Date post: | 07-Jan-2016 |
Category: |
Documents |
Upload: | gilangarya |
View: | 1 times |
Download: | 0 times |
7/17/2019 1-s2.0-S1878029615001073-main
http://slidepdf.com/reader/full/1-s20-s1878029615001073-main 1/5
Procedia Environmental Sciences 24 (2015) 303 – 307
1878-0296 © 2015 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license(http://creativecommons.org/licenses/by-nc-nd/4.0/ ).Selection and peer-review under responsibility of the LISAT-FSEM Symposium Committeedoi:10.1016/j.proenv.2015.03.039
Available online at www.sciencedirect.com
ScienceDirect
The 1st International Symposium on LAPAN-IPB Satellite for Food Security and EnvironmentalMonitoring
Geomorphology analysis of lava flow of Mt. Guntur in West Java
using Synthetic Aperture Radar (SAR) with fully polarimetry
Luluk D. Handayania,*, Bambang H. Trisasongko b, Boedi Tjahjono b
aCenter for Environmental Research, Bogor Agricultural University, Jl. Lingkar Akademik, Dramaga, Bogor, Indonesia 16680bSoil Science and Land Resource Departement, Bogor Agricultural University, Jl. Meranti, Dramaga Bogor, Indonesia 16680
Abstract
Mt. Guntur is one of active volcanoes in Indonesia, which has a very unique product of lava flows than other volcanoes. In order
to minimize the adverse effects of volcanic eruptions of Mt. Guntur, one mitigation measure that can be done is to map the
volcano landform through geomorphological analysis and identification of lava flow characteristics. It is very obvious that Mt.
Guntur lava flow is visible and clear so it is possible to do the landform lava flow mapping. However, this kind of study is very
rare especially in Indonesia. Thus, this research aims to analyse the geomorphologic and identify lava flows using high-resolution
optical imagery through characterization (signature) landforms by using L band SAR polarimetry backscattering in combination
with decision tree classification techniques using the QUEST algorithm (Quick, Unbiased, Efficient Statistical Trees). Moreover,
the accuracy of the classification results is calculated using matrix analysis accuracy and Kappa coefficient calculation. In
addition, the characterization of the object is also performed using analysis of spectral separation. The analysis showed that the
geomorphological analysis can be used for volcanic (V) landforms mapping based on morfocronologi aspect. The identification
of the image depends on the spatial resolution of the image. The objects were identified using IKONOS imagery, Google Earth,
and PALSAR imagery. The results of showed that the HV and VV polarization serves as the best combination to identify the lava
flows. Classification results showed a good accuracy value which is 51.80% with a Kappa coefficient of 0.43. This suggests that
the identification of a lava flow using linear polarization combined with decision tree classification has a good level of
confidence. © 2015 The Authors. Published by Elsevier B.V.
Selection and peer-review under responsibility of the LISAT-FSEM Symposium Committee.
Keywords: lava flow; backscattering; SAR; Mt. Guntur
* Corresponding author. Tel.: +62-81384780464.
E-mail address: [email protected].
© 2015 The Authors. Published by Elsevier B.V This is an open access article under the CC BY-NC-ND license
(http://creativecommons.org/licenses/by-nc-nd/4.0/ ).
Selection and peer-review under responsibility of the LISAT-FSEM Symposium Committee
7/17/2019 1-s2.0-S1878029615001073-main
http://slidepdf.com/reader/full/1-s20-s1878029615001073-main 2/5
304 Luluk D. Handayani et al. / Procedia Environmental Sciences 24 (2015) 303 – 307
1. Introduction
Indonesia is a country which is part of the world's ring of fire due to it has more than 400 volcanoes and 130 of
them were included in the category of active volcanoes. Volcano eruption is a natural phenomenon which very
harmful to the surrounding area due to the materials it produce which compose of gases, pyroclastic, and lava that
have a high temperature. Lava flow is very destructable that could damage everything on its way.. The distribution
of volcanoes in almost all region in Indonesia is a result of the tectonic plates collision that stretches from west toeast following the arc of volcano in Indonesia. Consequently, the activities of identification, field monitoring, and
mitigation of volcanic hazards require substantial time and cost. Therefore, geospatial technologies such as remote
sensing are urgently needed. A study showed that remote sensing approach is effective, efficient and costless [1].
