antenna line-of-sight (slant-range) direction using synthetic aperture radar (SAR) data taken at two separate acquisition times. The D-InSAR method

Transcription

1 Abstract (English) The availability of natural disaster database is still lacking in tropical regions, since mapping natural disaster over a large area requires a great deal of time and funding. In comparison, remote sensing data are now widely employed for various monitoring purposes, but they have not yet been utilized effectively and efficiently by end users. The use of remote sensing data can contribute to mapping disaster attributes on a variety of scales ranging from community, regional and global scale. The advancement of remote sensing technology has contributed to a significant level in reducing the amount of destruction caused by natural disasters. The objective of this research is to measure the earth surface changes due to the natural disaster occurrences such as landslide, volcano surface deformation and land subsidence. The target areas of the study are in tropical region of Jeneberang Watershed in Sulawesi Island, the city of Makassar, Indonesia, and Volcanic Island of Miyakejima in Japan. Indonesia is prone to natural disaster due to its position of being squeezed geologically by three major world plates and this fact makes Indonesia one of the most dangerous countries regarding natural disasters. Local governments and responsible bodies are not able to monitor the area because of the lack of spatial information supporting the decision making regarding the land condition. The lack of detailed and accurate susceptibility maps make it generally quite difficult to evaluate the extent of area affected by floods or landslides. Thus, establishment of a comprehensive database of disaster inventory is urgently required. Miyakejima volcano is one of the most active volcanoes in Japan that currently is still producing volcanic gas which therefore is significant to monitor. The method used in this research is Differential Synthetic Aperture Radar Interferometry (DInSAR). DInSAR is a technique useful for accurately detecting the ground displacement or land deformation in the 1

2 antenna line-of-sight (slant-range) direction using synthetic aperture radar (SAR) data taken at two separate acquisition times. The D-InSAR method is complementary to ground-based methods such as leveling and global positioning system (GPS) measurements, yielding information in a wide coverage area even when the area is inaccessible. This dissertation explains the ability and advantages of DInSAR as an efficient and cost-effective method compared to conventional ground survey for disaster monitoring in tropical area, especially landslides, volcanic surface deformation and related phenomena caused by natural disaster and human activities. Japanese Earth Resource Satellite (JERS-1) SAR and ALOS PALSAR images were used to generate DInSAR data in measuring the surface changes. The result of the image processing indicates slight movement of slope in Jeneberang Landslide area prior to the landslide event. This is confirmed with the crack measured in the field campaign. The continuous collapse of the caldera rim of Miyakejima Volcano was also picked by the image measurement validated by GPS measurement in the field. In particular sites of Makassar city, land subsidence of as much as 15 cm was shown from the analysis of detailed DInSAR processing. In summary, this study has demonstrated the capability of the DInSAR method in monitoring earth surface deformation for landslide, volcano as well as urban areas with a wide spatial coverage. The accuracy of the deformation detected can further be improved by increasing the data availability with shorter revisit cycle, more accurate geodetic data and better performance of processing software. 2

3 Abstract (Japanese) JERS-1 と ALOS PALSAR 合成開口レーダデータを用いた差分干渉解析によ る自然災害モニタリング リモートセンシングを用いることにより 地域から全球に至るさまざまなスケールにおいて災害情報をマッピングすることが可能になっている 本研究では 地すべり 火山噴火といった自然災害や 多くの場合に人為的にもたらされる地盤沈下の被害を減少させることを目指し リモートセンシングの応用がとくに遅れている熱帯地域において地表面変化を計測した インドネシアはその気候や環太平洋のプレート運動により 多くの自然災害を受けやすい熱帯地域に位置している ここでは スラウェシ島 Jeneberang における大規模な地すべりと Makassar における地盤沈下という二つの特徴的な災害を 合成開口レーダを用いた差分干渉法 (DInSAR) により解析した この方法は 2 時期に得られた SAR データの干渉画像からアンテナ視線方向の微小な地表面変化を検出する手法である GPS を用いた地上での計測に比べて 立ち入りが危険な地域であっても広域にデータ取得が可能である点に特徴がある JERS-1 と ALOS PALSAR データの解析から Jeneberang の地すべりでは イベントの発生前に斜面の移動が起こっていることが確認され これはクラックの現地調査の結果と一致した Makassar の都市域の解析では いくつかの地点において最大...cm におよぶ地盤沈下が明らかになった また 同様の手法を三宅島の雄山に適用し DInSAR 解析から得られる火口付近の崩落が現地の GPS 観測とよく一致することを見出した 本研究では以上のように DInSAR の手法が災害のモニタリングに有効な手法であることを実証した 3

4 Acknowledgement First of all, I would like to thank my supervisor, Professor Hiroaki Kuze, and my co-supervisor Associate Professor Josaphat Tetuko Sri Sumantyo for their continuous support, guidance and direction throughout my study. Their support and direction allowed me to learn and acquire so much knowledge and opportunity to international research environment. I would also like to express my gratitude to Prof Kazumi Hattori and fellow Dr. Ippei Harada for the support data and guidance given for Miyakejima Volcano research I belong academically to Laboratory of Sensor headed by Prof Kuze but also informally active in in Josaphat Microwave Remote Sensing Laboratory (MRSL) at Center for Environmental Remote Sensing Chiba University, therefore I also would like to thank you all those people in this 2 laboratories for their discussions and supports during my study. Especially to Mr. Bannu, Miyazawa san, Kaba san, Goto san, Saito san, Mr. Yuhendra and 2 most helpful persons during my study, Mr. Luhur Bayuaji and Mr. Bambang Setiadi for all his support and learning guidance. Institutionally, I would like to thank you Japan Aerospace Agency (JAXA) through Dr. Shimada for the cooperation of using SIGMASAR software that I was able to learn DInSAR processing and Hasanuddin University for the scholarship grant. Lastly I would extend my love and gratitude for my wife, Serliah Nur for all her support and love, to my son Zuhry for taking care of me in Chigusadai while studying in his Japanese Elementary School, my daughter, Faizah Almas Cantika and beloved Muhammad Zikry for their support and patience. To my parents and brothers and sisters in Indonesia for their encouragement and continuous supplication at any time. 4

5 Table of Contents Abstract (English)... 1 Abstract (Japanese)... 3 Acknowledgement... 4 Table of Contents... 5 Chapter One: Introduction Background Natural Disaster Management Japanese Earth Resource Satellite (JERS-1) ALOS PALSAR Research Objectives Structure of the dissertation Chapter Two: Background Theory Synthetic Aperture Radar Interferometry The Principle of a Radar Interferometer Radar interferogram and DEM Radar interferometry processing Data Selection Image coregistration Interferogram Coherence Flat earth fringes removal

6 2.4.6 Phase unwrapping Phase to Height Geocode Decorrelation and errors Decorrelation Spatial Decorrelation Temporal Decorrelation Volume Decorrelation Thermal Noise and Processing error Atmospheric effect Orbit error Geometric error Chapter Three: Methodology Basics of DInSAR Differential InSAR Chapter Four: Results and Discussion Mapping tropical landslides Study Site Dataset used Results and Discussion Conclusion Monitoring Miyakejima Volcano deformation Introduction

7 4.2.2 Datasets Used Result and Discussion Conclusion and further work Measuring Land subsidence of Makassar City Study area Data used Result and Discussion Conclusion Chapter Five: Conclusions and recommendations Conclusions Recommendations and future research References Publication Lists

8 Chapter One: Introduction 1.1 Background The term natural disaster has been brought as a common vocabulary to our daily life as news bringing information on natural disaster is occurring in any parts of our planet recently. Planet Earth provides us with the air, food, warmth, and materials we need to thrive. But Earth can also generate catastrophic disasters, from tsunamis and landslides to tornadoes and wildfires, that kill people, damage the environment, destroy property, and disrupt normal life. Such disasters may be sudden and violent, like an earthquake or flood, or gradual, like drought or the spread of a deadly disease. Today, scientists have shown that many such disasters are caused by the natural workings of our planet. There are more than 700 natural disasters every year, affecting around one person in 30. As definition, a natural disaster is the effect of a natural hazard such as flood, tornado, hurricane, volcanic eruption, earthquake, landslide, or even heatwave. It leads to financial, environmental or human losses. The resulting loss depends on the vulnerability of the affected population to resist the hazard. The term natural has consequently been disputed because the events simply are not hazards or disasters without human involvement. Many natural hazards are interrelated, e.g. earthquakes can cause tsunamis and drought can lead directly to famine. It is possible that some natural hazards are correlated, as well. A concrete example of the division between a natural hazard and a natural disaster is that the 2011 Tohoku Great Earthquake and Tsunami was a disaster, whereas earthquakes and tsunami are hazards. 8

9 1.2 Natural Disaster Management It is undeniable that recently these natural disasters occurred more frequently than ever before. This is also due to the access of information that people can receive other than the dynamic of our earth that changes as nature. Comprehending the nature of these disasters, we are hoping to be able to control what we can related to any negative impacts it may bring or at least we can mitigate and reduce the risk of the disaster. This is the term disaster management plays it roles. Therefore we need tools to be able to manage disasters. Disasters, both natural and man-made, are increasing in their frequency and catastrophic impact in the world specially in Indonesia. Earthquake, flood, landside, subsidence are only a few to mention (Tralli et.al, 2005., 2008; Chini et al., 2008; Abidin et al., 2009; U.S. Geological Survey, 2006; Mazzini et al., 2007; Smith et al., 2009). Disaster, in the true sense, always occur involving human beings, especially as victims (van Landewijk, 1994). Due to the lack or inadequate disaster prevention and mitigation strategies in Indonesia, the impact of disasters which they do occur is much greater than in other developed countries. Disaster detection and monitoring are essential aspect in disaster mitigation and it s conventionally done by ground survey. Nowadays, remote sensing, both in optical and radar sensor, plays important role on disaster mitigation. Satellite images are used for prevention and preparedness, to assess the scope of potential disasters and help mitigate damage that could occur. Upon occurrence of a disaster, the images are used to determine the extent of the damage and the location of refuge areas. In post-disaster management, they are used for reconstruction and sustainable development activities. Natural disaster management requires data acquired by remote sensing satellites. Information about potential dangers, such as climate, extreme weather, earthquakes, landslides, tsunami, volcanoes, and forest 9

