COSMO-SKYMED AND SENTINEL-1 DINSAR PROCESSING FOR GROUND INSTABILITY MONITORING IN INDONESIA

Fabio Bovengaa, Alberto Reficea, Antonella Belmontea , Raffaele Nutricatob, Davide Oscar Nittib, Maria Teresa Chiaradiab, Sotirios Valkaniotisc, Sofia Gkionic, Chrysanthi Kosmac, Athanassis Ganasc, Paolo

Manuntad, Elizar Elizare, Darusman Darusmane, Philippe Ballyf

aResearch National Council of Italy, ISSIA institute, Via Amendola 122, 70126 Bari, Italy; bDepartment of Physics “M. Merlin”, University of Bari, Italy; cNational Observatory of Athens (NOA), Athens,

Greece; dCollaborative Space Ltd, Dundrum, Ireland; eSyiah Kuala University, Banda Aceh, Indonesia; fEuropean Space Agency, ESA-ESRIN, Frascati, Italy

ABSTRACT Indonesia is periodically affected by severe volcanic eruptions and earthquakes, which are geologically coupled to the convergence of the Australian tectonic plate beneath the Sunda Plate. Multi-temporal SAR interferometry (MTI) can be used to support studying and modelling of terrain movements. This work is aimed at performing an analysis of ground displacements over Indonesian sites through MTI techniques. Two test sites in Sumatra and Java have been selected according to the availability of archived SAR data, GNSS networks, and geological data. Both COSMO-SkyMed (CSK) and Senitnel-1 data-sets have been processed through MTI algorithms. The derived displacement maps have been interpreted according to the available geological and geophysical information.

Index Terms— Multi-temporal SAR Interferometry, Ground displacement monitoring, Tectonics.

1. INTRODUCTION

Indonesia is periodically affected by severe volcanic eruptions and earthquakes, which are geologically coupled to the convergence of the Australian tectonic plate beneath the Sunda Plate [1]. Multi-temporal SAR interferometry (MTI) allows monitoring ground displacements over wide areas, thus supporting the modeling of both tectonic motion associated with faults and volcanic processes related to magma movement [2,3]. The main advantages of InSAR are the synoptic view of wide areas, and the periodic surveying that guarantees long time monitoring and time series analysis [4].

The study of tectonic phenomena poses challenges in MTI processing. The kinematics in this case is usually extremely slow, thus requiring high accuracy in the final deformation map, as well as a wide coverage in order to track deformations along the whole fault profile. Such large-

scale spatial analysis makes critical the removal of artefacts coming from atmosphere and orbital inaccuracies [3]. Finally, the presence of strong vegetation coverage, as in case of Indonesia islands, further adds hindering factors. In these cases, MTI processing can be improved by using additional information coming e.g. from geodetic data (GNSS networks) or atmospheric modeling.

This work is aimed at performing an analysis of ground displacements over Indonesian sites through MTI techniques. According to the abovementioned limits, the selection of the test sites has been performed based on the following requirements: i) presence of ground instabilities, possibly related to onshore active faults or volcanoes; ii) good expected interferometric coherence, according to both topography and land use; iii) availability of a reliable archived interferometric SAR dataset, iv) availability of ancillary geophysical data, such as from GNSS stations. According to these requirements, the following areas have been selected for MTI analysis: 1) the Banda Aceh region in Sumatra, hereafter referred to as Banda Aceh test site 2) the Yogyakarta area and Merapi Volcano in Java, hereafter referred to as Java test site.

For the Banda Aceh only CSK data were available. The processed stack is made of 50 images SAR images acquired between May 2011 and October 2016 in Stripmap mode, along descending orbits, with a mean incident angle of 32.2°. For the Java test site, reliable datasets are available from both COSMO-SkyMed and Sentinel-1 missions. The CSK data-set is mad by 50 images acquired between December 2011 and November 2016 in Stripmap mode, along ascending orbits, with a mean incident angle of 29.3°. The Sentinel-1 data is mage by 41 images acquired between 14.10.2014 and 31.05.2017 in Interferometric Wide Swath (IW) mode, along ascending orbits, and with a mean incident angle of about 33°.

