Sidiq 2014

Seri Deformasi LUSI:

TIME SERIES DEFORMATION ANALYSIS OF LUMPUR SIDOARDJO (LUSI) MUD VOLCANO USING INTERFEROMETRY SYNTHETIC APERTURE RADAR

T.P Sidiq, Y. Aoki, T. Kato, H.Z. Abidin

1. INTRODUCTION

The combination of mud, steam, fluid, and gas erupted on 29 May 2006 in sub district Porong, Sidoardjo East Java (Fig 1).

Term LUSI, stand for LUmpur (mud) SIdoardjo, was unofficially named to this eruption. This continuous eruption bring significant damage to the local social economy, infrastructures, and environmental.

Milkov (2005) said that instabilities of mud volcano characteristics represent it as a geo-hazard.

Four years after the first time observed eruption, about 12 million meter cubic of mud has been ejected. The mud is accommodated on about 620 hectare dike, some of the mud is released into the river near the dike and slowly shed the mud to the sea, around 10 km from the spill way. 14 villages buried by the mud, and about 30 thousand people should be relocated. The lost caused by LUSI reached 4 trillion rupiah (about $US 4 billion) by the fourth year (Republika 2010).

The lost is excluded from implicit lost due to degradation of transportation infrastructures, shattered of the local industries, and many other chain effects.

The rate and duration of LUSI mud volcano is very high and not usual (Davies, 2007). LUSI has contradictory behavior compared to most mud volcano exist in the present time. Among other mud volcano located in Java Island, none of

them have equal effect with LUSI. Figure 2 shows how LUSI grows from 3.6 m2 in August 2006 to 12 km2 in December 2009. While the photographs (Fig 2) show the condition and main feature of the mud volcano. A dike had been made and has been maintained to prevent the mud from more massive mud covering.

Four years after the first time observed eruption, about 12 million meter cubic of mud has been ejected. The mud is accommodated on about 620 hectare dike, some of the mud is released into the river near the dike and slowly shed the mud to the sea, around 10 km from the spill way. 14 villages buried by the mud, and about 30 thousand people should be relocated. The lost caused by LUSI reached 4 trillion rupiah (about $US 4 billion) by the fourth year (Republika 2010).

The lost is excluded from implicit lost due to degradation of transportation infrastructures, shattered of the local industries, and many other chain effects.

The rate and duration of LUSI mud volcano is very high and not usual (Davies, 2007). LUSI has contradictory behavior compared to most mud volcano exist in the present time.

Among other mud volcano located in Java Island, none of them have equal effect with LUSI. Figure 2 shows how LUSI grows from 3.6 km2 in August 2006 to 12 km2 in December 2009. While the photographs (Fig 2) show the condition and main feature of the mud volcano. A dike had been made and has been maintained to prevent the mud from more massive mud covering.

The data processing is performed using Gamma Software. Single complex look (SLC) image is first generated before the interferogram image. All images are processed without cropping, thus, one image will cover approximately 80x80 km.

The images are preferred not to be cropped before the displacement map generated because the eruption area is relatively small and featureless. Interferogram processing is performed using 3 range looks, and 9 azimuth looks, which result roughly 30 meter of spatial resolution. An adaptive filter is applied in time domain during interferograms generation to increase the signal-to-noise ratio (Goldstein and Welner, 1998).

Figure 3 : Data pairs for each satellite path

Processing all data pairs is preferred rather than selecting them based on perpendicular baseline length.

Figure 3 above show that the baseline can reach more than 2.5 km. Despite the perpendicular baseline is very large, some parts of the images are still having good coherence.

Large dataset is required to avoid the lack of data which can lead to underdetermined least square solution (Menke, 1989).

However, coherence based masking is still applied on the computed displacement maps in order to maintain good quality of inversion analysis.

Interferogram is generated by subtracting phase difference between two images with a simulated phase of the terrain. A 3-arc-second (approximately 90 meter) SRTM digital elevation model (DEM) is used as the terrain height information.

Linear absolute accuracy for SRTM DEM is less than 16 meter (Rodriguez, 2005; Berry et al., 2007), which result the phase error due to topography that can be neglected though for long baseline data pairs (Fukushima, 20009).