10 and land fires, is most relevant. By combining the satellite data with other information, Indonesia will be able to contribute to solving certain social problems, including disasters. Remotely sensed data play an integral role in reconstructing the recent history of the land surface and in predicting hazards due to flood and landslide events. Satellite data are addressing diverse observational requirements that are imposed by the need for surface, subsurface and hydrologic characterization, including the delineation of flood and landslide zones for risk assessments. Short- and long-term sea-level change and the impact of ocean-atmosphere processes on the coastal land environment, through flooding, erosion and storm surge for example, define further requirements for hazard monitoring and mitigation planning. The continued development and application of a broad spectrum of satellite remote sensing systems and attendant data management infrastructure will contribute needed baseline and time series data, as part of an integrated global observation strategy that includes airborne and in situ measurements of the solid Earth. (Tralli, et.al, 2005). Among the many types of disasters, there are 3 natural disasters are approached by using satellite remote sensing image processing technology, landslides, volcano deformation and land subsidence. Landslides are now also popular due to its occurrences not only in mountainous areas but also in cities which bordered with weak supported slope triggered by high intensity of rainfall. Landslides usually damage or block the road when it relatively small debris flow but can also catastrophic when it brings huge amount of debris downslope. Mountainous areas specifically vulnerable to sediment related disaster. High slopes with loose material support are areas fond of landslides incidence. Indonesia is the second highest fatalities according to International Landslides Committee, ILC, in 2004 amount to 441 fatalities after China with 449 (ILC, 2004). Recent database created in 2009 can be seen in figure 1. 10

11 Situated on the belt where rim of the tectonic edge elongate, term as ring of fire, Indonesia together with Japan have the biggest percentage of volcanoes. Therefore study about volcano has been a major research. One specific focus of this dissertation, it also research on the use of DInSAR to investigate one most active volcano in Japan in Miyakejima, a volcano just 180 kms south of Tokyo the capital of Japan. Land subsidence has detected in several major cities in Indonesia, also in the world and become major constrain to the development (Cascini et al., 2006; Marfai and King, 2007; Abidin, 2008; Catani et al., 2005; Chatterjee et al., 2006). It can turn into disaster by damaging infrastructures; result in loss of homes, cause injury or loss of life. There are many factors causing land subsidence, naturally or man-made. Conventionally, detecting and monitoring land subsidence are done by ground survey which cost a lot of time and budget, sometime endanger the surveyor. Remote sensing technology was demonstrated as alternative method. As for tropical region (high humidity and precipitation) such as Indonesia, the all-weather ability of SAR becomes great advantage as tool to monitor disaster corresponding to optical data. Synthetic Aperture Radar (SAR) has known as a precious data source for disaster monitoring (Colesanti et al., 2006; Nicoll and Gens, 2005; Kimura and Yaguchi, 2000). The all-weather operation advantage of SAR data has shown in numerous studies to help monitor fires, hurricanes, landslides, oil spill and earthquakes. Fusion between SAR and optical data has been applied extensively for flood monitoring (Chen et al., 2003). Interferometry SAR (InSAR) technology makes study on disaster monitoring more comprehensive on earthquake and volcanic eruption. Advance in Differential Interferometry SAR (DInSAR), further improvement of InSAR technology, makes the SAR ability on disaster monitoring far more useful. DInSAR technology has ability to detect earth surface deformation, leading 11

12 the ability to detect the glacier movement, land subsidence and volcanic eruption. In the last two decades, the usage of interferometry of Synthetic Aperture Radar (SAR) using satellite-based remote sensing has confirmed its ability for measuring the shape earth terrain and changing in the earth surface. Interferometric SAR (InSAR) is a technique to generate high quality digital elevation maps (DEMs) using pairs of high resolution SAR images. Pair of SAR images can be obtain at the same time using two separated antennas mounted in the same platform (referred as single-pass InSAR) or different observation time using one antenna (referred ad repeated-pass InSAR). When ground deformation occurred during repeatedpass InSAR, the amount of deformation in radar look direction (known as line-of-sight or range-direction ) can be measured using two SAR images and additional topographic information of the site. This technique is known as Differential InSAR (DInSAR). DInSAR is able to provide ground surface deformation at sub-centimeter accuracy. Detail information about SAR imaging and interferometry SAR are presented in Chapter Japanese Earth Resource Satellite (JERS-1) The Japanese Earth Resources Satellite-1 (JERS-1) was a joint project between the Japan Aerospace Exploration Agency (JAXA) and the Ministry of International Trade and Industry (MITI). It was launched in February 1992 and ceased operation on 11 October 1998, four years more than the original two year mission plan. It carried three instruments. a) An L-band (HH polarization) synthetic aperture radar (SAR); b) A nadir-pointing optical camera (OPS); c) A side-looking optical camera (AVNIR). The optical sensors collected information from eight spectral bands, while the SAR sensor operated in the L-band of the microwave wavelengths. 12

13 JERS -1 SAR Characteristics are frequency 1.3 GHz, Off Nadir Angle is 35 degrees, ground resolution of 18 meters with swath width 75 kms and HH (Horisontal Horisontal) polarization. The mission of JERS-1 was to last only two years, but in the end it was possible to obtain observation data on the earth (with regard to resources, disasters, the environment etc.) from the satellite for approximately six and a half years. Some representative examples of the research results achieved through use of these data are the retrieval of potential oil mine in Turpan basin of China, the clarifying of the reality of deforestation in the Amazon tropical rainforest and of the situation of diastrophism (deformation of the earth's crust) caused by the activity of the volcano Mt. Iwateyama. It has been proven that JERS-1 can be utilized to monitor volcano deformation which therefore having the availability of archive data can be used to invetstigate geological and tectonic condition of coverage area. 1.4 ALOS PALSAR The Japanese Earth observing satellite program consists of two series: those satellites used mainly for atmospheric and marine observation, and those used mainly for land observation. The Advanced Land Observing Satellite (ALOS) follows the Japanese Earth Resources Satellite-1 (JERS-1) and Advanced Earth Observing Satellite (ADEOS) and will utilize advanced land-observing technology. ALOS will be used for cartography, regional observation, disaster monitoring, and resource surveying. The Advanced Land Observing Satellite (ALOS) was launched with an H-IIA rocket from the Japanese Space Exploration Agency s (JAXA) Tanegashima Space Center on January 24, ALOS is the largest satellite developed in Japan. ALOS carries three remote-sensing instruments: Panchromatic Remote-sensing Instrument for Stereo Mapping 13

14 (PRISM), Advanced Visible and Near-Infrared Radiometer type 2 (AVNIR-2), and the Phased Array L-band Synthetic Aperture Radar (PALSAR). Detail about ALOS characteristic listed in Table 1.1. Figure 1.1 ALOS-PALSAR observation mode PALSAR is an enhanced version of the JERS-1 SAR, developed jointly by JAXA and the Ministry of Economy, Trade and Industry (METI). It is a fully polarimetric instrument, which operates in L-band with MHz (23.6 cm) center frequency and 14- and 28-MHz bandwidths. The antenna consists of 80 T/R modules on four panel segments, with a total size of m. shows a time sequence of the actual deployment of the PALSAR antenna in space, recorded by one of the onboard monitoring cameras mounted on the satellite body. PALSAR can be operated in five different observation modes: Fine Beam Single polarization (FBS), Fine Beam Dual polarization (FBD), Polarimetric mode (POL), ScanSAR mode, and Direct Transmission (DT) mode. Figure 1.1 shows illustration of the 14

15 ALOS-PALSAR observation mode. Table 1.1 and 1.2 shows the detail characteristics of ALOS PALSAR. Table 1.1 ALOS characteristics Item Characteristic Launch 24 January 2006 Orbit Sun-Synchronous Sub-Recurrent Altitude km (at Equator) Inclination deg. Recurrence Cycle Repeat Cycle: 46 days Sub Cycle: 2 days Orbital position accuracy 0.78 m Data rate 240Mbps (via Data Relay Technology Satellite) 120Mbps (Direct Transmission) Total weight at launch 4000 kg 1.5 Research Objectives In general, this research attempted to investigate the capability of Differential Interferometric Synthetic Aperture Radar (DInSAR) technique using Japanese L-Band SAR sensors satellite data to measure any changes occurred on earth surface hence can be used for disaster monitoring. The DInSAR is used for three types of disaster namely landslide, volcano deformation and land subsidence. In particular, this research is aimed to obtain one Geographic Information System GIS) layer from the DInSAR image processing result as an input for spatial analysis of landslide susceptibility map. 15

16 Table 1.2 Characteristic of ALOS PALSAR instrument Mode Fine ScanSAR Polarimetric Center Frequency 1270 MHz (L-band) Bandwidth 28 MHz 14 MHz 14,28 MHz 14 MHz Polarization HH+HV HH or or VV VV+VH HH or VV HH+HV+VH+VV Incidence 8 ~ 60 8 ~ ~ 43 angle degree deg deg 8 ~ 30 deg Range 14 ~ 88 7 ~ 44 m resolution m 100 m 24 ~ 89 m Swath 40 ~ ~ ~ 350 km km km 20 ~ 65 km Quantization 5 bits 5 bits 5 bits 3 or 5 bits Data rate , 240 Mbps Mbps Mbps 240 Mbps The integration of optical remote sensing data with SAR data along with ground measurement using GPS are expected to give good validation and compelentary data to achievement better result. There are number of sources of disturbance or noises in DInSAR processing, such as decorrelation and atmospheric delay. Decorrelation characteristic is specific accordance to the study site. It includes spatial, temporal and volume decorrelation. The level of decorrelation differs depending upon many factors such as slope of terrain, the local weather conditions, type of land cover, etc. The selection of suitable interferometric pairs is limited by temporal separations of the acquired SAR images as well as the characteristics of the site. 16

17 The changes happen on the earth`s surface are expected to be measured to give an accurate trend of the changes so we can predict and mitigate the possible measure of disaster in the future. The research output is expected to assist diasaster management planner to prepare disaster susceptibility map with the aid of Geographic Information Systems (GIS) specially the spatial analysis. 1.6 Structure of the dissertation This dissertation is organized in this following manner. Chapter 2 contains the history of SAR imaging and radar interferometry, as well as basic theory of DInSAR process. The noise and decorrelation that gives an effect to the DInSAR result will be included in this chapter. Chapter 3 contains the methodology of this study and brief explanation about satellite data will be discussed in this chapter. Chapter 4 focuses on the results and discussions. This result will be arranged according to types of natural disaster with Jeneberang Landslide, Miyakejima Volcano Deformation and land subsidence of Makassar City. Chapter 5 provides summary and conclusions. 17