Utilization of hyperspectral and multispectral remote sensing data has grown rapidly this recent year. However,
cloud cover is often found as an obstacle especially in tropical areas such as Indonesia. Cloud and shadow
disturbance cannot be repaired and by using masking process in the overall analysis to represents it as missing data
[2]. Therefore, SAR Radar could be used as an alternative way to overcome these problems. Radar is widely used to
obtain the natural resources data such as geographical analysis [3], surface flow pattern mapping [4], vegetation
ty pes mapping [5], mangrove forest mapping [6], the volcano mapping [7], and other applications. Recently, radar
utilization is also used to identify lava deposit trough the characterization of two objects in associated with aspect of
geomorphology [8].
SAR polarymetric can be used to determine the response of the object or backscatters using SAR with fully polarimetry imagery by the fourth polarization (HH, VV, HV, and VH). In addition, full polarization data analysis
can use polarization synthesis i.e. basic polarization signal processing technique either vertical (V) or horizontal (H)
into various derivated form of polarization (linear, elliptical, and circular) which are obtained in the phase of
modification. Therefore, it is very useful to present a graphical method in visualizing the response of a object as a
function of polarization come and backscatters. this visualization can be served as identifier polarization (signature)of an object.
Several studies using SAR with full polarization images generally produce sufficiently high classification
accuracy but often site-specific. This is due to backscatters influenced by the geometric characteristic associated
with typical surface roughness and dielectric properties of each object. Therefore, it is not necessary to use the same
method to other areas. Since 1980, optical data has been widely used for volcanic mapping. But this kind of data has
limited information due to cloud cover. This can be overcome by using radar data with SAR polarimetric band-L, so
we need a research to identify the characteristic of lava flow SAR polarimetry scattering, combined withgeomorphological analysis that produce a signature on object.
Mt. Guntur is one of active volcanoes in Indonesia that has very dominant of lava flow compared with the other
products. Lava flow on Mt. Guntur is very apparent that make the lava flow mapping is possible. However, this kind
of study is very rare in tropical countries such as Indonesia. In terms of radar data application development, it is
necessary to do geomorphological analysis using polarimetry SAR data to identify the characteristics of Mt. Guntur
lava flow.
2. Methodology
The research was conducted in Mt. Guntur area, Sinarjaya village, Garut Regency (Figure 1). This research was
conducted on September 2010 until March 2011. Primary data for this image analysis was ALOS PALSAR (Phased
Array L-band Synthetic Aperture Radar) fully polarimetric CEOS IA L1.1 with frequency band –
L (1270 MHz).Data acquired at March 30, 2009 with look angle 21,5 degree. We also used IKONOS imagery data acquired at 28thJune, 2006. The initial data processing is optic IKONOS imagery interpretation via open source Google Earth to get
landforms. Furthermore, the geocoding data used Shuttle Radar Topography Mission (SRTM) for Digital Elevation
Model (DEM) to obtain the data registered. This pre-process used ASF Mapready 2.3.6 and ENVI 4.5 softwares.
Registered data generates backscatter data consists of several Linear polarization which is VV (Vertical sent,
Vertical received) and HH (Horizontal sent, Horizontal received) which is parallel polarized, VH (Vertical sent,
Horizontal received) and HV (Horizontal sent, Vertical received) which is cross polarized. Imaging system used the
monostatic reciprocity theory where VH=HV, then used HH, HV, VV polarizations included in re, green, and blue
canal. Interpretation of ALOS PALSAR begins with filtering using JS Lee filter with 5x5 kernel size. After that,
sampling data collection were done for each landform type that is 100 pixel for training set and 75 pixel for
7/17/2019 1-s2.0-S1878029615001073-main
http://slidepdf.com/reader/full/1-s20-s1878029615001073-main 3/5
305 Luluk D. Handayani et al. / Procedia Environmental Sciences 24 (2015) 303 – 307
precision testing. Spectral separation with Transformed Divergence (TD) method were used for data analysis, while
landform classification is done using decision tree classification with Quick, Unbiased, efficient Statistical Trees
(QUEST) approach used ENVI 4.5 software by Kim and Loh [9]. In addition, we used Maximum Likelihood
classification for comparison. Analysis of ALOS PALSAR and IKONOS Imagery were then followed by field
surveys across the area in September 2010 to determine region of interest using GPS
Fig. 1. Research site
3.
Results and Discussion
SAR fully polarymetry band-L is one of the effective radar data used for the tropics area. Utilization of
backscatter method combined with decision trees classification techniques can be used to identify the signature of
lava flow in Mt. Guntur, West Java.
Identification of landforms is very dependent on the spatial resolution of the image. Object can be more identified
using IKONOS imagery (17 landform) compared with the generated from ALOS PALSAR imagery (7 landform).
