Application of GER-II data For Geomorphological Analysis

Application of GER-II data For Geomorphological Analysis

Cohen A
Nature Reserve Authority, Mizpe Ramon, Israel

Amarsaikhan D
Infformatics Centre, Mongolian Academy of Sciences. Ulaanbaatar-51, Mongolia

Dee Leeuw J, Andrews M
ITC P.O. Box 6, 7500AA Enschede, The Netherlands

Abstract
Various Remote Sensing (RS) platforms and sensors have been in existence for quite some time. These are Landsat (1-5) MSS, Landsat TM, SPOT and other airborne systems. However, they operate within few ranges of spectral bands. The GER-II airborne spectrometry aircraft developed by Geophysical Environment Research Corp. of New York has a 63 band multispectral scanner and offers tremendously increased opportunities. The aim of this paper is to illustrate the possibilities of the GER-II data in geomorphological analysis. For that purpose, GER-II and Landsat TM digital data of Ramon area, Israel has been processed and compared using the ERDAS – system which is installed in the HP-workstation.

Introduction
The GER-II Sensor (imaging spectrometer) measures light reflected from the earth’s surface, utilizing many narrow spectral bands to construct the image of the target. It consists of three spectrometers with three individual linear detector arrays and has 64 spectral channels through the visible, near and middle infrared between 0.43 and 2.5 micrometers. Of the 64 bands, 63 are spectral data channels while the remaining one is used for the gyroscope. Bands 1 to 24 recorded by spectrometer 1 cover the 0.433 to 0.972 micrometer region with a bandwidth of 23.4 manometer. Bands 25 to 31 are recorded by spectrometer 2 and span the 1.1 to 1.8 micrometer region with a band with of 120 manometers. Spectrometer 3 includes bands 32 to 63 and provides coverage between 1.98 and 2,5 micrometer with band width of 16.43 manometer. The images used in our study was taken from a flying height of 15,000 feet above sea level. It has a ground resolution of 13.44 m and radiometric resolution of 16 bits. The scanner makes a

scanning angle of 90 degree and samples 512 pixels in the across track direction. Having 63 different narrow channels this platform offers tremendous opportunities for all kinds of analysis. The illustrates one of its possibilities. For geomorphological analysis two generations of ancient dry riverbed basin have been analyzed and the result was compared with the result of Landsat TM data. to detect the studies features, various techniques used in the analysis of RS data have been applied.

Geomorphology in the Ramon Area
From the geomorphological points of view, the erosion crater, - Maktesh Ramon – is a closed basin. It is surrounded by cliffs all around, and at the south-east side there are some inlets and outlets of ancient rivers.

The Ramon dry river bed starts at 800m. above sea level in order one, and finishes at 350m. above sea level in order sic. The relative height of the cliffs is 330m. at Arod pass at the south-western side, 240m, at Astronaut pass and 195 m. at Machmal pass. The construction of the cliff varaite is Dolomite, limestone and clay layers, formed in the upper Cretaceous (Cenomanian). The lower part of this cliff is made from sandstone with one marine formation layer from the lower Cretaceous.

Methods and Analysis
For the analysis the following techniques have been applied and the results were analyzed and compared.

Principle Components (PC) analyses.
Combination of the bands selected after band correlation
Saturation enhancement
Statistical pattern recognition using maximum likelihood decision rule.

For the improvement of visual interpretation, PC analyses have been done. Because of the software limitation 63 band digital data was grouped into 6 groups, according to the responses in the electro-magnetic wavelength and each of which was compressed into 3 PC-s (see Appendix 1). Assigning red, green and blue colors respectively, various combinations of these PC-s have been made to detect studying geomorphological features and also they were individually analyzed. It is interesting to notice that PC-2 of the red range which contains only 0.27% of the total information is more interpretative and has better contrast than PC-1, which contains 99% of the information from 9 bands being in that group.

This indicates that, in PC analysis one should consider variance tails which may include useful information Combinations of the First PC-s gave nice results. But that was not enough to detect the features to be interested, because there is still some important information mission, although it contains the highest variances. To select uncorrelated bands, for each group which contains a number of bands has been calculated variable – covariance and further correlation matrices and chosen bands whose correlation is the lowest. After selection of the target bands, various combinations of these bands making different FCC images have been performed. Combination of the bands 1,30,48 and 7,15,48 gives good results. Specially, in the image obtained by combination of 7,15 and 45 bands, two generations of the ancient dry riverbed basin are more distinguished. In TM data, the best image has been made by combining bands 1, 4 and 7 although spectral differentiation of the studying features was not seen as in the image made from the GER-II. To make clearer boundaries between the geographical objects, the images were edge enhanced using 3 x 3 convolution filter, the results of which were slightly different from the originals. Among applied enhancement techniques the best result has been obtained by the use of saturation enhancement. To apply this technique at the beginning after sum normalization. 3D RGB data has been mapped onto 2D color triangle, thus removing the influence of