18 Chapter Two: Background Theory 2.1 Synthetic Aperture Radar Interferometry The usage of synthetic aperture radar technique to obtain fine resolution measurement in two dimensions and interferometry to obtain the third measurement was first introduced by Graham (1974). Currently the interferometric synthetic aperture radar (InSAR) technology has been developed and used successfully for topographic mapping and measurement of terrain deformation caused by earthquakes, subsidence, volcano deflation and glacial flow (Zebker and Goldstein, 1986; Gabriel, Goldstein and Zebker, 1989; Zebker et al., 1994a, 1994b; Massonnet and Adragna, 1993; Massonnet et al., 1993, 1994; Massonnet, Briole and Arnaud, 1995; Goldstein et al., 1993). An intuitive introduction to InSAR theory and its applications can be found in Gens and van Genderen (1996), Rocca et al. (2000) and Burgmann et al (2000). Differential InSAR (DInSAR) is the further development of InSAR technology which allows the usage of radar interferometry to measure small-scale movement in vertical direction. Gabriel et al. (1989) describe the DInSAR technique for mapping small elevation changes over large areas where SAR images was used to measure very small (1 cm or less) surface motions with good resolution (10 m) over large swath (50 km). DInSAR was used in various activities including volcanic activities, earthquakes and mining subsidence and land subsidence observation in populated area (Delaney and McTigue, 1994; Kobayashi et al., 1999; Tsuji et al., 2009; Yen et al., 2008; Berthier et al., 2007; Colesanti et al., 2005; Cascini et al., 2006; Chini et al., 2008; Abidin et al., 2005; Guoqinga and Jingqin, 2008) 18

19 2.2 The Principle of a Radar Interferometer Many remote sensing applications use multi-look SAR images as its data source. Multi-look SAR image represent image of averaged intensity (or amplitude) of multiple radar looks to reduce the speckles often exhibited in radar images. Basically the multi-look image contains the average intensity of several single-look complex (SLC) images which represents all the information from the return radar signals. SLC image record the pixel in complex format which consist of not only the intensity but also the phase of the signal. This phase information is determined by the distance between the target and the radar antenna. For a complex number of an SLC pixel, the magnitude of c is, which expresses the SAR intensity image, while the phase angle of c is. InSAR uses phase information which commonly ignored in SAR SLC images for Earth and planetary observations. SAR interferogram shows the phase differences between the corresponding pixels of the same object in two SAR images taken from nearrepeat orbits. It represents topography as fringes of interference. A radar beam is nominally a single frequency electromagnetic wave. Its properties are similar to those of monochromatic coherent light. When two nearly parallel beams of coherent light illuminate the same surface, an interferogram can be generated showing the phase shift induced by the variation of position and topography of the surface, as a result of the interference between the two beams. The same principle applies to the return radar signals. An SAR interferometer acquires two SLC images of the same scene with the antenna separated by a distance B called the baseline. 19

20 Figure 2.1 (a) The single pass SAR interferometer with both an active antenna, sending and receiving radar signals, and a passive antenna (separated by a distance B ) for receiving signals only. (b) A repeat-pass SAR interferometer to image the same area from two visits with a minor orbital drift B. For a single pass SAR interferometer, such as an SAR interferometer onboard an aircraft or the Space Shuttle, two images are acquired simultaneously via two separate antennas: one sends and receives the signals while the other receives only (Figure 2.1a). In contrast, a repeat pass SAR interferometer acquires a single image of the same area twice from two separate orbits with minor drift which forms the baseline B (Figure 2.1b); this is the case for ERS-1 and ERS-2 SAR, ENVISAT ASAR, RADARSAT and ALOS PALSAR. The purpose of InSAR is to derive an SAR interferogram which is the phase difference between the two coherent SLC images (often called fringe pair). Firstly, the two SLC images are precisely co-registered pixel by pixel at sub-pixel accuracy based on local correlation in combination with embedded position data in SAR SLCs. The phase difference between the two corresponding pixels is then calculated from the phase angles and of these two pixels through their complex numbers: (Equation 2.1) 20

21 Figure 2.2 The geometry of radar interferometry To understand the relationship between phase difference and the InSAR imaging geometry, let us consider an SAR system observing the same ground swath from two positions, A1 and A2, as illustrated in Figure 2.2. The ground point C is then observed twice from distance r (slant range) and r+. The distance difference between the return radar signals for a round-trip is 2 and the measured phase difference (interferogram) is (Equation 2.2) or 2 time s the round-trip difference, 2, in radar wavelength. From the triangle A1 A 2 C in Figure 2.2, the cosine theorem provides a solution for in terms of the imaging geometry as follows: [ ] or (Equation 2.3) where B is the baseline length, r the radar slant range to a point on the ground, the SAR look angle, and the angle of the baseline with respect to the horizontal at the sensor. The baseline B can be divided into two components which are perpendicular B and parallel B // to the look direction: 21

22 (Equation 2.4) (Equation 2.5) The InSAR data processing precisely calculates the phase difference between conforming pixels between a fringe pair of SLC SAR images to produce an interferogram. The applications of InSAR are largely based on the relationships between the interferogram, topography and terrain deformation, for which the baseline B, especially the perpendicular baseline B, plays a key role. 2.3 Radar interferogram and DEM One major application of InSAR is to generate a DEM (Digital Elevation Model). It is clear from Figure 2.2 that the elevation of the measured point C can be defined as (Equation 2.6) Where h is the height of the sensor above the reference surface (datum). This formula looks simple but the exact look angle is not directly known from the SLC images. We have to find these unknowns from the data which InSAR provides. From the SAR interferogram, we can express d by rearranging (2.2) as (Equation 2.7) Modifying (2.3) as a sine function of, (Equation 2.8) In this equation, the baseline B and slant range r are known and constants for both entire fringe pair images, while the only variable can be easily calculated from phase difference (SAR interferogram) using Equation (2.7). Thus sin( ) is resolved. Expressing cos In equation 2.6 as a function of and sin( ), 22

23 (Equation 2.9) In Equation (2.9), the angle of the baseline with respect to the horizontal at the sensor,,is a constant for the SAR fringe pair images and is determined by the imaging status, whereas sin( ) can be derived from the interferogram using Equations (2.7) and (2.8), and the elevation z can therefore be resolved. In principle, we can measure the phase difference at each point in an image and apply the above three equations based on our knowledge of imaging geometry to produce the elevation data. There is, however, a problem in this: the InSAR measured phase difference is a variable in the 2 period, or is 2 wrapped. Figure 2.3 shows an interferogram generated from a fringe pair of ERS-2 SAR images; the fringe patterns are like contour lines representing the mountain terrain but, numerically, these fringes occur in repeating 2 cycles and do not give the actual phase differences which could be n times 2 plus the InSAR measured phase difference. The phase information is recorded in the SAR data as complex numbers and only the principal phase values ( p) within 2 can be derived. The actual phase difference should therefore be (Equation 2.10) Expressed in terms of the slant range difference, (Equation 2.11) The interferometric phase therefore needs to be unwrapped to remove the modulo-2 ambiguity so as to generate DEM data. For a perfect interferogram modulo-2, unwrapping can be achieved accurately via a spatial-searching-based scheme but the various decorrelation factors mean 23

24 that SAR interferograms are commonly noisy. In such cases, unwrapping is an ill-portrayed problem. There are many well-established techniques for the unwrap-ping of noisy InSAR interferograms, each with its own merits and weaknesses, but the search for better techniques continues. There are also other corrections necessary, such as the removal of the ramps caused by the Earth s curvature and by the direction angle between the two paths as they are usually not actually parallel. 2.4 Radar interferometry processing Data Selection Interferometric pair can be formed by using two radar images. In order to get good result (low phase noise), short spatial and temporal separations are preferred. Also, the use of SAR images acquired during or after rain should be avoided as the dielectric property of the ground is different from the images acquired in drier conditions. Figure 2.3 shows an example of intensity image of SAR data Image coregistration In order to compute the phase difference between mater and slave SAR images, the two images have to be overlaid precisely at accuracy within a small fraction of a pixel. Massonnet and Feigl (1998) commented that there are three steps required for SAR image coregistration: Evaluating the geometric differences between the master and slave images. Conventional correlation of amplitude image patches can achieve accuracy near 0.03 pixels (Li and Goldstein, 1990). Modeling the geometric differences. A simple way is to use a least squares adjustment to approximate the distortion; Resampling the slave image into the master s geometry. It is commonly done by using bilinear (Lin et al., 1992) or bicubic (Leong Keong et al., 1994) resampling methods. 24

25 2.4.3 Interferogram The phase difference (interferogram) between the two images can be generated after the two SAR images have been overlaid precisely. The interferogram is calculated using the complex master multiplied by the conjugated complex slave. We can average over neighboring pixels to improve the signal-to-noise ratio, in a process called complex multilooking. An additional advantage of this filtering step is to obtain a square shape for the final pixel. In the interferogram, the parallel fringes along the range direction are caused by the constant height of the terrain. These parallel fringes are also referred to as flatearth fringes. Figure 2.4 shows the example of interferogram of area in Figure 2.3. ( Bayuaji,2010) Coherence Figure 2.3 Intensity Image of SAR data The signal in each pixel is the summation of the coherence addition of the elementary targets within the pixel. Coherence is another quantity to be estimated. It is a measure of local interferogram quality and is thus needed 25

26 in many interferometric signal processing steps. The coherence is the complex correlation coefficient of the two SAR images and can be calculated using (Equation Figure 2.5 shows the SAR image pair coherence of Figure Flat earth fringes removal E[ S S ] * E[ S ] E[ S ] 1 2 (Equation 2.12) Due to the geometry of InSAR system, there is a phase trend in range direction in the interferogram even if the terrain is ideally flat. This phase trend is commonly referred to as flat-earth fringes. The flatearth fringes are often subtracted from the interferogram before further processing. The resulting interferogram, after removing the flatearth fringes, is sometimes called a flattened interferogram (Rosen et al., 2000). Figure 2.6 shows the flattened interferogram of Figure 2.4. Figure 2.4 Interogram consisting flat earth fringes in slant range direction 26