Displacement maps have been obtained by processing the CSK and Sentinel-1 datasets though SPINUA algorithm [5], which performs Persistent Scattering (PS) analysis.

486978-1-5386-7150-4/18/$31.00 ©2018 IEEE IGARSS 2018

2. PROCESSING AND RESULTS ANALYSIS The tectonic analysis in Indonesia is difficult because the vegetation cover in the area causes lack of PS along and across the faults. The use of GPS data, which is important in such settings, was here hindered since the records available are incomplete and unreliable due to limited observation time span.

The use of two MTI processing chains and of datasets coming from both CSK and Sentinel-1 constellations, allows cross-validating final results. The processing of Sentinel-1 data has been more complex with respect to that of CSK data, as standard MTI displacement maps showed strong artifacts, likely due to residual atmospheric contributions and orbital errors. This was imputed to the lower number of images and the shorter timespan covered by Sentinel-1 with respect to the CSK data. In order to overcome this problem, an alternative processing scheme has been experimented, basically consisting in adding a step that filters the differential phase fields composing the stack in order to

preliminarily reduce atmospheric phase contributions. In particular, a high pass spatial filter has been designed, which works on the complex interferograms, thus avoiding phase unwrapping, and possible related errors. The size of the filter spatial window can be set according to the residual signal to be removed, and to the expected ground deformation. In the present case we used a size of 2 km x 2 km. This approach, supported by successful results obtained on other test sites, allowed to obtain displacement maps in Java from Sentinel-1 data consistent with those from CSK data.

The displacement maps derived over Banda Aceh and Java test sites are sketched respectively in Figure 1 (top panel) and Figure 2 (CSK result is in left panel, Sentinel-1 results is in the right panel).

In Banda Aceh, high resolution data from CSK provide good density of measurable targets on the ground, thus improving the delineation of the spatial deformation pattern, and increasing the chances to capture signals related to local

Figure 1. (Top) PS average LOS velocity maps over the Banda Aceh test site, obtained by processing CSK data. Areas

affected by local instabilities are indicated by white circles. (Bottom) PS average LOS velocity maps within the area labeled as A2 in the figure on the top. The inner panel shows the displacement temporal trend of a coherent scatterer.

487

instabilities. Nevertheless, due to the presence of vegetation in the region, several areas are still lacking PS.

In Java test site the PS point coverage is significant, due to the presence of urban and small settlements and agricultural activity in the area, resulting in the removal of most vegetation cover. Recent volcaniclastic deposits and areas scourged from recent eruptions around Mt Merapi volcano favour conditions for persistent scatterering.

Concerning the tectonic activity in Sumatra, due to the vegetation cover, PS targets are lacking just along the two main fault segments in the area (sketched as red lines in Figure 1), thus hindering a detailed analysis of tectonic activity. The Aceh segment (the western fault in Figure 2) is assumed to be presently inactive [6], and, at least where PSI data are available along this segment, the deformation rate is always between −2 mm/y and +2 mm/y, indicating that the area can be considered as stable. Concerning the Seulimeum segment (the eastern fault in Figure 2), by using the geodetic records reported in [6], we derived an expected LOS velocity related to the fault movement of about 6 mm/y. However, due to the scarce presence along and across this fault segment of PS measurements, it is not possible to provide a reliable evidence of the tectonic movements.

Besides the use of interferometric results for assessing the tectonic activity, ground displacements (mainly subsidences) have been identified both in Banda-Aceh region and in the Java test site, which basically reflect local effects. The causes of these displacements were investigated by using ancillary geological data, and, for the Banda Aceh test site, also by performing in situ inspections. The subsidence phenomena are mainly related to the presence of unconsolidated coastal/alluvial sediments and groundwater pumping.

Figure 1-B shows an example in Banda Aceh concerning a coastal area. The Inner panel shows a PS displacement temporal trend. This part of Banda Aceh was overrun and completely destroyed by the 2004 tsunami. Most subsidence points are positioned on port facilities structures and embankments. Extensive rebuilding and new constructions in the area add weight to the unconsolidated sediments. There is also an extensive presence of seasonally flooded crops (rice etc.) and salt production flats. This suggests that the subsidence occurring in the area is probably related to compaction of sediments and/or recent artificial fill.