Intensity, within the triangle, at first the data cloud has been shifted to the origin, to make the color balance and then spread the through the feature to use all possible colors. For the intensity of both Landsat TM and GER-II data, brightness values of the investigated bands have been taken, although in the case of the latter, radiance values of all or uncorrelated bands cloud be used. As seen from Fig. 3 in the saturation enhanced GER-II image. 2 generations of the dry riverbed basin are clearly distinguishable whereas they are not seen in the TM image (Fig. 4). To demonstrate spectral seperability of the geomorphological features a maximum likelihood classifier has been used. Training samples have been carefully selected through analysis. 3 bands (7, 15, 48) of GER-II data were chosen in the classification process and the image has been classified into 9 classes. In fig. 5 one can judge how the studied geomorphological features are spectrally distinguishable.

Conclusions
It has been shown that the GER-II data can successfully be used fro geomorphological analysis to distinguish the features which are spectrally not so easily distinguishable in other data. surely, the data has other applications, but it is necessary to mention the following limitations in processing.

The data is 16 bit, and requires a very large computer memory;
Not all software process so many bands simultaneously.

Therefore, there is a need for powerful hardware and software to handle it effectively and process such a large amount of data.

Acknowlegements
We are extremely greatly to Prof. Mazor Weizmann from the institute of Science and Mr. Ki Rohentson from the Inst. Geographic. Bogota for supplying materials. We also would like to thank Mr. W. Bakker for his help in the processing of the data and his useful comments.

References

Aslan H. Strahler, 1980, The use of Prior Probabilities in Maximum Likelihood Classification of RS Data. RS of Environment, 10, 135-163.
Mudler, N.J. and Middelkoop, Hans, 1990 Parametric Versus Non-Parametric Likelihood classification Proceedings of ISPRS, Wuhan, China
John A, Richards, 1986 Remote Sensing Digital Image Analysis, Springer, - Verlag Berlin Heidelberg New York. 275.
Paul M. Mather, 1987, Computer Processing of Remotely Sensed Images, John Wiley and Sons, Chichester, New York, Brisbane, Toronto, Singapore, 352.
Bakx, J.P.G 1991, Color Coding and Spectral Feature Extraction, ITC 42.
Ran Gerson, 1982 Talus Relicts Deserts, isr. J. Vol. 31, pp. 123-132.
J. Dan R. Gerson 1981 ARIDIC SOILS OF ISRAEL PP. 352-289.
Camson & Kinkby 1972 Hillislope, form & Process Wileg pp. 352-355.
Robert.E. Wallace 1977 Profils & ages of young fault scarps. North-central Nevada.
Ram B EN David & Emanoel Mazor 1988 Evolution of Mukhtesh Ramon isr. J. Earth Sci. Vol. 37. pp. 125-135.
H. Kaufman & W. Weisbrich 1991 Mineral identification Using GER-II S Data Acquired from Makhtesh Ramon isr.j.Earth Sci. Vol. 37 pp. 138-141.
H. Kauftan & W.Weisbrich 1991 Mineral Identification using GER-II Data Acquired from Mukhtesh Ramon EARSeL Vol. 1.no.1.pp.82092.

APPENDIX 1

1. Green (ands 1 to 10)
PC(a) = 68%
PC(b) = 27%
PC(c) = 4%

2. Red (bands 11 to 19)
PC(a) = 99%
PC(b) = .27%
PC(c) = .03%

3. IRI (bands 20 to 31)
PC(a) = 96.17%
PC(b) = 2.14%
PC(c) = .39%

4. IR2 (bands 32 to 35)
PC(a) = 82.5%
PC(b) = 9.44%
PC(c) = 5.75%

5. IR3 (bands 36 to 49)
PC(a) = 92.1%
PC(b) = 3.76%
PC(c) = 2.17%

6. IR4 (bands 50 to 63)
PC(a) = 78%
PC(b) = 14%
PC(c) = 2.4%

Basin	Stair (1)	Brown (2)Cuesta	M. Givat (3)Gash	M. Machmal (4)
Main altitude of third basin order	550	520	550	520
River basin order	1-4	1-4	1-3	1-4
Altitude basin first order.	670	640	675	580