Majalah Geologi Indonesia, Vol. 29 No. 2 Agustus 2014: 63-69

Extracting Linear Feature Density of SAR Data to Detect the Paths of a Geothermal System Ekstraksi Densitas Fitur Linier Data SAR untuk Mendeteksi Jalur Sistem Panas Bumi Asep Saepuloh1, Minoru Urai2, Suryantini1, Prihadi Sumintadireja1, Irwan Meilano1, Agus H. Harsolumakso1, and Emmy Suparka1 Faculty of Earth Sciences and Technology, Bandung Institute of Technology (ITB - Indonesia). Institute of Geology and Geoinformation, National Institute of Advanced Industrial Science and Technology (AIST - Japan).

1 2

Corresponding author: [email protected]; [email protected] Manuscript received: October 3, 2013, revised: April 8, 2014, approved: August 8, 2014

ABSTRACT The linear feature density related to geological structures is a key to predict fluid paths of a geothermal system. Manual tracing of the linear features is efficient but not effective because of time consuming, especially for a wide area with various Geomorphologic and Structural Features (GSF). An automatic extraction of linear feature density from Synthetic Aperture Radar (lifedSAR) was used by taking into account the statistical parameter to minimize the effect of man-made features and vegetation. The Jampang area in West Java, Indonesia was selected as a study site in accordance with the existence of Cimandiri Fault Zone (CFZ) and two active volcanoes: Mt. Salak and Mt. Gede. The existence of hot springs at CFZ indicated that the fault systems might be served as the path to the source of geothermal fluid. Thus, the heat source is aimed to be predicted through the Linear Feature Density (LFD). The geological structure map and DEM SRTM 90 m were used to validate the correctness of the lifedSAR. Keywords: lifedSAR, Geomorphologic and Structural Features, Cimandiri Fault Zone, geothermal fluid ABSTRAK Densitas fitur linier yang berhubungan dengan struktur geologi merupakan kunci untuk memprediksi jalur fluida sistem panas bumi. Penggambaran manual fitur linier efisien, tapi tidak efektif karena memerlukan waktu lama terutama untuk wilayah yang luas dengan berbagai Fitur Struktur dan Geomorfologi (GSF). Sebuah ekstraksi otomatis densitas fitur linier dari Synthetic Aperture Radar (lifedSAR) digunakan dengan memperhatikan parameter statistik untuk meminimalkan efek buatan manusia dan vegetasi. Daerah Jampang di Jawa Barat, Indonesia, terpilih sebagai lokasi penelitian sesuai dengan keberadaan Zona Sesar Cimandiri (CFZ) dan dua gunung api aktif: Gunung Salak dan Gunung Gede. Keberadaan mata air panas di CFZ menunjukkan bahwa sistem sesar mungkin berlaku sebagai jalan menuju sumber cairan panas bumi. Dengan demikian, sumber panas dapat diprediksi melalui Densitas Fitur Linier (LFD). Peta struktur geologi dan DEM SRTM 90 m digunakan untuk memvalidasi kebenaran lifedSAR. Kata kunci: lifedSAR, Fitur Struktur dan Geomorfologi, Zona Sesar Cimandiri, cairan panas bumi

63

Majalah Geologi Indonesia, Vol. 29 No. 2 Agustus 2014: 63-69

INTRODUCTION

about 0.5 to 1.7 cm/yr (Abidin et al., 2009). The existence of hot springs at the southern flank of the CFZ and their origin are supposed to be traced in this study.

The geological structures, especially active faults, usually control fluid flows and surface manifestation in a geothermal system (Soengkono, 1999; Rosenberg, 2009). Detecting the geological structures directly using optical sensor data is difficult due to vegetation cover, boundary of rock types, clouds, and/or erosion. Overcoming the problem, the microwave sensor data are used to detect geological structure presented by Geomorphologic and Structural Features (GSF) in the back-scattering intensity of Synthetic Aperture Radar (SAR) data. This paper is aimed to predict geothermal fluid paths along an active fault and to interpret the location of geothermal system using dual observation modes of SAR data.

The weak zones due to the faults and fractures might serve as fluid paths of a geothermal system (Demirel et al., 2004). According to the existence of two active volcanoes at the northern part from CFZ: Mt. Salak and Mt. Gede, the lifedSAR is applied to predict the possibility of continuity of the linear features connected to the heat source. The other source possibilities such as faults derived thermal in this study were excluded (e.g. Tianfeng, 1984; Serpen et al., 2009; Brogi, 2008). Therefore, the discussion is focused on the hydrothermal related Quaternary volcanism acting as geothermal occurrences in West Java (Carranza et al., 2008).

