Applications: Earthquakes and Tectonics (5)

CHARMING is a feasibility study funded by the European Space Agency's (ESA) Support to Science Element (STSE) Pathfinders 2013 project. The context of CHARMING is Probabilistic Seismic Hazard Assessment (PSHA), i.e. the scientific field which aims to quantify the probability that ground motion at a specified site will exceed some level of a given shaking parameter of engineering interest (e.g. Peak Ground Acceleration, PGA) during a specified future time frame. The end-users of PSHA represent a broad community including people concerned with land-use planning, seismic safety provisions of building codes (for the design of buildings, critical facilities, and lifelines), disaster preparation and recovery, emergency response, and organizations that promote public education for mitigating earthquake risk.

The project's main objective is to investigate whether surface deformation measurements, derived from GPS and Synthetic Aperture Radar data, can be successfully incorporated into PSHA models and improve their quality. In particular, we propose to investigate several aspects related to the marginal benefit provided by the integration of SAR compared to GPS alone, since to our best knowledge this project represents the first attempt worldwide to incorporate SAR measurements into hazard models.

As a test-case we consider a large portion of central and southern Italy. Within the project we shall generate interseismic velocity maps over this area, using a combination of ~100 permanent GPS stations and measurements derived from coast-to-coast strips of ERS-1/2 AMI, ENVISAT ASAR and ALOS PALSAR imagery. The latter are processed with state-of-the-art multi-temporal InSAR algorithms, including MERIS and ERA-Interim tropospheric delay corrections when applicable, and accounting for recently discovered error sources, such as the oscillator drift of the ENVISAT ASAR sensor. As a secondary project objective, the application of two novel SAR measurement techniques is also explored, namely the Intermittent Small Baseline Subset (ISBAS) technique, previously applied only in the context of landslides, and a multi-temporal extension of the Spectral Diversity or Multi Aperture InSAR technique, to measure also the flight-path (azimuth) deformation component. Measurements from individual SAR tracks are calibrated using GPS, whenever the spatial distribution of the latter allows a sufficiently robust estimation of long-wavelength error model parameters. Furthermore, a cross-calibration procedure to exploit the overlap of adjacent near-parallel tracks can be applied, provided the look-angle differences between near and far range are sufficiently small.

Within the area of interest, we identified a set of composite seismogenic sources, extracted at first from the DISS v.3.1.1 catalogue (Database of Italian Seismogenic Sources). For each source, the slip-rate and other key parameters shall be estimated inverting the surface displacement measurements. Well-established relations from seismology shall then be used to derive a recurrence model for each source, describing the number of earthquakes in a given time period above an established magnitude threshold. Finally, state of the art PSHA modelling techniques will be used to generate probabilistic models for PGA and other shaking parameters. These shall in turn be validated statistically, using data from national accelerometric and macroseismic intensity databases. The differences compared to existing PSHA models of Italy, based only on seismological data, shall be analyzed in detail.  

The Mid-Term Review results presented here include a demonstration and a validation of all the measurement and modelling techniques envisaged for the project, the only limitation concerning the extent of the area of interest, which shall be restricted to the Abruzzi region (200 km x 200 km). In particular, we shall discuss the requirements on SAR data and processing techniques, in terms of coverage, measurement accuracy, spatial resolution and sensitivity, to provide a meaningful contribution for this application in a challenging environment such as the Apennines.

We have investigated the post-seismic effects of the L’Aquila earthquake, that hit the city and the surround region on April 6, 2009. For this purpose, we have applied a multi-temporal InSAR technique and we have exploited the very high resolution SAR images acquired by the X-band sensor on board of COSMO-SkyMed (CSK) Italian satellite constellation. The series of CSK interferograms have allowed to retrieve the surface displacement field for a time span of 480 days after the main event.

In particular, we have focused our attention on the understanding whether the measured Line of Sight (LOS) deformation might have been produced only by tectonic afterslip or if other causes have occurred.

Data from previous studies and  literature have highlighted the reactivation of pre-existing deep seated gravitational slope deformations (DSGSD) after great earthquakes. For the considered case study, the deformation map has showed a clear pattern of surface movements on the Mt. Ocre (south-west from L’Aquila city centre; figure 1)  which, compared with geomorphological evidences on the Mt. Ocre ridge, has allowed to  interpret the result as gravitational processes triggered by past seismic activity,  that produced sackung type displacements. This suggests that part of the detected deformation is ascribable to gravitative processes triggered by the April 6 2009 mainshock.