Figure 3, shows examples of local displacements derived by using both CSK and Sentinel-1 data in Java. A subsidence can be observed in Yogyakarta urban area (panes (A) and (B)) induced by groundwater exploitation and soft sediment compaction, a result of major urban expansion and human activity during the last years. The high resolution of CSK data allows catching interesting details over urban structures as sketched in the inner panel of Figure 3-A. Interesting localised subsidence rates are also observed upon recent (2010) volcanic deposits at the flanks of Mt Merapi (panes (C) and (D)), a result of compaction of the soft explosive products (pyroclastic density currents and pumice). Panel (E) shows the deformation profile along the southern flank of Merapi volcano obtained from the COSMO-SkyMed data.

4. CONCLUSIONS

Multi Temporal Interferometry (MTI) in Indonesia is

difficult because the vegetation cover in the area causes lack of PS targets. Moreover, the GPS records available in the selected are, at the moment, quite limited and incomplete,

Figure 2. PS average LOS velocity maps over the Java test site, obtained by processing CSK data (A) and Sentinel-1 data

(B). Areas affected by local relevant instabilities are indicated by white circles. Red lines represent tectonic faults.

488

thus resulting unreliable for supporting MTI processing. In this work, MTI techniques have been refined and used for processing both high resolution X-band CSK data and medium C-band resolution Sentinel-1.

MTI displacement analysis seems to confirm the inactivity of the Aceh fault segment, as foreseen by geodetic studies. Ground displacements (mainly subsidence) have been identified both in Banda-Aceh region and in the Java test site, which reflect local effects basically related to compaction of unconsolidated sediments of different natures. Results have been validated by geological experts, and in situ inspections.

4. ACKNOWLEDGMENTS

Work supported by ESA project titled “Integrating SAR interferometry and GNSS for studying tectonic processes in Indonesia” (contract 4000114611/15/F/MOS), ESA ITT AO/1-7864/14/F/MOS, Alcantara Study reference 14-P16 “Alcantara Study Enhanced Tectonic Characterization for Indonesia”. COSMO-SkyMed data are provided in the framework of the ESA CAT-1 Third Party Mission (TMP) proposal ID 33378.

5. REFERENCES

[1] W. Hamilton, “Tectonics of the Indonesian region,” U.S. Geological Survey Prof. Paper 1078, 1979.

[2] E. Sansosti, F. Casu, M. Manzo, and R. Lanari, “Space-borne radar interferometry techniques for the generation of deformation time series: an advanced tool for Earth’s surface displacement analysis,” Geophys. Res. Lett, 37(20), L20305, 2010.

[3] A. Hooper, D. Bekaert, K. Spaans, K. and M.T. Arıkan, “Recent advances in SAR interferometry time series analysis for measuring crustal deformation,” Tectonophysics, 514-517, pp. 1-13, 2012.

[4] R. Hanssen, R., Radar Interferometry: Data Interpretation and Error Analysis, Kluwer Academic Publishers, Dordrecht, The Netherlands, 2001.

[5] F. Bovenga, A. Refice, R. Nutricato, L. Guerriero, and M.T. Chiaradia, “SPINUA: a flexible processing chain for ERS / ENVISAT long term Interferometry,” Proc. of ESA-ENVISAT Symposium, September 6-10, 2004, Salzburg, Austria. ESA Special Publication SP-572, 2005.

[6] J. F. Genrich, Y. Bock, R. McCaffrey, L. Prawirodirdjo, C.W. Stevens, S.S.O. Puntodewo, C. Subarya, C. and S. Wdowinski “Distribution of slip at the northern Sumatran fault system,” J. Geophys. Res., 105(B12), pp. 28327–28341, 2000.

Figure 3. (Top) PS average LOS velocity maps around the Yogyakarta city (area labelled as A1 in Figure 3) in the Java test site, obtained by processing CSK data (A) and Sentinel-1 data (B). (Bottom) PS average LOS velocity map on the flanks of the Merapi volcano (area labelled as A2 in Figure 3) in the Java test site, obtained by processing CSK data (C) and Sentinel-

1data (D). (E) Profile along the southern flank of Merapi volcano.

489