27 2.4.6 Phase unwrapping The interferogram shows the phase value in the range between 0 and 2π. This discrete phase is known as "wrapped phase", which is ambiguous by integer multiples of 2π. The phase unwrapping technique is applied to compute the continues phase difference so the 2π ambiguity inherent in the phase measurements can be solved. Figure 2.6 shows the idea of phase unwrapping process. Figure 2.5 Coherence image of SAR image pair in slant range direction 27

28 Figure 2.5 Flattened interogram Figure 2.6 Phase unwrapping process Once the absolute phase of each pixel of the interferogram is known, the height or deformation of the terrain can be computed. The commonly used phase unwrapping techniques are path-following methods, e.g. Branch- Cut algorithm by Goldstein and Werner (1998), and Minimum-Norm methods, e.g. Least- Squares estimation algorithm (Takajo and Takahashi, 1988; Eineder et al., 1998). Figure 2.7 shows the result of phase unwrapping process of Figure

29 Figure 2.7 Unwrapped phase image Phase to Height After unwrapping the phase of the interferogram, the height of the terrain can be computed using Equation (2.13); or the displacement of the land surface along the line-of-sight of the radar can be calculated using Equation (2.6). Radar interferometry measures the relative height differences between the pixels. Thus, a tie-point is required to get the absolute height information Geocode Generally SAR image is generated by recording the backscattered signals in range-azimuth coordinates from radar viewpoint, this projection called slant-range projection. Until now, the processing are taken in slantrange projection. In order to make the radar interferometry results more having an important effect, they should be re-projected into a standard geographic coordinate system. The process of converting the image from slant-range projection into a standard geographic coordinate system is called geocoding or georeferencing. After geocoding, the results can be 29

30 further interpreted against other spatial data. Figure 2.8 shows the geocoded image of Figure 2.7. Figure 2.8 Geocoded of unwrapped phase image 2.5 Decorrelation and errors As described in previous section, the radar interferometry process exploits the phase information between two SAR images in order to retrieve the terrain and height information. The result of radar interferometry result may consist of phase information from many sources such as topography, land surface displacement, atmospheric heterogeneity, noises, and so on. It is represented in the interferogram result. Some of these phase components may be more dominant than others depending on the modes and characteristics of the interferometric pair used. For example, the phases due to land displacement and atmospheric heterogeneity are considered negligible in single-pass radar interferometry systems, but may be significant in repeat-pass systems. 30

31 The phase noises and errors degrade the accuracy of radar interferometry. The errors include orbit and geometric errors. There are other phase disturbances which are generally considered as error sources, such as atmospheric effect and topographic residuals if DInSAR is performed. Therefore, the quality of radar interferometry is based on whether the phase noise and the residuals of the unwanted phase components can be eliminated properly. The interferometric quality of an identical imaged pixel in two SAR images can be assessed by measuring the correlation between the pixels. The amount of correlation, or so-called coherence, can be considered as a direct measure for the similarity of the dielectric properties of the same imaged pixels between two SAR acquisitions. The backscattered signal from each pixel area is the summation of the coherence addition of the elementary targets within the pixel. The formula for calculating the coherence is listed Equation A coherence value of 1 means the two pixels are totally correlated and there is no phase noise presented; a coherence value of 0 means two pixels are totally decorrelated and there is only phase noise presented. In other words, decorrelation leads to low coherence and high phase noise in images Decorrelation Decorrelation occurs according to the study site area. It depends upon many factors such as vegetation cover, local climate conditions, complexity of the terrain, land use, etc. As a rule of thumb, decorrelation is more severe for an area having heavily vegetated cover, complex terrain or constantly changing climate conditions. However, high coherence can still be preserved over 4 years in desert due to low levels of precipitation and sparse vegetation (Fialko et al.). Over a vegetated area, The L-band signal has higher correlation than C-band (shorther wavelength). This is because the wavelength of C-band is more comparable to the size of tree leaves. 31

32 Decorrelation can be classified into three main categories: spatial, temporal and volume decorrelation. They can be presented as (Zebker and Villasenor, 1992): Spatial Decorrelation SNR spatial (Equation 2.13) temporal Spatial decorrelation is caused by the physical separation between the locations of the two SAR antennas. This physical separation is commonly referred to as the baseline, which has been illustrated in Figure 2.2. It is also referred to as the baseline decorrelation error (Gatelli et al., 1994) Temporal Decorrelation Temporal decorrelation is caused by the variation of the dielectric properties of ground objects over time between two repeat-pass acquisitions. In repeat-pass interferometry the sub-resolution properties of the imaged scatterers (sub-scatterers) may change between surveys, e.g. by the movement of leaves and branches, water surface, or vegetation growth. If the random scattering phase contributions 1scattering and 2scattering shown in Equations (2.4) and (2.5) in the two images no longer cancel each other out, the two SAR images start to decorrelate Volume Decorrelation Volume decorrelation is also about the phase stability of the imaging cell (pixel) over time, especially over vegetated regions. The dielectric properties and alignments of tree leaves and branches in the canopy may result in the variation of the backscattered radar signals. They are also affected easily by the local weather conditions, such as rain and wind. 32

33 Therefore, it is desirable to have a finer SAR imaging resolution in order to minimize the variation of the random scattering phases Thermal Noise and Processing error Besides the noises caused by decorrelation, there are other phase noises, such as system thermal noise, sampling and processing artifacts, and statistical correlation of the individual radar echoes before they are combined to form interferograms (Zebker and Rosen, 1994) Thermal noise is minimised by using the greatest possible transmitter power and lowest noise receivers. Sampling and processing artifacts are a trade-off between data system complexity and cost (Zebker and Rosen, 1994) Atmospheric effect Interferometric phases vary with the distance of the path between the radar antenna and ground objects. Electromagnetic radiation travels at a slightly different speed in mediums with different values of refractive index. So, when the microwave signal of radar travels through the atmosphere the variation of the density of electrons in the ionosphere (Massonnet and Feigl, 1998) or water vapour in the troposphere, causes a change in the refractive index of the atmosphere. As a result, the radar signals may travel at different speeds through the atmosphere at different acquisitions. It is shown as phase variation in interferograms even for a flat terrain where no land deformation occurs between the two radar acquisitions Orbit error A horizontal or vertical shift of the entire interferometer will cause the same shift of the reconstructed digital elevation model. Incorrect orbit information of the SAR platform may rotate the baseline between the interferometric pair. 33

34 For SAR interferometry, along-track errors are usually accounted for during the coregistration of the two SAR images (Hanssen, 2001). They can also be regarded as timing errors (Massonnet and Vadon, 1995). The acrosstrack error causes orbital fringes parallel to the flight path or azimuth direction. The fringes perpendicular to the azimuth direction, i.e. parallel to the range direction, are the result of an incorrect estimation of the radial and across-track velocities (Kohlhase et al., 2003). The orbital fringes may not always lay parallel due to the baseline vector varying as a function of time in the two separate acquisitions. In this case, the orbital fringes are more difficult to model and remove accurately Geometric error When a SAR antenna looks towards a steep terrain, the slope facing the antenna appears brighter and shorter in the image, and the slope facing away from the antenna appears darker and longer. This is because the radar recognizes the distance of an imaged object by timing its radar echo. Normally, an object which is closer to the antenna in the ground range direction would return the signal to the antenna earlier than the objects farther away from the antenna. But objects with a higher altitude may be considered closer to the antenna than it should be geographically with respect to its surrounding objects. This effect is called foreshortening. It occurs when the slope facing the antenna is less steep than the line perpendicular to the look direction. In other words, it is when the steepness of the slope is less than the incidence angle of the radar. This effect becomes more severe as the angle of slope approaches the incidence angle of radar. Layover which is an extreme case of foreshortening occurs when the slope facing the antenna is steeper than the radar incidence angle. It causes an inversion of the image geometry, i.e. the tops of the slopes will be imaged before their bases. Figure 2.9 shows the geometric error caused by radar mapping. 34

35 Figure 2.9. Geometric error of radar image 35

36 Chapter Three: Methodology 3.1 Basics of DInSAR Satellite differential InSAR (DInSAR) is an effective tool for the measurement of terrain deformation as caused by earthquakes, subsidence, glacial motion and volcanic deflation. DInSAR is based on a repeat-pass spaceborne radar interferometer configuration. As shown in Figure 2.5, if the terrain is deformed by an earthquake on a fault, then the deformation is translated directly as the phase difference between two SAR observations, made before and after the event. If the satellite orbit is precisely controlled to make the two repeat observations from exactly the same position, or at least B, as illustrated in Figure 2.4, the phase difference measured from InSAR is entirely produced by the deformation in the slant range direction. This ideal zero-baseline case is, however, unlikely to be the real situation in most earth observation satellites with SAR sensor systems. In general the across-event SAR obervations are made with a baseline B 0, and as a result the phase difference caused by terrain deformation is jumbled with the phase difference caused by the topography. Figure 3.1 Illustration of phase shift induced from terrain deformation, measured by differential InSAR 36

37 Logically, the difference between the interferogram generated from two SAR observations before terrain deformation and that from two SAR observations across the deformation event should cancel out the topography and retain the phase difference representing only terrain deformation. Topographic cancellation can be achieved from the original or from the unwrapped interferograms. The results of DInSAR can then be presented either as differential interferograms or deformation maps (unwrapped differential interferograms). For simplicity in describing the concepts of DInSAR processing, the phase difference in the following discussion refers to the true phase difference, not the 2 wrapped principal phase value. With two pre-event SAR images and one postevent image (within the feasible baseline range), a pre-event and a cross-event fringe pair can be formulated, so a differential interferogram can be directly derived. The DInSAR formula, after correction for flattening of the Earth s curvature, is as follows (Zebker et al., 1994a): (Equation 3.1) where is the look angle to each point in the image assuming zero local elevation on the ground (the reference ellipsoid of the Earth); 1flat and 2flat are the unwrapped pre-event and cross-event interferograms after the flattening correction for the Earth s curvature; and D represents the possible displacement of the land surface. The ratio between the perpendicular baselines of the two fringe pairs is necessary because the same 2 phase difference represents different elevations depending on. The operation cancels out the stable topography and reveals the geometric deformation of the land surface. 37