An automatic extraction of the linear features density from Synthetic Aperture Radar termed as lifedSAR is used to quantify the geological structures around Cimandiri Fault Zone (CFZ) (Saepuloh et al., 2013). The CFZ located in West Java Province, Indonesia is an active fault elongated from Pelabuhan Ratu, through Sukabumi, Cianjur, and Padalarang (Figure 1). This fault system has a horizontal displacement rate

DATA AND METHOD The SAR back-scattering intensity data represents the power return to the satellite as a function of several surface parameters such as roughness, gradient slope, magnetic, and dielectric property of the ground (Saepuloh et al., 2012). The parameters are presented

o

-4 S

Mt. Salak o

-6 S

Mt. Gede

JAWA

CFZ - 8o S 170 km o

10 S

o

110 E

o

115 E

Figure 1. Location of studied area in West Java, Indonesia (black square) and Cimandiri Fault Zone (blackdashed polygon). Mt. Salak and Mt. Gede (black triangles) were overlaid to the map. 64

Extracting Linear Feature Density of SAR Data to Detect the Paths of a Geothermal System (A. Saepuloh et al.)

by the GSF in the SAR back-scattering intensity images (Saepuloh et al., 2012). According to the GSF, the ridges, valleys, hills, and rivers are the crucial components for detecting the linear features. Their geomorphology features are characterized by high back-scattering intensity in general.

multilook processing with 3×6 factor to keep the spatial resolution of image along the range and azimuth directions as 28 m and 23 m, respectively. Finally, the multilook image in a slant range was transformed to ground range direction based on simulated-DEM derived from the SRTM 90 m data.

The Phased Array type L-band Synthetic Aperture Radar (PALSAR) onboard the Advanced Land Observing Satellite (ALOS) with two orbits mode: toward north (ascending) and south (descending) flight directions are used in this study. Detail information about data acquisition is listed in Table 1.

A problem arising at the edge of detection step due to geometrical distortions in the SAR data that the rough surface with high gradient topography produced double edges in the SAR back-scattering intensity images. The double edges originated from back- and fore-slope effects are detected as well. Thus, selecting only the fore- or back-slope is sufficient for this case. A statistical analysis is applied to separate the fore- and back-slope effects. An optimum threshold for selecting the fore-slope is a third quartile of the histogram. Therefore, this threshold is used to separate the fore- and back-slop effects.

Table 1. ALOS PALSAR Data used in the Study No

Date

Orbit

Path/Row

1

Oct. 10, 2009

Asc

438/705

2

Dec. 7, 2010

Des

103/376

Figure 2 shows the SAR back-scattering intensity images at the CFZ in ascending and descending modes. The two orbits provided

The raw data of ALOS PALSAR were transformed into Single Look Complex (SLC). Then, the SLC data were quantified by a

Descending Image

Ascending Image -6.6

-6.6

Mt. Salak

-6.65

-6.8

-6.8

-6.85

-6.85

-6.9

CFZ

-6.95

Mt. Gede

-6.75

Lat

Lat

-6.7

Mt. Gede

-6.75

Mt. Salak

-6.65

-6.7

-6.9

CFZ

-6.95

-7

-7

-7.05

-7.05

-7.1

-7.1 106.4

-20

106.5

0

106.6

20

106.7 Lon 40

106.8

60

80

106.9

106.4

-20

106.5

0

106.6

20

106.7 Lon 40

106.8

60

80

106.9

100

Figure 2. SAR back-scattering intensity images of ALOS PALSAR in ascending (a) and descending (b) orbits. The colour bar indicates the intensity values in dB and the arrow indicates Line of Sight (LOS) of satellite. Mt. Salak and Mt. Gede are located in the north-northeastern part of the images and presented by black triangles. 65

Majalah Geologi Indonesia, Vol. 29 No. 2 Agustus 2014: 63-69

GSF in opposite points of satellite views. Therefore, the use of dual orbits is useful for identifying complete linear features in the SAR back-scattering intensity images. The Laplacian of Gaussian (LoG) filter was adopted for detecting the edge of the GSF. This filter has an advantage to minimize loss information and to decrease processing time (Sotak and Boyer, 1989). The LoG kernel filter was applied to the SAR back-scattering intensity image, β, as follows: x  y   2  x, y  

x

2

2

 y  2

2

6

2

e x

e x

2

2

2

 2 2

.....(1)



 y2

 2 2

y

where x and y are spatial coordinates, and σ (=0.5) is a selected constant of the Gaussian standard deviation. Finally, the Linear Feature Density (LFD) could be calculated for each window size at 10×10 pixels as follows: LFD 

w

 Ax n 1

n

, y n Q75  p ............... (2)

where w is window size, A is pixel of linear features, Q75 is a selected threshold using the Detected Linear Features

