Lineament extraction and analysis, comparison of LANDSAT ETM and ASTER imagery.

Case study: Suoimuoi tropical karst catchment, Vietnam

Hung L.Q.1,2, Batelaan O.1, and De Smedt F.1

1) Department of Hydrology and Hydraulic Engineering, Vrije Universiteit Brussel, Pleinlaan 2, 1050

Brussels, Belgium. Email to: quochle@vub.ac.be. 2) Department of Remote Sensing and Geomatic, Research Institute of Geology and Mineral

Resources, Thanhxuan – Hanoi – Vietnam

ABSTRACT Vast areas of the world consist of hard rocks (basement complexes), where water is restricted to secondary permeability, and thus to the fractures and the weathered zones. As the success ratio of drilling in hard rock terrain may be low, and the use of geophysics is often judged as too expensive, the study of lineaments from remote sensed imagery offers an attractive alternative analysis technique. High production areas in hard-rock aquifers are generally associated with conductive fracture zones. An effective approach for delineation of fracture zones is based on lineament indices extracted from satellite imagery. Together with a detailed structural analysis and understanding of the tectonic evolution of a given area it provides useful information for geological mapping and understanding of groundwater flow and occurrence in fractured rocks. The accuracy of extracted lineaments depends strongly on the spatial resolution of the imagery, higher resolution imagery result in a higher quality of lineament map. The ASTER sensor provides imagery with a higher resolution (15m) than the LANDSAT sensor (30m). It is tested and shown here that extracted lineaments from the VNIR ASTER imagery are considerably less noisy and show a higher accuracy than lineaments extracted from other imagery. Keywords: Lineament, remote sensing, LANDSAT, ASTER

1. INTRODUCTION Linear features on the earth surface have been a theme of study for geologists for many years, from the early years of the last century (Hobbs, 1904, 1912) up to now. From the beginning, geologists realized that linear features are the result of zones of weakness or structural displacement in the crust of the earth. A lineament is a mappable linear or curvilinear feature of a surface whose parts align in a straight or slightly curving relationship. They may be an expression of a fault or other line weakness. The surface features making up a lineament may be geomorphological, i.e. caused by relief or tonal, i.e. caused by contrast differences. Straight stream valleys and aligned segments of a valley are typical geomorphological expressions of lineaments. A tonal lineament may be a straight boundary between areas of contrasting tone. Differences in vegetation, moisture content, and soil or rock composition account for most tonal contrast (O’Leary et al. 1976). In general, linear features are formed by edges, which are marked by subtle brightness differences in the image and may be difficult to recognize. On the earth, lineaments could be (1) straight stream and valley, (2) aligned surface depressions, (3) soil tonal changes, (4) alignments in vegetation, (5) vegetation type and height changes, or (6) abrupt topographic changes. All of these phenomena might be the result of structural phenomena such as faults, joint sets, folds, cracks or fractures. The old age of many geological lineaments means that younger sediments commonly cover them. When reactivation of these structures occurs, this results in arrays of brittle structures exposed on the surface topography. Similarly, the surface expression of a deep-seated lineament may be manifested as a broad zone of discrete lineaments (Richards, 2000). In order to map structurally significant lineaments, it is necessary first, by careful and critical analysis of the image, to identify and screen features not caused by faulting (Sabins, 1997). The study of lineaments has been applied successfully to structural geology studies and their applications such as ore-forming systems, mineral exploration, petroleum, nuclear energy facility sittings (Lalor, 1987; Woodall, 1993, 1994;

Remote Sensing for Environmental Monitoring, GIS Applications, and Geology V,edited by Manfred Ehlers, Ulrich Michel, Proc. of SPIE Vol. 5983, 59830T, (2005)

0277-786X/05/$15 · doi: 10.1117/12.627699

Proc. of SPIE Vol. 5983 59830T-1

Study area-'H I N A

*I9(7> TonW-v-- . '--'.t. 4

THAILAND'

CAM8ODL?