38 Deformation D is directly proportional to the differential phase difference, thus DInSAR can provide measurements of terrain deformation at better than half the wavelength of SAR at millimeter accuracy. For instance, the wavelength of the C-band SAR onboard ENVISAT is 56mm and thus 28mm deformation along the slant range will be translated to 2 phase difference in a cross-event C-band SAR interferogram. As an alternative approach, if a high-quality DEM is available for an area under investigation, 1 can be generated from the DEM with an artificially given baseline equal to the baseline of the across-event fringe pair and simulated radar imaging geometry (Massonnet et al., 1993, 1994; Massonnet, Briole and Arnaud, 1995). In this case, DInSAR is a simple difference between the crossevent SAR interferogram 2 and the simulated interferogram of topography 1sim: (Equation 3.2) The advantages of using ademto generate f1sim are that it is not restricted by the availability of suitable SAR fringe pairs and that the quality of f1sim is guaranteed without the need for unwrapping. As further discussed in the next section, the quality of an SAR interferogram is often significantly degraded by decoherence factors and, as a result, unwrapping will unavoidably introduce errors. Obviously, one crucial condition for DInSAR is that the satellite position for SAR imaging is controlled to a high precision and is frequently recorded to ensure accurate baseline calculation. If the satellite can be controlled to repeat exactly, providing an identical orbit, then the so-called zerobaseline InSAR is achieved, which, without topographic information, is in fact the same as a DInSAR measurement of deformation directly. In many applications, it is not always necessary to go through this rather complicated process to generate differential SAR interferograms. Since the 38

39 fringe patterns induced from topography and from terrain deformation are based on different principles, they are spatially very different and can often be visually separated for qualitative analysis. Also, since the terraindeformation-induced fringes are a direct translation from deformation to phase difference, they are often localized and show significantly higher density than the topographic fringes, especially in an interferogram with a short B?. For a flat area without noticeable topographic relief, any obvious fringe patterns in a cross-event SAR interferogram should be the result of terrain deformation. In such cases, we can use InSAR for the purpose of DInSAR with great simplicity. Figure 2.6 is an ENVISAT ASAR interferogram (draped over a Landsat-7 ETM (true colour composite image) showing the terrain deformation caused by an earthquake of moment magnitude 7.3 which occurred in the Siberian Altai, on 27 September The high-quality fringes are produced mainly in the basin where elevation varies gently over a range of less than 250 m. With B =168m for this fringe pair, of wavelength =56mm, orbital attitude h=800km and look angle =23, we can calculate from Equation (2.19) that each 2 phase shift represents 56.6m elevation. Thus the 250m elevation range translates to no more than five 2 fringes. The dense fringe patterns in this interferogram are dominantly produced by the earthquake deformation. Certainly, when we say that DInSAR can measure deformation on a millimetre scale, we mean the average deformation in the SAR slant range direction in the image pixel, which is about 25m spatial resolution for ERS InSAR. 3.2 Differential InSAR Interferometry SAR (InSAR) and DInSAR techniques are based on the combination of two radar images taken at different acquisition times. The phase information that recorded in SAR images are analyzed to derive 39

40 the local topography (InSAR) or detect and quantify the ground displacement that has occurred between the two acquisitions (DInSAR) (Raucoules et al., 2007). The phase difference between an InSAR data pair ( ) can be expressed as (Chatterjee et al., 2006) (Equation 3.3) Here, disp, atm, noise, topo, and flat refer to the phase difference originating from ground displacement along the slant-range, atmospheric noise, noise from radar instrument and temporal deceleration, error due to topographic height, and error due to the assumption of ideally flat earth terrain, respectively. In the process of extracting the ground displacement, the topographic ( topo) and flat earth phase difference ( flat) can be removed using independently derived digital elevation model (DEM) data and precise satellite orbital data, respectively. The result of this removal process is generally called DInSAR. In this study, JAXA SIGMA-SAR software developed by Shimada (1999) is used to obtain the interferogram. In order to remove the noise from radar instrument and temporal deceleration ( noise ), the Goldstein-Werner filtering process is applied to the noisy interferogram (Goldstein and Werner, 1998). The resulting DInSAR interferogram is in the form of phase cycle (phase difference), each cycle being correlated to ground displacement along the slant-range direction. In the case of ALOS/PALSAR, the wavelength is 23.6 cm (L-band) and hence, each cycle in interferogram represents ground displacement of 11.8 cm. Figure 3.1 illustrates the flowchart of DInSAR processing, where an external or pre-existing DEM is used to simulate the topographic phase. In this study DEM 90 m of Shuttle Radar Topography Mission (SRTM) is applied. The DInSAR process is referred to as 3-pass or 4-pass mode if the DEM is generated by InSAR. Otherwise, it is referred to as a 2-pass mode if the DEM is generated using 40

41 the methods other than InSAR, such as aerial photogrammetry and airborne laser scanning. The isolated displacement phase information has to be unwrapped first before it can be converted to the height displacement along the line-ofsight, or the slant range, of the radar system using Equation (2.6). After geocoding or geo-referencing the result, the final height displacement map is generated. For simplicity reason, in this chapter it is arbitrarily assumed that surface displacement occurred in this dissertation is solely caused by the vertical surface displacement. The artefacts of phase variation in interferograms can be explained by the turbulence or heterogeneity in the atmosphere. Variances in the density of water vapour in the troposphere and electrons in the ionosphere can lead to changes of refractive index of the atmosphere. Therefore, the path along which the radar signals travelled between the satellite and the ground objects may seem to be different at the same site but at different times due to the different atmospheric condition. The atmospheric delay can be identified using the fact that its fringe structure is independent over several interferograms (Massonnet and Feigl) or can be modelled by using GPS networks (Li et al., 2006). 41

42 Figure 3.1. Flowchart of DInSAR processing There are two kinds of resolution on DInSAR result, spatial resolution and deformation resolution. Spatial resolution of DInSAR is based on the ability of SAR sensor and the observation mode in range and azimuth direction. Spatial resolution of ALOS-PALSAR is varying depend on the observation mode, for Fine Beam Single (FBS) observation mode, the maximum resolution in range direction is 7 meter. The resolution in azimuth direction is depend on the number of look during processing. Deformation resolution is depend on the wavelength of the sensor. Short wavelength will give better resolution in some cases. 42

43 Chapter Four: Results and Discussion In this dissertation the Differential Interferometric Synthetic Aperture Radar (DInSAR) using JERS-1 and ALOS PALSAR data is applied in three types of natural disaster, namely landslides, volcano deformation and land subsidence. 4.1 Mapping tropical landslides Recently, natural disasters are increasing in terms of frequency, complexity, scope and destructive capacity. They have been particularly severe during the last few years when the world has experienced several large-scale natural disasters such as the Indian Ocean earthquake and tsunami, hurricanes and typhoons; the Kashmir earthquake in Pakistan; floods and forest fires in Europe, India and China; and drought in Africa. (Sassa, 2005). Mapping such natural disaster areas is essential to prevent and mitigate people from further damage that might occur before and after such event. Mountainous areas specifically vulnerable to sediment related disaster. High slopes with loose material support are areas fond of landslides incidence. Indonesia is the second highest fatalities according to International Landslides Committee, ILC, in 2004 amount to 441 fatalities after China with 449 (ILC, 2004). In Indonesia in particular, in these recent years natural disasters have occurred more often statistically compared to a decade ago (BNPB, 2009). Once within a month, in three different island, Indonesia had been stricken by earthquake, tsunami, flash floods and volcanic eruptions with severe fatalities to the people and environment. It is obvious that Indonesia is prone to natural disaster due to its position of being squeezed geologically by three major world plates and this fact makes Indonesia one of the most dangerous countries regarding natural disasters. 43

44 Effort to make landslide inventory in Indonesia has been initiated by National Bureau of Disaster (Matigation, Badan Nasional Penanggulangan Bencana, BNPB) but has not been able to cover nationally because not all provinces branch can do the inventory. Therefore effort to create landslide inventory in Indonesia has been handled by local research institute bodies cooperating with international research bodies or research grant. The lack of detail satellite imagery has prevented the Indonesian government to fully support this effort. Figure Maps of Global Landslide Inventory, (ILC, 2009). Unfortunately, local governments and responsible bodies are not able to fully monitor the area because of the lack of spatial information supporting the decision making regarding the disaster area`s condition. The lack of detail and accurate susceptibility maps make it generally quite difficult to evaluate the extent of area affected. Thus, establishment of a comprehensive database of disaster inventory is urgently required. The availability of remote sensing data is now promising but yet has not been effectively and efficiently utilized by end users. The use of remote sensing data can contribute to mapping disaster attributes on a variety of scales ranging from community, regional and global scale (Westen, 2000, 2008). Various technique and models have also been developed to specifically map landslides. The use of remote sensing data, whether air-, satellite- or ground-based varies according to three main stages of a landslide related 44

45 study, namely a) detection and identification; b) monitoring; c) spatial analysis and hazard prediction(singhroy et.al, 2004; Matternicht et.al, 2005). Rain induced landslides are one of the most common types of natural disasters and they frequently happen in Indonesia as well as in the Asia Pacific Region. During the period of landslides occurred in Indonesia and due these landslides 2,886 people lost their lives 1,215 people were injured and 14,849 lost their homes. Normally, landslides occur during the rainy season, bringing a sudden flow of debris leaves many victims along its path. On the other hand, Synthetic Aperture Radar Interferometry (InSAR) is an established method for the detection and monitoring of earth surface processes. This approach has been most successful where the observed area fulfills specific requirements, such as sufficient backscattering, flat slope gradients or very slow changes of vegetation. The emerging Differential SAR Interferometry technique is able map slight surface deformation for a specific type of landslide. This technique can be utilized to create an inventory database. In this research we focus on a large scale mass movement occurred on 26 March 2004 in Jeneberang Watershed, Indonesia. A research on this landslide event has been conducted by Hasnawir (2006), and JICA Sabo Urgent Investigation Group led by Prof Tsuchiya from Kochi University (2009). The landslides area has also been recently revisited for ground data collection in August We also compared this DInSAR technique on the other area of landslides incidence in Tenjolaya, West Java Study Site The study area covers the whole Jeneberang watershed in South Sulawesi Province (Figure 2). The climate of Sulawesi Island is tropical. The northeast monsoon gives rise to the rainy season between November and 45