3rd quartile of the histogram produced by eq. [1], and Δp is pixel resolution. The extracted linear features in ascending and descending modes are depicted by red and blue lines in Figure 3. The addition procedure of both linear features was used as a basis for density calculation. RESULT AND DISCUSSION Figure 4a shows the LFD originated from combination of two SAR back-scattering intensity images as shown by Figure 2. This combination is useful to reduce SAR geometrical problem such as shadowing and layover effects. The colour bar indicates the density in percent grid from window size 10 × 10 pixels. The areas with 100% LFD are located at a grid which is fully filled by the linear features. On the contrary, the areas without linear features will be 0 LFD. For this studied area, the minimum and maximum linear features are located in blue and red portions, respectively. The LFD values less than 0.3% and higher than Descending

Ascending

-6.6 -6.65 -6.7 -6.75

Lat

-6.8 -6.85 -6.9 -6.95 -7 -7.05

125 km

-7.1

106.4 106.45 106.5 106.55 106.6 106.65 106.7 106.75 106.8 106.85 Lon 0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

106.9

0.9

Ascending+Descending 1

Figure 3. The detected linear features using lifedSAR was shown by white pixels (left). The shadowing and layover effects were compensated using SAR dual observation mode (right). 66

Extracting Linear Feature Density of SAR Data to Detect the Paths of a Geothermal System (A. Saepuloh et al.) ALOS PALSAR LFD

Eror Difference

LSTM LFD

-6.6

-6.6

-6.6

Mt. Salak Mt. Gede

Mt. Gede

-6.8

-6.9

-6.9

CFZ

-7

H-zone

106.4

H-zone 106.4

Lon

0

0.5 % Grid

CFZ

-7

-7.1 106.8

106.6

Mt. Gede

-6.8 -6.9

CFZ

-7

-7.1

Mt. Salak

-6.7

Lat

Lat

-6.8

Mt. Salak

-6.7

Lat

-6.7

-7.1 106.8

106.6

8 km

H-zone

Lon

1

0

0.5 % Grid

106.8

106.6

106.4

Lon

1

10

20

%

30

40

50

Figure 4. The comparison of the detected LFD using ALOS PALSAR (a), simulated SRTM 90 m (b), and their error differences (c). The hot springs were presented by white dotted arrows at the H-zone in the southern part of images (red oval).

0.6% are classified as low and high densities, respectively. To estimate the effectiveness of the SAR back-scattering intensity, the same procedure was applied to the SRTM 90 m data. The elevation values of SRTM 90 m were simulated following the ALOS PALSAR Line of Sight (LOS). Then, the simulated image was presented by shading with LOS angle as sun illumination. The same colour coding was used for a spatial mapping of the LFD values. Figure 4b shows the LFD using simulated SRTM 90 m data for the same boundary study. Comparing both data, we could see that the LFD originated from SAR back-scattering intensity provides more detail contour map than SRTM 90 m. The LFD features could be detected clearly at the CFZ using SAR back-scattering intensity, but less effective in the SRTM 90 m. This condition is supposed to be caused by the higher spatial resolution and/or sensitivity of SAR back-scattering intensity data to the surface condition. Complete differences are presented by error differences in Figure 4c. The high differences of more than 30% are located between Mt. Salak and Mt. Gede and also at northern part of Mt. Salak. Based on

the elevation map, the high error differences are located at a flat terrain with an elevation lower than 500 m. Comparing to the published geological map of Bogor Quadrangle (Effendi et al., 1998), the multi-look factor 3×6 of the ALOS PALSAR data producing a pixel spacing 28×19 m in range and azimuth directions is plausible to map the density of linear features for map scaling higher than 1:100.000. The fault and fractured zones are located at high LFD values. Moreover, the high values of LFD at N 65°E are also consistent with the field investigation of the sinistral strike-slip fault of CFZ at N 70°E (Dardji et al., 1994). The hot springs located in CFZ coincide with high LFD at the southern part termed as H-zone (Figure 4a). The fault and fractured zones presented by high LFD are interpreted as the path of the hydrothermal to the surface. It may infer that the H-zone serves as a discharge area of a geothermal system. An interesting phenomenon could be seen that the high LFD values using SAR back-scattering intensity and SRTM data continue from Mt. Salak toward south 67

Majalah Geologi Indonesia, Vol. 29 No. 2 Agustus 2014: 63-69

and then connected to the CFZ. However, this continuity is disconnected to the Hzone. On the contrary, the high LFD values around Mt. Gede are isolated and bordered by low LFD values. The localized faults and fractures might be more dominant at Mt. Gede than Mt. Salak. In addition, the LFD values around Mt. Gede from SAR back-scattering intensity data are low due to out of scene area in ascending data (Figure 2a). Based on this condition, it can be interpreted that the hot springs at H-zone might be independent to the heat source beneath Mt. Salak and Mt. Gede. Therefore, the other possibility out from volcanism should be taken into account to know the source of the geothermal system such as fault derived- thermal and/or deep reaching water through the fault system. Further investigations and measurements are necessary to correlate the subsurface structures beneath H-zone and CFZ.

as a discharge area of a geothermal system which independent to the high LFD values at Mt. Salak and Mt. Gede. Further investigation and study is necessary to know the hydrothermal origin at CFZ.

