the triangulation report, we launch the DEM generation in raster format, with an accuracy of 5m (Fig.3).
Fig. 3 ALOS PRISM DEM generated In the table below, showing the residues in the report of the extraction of DEM, we see that the total RMSE for all points (GCPs, Check Points and Tie Points) is about 2.5 meters.
GCPs Check Points Tie Points Global
Number of points 6 5 27 38
Min Error -6.4419, -6.6481 -2.0227 -6.6481
Max Error 3.3629 -1.0970 2.3423 3.3629
Mean Error -2.1222 -4.7019 -0.0509 -0.9899
RMSE 3.7492 5.1186 1.1517 2.5708
The PRISM DEM (P-DEM) is further evaluated using a DEM developed from a topographic map at scale 1:50000 (T-DEM), by digitizing the contour lines. The first check is to see if the relative position of the terrain and the elevation values match (Fig.4). As shown in Figure 4, the P-DEM and the T-DEM are almost comparable and the elevations are in the same color bar ranges. A difference map is made between P-DEM and T-DEM. Most of areas are colored in yellow and sky blue. The extreme outliers are obvious where there are red and dark blue with extreme values located in areas with steep slopes.
Fig. 4 Comparison between the two DEM Colloque International des Utilisateurs de SIG, Taza GIS-Days, 23-24 Mai 2012 Recueil de Proceeding
99
4. Conclusions In this study, a DEM of 5m resolution, covering the northern region of Sahel-Oualidia (Moroccan Atlantic coast), was generated from a stereo pair ALOS-PRISM with a RMSE of ± 2.5 m. The accuracy of the DEM obtained remains the best in our study area and is sufficient for the production of orthoimages or for digital terrain analysis for Hydrogeology, but precision of the generated DEM needs to be improved by correcting or changing a few control points. The use of points from surveys using DGPS can correct these errors. Acknowledgements The work described in this publication has been carried out as a part of TIGER Initiative supported by the European Space Agency (ESA). References [1] H. Yu, Monitoring glacier elevation changes over the tibetan plateau using ALOS PRISM and ICESat. MSc Geomatics, Deldt Universitu of Technology, The Netherlands (2010). [2] L. Renouard, “Extraction automatique de MNT à différents resolutions”, ISPRS Commission IV, USA, 1992, pages (pp. 886-893). [3] B. Trisakti, I. Carolita and A.A. Pradana, “Digital elevation model from PRISM-ALOS and ASTER stereoscopic data”, International Journal of Remote Sensing and Earth Sciences, 2009, pages (pp.29-38). [4] Y. Osawa, “Optical and microwave sensors on Japanese mapping satellite – ALOS”, ISPRS, 2004, pages (pp.309-312). [5] T. Igarashi, “ALOS mission requirement and sensor specifications”, Adv. Space Res. Vol, 28, No. 1, 2001, pages (pp.127131).
International Conference of GIS-Users, Taza GIS-Days, May 23-24, 2012 Proceeding Book