46 May (with the maximum rainfall being in December and January), and the southwest monsoon causes the dry season between June and October. The monthly rainfall is more than 700 mm per month from December to February, and reaches 900 mm in January. The average annual rainfall is 4,424 mm. Under these conditions, the outlet valley from the caldera can maintain its dominant down-cutting position by capturing the runoff from the primary depression to enlarge its drainage basin (Tsuchiya et.al, 2009). The March 2004 landslide impacted significant damage with special concern on Bili-Bili Dam located downstream, which supplies water to the city of Makassar (capital city, population of 1.2 million). Approximately m 3 of earth material consisting of volcanic fragmental rocks and debris (Bawakaraeng Formation) slid to the upper part of River Jeneberang, covering one village, one primary school with 32 casualties (Tsuchiya.S, et al., 2009). Some of the mechanical factors that enhanced this landslide are the tremendous height of the side wall of the caldera, fragility of the bedrock of the side wall, and susceptibility to erosion of the accumulated sediment inside the caldera (Hasnawir, 2006). This landslide is categorized as rock avalanche and the movement type is rotational and triggered by heavy rainfall. 46

47 Figure Study sites (a). Upper part of Jeneberang River and (b). Tenjolaya site, West Java. Darker violet color indicates high incidence of landslides (BNPB, 2009) The second site of the landslides study located in West Java, at the village of Ciwidey subdistrict of Pasir Jambu, 30 kms southwest of Bandung City. The landslide stroked on Feb 22, 2010 at 8.30 where most of the tea plantation workers were on location. Materials were mainly debris and loose limestone, transported from Waringin Mountain with elevation of 3910 m above sea level. Fatalities recorded 6 died and less than 60 buried unfound. The area is very prone to landslides as the site location surrounded by high slope terrain and worsened by the tea cultivating land Dataset used In order to visually identify the affected area of the landslides, we utilized optical satellite images. Landsat 7 ETM visible/infrared image acquired on 28 September 2002 and an ASTER image of 2005 acquired after the landslide on 7 September were used to compare the pre and post event of the Jeneberang landslide. Synthetic Aperture Radar (SAR) images of Japanese Earth Resource Satellite (JERS-1) acquired in are used in this study. All the SAR images are on the descending mode.. Figure 1 shows the Landsat 7 ETM image with composite band of 432 (a) before the incidence (2002) and (b) ASTER composite image band 432 after (2005) the Jeneberang landslide, indicating the extent of the unvegetated area (encircled). Synthetic Aperture Radar Interferometry (InSAR) is an established method for the detection and monitoring of earth surface processes. This approach has been most successful where the observed area fulfills specific requirements, such as sufficient backscattering, flat slope gradients or very slow changes of vegetation. Generally two SAR data acquisitions, called 47

48 scenes or images, of the same area are required to generate interference fringes resulting from phase differences that can be interpreted as heights or displacements. The typical SAR geometry of master and slave scene with a short baseline for the detection of earth surface changes (Riedel and Walther, 2008; Colesanti and Wasowski, 2006). DInSAR processing has been applied to 6 different level 0 data of JERS-1 SAR from different acquisition years of with the JAXA/SIGMA SAR processing software (Shimada, 1999). Each pair of data went through the same procedure from image coregistration, interferogram generation until phase unwrapping. Before phase unwrapping the DInSAR image was filtered using Goldstein and Werner filter with preconditioned conjugate gradient (PCG) (Singhroy.V and Molch.K, 2004). At the end all images were geocoded using cubic convolution with UTM transformation by resampling the DInSAR data into 12.5 resolution. The same procedure was applied to JERS-1 SAR of 3 different acquisition years of the Tenjolaya site. We processed 6 scenes of path 106 row 312 ranging from acquisition year and 3 scenes at the lower part of scene 106/312 in All scenes are on the descending mode. The Tenjolaya site is located at the south part of Bandung Basin in a valley of mountainous area as shown in Figure 4 on an aerial photograph taken on July 3, 2007 before the landslides happened and Figure 5 an aerial photo taken on March 10, 2010 showing after the landslides. Change detection analysis and thematic classification have been implemented to both the visible and SAR images over the study area to generate landslide susceptibility map. Under one GIS platform all data were integrated in ArcMap GIS 9.3 software with one UTM projections. All spatial data layers such as geology, structure, landcover, aspect and slope data as well as Digital Elevation Model (DEM) including field observation data is used to create a GIS-based landslide inventory database. This study required us to locate the exact location of the landslides coverage. On the 48

49 ASTER image we could delineate the coverage and make as vector file to overlay the SAR processed images. Field survey were conducted twice in 2009 and 2010 by visiting the affected area specially for the GPS measurement of the crack incidences along the weak zone of prone areas that considered to be triggering slides in the future. (a) (b) Figure Landsat 7 ETM composite image (432) (a) before (2002) and (b) ASTER composite image (432) after the landslide (2005). Yellow circle indicate the landslide area. The discussions with the staffs of The Jeneberang River Sediment Control Project under the coordination of public works office in South Sulawesi gave us significant overview of the post landslide monitoring. Field survey was conducted in August 2009 by taking GPS measurements and sample material and the second visit was to obtain the crack locations and dimensions. The project started monitoring the cracks along the weak zone of upper head of the landslide since 2006 and continued until now. The crack location was measured regularly on monthly basis by staffs assigned to do the task. The measurement was done by measuring the distance of the crack. Figure 5 shows the situation in 2 locations of the crack. Table below describes the geographic location, taken by geodetic GPS using Universal Transverse Mercator (UTM) and dimension of the openings. 49

50 Table Geographic location (UTM) and dimension of the openings. LOCATION X (m) Y (m) Length (cm) DISPLACED Crack Crack Crack Crack Crack Crack Crack Crack Crack Crack Crack Crack Crack Crack Crack Crack In the past, prior to the landslide, the crack was not monitored on a regular basis. Local people used to notice the gap and had to jump over to the other side of the crack. Based on the report from mountain climbers, several sites of the crack initially just showed small gap and gradually became larger and larger. Unfortunately the scientific records for these cracks were not found. After the landslide occurred people began to realize that the cracks happened in the past were no longer found. This is also happening to the new cracks that are now being recorded. Sometimes the displacement was nowhere to be found because it was already collapse. 50

51 Scale 1: Scale 1: Figure Tenjolaya site. Aerial picture taken before July 3, 2007 (a) and after 3 March 2010(b) 51

52 Figure Pictures of the crack measurement situation in the field Results and Discussion Based on the Geographic Information Systems (GIS) data processing using ArcGIS 9.3 and ENVI 4.5 software, visual delineation of Landsat image overlaid with vector data shows that an area of approximately 647 ha has been affected at the upper part of River Jeneberang, with indications of significant mass movement prior to the landslide. On the other hand, the SAR processing using SIGMASAR yielded intensity images for each level 0 JERS-1 SAR data (Fig.2a). Generating DInSAR images require a pair of different acquisition images. Among the 6 pairs being processed, only 2 pairs show reasonable good coherence namely the pair images of 1995 March/1996 May and 1996/1997 (Fig. 2b). Based on the theory, coherence in DInSAR images are partly caused by the baseline of the two different acquisition time of the satellites (Figure 7). In Table 1 both 1995/1996 and 1996/1997 pairs have less than 500 m baseline. The DInSAR image pair of 1995/1996 shows a linear movement of 5 cm around the landslide area, suggesting the occurrence of cracks/gaps related to subsidence before the landslide event (Fig 6a). Ground validation using high resolution differential GPS from the field also supports this interpretation of the DInSAR image. 52

53 1a 2a 3a 1b 2b 3b 4a 5a 6a 4b 5b 6b -5.9 cms Figure JERS-1 SAR DInSAR Processing Images: (a) coherence image of the six pairs of DInSAR processing result. Yellow circle indicate the affected area. and (b) DInSAR Images from 6 different pairs. 1-6 with pairs of 93/94, 94/95, 95/96, 96/97, 97/98 and 98/98. List of dates can be seen on Table 2. 53

54 Figure DInSAR Image showing the linear displacement (a), black dots are points of GPS Field measurement of the crack locations (b) From project report of Jeneberang Sabo Dam, we managed to obtain the GPS location of the cracks that occurred before and after the landslides. Overlaying this points allow us to confirm the throne and the head of the landslides (Figure 4.1.7). Table 2. Baseline information for each pair of DInSAR Processing of Jeneberang Site Week Pair (RSP 77/309) Differ Baseline (m) Bp (m) Bh (m) / / / / / / From all of the DInSAR images of Tenjolaya site (Figure 8), the upper scene of the DInSAR image showed a good coherence hence the subsidence in Bandung city can be obviously identified (dotted circle) and the 54

55 phenomena is consistent in other pairs of the processed images. Unfortunately, this didn`t happen on the lower scene where the exact location of Tenjolaya site (yellow circle). The matching process didn`t give good coherence hence we couldn`t identify any surface displacement. Figure DInSAR Image of Tenjolaya site (yellow circle). Dotted circle Conclusion Bandung subsidence DInSAR processing image obviously will give variety of results on surface displacement mapping. This study has shown that not all JERS-1 SAR pairs of the designated areas give good coherence image which therefore will give better result on the DInSAR images. Integrating optical satellite image (Landsat) with SAR image processing can complement the change detection analysis and Differential Interferometric SAR (DInSAR) is proven to be one of the effective methods in mapping surface displacement especially for landslides. Integration of remote sensing and GIS can provide information on pre- and post-landslide situations. 55

56 Further works must be carried out to validate the accuracy of the DInSAR image as it is still an estimation result. Validation with ground checked must also be carried out. Applying DInSAR method with ALOS PALSAR images will be promise a better result considering ALOS PALSAR has better accuracy for the device specifications compared to JERS-1. The next step is to input DInSAR image to the GIS map to create landslides susceptibility map hence creating landslides inventory. 4.2 Monitoring Miyakejima Volcano deformation Introduction Japan, as one country lies at the edge of the ring of fire, is believed to have approximately 10% of world`s on-land volcanoes. The crust of the earth in this Japanese archipelago is covered by 10 huge tectonic plates. Slight movement of these plates can cause major natural disasters especially earthquake and volcanic eruption (AIST, 2005). Volcanic eruption is one of a kind where the dangerous environment surrounding it causes great difficulties for close-up monitoring and surveillance. In order to understand the status of a volcano, a complete volcanomonitoring system comprising seismic observation, geodetic observation and geochemical observation is required (Dzurisin et.al 2003) The seismic observation is designed to monitor the volcanic activity beneath the surface; geodetic observation to detect deformation on the surface; and geochemical observation to monitor the chemical properties of substances that emerge to the surface and the atmosphere. One technique in geodetic observation is Interferometric Synthetic Aperture Radar (InSAR) that has been used since 1995 to monitor surface displacement related to volcanic activity (Tralli et.al, 2005 and Geshi et.al, 2003). Miyakejima is a volcanic island of about 8 km across, located on latitude and longitude of N: 34.08, and E: , about 180 km south of Tokyo, on the volcanic chain of Izu-Ogasawara (or -Mariana). Including the 56