0 200 40000

LIZ

Ell I

g

Fg

O'Driscoll and Campbell, 1997; Kutina, 1980; Karnieli et al., 1996; Mostafa and Zakir, 1996; Arlegui and Soriano, 1998; Suzen and Toprak, 1998; Zakir et al., 1999), and water resource investigations, groundwater studies (Lattman and Parizek, 1964; Mabee et al., 1994; Magowe and Carr, 1999; Fernandes and Rudolph, 2001; Hung et al., 2003). Lineament identification via remotely sensed data is achieved by using two principal techniques. First, lineament data can be visually enhanced using image enhancement techniques (image ration, image fusion, directional edge-detection filters) and a lineament vector map can be produced using manual digitizing techniques (Arlegui and Soriano, 1998; Suzen and Toprak, 1998). Second, a lineament map may be produced using computer’s software and algorithms (Burdick and Speirer 1980, Karnieli et al., 1996; Baumgartner et al. 1999, Hung et al. 2002, 2003, Kim et al. 2004). Study area The study area, the Suoimuoi river catchment, is situated northwest of the city of Son La, between longitude 103°33’ E and 104°00’ E and latitude 21°20’ N and 21°29’ N, covering 284 km² (Fig. 1). This area consists mainly of two sub-areas. The south-western parts are composed of karst water-bearing carbonate rocks of Paleozoic age, Banpap (D2bp) and Early Permian-Carboniferous Chienpac (C-P1cp) Formations. The north-eastern part is of Middle Triassic age with the Dong Giao Formation (T2dg). In the Suoimuoi catchment, karst occurs in lime stones and dolomites of Late Cambrian, Middle Devonian, Carboniferous - Early Permian and Middle Triassic age. These carbonates have a favourable composition, texture, and structure for karstification (Hung, 2001).

Many different tectonic phases and neotectonic movements have intensively affected these rocks. The present Son La karst highland is dissected by NW-SE, SW-NE, sub-latitudinal, and sub-meridian trending faults. The NW-SE fault system resulted from the collision of continental crust in the Precambrian. The region then was affected by NW-SE oriented folding due to Indosinian closure of the Paleotethys, starting in Late Permian and culminating during Middle Triassic. All the deposits were finally uplifted during the Neogene Himalayan collision event (Tri and Tung, 1979; Tien et al., 1991). Two main types of conjugate fracture patterns are recognized for the Suoimuoi catchment, depicted by the diagrams in Fig. 2. Hop (1997) supposed that both fracture types (E-W and NW-SE) generated faults (shear fractures) and extensional fractures, which affected the rocks from Precambrian up to Middle Triassic. The E-W fracture type is accompanied by shear fractures with a NW-SE and NE-SW direction and extension fractures with a sub-E-W

direction. The NW-SE fracture type is accompanied by shear fractures of sub-N-S and sub-EW direction and extension fractures with a NW-SE direction. The correspondence of the NW-SE shear and extension directions and the E-W correspondence of shear and extension directions, make these two directions most favourable for groundwater development. Landsat ETM and ASTER imagery are used to improve the quality of lineament map by their spectral and resolution properties. The images were resized to the exact coordinates of the study area by a subset from the original scenes. The Landsat ETM image was a subset from an original scene of a path 128 and row 045 and acquired on 27/1/2000. The ASTER image was a subset from an original ASTER scene of L2 processing level, acquired on 6/10/2003. All images are registered to the UTM WGS84 N48 projection. The image resolution of LANDSAT and ASTER are sufficient for the study at map scale of 1/50.000 to 1/100.000.

Figure 1: The Suoimuoi catchment

Proc. of SPIE Vol. 5983 59830T-2

U3 UI* 'ILUI N ____=,. — U3

(NW-SE)

A'

Where: σ1 is the direction of shear fracture, and σ3 is the direction of extension fracture Figure 2: Rose diagram depicting the distribution of faults on the geological map and the orientations of shear and extension fractures

2. DEVELOPMENT OF METHODOLOGY The aim of this paper is to create a methodology for automatically and digital lineament analysis. The processing is developed based on the map scale of 1/50.000. Figure 3 shows the major steps which are applied for the lineament analysis. The methodology aims at incorporating remote sensed data and principles of processing these data (Drury, 1993 Schowengerdt, 1997; Hung et al. 2002 and 2003; Kim et al. 2004). Lineament extraction Lineament is extracted

from image

Lineament correction

Statistical computation

Griding by indice

Classification

Filter lineament, breaking, joining, removing, to create the basic map of lineament

Classifying by density.

