Templeta generale

2.4.WP 1310 Application to co-seismic test-cases (Pakistan)

In Annex A to [DA4], a draft paper submitted to Geophysical Journal International was provided. Following the reviewers comments this paper was accepted for publication, and improved significantly concerning the aspects discussed below.

We modified our model, which now includes two NE-SW faults, in addition to the two NW-SE faults we had assumed in our first submission. The latter are still found to contain most of slip (peak values of 120-130 cm), although the former are indeed non negligible (peak values of 50-60 cm).

We came to this complex model after analyzing three "basic geometries", namely two main NW-SE faults (our first submission), two main NE-SW faults and two conjugate faults. We added an Appendix B in the manuscript, in which we analyze in detail the performance of these models in terms of: residual RMS to the SAR observations (Table S1, Fig. S11); ability to reproduce the surface deformation patterns (Figs. 4, S8, S9, S10); comparison with the two available GPS stations closest to the epicentral area (Table S2).

Each of the basic models has one or more significant limitation in fitting the SAR data, as detailed in Appendix B, whereas GPS and seismicity alone provide only loose constraints to support one or the other solution. The complex faulting model we propose in our revised paper overcomes most of the limitations concerning fitting of the SAR measurements and is in better agreement with the few GPS measurements.

Moreover, we compared our SAR data and our models with the coseismic GPS data from Khan et al. 2008. We modified the modelling paragraph (section 5) accordingly, and added Tables 2 and S2.

In addition, in the final version of our paper we compare our model with the seismicity. In particular we used the seismic data of W. Szeliga, who carried out double-difference relocations of the main events of the sequence. We now plot alongside our modelled fault traces in Fig. 8 and our slip distributions in Figs. S4 and S5 and comment on agreements/disagreements in the text.

As shown in Fig. S3 and Fig. 7, the mainshock hypocentres are located quite close, only one/two kilometres away from our main (NW-SE oriented) fault planes. However a significant discrepancy is present concerning the depths: in fact the relocated hypocentres are located at 21 and 23 km depth. In contrast our modelled faults slip at shallower depths (about 3-13 km). Regarding the December aftershock (Fig. S4), the relocations are in very good agreement with our fault plane (Fig. 7), however, also in this case, our seismic source slips at shallower depths (1-5 km) with respect to the hypocentre location (10 km).

Finally we improve all plots, especially the 3D ones, following the reviewers suggestions.

2.5.WP 1310 Application to co-seismic test-cases (Lunigiana)

On 2013 June 21 a Mw 5.1 earthquake struck the Lunigiana area (North-western Italy). Although this area is not present in the project test case list, we applied the MUSA project experience and interferometric tools to study the Lunigiana earthquakes. The mainshock is characterized by quasi-pure normal dip-slip mechanism and occurred along a ~45° N-dipping fault. Given the fault geometry and orientation, the expected hanging wall coseismic displacement can be expected to be equally distributed in its vertical (downward) and horizontal (Northward) components, both consisting of a few centimetres of displacement. Despite the moderate magnitude of the earthquake, we applied both DInSAR and MAI to retrieve the LoS and azimuth displacement components. While a 3 cm line-of-sight displacement was measured with DInSAR, the MAI noise floor was too high to observe any deformation pattern potentially of seismic origin. A scientific paper based on this work was submitted to the Tectonophysics journal and is currently under review. All details about the DInSAR and MAI processing as well as the major findings of this work are discussed in detail in the submitted manuscript, reported in Annex A.

2.6.WP 1310 Application to co-seismic test-cases (Kefalonia)

On 2014 February 03, a Mw 6.0 earthquake struck the Kefalonia Island (Western Greece), only 8 days after a Mw 6.1 mainshock located slightly more to the east. Although this area is not present in the project test case list, we applied the MUSA project experience and interferometric tools to study the Kefalonia earthquakes. The focal solutions shows a righ-lateral strike-slip mechanism along a high angle ~N-S oriented fault. The kinematics and geometry of the supposed fault plane, together with the quite large magnitude of the seismic event, suggested the main deformation component to be N-S, and above the MAI and offset tracking measurement noise floors. Indeed this proved to be the case, and the MUSA developments are proving essential to correctly interpret this complex deformation pattern. All details about the DInSAR processing and the interpretation of measured signals are discussed in an internal report, attached in Annex B. The work is a joint effort with the National Observatory of Athens, and shall lead to the preparation of a scientific paper in the next two months.

2.7.WP 1320 Application to inter-/post-seismic test-case 1 (L'Aquila)

The TD3.3 dataset, consisting of 29 ascending COSMO-SkyMed images, was processed with the StaMPS software [Hooper et al., 2007]. 28 interferograms were formed respect to a master acquisition on Aug. 19, 2009. Due to the very high temporal sampling, cfr. Figure , a very high density of Persistent Scatterers was found throughout the image frame.