57 body under the sea level, the base of the volcano has the diameter of about 13 kms and the height is about 1,000 m (Figure.1). It is a stratovolcano, consisting largely of basaltic rocks. The summit area is characterized by double calderas, 3.5 and 1.5 km across, in which a scoria cone Oyama is located. Near the coastal line, there are many craters formed by phreatomagmatic eruptions. (Geshi et.al, 2006). Tectonic setting of Miyakejima is complex. Magmatic activity of Miyakejima relates the subduction of the Pacific Plate at Izu-Bonin trench running the east side of the volcanic chain. The edifice of Miyakejima volcano is building on the Philippine Sea Plate, which is moving to NNW direction and subducts to Eurasian Plate and North-American Plate at Suruga trough and Sagami trough at the southern off of the Japan Islands. Northern boundary of the Philippine Sea Plate is locally bending northward by the collision of Izu-Peninsula about 100 km north of Miyakejima and forms compressional tectonic setting around the peninsula (Geshi et.al, 2005). This tectonic setting can be seen in Figure The history of Miyakejima eruption was dated back to 500 hundreds years ago where 12 big eruption had occurred at interval of years. In 20 th century it periodically erupted roughly every 20 years in 1940, 1962, and 1983 and the latest one in the year The manner of the 2000 eruption was largely different from what had been experienced during the last hundred years. Since June 26, 2000, vigorous earthquake swarm began. The submarine eruption was observed on 27 June. The hypocenters of the earthquakes migrated westward and the swarm started near Niijima- Kodujima islands. The first summit eruption occurred on 8 July. 57

58 Figure Site Map showing Miyakejima Island and the tectonic setting of Izu Ogasawara Volcanic chain. The eruption was small but a large collapse crater was observed after the eruption. Summit collapsed continued also after the event. Summit eruptions with volcanic ash were observed on 10 August. The largest eruption occurred on 18 August. At present situation, seismic activity of Miyakejima has declined but very large flux of SO2 is measured and the collapse of the inner rim of crater is apparent (Figure ) 58

59 Figure Southwestern view of the summit subsidence at Mt. Oyama, Miyakejima volcano. Initial stage of subsidence (1.0x0.8 km with 0.2 km depth), Smoke including abundant sulfur dioxides were emitted from the Oyama Caldera (1.6 km across and 0.5 km deep), taken on 4 June 2001 (Geshi et.al, 2002). Prior to the 2000 eruption, various kinds of geophysical observations were installed on Miyakejima including seismic array, gravity surveys, and Global Positioning System (GPS) as well as borehole tilt measurements. After the event of the 2000 eruption, many publications have been filed (Geshi et.al, 2003; Irwan et.al, 2003 and Nakada et.al, 2005). The advancement of remote sensing technology on the other side has contributed significant level of assistance in monitoring these natural disasters so we can reduce nature s calamities and threat to human lives and property (Tralli et.al 2005). Differential synthetic aperture radar interferometry (DInSAR) is a technique useful for accurately detecting the ground displacement or land deformation in the antenna line-of-sight (slant-range) direction using synthetic aperture radar (SAR) data taken at two separate acquisition times (Bayuaji et.al, 2010). Despite the capability of DInSAR in mapping precise surface displacement, there are also some hindrances that must be taken into consideration especially in the application of volcano monitoring. The 59

60 D-InSAR method is complementary to ground-based methods such as leveling and global positioning system (GPS) measurements, yielding information in a wide coverage area even when the area is inaccessible (Massonet et.al 1998). The potential precision of DInSAR depends on many things (e.g., the accuracy of the orbit determination for the two satellites passes, atmospheric conditions that affect the phase delay) but in principle the surface displacement measurement can have a precision of 2%-5% of the SAR wavelength. Typical SAR wavelengths are in the range 3-30 cm, implying millimeter to centimeter precision in surface displacement between two satellite passes. Due to its precision measurement, this method has been widely the application of land and mine subsidence, earthquake analysis and 1andslides (Bayuaji et.al, 2010; Sri Sumantyo et.al, 2011; Perissin and Wang, 2011 and Alimuddin et.al, 2011). In this paper we explore the capability of DInSAR method using ALOS PALSAR images in mapping the surface displacement occurred over time on the inner rim of Miyakejima Volcano Datasets Used The ALOS (Advanced Land Observing Satellite) was launched by the Japan Aerospace Exploration Agency (JAXA), in 2006 and had been continuously sending images of earth observation until recently, the satellite has been decommissioned due the battery life. The PALSAR sensor carried onboard the ALOS is an L-band SAR imager that can provide about ten meter ground resolution when operated in the Fine Beam Single FBS mode (the range resolution is twenty meters for the Fine Beam Dual FBD mode Rosenqvist et.al, 2010). ALOS PALSAR (Phased Array type L-band Synthetic Aperture Radar) data/images are used to derive the DInSAR results presented in this research. Howard A. Zebker, a long time volcano researcher using InSAR technique stated recently that the dawn of the 21 st century has witnessed an 60

61 explosion in the development of reliable interferometric imaging of active volcanoes, and its application to most of the 600 potentially active volcanoes. Several research on the use of InSAR has implemented this technique on tropical volcano, Ibu and Merapi Volcano in Indonesia(Agustan et.al, 2010) and Merapi Volcano ( Saepuloh et.al, 2011), Kilauea in Hawaii ( Zhe, 2009) and many more. Therefore in this research we try to combine the InSAR technique with the validation of ground measurement. In SAR interferometry (InSAR), the phase data of SAR images are analyzed to derive the local topography (original InSAR) or detect and quantify the ground displacement that has occurred in the slant-range direction between the two acquisitions data (Differential-InSAR, DInSAR) (Agustan et.al, 2010). This technique has been used since 1995 to monitor surface displacement related to volcanic activity (Tralli, et.al, 2003). The phase difference between an InSAR data pair can be ( Int,P1-P2) expressed as follows: Int,P1-P2 = Disp,P1-P2 + atm, P1-P2 + noise,p1- P2 + topo, P1-P2 + flat,p1-p2 (1) where Disp,P1-P2 + atm, P1-P2 + noise,p1- P2 + topo, P1-P2 and + flat,p1-p2 refer assumption of ideally flat earth terrain, respectively. In the process of extracting the ground displacement, the topographic condition ( topo,p1- P2 ) and flat earth ( flat,p1-p2 )phase differences should be removed using digital elevation model (DEM) data and precise satellite orbital data, respectively. The result of this process is generally called D-InSAR, which estimates the ground displacement in the slant-range direction (Liu and Philippa, 2009). Below is the list of the data we used to produce the DInSAR with mostly 46 days interval of temporal observation. The PALSAR images acquired in 5 different dates ranging from July, October, December in year 61

62 2007 and January and March acquired in All the data sets have the same observation parameters with reference system for planning (RSP) number 407, path number 670 and off nadir angle of 34.3 o. Of the 5 scenes ALOS PALSAR image available, 4 pairs of InSAR could be generated. Details of the pair can be seen in Table We utilized SIGMASAR software developed by Dr. Shimada of JAXA to create DInSAR images and a DInSAR two-pass method is used (Shimada, 1999). In this approach, two different images in an interferometric pair are needed to calculate one interferogram (Rosenqvist et.al, 2007). After the slave image is coregistered to the master image, the phase difference (raw interferogram) between them can be calculated using conjugated multiplication. An external DEM (Digital Elevation Model) is introduced so as to compute the topographic phase. The schematic flow for this processing can be seen Figure 3. Table ALOS PALSAR pair and baseline information Pair Date 1 Date 2 Δt days Baseline P (m) Baseline Perp (m) Mode Ascending Ascending Ascending Ascending Since 90m resolution SRTM (Shuttle Radar Topography Mission) DEM is used in this study, it must be oversampled to correspond to the 10m or 20m resolution of the PALSAR images. Based on the oversampled DEM and baseline information of the master and slave, a simulated SAR image can be generated. Then topographic phase can be obtained, and the differential interferogram can be obtained by removing the topographic phase from raw interferogram. The phase unwrapping step calculates the 62

63 absolute deformation in the satellite s line of sight (LOS) direction, based on the differential interferogram. Below is the schematic diagram of the flow. Along with the SAR data, we also used optical image data for visual observation and landcover time series change analysis of Landsat ETM acquired on 21 October 1999 before the 2000 eruption and ALOS AVNIR2 image acquired on 1 March We ran an NDVI image comparison before and after the eruption to delineate the border of the affected area of the volcanic material erupted (Figure 4.2.3). A B Figure Optical Satellite Image of the Island. A. Landsat Image acquired on B. ALOS AVNIR2 acquired on Differential synthetic aperture radar interferometry (DInSAR) has been proven to be able to detect slight surface deformation occurred on volcanic environment monitoring. The result of image processing must be confirmed with GPS data and must be validated with ground checking as well as the support of seismic activity data. GPS data obtained from the device installed in 4 locations on the island labeled A,B,C and D on optical image figure. (Figure 4.2.3). The GPS data range from taken on daily bases at For the purpose of validation, the data used only in correlation with the ALOS PALSAR image temporal range. DInSAR images were created and resampled to 10 meter resolution pixel to be overlaid with vector data to confirm the actual location where 63

64 surface of the caldera rim has changed using geocoded cubic convolution to UTM projection. The image was filtered using Goldstein and Werner filter and the unwrapping method used Preconditioned Conjugate Gradient (PCG) Result and Discussion We compared the intensity image from each scene of PALSAR with the same image acquisition time of ALOS AVNIR2 image focusing on the inner rim of the caldera (Figure 4.2.3). Each of the images created from the pair of 2 acquisitions time. Changes on the collapse of the inner rim can be detected from the time series change analysis using both the optical images and the SAR intensity images (Figure 4.2.4). There are four pairs of geocoded DInSAR images produced. Not all pairs show good coherence due to the atmospheric condition of the acquired images but specifically on the caldera area the four pairs indicate surface displacement has occurred. This can be seen on the comparison of the 4 pairs of the coherence data generated from the InSAR processing along with the DInSAR images after filtering (Figure 4.2.6). Figure Intensity image created from master pair of /