In order to confirm such hypothesis, a Finite Element Analysis of two cross sections along Mt.Ocre ridge has been performed. Two different analyses have been conducted considering both the elastic and the elastoplastic rheology of the involved materials.

In particular, the numerical results from the elastoplastic rheology model are in good agreement with the measured data, highlighting also a nonlinear deformation of the Mt. Ocre ridge because of the yielding material. These outcomes enforce the hypothesis that the part of the postseismic deformation of Mt.Ocre ridge is caused by reactivation of pre-existing deep seated gravitational slope. 

The creeping segment of the San Andreas Fault extends from the Cholame plain in the South to the San Francisco Bay Area in the north. This segment is bounded to the north by the southern end of the rupture of the 1906 M7.9 San Francisco earthquake and to the south by the 1857 M7.9 Fort Tejon rupture. In between, no significant earthquake (M>7) has been recorded or inferred from geological records, although moderate ruptures have been reported in the past 200 years. As creeping sections of active faults are considered as key in the initiation, propagation and termination of seismic rupture, careful assessment of the degree of fault coupling is needed. Although several models have already suggested a certain amount of elastic strain accumulation along the creeping segment, choices of modeling, including unphysical regularization of the inverse problem, obscure our estimates of the uncertainties derived from geodetic data. We use high resolution velocity fields derived from Synthetic Aperture Radar (SAR) and Global Positioning System (GPS) to infer a probabilistic estimate of fault coupling along the central San Andreas Fault using a fully Bayesian approach.

We process 4 sets of SAR images from the ALOS satellite covering the creeping section between 2006 and 2010 to generate 4 Line-Of-Sight (LOS) maps of ground displacement rate. We generate interferograms using the Stanford MoComp processors. We remove atmospheric delays by estimating an empirical relationship between phase and topography among multiple spatial wavelengths. Finally, we generate 4 maps of the average ground velocity using the MInTS algorithm, solving for a secular term and seasonal oscillations in the wavelet domain. We extract a subset of GPS-derived velocities from the Unified Western US Crustal Motion Map. GPS data have been collected from 3700+ campaign sites and 1300+ continuous sites over the 1993-2010 period from different data centers, including UNAVCO, SCEC, NCEDC and USGS. Daily solution were produced with GAMIT and processed with QOCA to extract the secular velocities.

We sample the posterior Probability Density Function of slip along the San Andreas and Calaveras Faults, using a massively parallel implementation of CATMIP optimized for the use of Graphics Processing Units. Our error model includes a careful description of correlated noise in the InSAR data and an estimate of the effects of uncertainties on the stratification of the crust. Our results show a 70-km-long rapidly creeping section (>2 cm/yr), embedded between two transition zones with average creep rates (~1 cm/yr). These transition zones are characterized by an intense micro- and moderate seismic activity and by the presence of locked asperities, consistent with the occurrence of M6 earthquakes in the past. The rupture areas of the 1857 and 1906 appear locked. The main improvement brought by the use of a Bayesian approach is the ability to quantify the robustness of our model based only on the data and the prior knowledge. We are able to provide probabilistic answers to simple questions. For instance, there is 50% chance that the equivalent slip deficit along the rapidly creeping section exceeds 1 cm/yr. In addition, there is a 70% probability that fault coupling is higher than 0.5 along the small coupled regions identified north of the creeping section. 

Our fault coupling model suggests a gradual transition from a rapidly creeping segment in the center to locked regions to the north and the south through transition zones that exhibit a certain roughness including moderate size locked asperities and high microseismic activity. As these zones may have been involved in the initiation of past major ruptures and because of the significant strain accumulation along the entire section, our results calls for a re-assessment of the role of the creeping segment on the seismic history and on the future ruptures of the San Andreas Fault.

The 1992 Landers, California earthquake was the first of over 100 earthquakes to date to be studied with InSAR. There are many benefits to the use of InSAR data for obtaining source parameters of earthquakes, including the finding of accurate, effectively ‘ground-truthed’ locations, sensitivity to fault geometry, the capability to map the distribution of slip on a fault at high resolution, and the ability to measure certain parameters, such as the length of fault rupture, directly. We have compiled such earthquake source parameters, including details of fault geometries, locations and slip distributions, from over 130 published studies. We demonstrate the utility of the resulting earthquake catalogue, named the InSAR Centroid Moment Tensor (ICMT) catalogue, for addressing current problems in geophysics, by applying it to three distinct problems.