Distribution function, rose diagram, histogram

Calculating lineament indice by area unit (derectional or non-directional)

Figure 3: The major steps of lineament analysis

2.1. Lineament extraction From previous studies we can conclude that lineaments usually occur as edges with tonal differences in satellite images. According to literature there are two common methods for the extraction of lineaments from satellite images: 1. Visual extraction: At which the user first starts by some image processing techniques to make edge enhancements, using the directional and non directional filters such as the Laplacian, and Sobel, then the lineaments are digitized manually by the user. 2. Automatic (or digital) extraction: various computer-aided methods for lineament extraction have been proposed. Most methods are based on edge filtering techniques. The most widely used software for the automatic lineament extraction is the LINE module of the PCI Geomatica. A comparison between the visual and automatic methods for lineament extraction is shown in Table 1, after Hung 2001. The ETM image used in this study is composed of 6 VNIR and SWIR bands with a resolution of 30 m, and the panchromatic band 8 with a resolution of 15 m. The ASTER image has 3 VNIR bands with a resolution of 15 m, and 6 SWIR bands with a resolution of 30 m (Table 2). Lineaments were automatically extracted from these bands with the LINE module and parameters as shown in Table 3. In this study, the thermal bands of both ETM and ASTER will not be

NW-SE Sub N

NW-SE

Sub E

Proc. of SPIE Vol. 5983 59830T-3

taken into account because of too low resolution, 120 and 90m respectively. Table 1: Comparison between the Visual and the Automatic (Digital) lineament extraction methods

Visual process Digital process - Depend on the quality of the performance of the image (on paper and/or screen)

- Depend on only the quality of the image

- Partly depend on the complexity of the research area - Totally depend on the complexity of the research area - Strongly depended on human experience and ability - Totally depend on the mathematical function of software - Takes a lot of time - Very quickly - Strong effect of human subjectiveness - Little effect of human subjectiveness - Easy to distinguish the kind of lineament (tectonic setting, manmade, …)

- Can not recognize the kind of lineament, so the result may be confused.

- Simple but subjective method - Complex but objective method Table 2: VNIR and SWIR Landsat ETM and ASTER imagery

ETM+ ASTER Subsystem Wave length (µm) Resolution (m) Wave length (µm) Resolution (m)

Blue 0.45-0.52 30 Green 0.53-0.61 30 0.52-0.60 15 Red 0.63-0.69 30 0.63-0.69 15 V

NIR

NIR 0.75-0.90 30 0.78-0.86 15 Panchromatic 0.52-0.90 15

1.55-1.75 30 1.60-1.70 30 2.10-2.35 30 2.145-2.185 30

2.185-2.225 30 2.235-2.285 30 2.295-2.365 30

SW

IR

2.360-2.430 30 Table 3: The used parameters for the PCI - LINE Module, for each data type. RADI - Radius of filter in pixels, GTHR - Threshold for edge gradient, LTHR - Threshold for curve length, FTHR - Threshold for line fitting error, ATHR - Threshold for angular difference, DTHR - Threshold for linking distance (PCI Geomatica Manual, 2001).

Parameter Values for ASTER and the fused ETM image

Values for the TM image Parameter

Values for ASTER and the fused ETM image

Values for the TM image

RADI 5 5 FTHR 3 3 GTHR 10 10 ATHR 7 7 LTHR 7 3 DTHR 3 3

The main advantage of ETM is that the panchromatic band can be used to have an image with the resolution of 15 m, after the process of image fusion. The method of intensity–hue–saturation (IHS) fusion is applied to enhance the resolution of LANDSAT image bands 3, 4 and 5, which is reported to have smallest spectral error (Schowengerdt, 1997). The number of lineaments from each band in each data type, before and after correcting, is given in Table 4. There are two groups of lineaments, Group A consists of lineaments derived from images of 30 m resolution and B of 15 m resolution (Fig. 4). The higher number of lineaments in group B is derived from the three ASTER bands 1, 2 and 3, and three ETM fused band 3, 4, and 5. The number of lineaments in group B is proximately twice the amount in group A. While, the number of lineaments in each image in group B is almost the same, (Table 4 and Fig. 4B), the number of lineaments in each image in group A is a function of the wave length. The highest number of lineaments in both groups is in the VNIR image, band 3 of ASTER in group B and band 4 of LANDSAT in group A. The error of number of lineament before and after correcting is smallest in VNIR band of both groups (Table 4). By visual comparison, it is clear that VNIR is the best band for the automatic lineament extraction in the study area, due to its high lineament frequency and accuracy. Schowengerdt (1997) also observed that the VNIR band has smallest spectral error. The distribution of number of lineament for LANDSAT and ASTER image shows the same trend of alteration (Table 5, 6 and Fig. 5, 6) by

Proc. of SPIE Vol. 5983 59830T-4

direction. It proves that the parameters which were chosen for LINE module are suitable for the research area.