This is because the spatial resolution IKONOS imagery is higher than PALSAR. However, the image can be
quantitatively measured the backscatter of objects using third linear polarization variable (Figure 2).
(a)
(b)
Fig. 2. (a) IKONOS imagery; (b) result of interpretation Mt. Guntur
7/17/2019 1-s2.0-S1878029615001073-main
http://slidepdf.com/reader/full/1-s20-s1878029615001073-main 4/5
306 Luluk D. Handayani et al. / Procedia Environmental Sciences 24 (2015) 303 – 307
The analysis of spectral separation showed that Mt. Guntur lava flows and crater landform can be separated
through spectral separation. The spectral separation can be done using Transformed Divergence Method (TD) in the
training data. As a result, crater landform has a high separation (value close to 2) and landform as well as the
youngest lava flows that can be easily identified. While landform lava flow 1, lava flow 3, and old lava flow could
not be separated as good as young lava (values close to 0).
In addition, characterization landforms could also be done through statistical separability analysis presented in
boxplot and scatter diagram (Figure 3). Figure 3 showed that the landform of youngest lava flow has the highestaverage value of HV polarization backscattering than other lava flow landform. It was could be seen that the HH and
VV polarization backscattering showed similar characteristics. Thus the use of a combination of polarization is not
particularly recommended for the identification of two types of lava flow due to parallel polarization. The best grade
separation is indicated by a combination of VV and HV.
(a) (b)
Fig 3. (a) Landforms backscattering boxplot; (b) landforms backscattering scatter diagram
(a) (b)
Fig. 4. (a) decision trees construction; (b) interpretation of ALOS PALSAR Imagery
Description:
K = Crater
AL1 = Lava flow 1
AL2 = Lava flow 2
AL3 = Lava flow 3
ALA = Youngest lava flow
ALM = Young Lava flow
ALT = Old Lava flow
7/17/2019 1-s2.0-S1878029615001073-main
http://slidepdf.com/reader/full/1-s20-s1878029615001073-main 5/5
307 Luluk D. Handayani et al. / Procedia Environmental Sciences 24 (2015) 303 – 307
(a)
(b)
Fig. 5. (a) decision tree classification, (b) maximum likelihood classification
4.
Conclusion
We have shown that the identification of lava flow landform using linear polarization in combination with
decision tree classification has a good level of confidence and relevant enough to be used as the initial identification
of lava flows signature.
References
1.
Crowley JK, BE Hubbard, JC Mars. Analysis of potensial debris flow source areas on Mount Shasta, California, by using airbone and
satellite remote sensing data. Remote Sensing of Environment 2003; 87 : 345-358.
2. Tjahjono B, AHA Syafril, DR Panuju, A Kasno, BH Trisasongko, F Heidina. Pemantauan lahan sawah menggunakan citra Alos AVNIR-2.
Jur nal Ilmiah Geomatika. 2009; 15 (2) : 1-8.
3.
Weeks R, M Smith, K Pak, A Gillespie. Inversion of SIR-C and AIRSAR data for the roughness of geological surfaces. Remote Sensing of
Environment 1997; 59 : 383-396.
4. Vieux BE, BP Bedient, E Mazroi . Real-time urban runoff simulation using radar rainfall and physics-based distributed modeling for site-
specific forecasts. Di dalam : Simulation using Radar Rainfall and Physics-Based Distributed Modeling. Proceedings of Symposium 10th
International Conference on Urban Drainage, Copenhagen.. Denmark, 1-8; 2002.
5. Simard M, GD Grandi, S Saatchi, P Mayaux. Mapping tropical coastal vegetation using JERS-1 and ERS-1 radar data with a decision tree
classifier. International Journal of Remote Sensing . 2002;23 : 1461-1474.
6. Trisasongko BH. Tropical mangrove mapping using fully-polarimetric radar data. ITB Journal of Science. 2009; 41(A) : 98-109.
7.
Weissel JK, KR Czuchiewski, Y Kim. Synthetic aperture radar (SAR)-based mapping of volcanic flows: Manam Island, Papua New Guinea.
Natural Hazard and Earth System Science. 2004; 4 : 339-346.
8.
HandayaniDWH, BH Trisasongko, B Tjahjono. 2011. Identifikasi aliran lava menggunakan metode hamburan balik radar polarimetri band L.
2011. Seminar Nasional Geomatika Pengelolaan Sumberdaya dan Penanggulangan Bencana Alam Bakosurtanal, Cibinong. 5-6 April 2011.
ISBN 978-979-26-69961
9.
Kim H, WY Loh. Classification trees with unbiased multiway split. J ournal of American Statistical Association. 2001;96 : 598-604.