ALOS PALSAR images come with good orbital information with 1 meter accuracy.

However, there still error caused by orbit inaccuracy which is subtracted from calculated displacement map using estimation of 2-dimension quadratic model phase function.

Atmospheric errors are mainly caused by ionospheric and tropospheric influences. The ionospheric influences are frequency-dependent which is hard to be modeled.

While the tropospheric influences is usually divided into hydrostatic, wet and liquid components (Hanssen, 2001).

The hydrostatic component is mainly related to refractive index of air pressure (e.g. altitude) and wet delay on the water vapor pressure (Jehleet.al, 2008).

The hydrostatic component which has height dependent characteristic is reduced using linear model consisting a phase constant and phase slope.

While the hydrostatic component is reduced, other components that lead to several centimeters of error are still remaining on the displacement map (Fukushima, 2009).

RESULT

In general, the deformation could be revealed very well using InSAR method in the case study area. An ellipsoidal pattern is clearly visible from the interferograms (Fig 6) with the subsidence rate in the first 8 months is much greater than after.

The maximum deformation reach ~32 cm from June to October 2006, while the total maximum deformation reach ~60 cm in more than three years. An uplift pattern is observed in the north side of the eruption. The uplift is considered as a buoyancy effect.

Another deformation pattern is also observed in the north east direction of the main crater with maximum LOS deformation reach ~30 cm in more than three years. The cause of this phenomenon still needs further analysis, which is not included on this study.

The GPS measurement shows much lower value in a point adjacent to the center of the crater (Fig 7). Smoothing factor is expected caused this significant difference.

Although the smoothing process reduce some errors in the processed interferograms, but it also effect the signal itself, and specially the big signal occur in the first months in this case.

Figure 6 : GPS and InSAR data comparison

CONCLUSION

InSAR method reveals a big deformation in ellipsoidal form, cover approximately 3km x 6km (18km2).

The ellipsoidal form is confirmed by the fact of how the cracks appear around eruption area (Fig 8).

Two different pattern of subsidence is clearly visible. Deformation outside the main eruption has almost similar rate with the main eruption, but remain unexplained Linear inversion perform well although no further parameterize method was used.

Smoothing factor that is used on this study caused some important signal lost. Because of that, the study about smoothing factor determination is still needed to get the better result.

The time series analysis also reduces the atmospheric error, and can be used as a good tool in large data set interpretation.

A better understanding of the temporal variation can be given by the method, although a large data set is needed to perform good analysis.

PLAN IN THE FUTURE

As the smoothing factor has not well performed in this case, we plan to use other temporal filtering methods.

The temporal filtering becomes important issue because the atmospheric error varies through time.

Temporal filtering can reduce the atmospheric error very well, and also provide better time series analysis.

More GPS data comparison will also be done in the future. Only one GPS point out of 52 GPS points is compared with the InSAR data processing result right now.

The final goal to achieve is the source modeling. Right now, many scientists try to model the source of the mud volcano using many data, but none of them had modeled it using InSAR data. As the data processing finish, a paper is alsoplanned to be published in the near future.

JOINT WORK

This research is a joint work between Earthquake Research Institute (ERI) of Tokyo University, and Institute Technology Bandung through JICA program. The pre-result of this research had been presented in poster session of AGU

meeting last December, San Francisco USA.

REFERENCES

Abidin H. Z, R. J. Davies, M.A. Kusuma, H. Andreas, T.

Deguchi}, Subsidence and uplift of Sioardjo (East Java) due

to the eruption of the Lusi mud volcano (2006-present),

Environmental Geology, 2008.

Takeuchi, Y. Oguro, A comparative study of coherence

patterns in C-Band and L-Band Interferometric SAR from

tropical rain forest areas, Advanced Space Res Vol. 32

p.2305-2310, 2003

Schmidt D, A, Burgmann, R, Time-dependent land uplift and subsidence in the Santa Clara valley, California, from a large interferometric synthetic aperture radar data set, Journal of Geophysical Research Vol 108, p. 1-13, 2003