0

1000

2000

3000

4000

5000

6000

7000

8000

9000

10000

TM1 TM2 TM3 TM4 TM5 TM7 AST4 AST5 AST6 AST7 AST8 AST9

Image

Num

ber o

f lin

eam

ent

BeforeAfter

A

0

5000

10000

15000

20000

25000

ETM3_F ETM4_F ETM5_F AST1 AST2 AST3

Image

Num

ber o

f lin

eam

ent

BeforeAfter

B Figure 4: The number of lineaments before and after correction; A. 30m resolution images; B. 15m resolution images Table 4: The number and max length of lineaments before and after correction

Before correcting After correcting Band number

Total number Max length (m) Total number Max length (m) The deferent of number of lineament, before and after correcting (%)

TM1 9819 1897 8833 1897 11.2 TM2 9656 2058 8718 2058 10.8 TM3 9764 1916 8806 2029 10.9 TM4 9653 2670 9156 2670 5.4 TM5 9926 2643 8989 2643 10.4 TM7 9826 2643 8920 2643 10.2 AST4 9119 2175 8342 2175 9.3 AST5 9203 2343 8356 2343 10.1 AST6 9357 2238 8522 2238 9.8 AST7 9350 2451 8539 2451 9.5 AST8 9390 2125 8512 2125 10.3 AST9 9436 2099 8618 2099 9.5 ETM3_F 21834 1340 18444 1340 18.4 ETM4_F 21985 1452 18667 1452 17.8 ETM5_F 21938 1158 18880 1158 16.2 AST1 22431 1423 19104 1423 17.5 AST2 22428 1461 18973 1461 18.2 AST3 22785 1468 19369 1468 17.6

Table 5: Distribution of number of lineaments by direction interval for SWIR image.

Interval AST4 AST5 AST6 AST7 AST8 AST9 TM1 TM2 TM3 TM4 TM5 TM7 <13.06 912 905 886 931 896 935 946 942 947 1015 955 949 13.06 - 26.11 739 657 720 706 738 686 771 781 708 822 797 789 26.11 - 39.17 1058 986 994 990 961 963 950 1024 982 1156 1085 1072 39.17 - 52.23 947 933 968 918 932 960 1004 1015 945 1074 1015 1009 52.23 - 65.29 857 824 836 819 777 799 833 892 881 1047 919 900 65.29 - 78.34 659 597 642 617 595 521 623 687 588 640 645 649 78.34 - 91.40 385 429 406 394 451 518 473 416 528 560 497 492 91.40 - 104.46 291 307 322 307 321 322 348 280 318 256 325 325 104.46 - 117.52 236 333 304 349 338 350 313 282 331 266 326 324 117.52 - 130.58 289 326 359 332 339 374 386 317 350 278 284 295 130.58 - 143.64 444 553 527 530 540 519 522 484 550 378 494 476 143.64 - 156.69 577 555 545 565 605 608 594 490 586 532 516 514 >=156.69 948 951 1013 1081 1019 1063 1070 1108 1092 1132 1131 1126

Proc. of SPIE Vol. 5983 59830T-5

0

200

400

600

800

1000

1200

1400

1 2 3 4 5 6 7 8 9 10 11 12 13Interval

Num

ber o

f lin

eam

ent

AST4 AST5 AST6 AST7 AST8 AST9 TM1 TM2 TM3 TM4 TM5 TM7

Figure 5: Distribution of number of lineaments by direction interval for SWIR image.

0

500

1000

1500

2000

2500

1 2 3 4 5 6 7 8 9 10 11 12 13 14Interval

Num

ber o

f lin

eam

ent

ETM3ETM4ETM5

A

0

500

1000

1500

2000

2500

1 2 3 4 5 6 7 8 9 10 11 12 13 14

Interval

Num

ber

of li

neam

ent

AST1AST2AST3

B Figure 6: Distribution of number of lineament by direction interval for VNIR image. A. Fused ETM; B. ASTER Table 6: Distribution of number of lineament by direction interval for VNIR image.