65 It is clear the least coherence pair atmospheric disturbance can be in the form of weather condition (Massonnet et.al, 1998). This can be seen on the pair acquired in the fall season compared to the pair taken in the spring time where the date of acquisition data on the area experience a lot of rainfall in turn will affect the response from dielectric constant property of the object. This is called seasonal change. The other reason for the inconsistency of the DInSAR images is the baseline distance of the acquisition position of the different date images; baseline data can be seen in Table The four pairs of DInSAR images can be seen in Figure The obvious scene indicated the collapse of the rim can be observed in the pair of images where the fringes (cycle of colors) repeatedly concentrated on the rim of the caldera. The collapsed area is consistently apparent in the other DInSAR pair images confirming the actual observation in the field. The pyroclastic material consisting the rock and the tectonic activity are two major factors causing the collapse of the inner rim of the caldera. The pair images is indicating the collapse of the inner rim wall of the Oyama caldera approximately 5 meters as observed in the line of sight (LOS) with the highest rate subsidence of 7 cm for the 3 month period of different acquisition time. Ranges of the subsidence rate vary for each pairs from 2-7 cm with accumulation rate of 20 cms within the period of the observation time (Figure 4.2.6). 65

66 A B C D Figure Coherence SAR Images pairs generated from SIGMASAR processing. A / B / C / D / With the GPS data obtained from the installed unit, we managed to confirm if there are actually some fluctuations on the changes of the surface on Miyakejima Volcano especially on the caldera. Overlaying the exact location on the satellite image enable us to see which part of the island is experiencing highest changes. We have 4 GPS location on each side of the island, Northwest (A), Southeast (B), East (C) and West (D). The locations can be seen in Figure on the optical image. We cannot have any GPS installed on the caldera rim due to the risk of falling. 66

67 A B Line -5.9 cm Line Of Sight (LOS)-5.9 cm C D Line -5.9 cm 0 Line -5.9 cm Figure Differential Interferometry SAR Images pairs generated from SIGMASAR processing. A / B / C / D / The rim of the volcanic caldera is indicated by yellow circle Clearly from the DInSAR images, we can observe that the rim of the caldera is collapsing. These phenomena are also supported by the field pictures taken from certain point of the caldera rim (Figure 4.2.6). The pictures were taken in four different dates in 2 different years, , , , and the most recent year From the line graph generated of the GPS data ranging from July 2007 until March 2008, it is obvious that the GPS shows slight fluctuations in every interval day which assumed to be noise. But bigger fluctuations on the data can be considered the actual movement of the expansion and the deflation of the magma chamber Figure

68 Picture taken on 18 May 2007 Picture taken on 10 October 2007 Picture taken on 9 Picture taken on 5 Picture taken on 8 Picture taken on 8 Picture taken on 8 Figure Images from Miyakejima Volcano Caldera taken in different dates. 68

69 This fluctuation of inflation and deflation of the surface changes that was picked up by the DInSAR processing, is able to be observed in some areas where the changes occurred in Figure 8. Based on the line graph we can measure in the northwest location the highest range of changes is 5 cm, on the southeast 14 cm, on the east 12 cm and on the west 6 cm. The average of changes during the whole 9 month observation is 3-5 cms Conclusion and further work Differential synthetic aperture radar interferometry (DInSAR) has been proven to be able to detect slight surface deformation occurred on volcanic environment monitoring. The result of image processing has been confirmed with GPS data and validated with ground checking. Ideally this findings can also be confirmed with seismic activity data. During the period of 9 months of ALOS PALSAR image data observation, the collapse of the caldera rim accumulated to 5 meters with gradual surface changes from 2-7 cms for each DInSAR pair images. Thorough atmospheric environment particularly detail weather condition on the acquisition time will be further investigated to reduce the inaccuracy factor of the DInSAR processing result. Seismic activity data is to be confirmed with the image acquisition data. For future research, we would like to correlate the changes of the caldera rim with the emission of sulphuric gases from the caldera. These two parameters are closely related and important for volcano monitoring and mitigation. For future research, we would like to correlate the changes of the caldera rim with the emission of sulphuric gases from the caldera. These two parameters are closely related and important for volcano monitoring and mitigation. 69

70 Figure Graph comparison on the GPS measurement taken in 4 different locations 70

71 4.3 Measuring Land subsidence of Makassar City Most major cities in the world have experienced land subsidence phenomena on some parts of them due to the load of development and modernization. Cities in Indonesia experienced the same condition. The excessive extraction of groundwater for the needs of industry has forced the condition where the table drops which can trigger subsidence, can be one of the culprit responsible for this phenomena. Study on the subsidence on the city of Jakarta, capital of Indonesia has been initiated since 1970s using various method of measurement with the latest method using this differential interferometric synthetic aperture radar (DInSAR). Other cities have been investigated using this technique as well including Semarang, Surabaya and the latest one Bandung, West Java province. Part of this dissertation tried to measure the dimension of land subsidence phenomena that has occurred in this city Study area The city of Makassar as the capital of South Sulawesi Province is the 4 th largest city in Indonesia, considered to be the gateway of Indonesia from the eastern part. Situated at the southwest part of Sulawesi Island, Makassar city covers an area of km2 divided into 14 sub districts. Makassar city lies on the geographic coordinate of '27,97" '31,03" East Longitude and 5 00'30,18" -5 14'6,49" South Latitude. The landform is relatively flat, classified as alluvial plain and topography level from 0-21 meter above the sea level. Geologically, the city is covered by 3 types of formation, Camba Volcanic Formation, Baturape Volcanic formation which mainly consist of fine sediment clastic of volcanic eruptive rocks but mostly eroded and alluvium formation deposit as recent weathered material. In general, we can find 3 types of rock units, basalt, tuff and breccia which derivated from the volcanic origins and sediment deposit like fine to coarse sand. Geology Map can be seen in figure

72 INDONESIA Figure Study area of Makassar City Land Subsidence with 14 subdistricts. Based on the statistic data of Makassar City, population has been increasing due to the development and modernaisation that create urbanisation. From in 1990, the city population has increased in just 10 years. 72

73 Table Total population distributed from each subdistrict of Makassar (BPS Kota Makassar from Total Population Growth No Subdistrict Mariso 55,607 51,003 51,98 55, Mamajang 67,929 58,85 56,988 61,294-1,46-0,91 3 Tamalate 199,65 253, , ,464 2,49 2,21 4 Rappocini 133,66 145, Makassar 92,513 80,127 79,362 84, Ujung Pandang 38,182 27,765 27,279 29, Wajo 44,381 34,114 32,519 35, Bontoala 64,56 56,875 54,671 62, Ujung Tanah 45,229 44,055 45,156 49, Tallo 111, , , , Panakkukang 150, , , , Manggala 100, Biringkanaya 73, ,934 89, , Tamalanrea 90, Makassar ,100,019 1,060,011 1,272, Makassar as the capital city of the South Sulawesi Province is the target of urbanization. Hence the situation will continuously grow alongside the needs for the people for development to have a better life. As the number of population increase, industries has been triggered to provide new open areas for business, construction and this situation demand spaces. When the government can not be pushed in the rural area, the spaces in the city will eventually decrease for particular landuse creating land degradation and could generate land subsidence in the future due to the extraction of water well in the surrounding areas. 73

74 Camba Volcanic Formation Baturape Volcanic Formation Alluvium Deposit (a) (b) Figure 4.3.5, (a) Geology Map of Makassar and (b) Intensity Map of the path- row JERS-1 SAR Map dated Data used Below is the summary of 8 scenes of JERS-1 SAR data acquired from with one scene for one particular year except 1998, we have 2 scenes. All JERS-1 SAR are in descending modes with 35.5 degree of incident angle. The subsidence analysis is also supported with optical image data, SPOT 4 acquired Field campaign was conducted in September 2009 and January 2011 with each measurement was taken by Global Positioning System (GPS). 74

75 Table Pairs of JERS-1 SAR data used to monitor subsidence with baseline information acquired from Pair (RSP 78/309) Week Difference Baseline (m) Bp (m) Bh (m) / / / / / / /

76 1a 2a 3a 1a 2b 3b -5.9 cms -5.9 cms 4a 5a 6a 4b 5b 6b -5.9 cms -5.9 cms Figure DInSAR processing images a. Coherence images b. DInSAR images Pair of 1993/1994, 2. Pair of 1994/1995, 3. Pair of 1995/1996, 4. Pair of 1996/1997, 5. Pair of 1997/ Pair of 1998/1998. Later image is master and earlier image is slave. 76

77 -5.9 cms Figure DInSAR Processing images of Makassar City. A.Coherence image of 1995/1996. B. DInSAR Image C. SAR, real image D. Deformation Image 77

78 Figure Focus on deformation image of Tamalate and surrounding overlaid with Ikonos image acquired on The yellow circle line indicates subsidence occurrence. Field campaign conducted in September 2009 revealed some locations that indicate the incidence of land subsidence as well as the fact that some parts of the city are having load of building construction that make the city experience of slight movement of its earth surface. New building construction of warehouses can be seen in picture P1 taken in the area of 78

79 Tallo, New housing and modern apartment as well as community business complex in P2. Evidence of subsidence can be seen from P4 in Paotere, P5 in Panakkukang, P6 in Mariso and P7 in Tamalate. On of the main road, the soil load can be a thivkness of centimetres, P Result and Discussion We have shown that the application of DInSAR technique using JERS-1 data can reveal subsidence conditions in the study area. Mostly the subsidence occurred in the northern part of Jakarta city during the time interval studied here, though the population density in northern part is lowest among the entire city regions. Industrial district, reclamation area, trading center area, international airport and the seaport are built in this region. The center of the subsidence with the subsidence-affected coverage area can also be estimated easily. It has been found that the subsidence occurred in separated regions with different land usage. Nevertheless, the ground survey has indicated that high human activity exists in every point of subsidence. Various human activities such as ground water pumping and construction working should have affected the local subsidence phenomena in Jakarta, as in the case of other large-scale cities (Raucoules et al., 2003; Abidin et al., 2008). The main cause of subsidence in Makassar has not been revealed because of the complex feature of the phenomena. Nevertheless, the result of the present study strongly suggests that the human activity and land use alteration are influencing the geomorphological changes in this city. 79

80 Figure Images from the observation area taken during the field campaign in