First, we use the catalogue as a means of testing the robustness of earthquake source parameters obtained by seismological means. In general, we find good agreement between fault geometric parameters (strike, dip and rake) between the ICMT catalogue and the Global Centroid Moment Tensor (GCMT) catalogue, derived from long-period teleseismic data. Estimates of seismic moment agree well between the ICMT and GCMT catalogues, with no evidence of any systematic bias or dependences on region or mechanism type.

Second, we test the robustness of different tomographic Earth velocity models by using them to model long-period teleseismic waveforms, and comparing the source locations, geometries and moments thus generated to the ICMT values. We find that there are significant differences in performance between several of the different Earth models in terms of the similarity of the fault geometry and moment produced, but also that no current Earth model is capable of matching the ICMT locations, suggesting that all of the models are failing to capture some degree of heterogeneity in the crust and/or mantle. Preliminary tests suggest that the ICMT location can be reconciled with the GCMT location by adding additional time shifts to the seismic waveform data for specific source-receiver paths, potentially providing a means for constraining mantle heterogeneities that affect those paths.

Finally, we use the information on fault dimensions, slip and moment in the ICMT catalogue to address problems of earthquake scaling. Considering events of all mechanisms together, we find a scaling relationship between moment (M₀) and fault length (L), such that. M₀ is proportional to L^1.81. We find differences in this power law exponent with mechanism type, with thrust events showing an exponent of ~2, consistent with L-model scaling (e.g. Scholz, 1982). We additionally find a tentative relationship between the slip-to-length ratio for events and fault slip rate. Low slip-to-length events (0.4–4 x 10^-5) include subduction earthquakes and events occurring on strike-slip faults with fast slip rates (> 5 mm/yr). The high slip-to-length events (1–3 x 10^-4) include several events on faults with slow slip rates (< 2 mm/yr).

With the Sentinel-1 mission promising rapid post-event coverage and limited temporal decorrelation, study of moderate and large magnitude continental earthquakes with InSAR is likely to become increasingly possible, and increasingly routine. We suggest that this expected increase in earthquake models derived from InSAR data will provide a rich dataset for validating earthquake catalogues and studies of Earth structure and earthquake scaling in future.

We have investigated the liquefaction phenomenon occurred on May 20, 2012 Emilia earthquake (ML 5.9), induced by the water pressure increase in the buried and confined sand layers, exploiting the Differential Interferometry Synthetic Aperture Radar (DInSAR) capability to detect and estimate their surface displacements. The level-ground liquefaction is the result of a chaotic ground oscillation caused by the earthquake shaking and the observed failures are due to the upward water flow caused by the excess of pore pressures. Here, using a SAR dataset composed of four COSMO-SkyMed X-band SAR images, covering the period April 1 – June 6, 2012, amultidisciplinary (remote sensing,geological and geotechnical) approachhas been appliedin order to relate local changes detected by DInSAR measurements to the observed liquefaction phenomena.

The detection of induced subsidence measured by the DInSAR phase, and of surface changes revealed by SAR derived features (i.e., interferometric coherence and intensity correlation), allowed the identification of zones affected by differential compaction occurred in urban areas, probably associated to a volumetric deformation of the ground induced by liquefaction of saturated sandy deposits. Complex coherence and intensity correlation revealed themselves capable to identify different sources of changes, respectively small random changes in urban settlements and abrupt changes due to urban damage or soil ruptures.

Geological-geotechnical analysis has been also performed in order to assess if the detected SAR-based surface effects could be due to the compaction induced by liquefaction of deep sandy layers. With this regard, the results obtained from 13 electrical cone penetrometer tests show the presence of a fine to medium sandy layer at depths ranging between 9 and 13 m, which probably liquefied during the earthquake, thus inducing vertical displacements between 3 and 16 cm. The quantitative results from geological-geotechnical analysis and the surface punctual effects measured by DInSAR are in good agreement, even if some differences are still present, probably ascribable to the local thickness and depth variability of the sandy layer, or to lack of deformation detection due to DInSAR decorrelation.

One of the most relevant results is the capability of this approach to detect settled areas where no surface effects, sandy boils or cracks, can be observed through ground surveys. Moreover, with the already operational SAR satellite constellations, such as COSMO-SkyMed and TerraSAR-X, and the next launch of the second Sentinel-1 satellite (two satellites working for 7 years with a revisit time of six days), more outcomes are expected. A faster processing of data might also provide a fundamental contribution to the management of seismic crisis, giving a more complete and detailed information on the “event scenario” in order to provide authorities with elements for a rational planning of the emergency.

During the round table, seed questions proposed by the chairs will be discussed with the audience.