ETM3 ETM4 ETM5 AST1 AST2 AST3 <12.15 1677 1769 1897 1820 1666 1938 12.15 - 24.30 1563 1625 1667 1553 1543 1704 24.30 - 36.45 1868 1991 2004 1974 1876 2060 36.45 - 48.60 2022 2204 2336 2100 2032 2238 48.60 - 60.75 2031 2103 2282 2054 1916 2171 60.75 - 72.90 1582 1665 1740 1546 1377 1793 72.90 - 85.06 1131 1238 1215 1134 1168 1227 85.06 - 97.21 764 787 750 730 785 777 97.21 - 109.36 508 455 393 509 564 448 109.36 - 121.51 580 444 374 556 677 530 121.51 - 133.66 616 488 375 634 761 518 133.66 - 145.81 866 677 639 931 1085 761 145.81 - 157.97 1132 964 961 1288 1199 972 >=157.97 2104 2257 2247 2275 2324 2232

Proc. of SPIE Vol. 5983 59830T-6

a

r

h

a

2.2. Lineament correction The lineament correction is an automatic process, which is based on the statistics of the extracted lineaments (Fig. 7). A correction tool is developed, which allows users to redefine the lineament map. There are several kinds of lineament errors, such as, not connected (because of discontinuous reflectance on the image), overlapping or too close (appearing in the streams or image shadow), abnormal lineaments (too short or too long).

Filter lineament –Break polyline function

Filter lineament –Connect function

Filter lineament –Remove line function

Import lineament map from GIS software (*.dxf, *.text, *.mif)

Menu File –Import

Sort lineament by coordinate Break polyline to line

Connect/replace line

Remove line

The corrected lineament map

The statistical test

Filter lineament – Sort by coordinate

Filter lineament – Common statistic function

Figure 7: The major steps of lineament correction In order to save time for running the correction process, all lineaments are sorted by their coordinate. The lineament map, which is a result from the vectorization software, contains lines and polyline; polylines have first to be broken up in lines. In Fig. 8 shows three cases of correction of lineaments. The resulting lineament map contains only straight line, with clear statistical characteristics.

A B C Where: - a, b, c are different extracted lineaments, r is the resulting corrected lineament

- α, β are the angle between two lineaments Figure 8: Three cases of incorrect lineaments and the resulting correction. There are three major cases, which occur when lineaments have to be joined. The most usual case is presented in Fig. 8a. The two lineaments, a and b, are too close and intersect each other. If the angle α is small enough, the two can be joined into a new lineament r. The length of r is calculated through the maximum and minimum coordinate of a and b. While the direction of c is calculated as the average direction of a and b. If α equals to 0, and if the distance between a and b is also equal to 0 (overlap) or closed to 0 (parallel) then the length of c is calculated as the maximum and minimum coordinate of a and b and the same direction. Fig. 8b shows a less common case but very important one. In this case, two

The original lineament

The corrected lineament

The extended lineament by its direction

Proc. of SPIE Vol. 5983 59830T-7

t

lineaments, a and b, are too close and do not have an intersection. If the angle α is small enough and the distance is small so it could be combined as r. The length of r is calculated through the maximum and minimum coordinate of a and b. The direction of r is calculated from the average direction of a and b. This case happens in complicated terrain with straight valley or wide rivers. The complexity is shown in Fig. 8c. In this case, we have a combination of the situation in Fig. 8a and 8b, the lineaments, a, b and c, are too close. Angles and lengths are used to create a new lineament.

Figure 9: Extracted lineament error by topography. Another error that could effect the lineament extraction is the boundary of the sub catchment (Fig 9). In order to remove this kind of error, the DEM of the research area is used to define the sub catchment. All lineaments within a buffer zone of the sub catchment boundary are removed. By using the statistical descriptors of the lineament map, all of the abnormal lineaments, such as, too short or too long can be removed. This tool is very helpful for combining lineament maps from difference sources, because each technique or type of data can provide their own result.