The mean deformation rate obtained with a conventional processing run in shown in Figure , top-left. On one hand this map shows a relatively strong subsidence (shown in red) in areas which have been previously associated to the post-seismic displacement of the Mw 6.3 L'Aquila event, of Apr. 6, 2009 [Bovenga et al., 2010], [Lanari et al., 2010], [Reale et al., 2011], [D'Agostino et al., 2012]. On the other hand, the strong positive line-of-sight component measured in the top-right and bottom-left image corners is unexpected and appears to be correlated with topography.

It could be possibile for this signal to be entirely due to seasonal changes in atmospheric stratification. To investigate this the methods discussed in section 2.3 were applied.

The SAR-based correction method, section 2.7, was applied by generating an estimation mask based on the uncorrected mean velocities map and multi-temporal coherence. All points with mean uncorrected velocities > > 0 mm/yr (i.e. not subsiding) and with a _x > 0.6 were initially selected for model parameter estimation. These were then assigned to height bins of 50 m between minimum and maximum height s of 0 m and 3000 m respectively. The 40 most coherent points were retained for each bin.

The ECMWF approach of section 2.8 was applied, extracting pressure-level products at 06:00 UTC (SAR acquisitions are at around 05:00 UTC), for a large are covering the whole of central Italy.

Finally also a cascade approach was attempted, in which the results of the ECMWF correction were provided as an input to the SAR-based correction approach.

The results of these three approaches are shown in Figure : top-right is the SAR-based correction approach, bottom-left the results of the ECMWF correction method, bottom-right the outcome of the cascade approach.

All correction procedures greatly reduce the strong uplift signals seen in the uncorrcted mean-velocity map. The fact that this signal indeed represents an artifact is confirmed by a comparison with 10 GPS stations, for which SAR measurements could be derived and for which a complete GPS time-series is available, spanning the SAR acquisition intervals. Table summarizes the results of the SAR vs. GPS comparison. Figure through Figure show the time-series at each of the GPS stations.

For every single station, the residuals with respect to GPS are smallest for the SAR-based approach. On average a 6.7 mm residual is observed, compared to the 10.1 mm of the uncerrected time-series. The ECMWF and cascaded EMWF+SAR-based approaches show actually slightly worse residuals compared to the uncorrected results, of 11.7 mm and 13.1 mm. This is probably due to a combination of bias-errors, apparent in Figure , and to a greater amount of uncorrelated noise at each epoch. This can be seen at all slowly-moving stations, e.g. COCA in Figure .

However, a qualitative comparison with the SAR-based correction results (Figure top-right) reveals that some important deformation gradients, hidden in the uncorrected results, are indeed revealed also in the ECMWF-corrected maps, albeit with a lower accuracy compared to the SAR-based approach. However, this result is of great interest considering that the ECMWF correction procedure is operator-independent and can in principle be completely automated. Finally, the cascade approach does not provide any advantage over the SAR-based one, while loosing at the same time the main advantage of the ECMWF approach (i.e. automation). It should therefore not be pursued further.

Figure TD3.3 Perpendicular baseline (y-axis) vs. temporal baseline (x-axis).

Figure (Top-left) Uncorrected mean post-seismic velocities (Top-right) After application of the SAR-based correction method of section 2.3.1 (Bottom-left) After application of the ECMWF correction method of section 2.3.2 (Bottom-right) After application of the ECMWF correction followed the SAR-based correction.

GPS station	Uncorrected (mm)	SAR-based correction (mm)	ECMWF correction (mm)	ECMWF + SAR-based correction (mm)
AQRA	5.9	5.6	5.7	6.2
CADO	10.4	6.8	11.4	12.9
COCA	13.1	7.0	16.0	18.9
LNGS	13.4	9.9	15.1	16.9
MAMA	9.6	8.1	11.5	13.5
PAGA	8.4	5.6	8.8	9.6
RDCA	14.6	8.2	17.8	18.8
ROFA	12.8	5.1	17.2	18.1
ROIO	6.6	5.2	7.9	9.8
SMCO	6.6	5.9	5.9	6.2
Mean	10.1	6.7	11.7	13.1

Table Standard deviation of SAR - GPS displacements.

Figure GPS vs. SAR time-series at the AQRA station.

Figure GPS vs. SAR time-series at the CADO station.

Figure GPS vs. SAR time-series at the COCA station.

Figure GPS vs. SAR time-series at the LNGS station.



Figure GPS vs. SAR time-series at the MAMA station.

Figure GPS vs. SAR time-series at the PAGA station.

Figure GPS vs. SAR time-series at the RDCA station.



Figure GPS vs. SAR time-series at the ROFA station.



Figure GPS vs. SAR time-series at the ROIO station.

Figure GPS vs. SAR time-series at the SMCO station.