CONCLUSIONS

Carranza, E.J.M., Wibowo, H, Barritt, S.D., Sumintadireja, P., 2008. Spatial data analysis and integration for regional-scale geothermal potential mapping, West Java, Indonesia. Geothermics, 37, p.267-299.

The Linear Features Density (LFD) based on Geomorphologic and Structural Features (GSF) of SAR data could be calculated effectively using a linear feature density from Synthetic Aperture Radar (lifedSAR). A 10 × 10 grid size for multi-look factor 3×6 of ALOS PALSAR back-scattering intensity data is plausible to map the density of linear features for maps scale higher than 1:100.000. The LFD originated from ALOS PALSAR back-scattering intensity provides more detail density map than the simulated SRTM 90 m. The error differences more than 30% were located at a low elevation. The high LFD with contour pattern at N 65° E was detected, consistent to the field investigation of the sinistral strike-slip fault. The hot springs associated with Cimandiri Fault Zone (CFZ) are located at H-zone with high LFD values. This zone was interpreted 68

ACKNOWLEDGEMENTS The authors thank the National Institute of Advanced Industrial Science and Technology-Japan (AISTJapan) for providing ALOS PALSAR data under a collaborative research work with ITB. REFERENCES Abidin, H.Z., Andreas, H., Kato, K., Ito, T., Meilano, I., and Kimata, F., 2009. Crustal Deformation Studies in Java (Indonesia) Using GPS. Journal of Earthquake and Tsunami, 3 (2), p.77-88. Brogi, A., 2008. Fault zone architecture and permeability features in siliceous sedimentary rocks: Insights from the Rapolano geothermal area (Northern Apennines, Italy). Journal of Structural Geology, 30, p.237-256.

Dardji, N., Villemint, T., and Rampnoux, J.P., 1994. Paleostresses and strike-slip movement: the Cimandiri Fault Zone, West Java, Indonesia. Journal of Southeast Asian Earth Sciences, 9(1/2), p.3-11. Demirel, Z., Yildirim, T., and Burcak, M., 2004. Preliminary study on the occurrence of geothermal systems in the tectonic compressional regions: an example from the Derman geothermal field in the Biga Peninsula, Turkey. Journal of Asian Earth Sciences, 22(5), p.495-501. Effendi, A.C., Kusnama, and Hermanto, B., 1998. Geological Map of the Bogor Quadrangle, Jawa, scale 1:100.000. Geological Research and Development Centre of Indonesia, Bandung. Rosenberg, M.D., Bignall, G., and Rae, A.J., 2009. The geological framework of the Wairakei-Tauhara Geothermal System, New Zealand. Geothermics, 38 (1), p.72-84.

Extracting Linear Feature Density of SAR Data to Detect the Paths of a Geothermal System (A. Saepuloh et al.) Saepuloh, A., Koike, K., and Omura, M., 2012. Applying bayesian decision classification to PiSAR polarimetric data for detailed extraction of the geomorphologic and structural features of an active volcano. IEEE Geoscience and Remote Sensing Letters (GRSL), 99(4), p.554-558. doi.10.1109/ LGRS.2011.2174611 Saepuloh, A., Urai, M., Meilano, I., and Sumintadireja, P., 2013. Automatic extraction and validation of linear features density from ALOS PALSAR data for active faults and volcanoes. Proceeding of the International Symposium on Remote Sensing, Chiba, Japan. Serpen, U., Aksoy, N., Ongur, T., Yucel, M., and Kayan, I., 2009. Geoscientific investigations on north

of Balçova geothermal system in Turkey. International Journal of Geology, 3(4), p.87-96. Soengkono, S., 1999. Te Kopia geothermal system (New Zealand) - the relationship between its structure and extent. Geothermics, 28(6), p.767-784. Sotak, G.E. and Boyer, K.L., 1989. The Laplacianof-Gaussian kernel: a formal analysis and design procedure for fast, accurate convolution and fullframe output. Viz. Comp. Graph., Image Processing, 48, p. 147-189. Tianfeng, W., 1984, Recent tectonic stress field, active faults and geothermal fields (hot-water type) in China. Journal of Volcanology and Geothermal Research, 22, p.287-300.

69