2.3. Lineament density

2.3.1. Lineament indices Lineaments are commonly analyzed using frequency or length against azimuth histograms (Mostafa and Zakir, 1996; Zakir et al., 1999), rose-diagrams (Karnieli et al., 1996), and/or lineament density maps (Zakir et al., 1999). The most common method is to calculate lineament density based on the number of lineaments per unit area (number/km²), or the total length of lineaments per unit area (km/km²) or combining all. However, there are 6 indices, 3 main indices and 3 modified indices. The first main index is the number of lineaments per unit area ( N -number/km²), the second main

index is the total lengths of lineaments per unit area ( ∑=

=N

iixL

1

-km/km², while ix is the length of lineament number i),

the last main index is the total number of the intersections of lineaments per unit area (NI - number/km²). There are two modified indices, the ratio of the intersections (NI) versus the number of lineaments (N) and the ratio of the intersections (NI) versus the total length of lineaments (L), which are usually used for studying fracture zones in specific direction. The last modified index, the average length of lineaments ( NL ) is usually used when the scale of study is small (smaller than 1/100.000). By using this index, we can minimize the volume of calculation because there is no difference in the development trend of the number of lineament and total lengths of lineament. The lineament density indices are calculated for a raster with a certain grid cell resolution, which can be defined based on the statistical descriptors and can be reclassified into ranges of values and presented as an isopleths map.

2.3.2. Area unit The most important step is to convert the lineament map with only line objects to a lineament map with only point objects. The point location and values are defined following Fig. 10. The point location is located at the grid cell centre.

Sub catchment boundary

Stream

Lineament

Proc. of SPIE Vol. 5983 59830T-8

The point value is calculated based on the lineaments inside this grid cell of size dx and dy (Fig. 10). User defined block sizes are always better, because the block size is depended not only the statistical descriptor of lineaments but also the geological condition of research area. The value for each lineament index is calculated by the part of lineament inside this grid cell (Fig. 10).

2.3.3. Gridding The lineament indices densities are created by simple kriging interpolation tool of the software Surfer. However, the lineament density map will be more useful when it is combined with other information. Surfer allows the lineament density maps to be integrated with other data like geophysical and geochemical

data that can improve the lineament interpolation. Figure 11 shows the results of the lineament indices for a unit area of 1 by 1 km. Higher densities are assumed to represent a higher level of fracturization of the rock. In general, the major characteristics, (local peak, direction of peak) of the lineament indices density are the same. However, there are differences in detail. Comparing the map of number of lineaments and the number of intersections of lineaments, it is observed that the locations of lower values of the number of lineaments (Fig. 11A) are not always the same as that in the map of number of intersections (Fig. 11B). This can be explained by the complicated tectonic regime in the area. Fig. 11D gives the ratio of Fig. 11C and Fig 11A, i.e. the average lineament length, which can be used to define the fractured zone of the study area. The corrected lineament map of the study area is shown in Fig. 12A. By overlaying the fault map of the study area on average length of lineament map (Fig. 11B); we can see that most of the faults are located very close to locations with high values of density. The directions of the faults are corresponding with the directions of the high density patterns.

355000 360000 365000 370000 3750002360

000

2365

000

2370

000

2375

000

2380

000

55

65

70

75

80

85

90

95

100

0 5 10

k i l o m e t e r s

Contour line and value Catchment boundary75 A

355000 360000 365000 370000 3750002360

000

2365

000

2370

000

2375

000

2380

000

2

4

6

8

10

12

14

16

18

20

22

0 5 10

k i l o m e t e r s

Contour line and value Catchment boundary75 B

A dy

dx Where: A is centre point of the grid cell, dx and dy are the x and y size of the grid cell Fig. 10: Defining a grid cell with intersecting lineaments.

Proc. of SPIE Vol. 5983 59830T-9

355000 360000 365000 370000 3750002360

000

2365

000

2370

000

2375

000

2380

000

9000

12000

13500

15000

16000

17000

18500

20000

21500

24000

25000

25500

0 5 10

k i l o m e t e r s

Contour line and value Catchment boundary75 C

355000 360000 365000 370000 3750002360

000

2365

000

2370

000

2375

000

2380

000

3500

4625

5750

8000

10250

12200

13500

14500

15500

16500

17500

18500

19500

0 5 10

k i l o m e t e r s

Contour line and value Catchment boundary75 D Figure 11: Interpolated density map of lineament indices. A – Number of lineament; B – Number of intersection of lineament; C - Length of lineament; D – Average lineament length

