Extracting linear features density of SAR data to detect the paths of a geothermal system

HAGI-IAGI Joint Convention Medan, 28-31 October 2013

1. Extracting linear features
density of SAR data to detect
the paths of a geothermal
system
A. Saepuloh
1
, M. Urai
2
, Suryantini
1
, P.
Sumintadireja
1
, I. Meilano
1
, A. H.
Harsolumakso
1
, and E. Suparka
1
1
Faculty of Earth Sciences and Technology, ITB –Indonesia
2
Institute of Geology and Geoinformation, AIST –Japan
HAGI-IAGI Joint Convention Medan, 28-31 October 2013

2. Background
oThe linear features density (LFD) related to
geological structures is a key to predict the fluid
paths of a geothermal system.
oManual tracing of the linear features is efficient
but not effective because of time consuming.
oAn automatic extraction of linear feature density
from Synthetic Aperture Radar (lifedSAR).
oInterpret the patternof LFDas the path to the
source of geothermal fluid.

3. Microwave Remote Sensing
Optic Sensor:
1.Landsat
platforms
2.ALI
3.Hyperion
4.ASTER
5.PRISM
6.AVNIR
Microwave Sensor:
1.TerraSAR-X
2.RADARSAT platforms
3.ERS platforms
4.SIR-C
5.JERS-1
6.ALOS/PALSAR
Synthetic ApertureRadar
(SAR) data part of
microwave sensorwere
used in this study.

4. Interaction of Microwave Signal to Surface
??????
??????=??????
????????????
0
??????
??????
2
??????
2
4??????
3
??????
4
The radar equation
Incident
angle
Roughness funct.
σ
0
=4??????
4
ℎ
0
2
cos
4
??????
????????????
2
??????

5. Dual SAR Observations
Range
Depression angle
Ascending Orbit Descending Orbit
Weak
Radar
ground
range
image
oDual SAR observation
were used: Ascending
satelite heading from south
toward North and
Descending in vice versa.
oProvide surface information
in two look directions for
the same object.
StrongStrong

6. Study Area
oThe Jampangarea in
West Java –Indonesia.
oThe existence of
CimandiriFault Zone
(CFZ).
oTwoactive volcanoes: Mt.
Salakand Mt. Gede.
oThe CFZhas a horizontal
displacement rate about
0.5 to 1.7 cm/yr(Abidinet
al., 2009).

7. Purpose
oExtract the linear features density (LFD) in
SAR images automatically.
oCorelatethe LFD to the geothermal fluid
paths along an active fault.
oInterpret the source of the hot springs at
the southern flank of the CFZ.

8. Data
The Phased Array type L-band Synthetic
Aperture Radar (PALSAR) onboard the
Advanced Land Observing Satellite (ALOS).
No Date Orbit Path/Row
1 Oct. 10, 2009 Ascending 438/705
2 Dec. 7, 2010 Descending103/376

9. Method
oAn automatic extraction of the linear features density
from Geomorphologic and Structural Features (GSF) in
the SARimages.
oThe raw data of ALOS PALSAR were transformed into
Single Look Complex (SLC).
oThe SLC data were quantified by a multi-look 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.
oThe multi-look image was transformed to ground range
direction based on simulated-DEM derived from the
SRTM 90 m data.

10. SAR: Geomorphologic and Structural
Features (GSF)
Geomorphologic Detection
Saepulohet al., 2012. IEEE-GRSL, Vol.
PP, No. 99.
CB
A
2008
2010
Saepuloh, et al., 2013. JVGR, Vol. 261, pp.
130-143.
Structural Detection

11. Linear Features Density SAR
(LifedSAR)
oThe Laplacianof Gaussian (LoG) filter was
adopted for detecting the edge of the GSF
oLinear Feature Density (LFD) could be
calculated for each window size()
( )
( )
( )

−
+
−
+
−+
=
xy
yx
yx
e
eyx
yx
2
22
2
22
26
2222
2
2
2
,




 ( ) 
=
=
w
n
Qnn
pyxALFD
1
75
,

12. LFD Calculation
oThe rough surface with high gradient
topography produced back-and fore-slope
effects in the SAR images.
oSelecting only the fore-or back-slope is
sufficient for this case.
oOptimum threshold for selecting the fore-
slope is a third quartile of the histogram.

13. SAR back-scattering intensity
SAR back-scattering intensity in ascending and descending modes
shows the location of Hot Springs area in CFZ, Mt. Salak, and Mt.
Gede.
Hot springs area
Fluid path ?
Fluid path ?

14. Descending Ascending
Detected Linear Features
12.5 km
Combining ascending and
descending provides broader
linear features.

15. Linear Features Density
+ = Total LFD
The LFD was calculated from the linear features by applying
high frequency filtering to remove uncorelated features.

16. Total LFD
oThecolorbarindicatedthe
densityinpercentgridfrom
windowsize10×10pixels.
otheminimumandmaximum
linearfeaturesarelocatedin
blueandredportions,
respectively.
oTheLFDvalueslessthan
0.3%andhigherthan0.6%
areclassifiedaslowand
highdensity,respectively.

17. Total LFD
oThehotspringslocated
withinCFZcoincidedwith
highLFDatsouthernpart
termedasH-zone.
oThefaultandfractured
zonespresentedbyhigh
LFDareinterpretedasthe
pathofthehydrothermalto
thesurface.
oItmayinferthattheH-zone
servesasadischargearea
ofageothermalsystem.

18. Total LFD
oThehighLFDwithcontour
patternatNW-SEandNE-
SWweredetected.
oThehighLFDvalues
continuefromMt.Salak
towardsouthandthen
connectedtotheCFZ.
oThis continuity is
disconnectedtotheH-zone.
oThelowLFDvaluesaround
Mt.Gedeareisolatedand
borderedbyhighLFD
values.

19. Discussion
oThehighLFDvaluesaroundMt.Gedeareisolatedand
discontinuetotheH-zonemightbecausedbylocalized
faults.
oThediscontinuityofhighLFDvaluesfromMt.Salak
towardH-zonewasinterpretedthatthehotspringsat
H-zonemightbeindependenttotheheatsource
beneathMt.SalakandMt.Gede.
oTheotherpossibilityoutfromvolcanismshouldbe
takenintoaccounttoknowthesourceofthe
geothermalsystemsuchasfaultderivedthermal
and/ordeepreachingwaterthroughfaultsystem.

20. Comparison
The comparison of the detected LFD using ALOS PALSAR (left), simulated SRTM
90 m (middle), and their error differences (right). The hot springs were presented
by white dotted arrows at the H-zone in southern part of the images (red oval).
The LFD originated from SAR back-scattering intensity provided more detail
contour map even though at flat terrain.

21. Conclusion
oThe Linear Features Density (LFD) relatedto GSF could be
calculated effectively using linear featuresdensity from
Synthetic Aperture Radar (lifedSAR).
oThe LFD originated from ALOS PALSAR back-scattering
intensity provided more detail density map than SRTM data
because of spatial resolution and sensitivity of back-scattering
intensity.
oThe high LFD with contour pattern at NE-SWwas detected
and consistent to the field investigation of the sinistralstrike-
slip fault.
oThe hot springs associated with CimandiriFault Zone (CFZ)
was interpreted as discharge area of a geothermal system
which is independent to the high LFD values at Mt. Salakand
Mt. Gede.

22. Acknowledgement
oThe authors would like to express the
deepest appreciationto the Graduate
Research onEarthquake and Active
Tectonics (GREAT-ITB) for fully support to
attend the JCM-2013.
oThe ALOS PALSAR used in this study
were provided by Geo-Grid AIST under
collaboration research activity.