StreamLineament

Catchment boundary

375000° E

UTM (WGS84) Zone 48N

370000° E

105

kilometers

0

2365

000°

N23

6000

0° N

355000° E 360000° E 365000° E

2380

000°

N23

7000

0° N

2375

000°

N

A

355000 360000 365000 370000 3750002360

000

2365

000

2370

000

2375

000

2380

000

0 5 10

k i l o m e t e r s

Contour line and value

Catchment boundary

75

3500

4625

5750

8000

10250

12200

13500

14500

15500

16500

17500

18500

19500

Fault

B Figure 12: A – Map of lineament in the study area; B – Average lineament length and local fault Rose diagrams are used to test the distribution of lineaments by direction (Fig. 13A). The dominant direction of the fault system in the study area does not correspond with the main direction found in Fig.2. The reason is that a fault can be represented by a number of lineaments with the same direction. However, testing the distribution of lineament directions in a sub-area of the lineament map reveals that the rose diagram of faults and lineaments are very similar (Fig. 13B).

A

Proc. of SPIE Vol. 5983 59830T-10

'p

27OO 8 1CO

7

A B Figure 13: A - Rose diagram depicting the distribution of lineaments in Fig. 11A; B - Rose diagram depicting the distribution of lineaments at the point A in Fig. 11B

3. CONCLUSION In this study the lineaments in the area of Suoimuoi were automatically extracted using the module LINE of PCI Geomatica from two different sensors, ASTER and LANDSAT ETM. A comparison was made between the different bands to select the best bands for lineament extraction. Results indicate that the best bands for automatic lineament extraction are the VNIR band3 of ASTER and the fused band 4 of the LANDSAT ETM. Results also indicate that ASTER data are the best for accurate detailed lineament analysis at the scale of the 1:50,000. A methodology for lineament extraction and analysis is developed. This methodology has been proven as a suitable method for lineament extraction and analysis by comparison with the geological data for the Suoimuoi catchment. The density map of lineament indices and rose diagrams of the extracted lineaments indicate the same geological structure as reported before. The density map of average length of lineament can be used to define the facture zone in the study area.

REFERENCES 1. Arlegui, L. E., Soriano, M. A., “Characterizing lineaments from satellite images and field studies in the central Ebro

basin (NE Spain)”, International Journal of Remote Sensing 19 (16), 3169-3185, 1998. 2. Baumgartner, A., Steger, C. Mayer, Eckstein, W. and Ebner, H., “Automatic road extraction based on multi-scale

grouping and context”, Photogrametric Engineering and Remote Sensing, Vol. 65, PP. 777-785, 1999. 3. Burdick, R. G. and Speirer, R. A., “Development of a method to detect geologic faults and other linear features from

LANDSAT images”, U. S. Bureau of Mines Report Inv., 8413:74, 1980. 4. Drury, S., Image interpretation in Geology, Chapman and Hall, London, 1993. 5. Fernandes, A.J., Rudolph, D.L., "The influence of Cenozoic tectonics on the groundwater production capacity of

fractured zones: a case study in Sao Paulo, Brazil”, Hydrogeology Journal 9, 151–167, 2001. 6. Hobbs, W.H., “Lineaments of the Atlantic border region”, Geological Society of America Bulletin 15, 483–506,

1904. 7. Hobbs, W.H., “Earth Features and Their Meaning: An Introduction to Geology for the Student and General Reader”,

Macmillan, New York, NY (347pp), 1912. 8. Hop, N.D., “Geological map of Thuan Chau – Son La (Vietnam)”, The Geological Survey of Vietnam, 600 pp,

1997. 9. Hung LQ, Dinh NQ, Batelaan O, Tam VT, Lagrou D, “Remote sensing and GIS-based analysis of cave development

in the Suoimuoi catchment (Son La-NW Vietnam)”, Journal of Cave and Karst Studies 64(1):23-33, 2002. 10. Hung, L.Q. and Batelaan, O., “Environmental geological remote sensing and GIS analysis of tropical karst areas in

Vietnam”, Proceedings of the IEEE International Geoscience and Remote Sensing Symposium (IGARSS), Toulouse, France, 21-25 July 2003, Volume IV, p. 2964-2966, 2003.

11. Hung, L.Q., “Remote sensing based hydrogeological analysis of Suoimuoi catchment Vietnam [MSc thesis]”, Vrije Universiteit Brussel, 87p, 2001.

12. Hung, L.Q., Batelaan, O. and San, D.N., “Lineament analysis in fractured rocks, methodology and application to the Suoimuoi karst catchment (full paper on CD)”. Proceedings of the International Conference on Groundwater in Fractured Rocks, 15-19 September 2003, Prague, eds. Krasny , J., Hrkal, Z. and Bruthans, J., 2003.

Proc. of SPIE Vol. 5983 59830T-11

13. Karnieli, A., Meisels, A., Fisher, L., Arkin, Y., “Automatic extraction of geological linear features from digital

remote sensing data using a Hough Transform”, Photogrammetric Engineering & Remote Sensing, 62, 525–531, 1996.

14. Kim, G.B., Lee, J.Y., Lee, K. K., “Construction of lineament maps related to groundwater occurrence with ArcView and AvenueTM scripts”, Computers & Geosciences, Vol. 30, 1117–1126, 2004.

15. Kutina, J., “Are lineaments ineffective guides to ore deposits?”, Global Tectonics and Metallogeny, 1, p. 200–205, 1980.

16. Lalor, J. H., “The Olympic Dam copper-uranium-gold deposit, South Australia”. Transactions Fourth Circum-Pacific Energy and Mineral Resources Conference, Singapore. - American Association Petroleum Geologists. Tulsa, Oklahoma: 561-567, 1986,

17. Lattman, L.H., Parizek, R.R., “Relationship between fracture traces and the occurrence of ground water in carbonate rocks”, Journal of Hydrology 2, 73–91, 1964.

18. Mabee, S.B., Hardcastle, K.C., Wise, D.U., “A method of collecting and analyzing lineaments for regional-scale fractured-bedrock aquifer studies”, Ground Water 32 (6), 884–894, 1994.

19. Magowe, M., Carr, J.R., “Relationship between lineaments and ground water occurrence in western Botswana”, Ground Water 37 (2), 282–286, 1999.

20. Mostafa, M., Zakir, F., “New enhanced techniques for azimuthal analysis of lineaments for detecting tectonic trends in and around the Afro-Arabian Shield”, International Journal of Remote Sensing, 17, 2923–2943, 1996.

21. O’Leary, D.W., Freidman, J.D., Pohn, H.A., “Lineaments, linear, lineation-some proposed new standards for old terms”, Geological Society of America Bulletin 87, 1463–1469, 1976.

22. O'Driscoll, E. S. T. & Campbell, I. B., “Mineral deposits related to Australian continental ring and rift structures with some terrestrial and planetary analogies. Global Tectonics and Metallogeny 6: 83-101, 1997.

23. PCI Geomatica Users’ Manual, 2001. 24. Richards, J. P., “Lineaments Revisited”, SEG newsletter, 42, 14–21, 2000. 25. Sabins, F., Remote Sensing: Principles and interpretation, 2nd edn (New York, NY: Freeman), 1997. 26. Suzen, M. L., Toprak, V., “Filtering of satellite images in geological lineament analyses: an application to a fault

zone in Central Turkey”, International Journal of Remote Sensing, 19, 1101–1114, 1998. 27. Schowengerdt, Remote Sensing 2nd Ed, Elsevier(Academic Press), 522pgs, 1997. 28. Tien, P. C., “Geology of Cambodia, Laos and Vietnam. Explanatory note to the Geological map at 1:1 000 000

scale”, Geol. Survey Vietnam, 1991. 29. Tri, T. V., Tung, N. X., “Geology of Vietnam, the north part”, Science-Techn.Publishing House, Hanoi, 1977. 30. Woodall, R., “The multidisciplanary team approach to successful mineral exploration”, Society of Economic

Geologists Newsletter 14: 1-6, 1993. 31. Woodall, R., “Empiricism and concept in successful mineral exploration”, Australian Journal of Earth Sciences

41(1): 1-10, 1994 32. Zakir, F., Qari, M., Mostfa, M., “A new optimising technique for preparing lineament density maps”, International

Journal of Remote Sensing, 20, 1073–1085, 1999.

Proc. of SPIE Vol. 5983 59830T-12