Ecv ice Sheets Technical Proposal

Applicable documents

ID	Document	Reference	Version
DA 1	ESA Invitation to Tender ESRIN/AO/1-7616/13/I-BG. Support to Science (STSE) - PATHFINDERS	PFL-POE/2013/978/BG/cb	20 Sep. 2013
DA 2	Statement of Work (SoW), Appendix I to [DA 1]	EOP-SA/241/DFP-dfp	Issue 1, Rev. 0, Sep. 2013
DA 3	Draft contract, Appendix II to [DA 1]
DA 4	Special Conditions of Tender (SCT), Appendix III to [DA 1]

Chapter 2: Technical Proposal

1Introduction

1.1Understanding of technical requirements

1.1.1Background

The fundamental element that drives all preparedness and prevention activities in seismic risk management, is seismic hazard. The latter is commonly evaluated using probabilistic procedures (e.g., Cornell, 1968; Senior Seismic Hazard Analysis Committee [SSHAC], 1997). Probabilistic seismic hazard assessment (PSHA) provides 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, peak ground velocity, PGV, spectral acceleration, SA, at different periods) during a future time frame (exposure time). The end-users of PSHA include 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.

PSHA is essentially performed in two main steps: first, a seismic source (or earthquake rupture forecast, ERF) model has to be defined to estimate the spatial distribution and rates (or probability) of occurrence of forthcoming earthquakes; then, a ground motion predictive model (GMPM) is needed to evaluate the expected ground shaking at the site due to the occurrence of a given earthquake.

Besides classical seismic source models, constituted by area source zones where seismicity is uniformly distributed and activity rates are derived from earthquake catalogues only, in recent years, also ERF models based on fault sources have been developed. In particular, fault-based PSHAs were performed in California (Field et al., 2009), New Zealand (Stirling et al., 2012), Europe (SHARE http://www.share-eu.org and central Italy (Peruzza et al., 2011). A fault-based PSHA model aims to parameterize the larger and known faults of a region in terms of physical geometry and slip rates to evaluate the long-term rate of all earthquakes above a specified threshold of magnitude. Generally, the activity rates of fault-based PSHA models are derived primarily from geologically-estimated fault slip rates. For example, in the framework of the EU project SHARE, as well as in central Italy, average long-term slip rates for individual faults were selected from a wide range of published rate values, including paleoseismic datasets and topographic offsets. In few cases, geodetically-constrained slip rates were considered. This is, for example, the case of the Working Group California Earthquake Probability (WGCEP, 2007), which also used geodetic data to constrain the strain rates for the crustal shear zones and the slip rates for some active faults (Field et al., 2009).
Crustal deformations measured by geodetic techniques have the potential of reducing the large uncertainties of geologically–estimated slip rates (including paleoseismology), thus leading to a significant improvement of seismic hazard assessment. In particular, the space geodesy techniques of GNSS and InSAR are the only means to measure crustal deformation over large areas with the precision, coverage, temporal and spatial sampling required for incorporation into PSHA models. Over the past 20 years these techniques have repeatidly demonstrated their capability of measuring the crustal deformation associated with all phases of the seismic cycle (Salvi et al., 2012): from the preparatory, stress-loading, inter-seismic phase, to the co-seismic phase, when a portion of the stress is suddenly released during seismic fault dislocations, to the post-seismic phases, in which the stress changes are redistributed throughout the crust, triggering aftershocks and other effects. Since the deformation occurring in the interseismic phase can be exploited by PSHA models, it is the one of greatest interest for this project.

GPS techniques are more mature than InSAR, and the current state-of-the-art can provide the horizontal components of the mean interseismic velocity with sub-millimetric accuracy on a nation-wide basis (Devoti et al., 2011), whereas a lower (centimetric) accuracy can be attained for the vertical component. Their spatial resolution however, is limited by the density of the network, which in the Italian case for instance is in the order of 30 km. Current multi-temporal InSAR techniques (e.g. Ferretti et al., 2011, Hooper et al., 2008, Berardino et al., 2002) are complementary to GPS in these aspects, offering higher spatial resolutions, and improved sensitivities to the vertical deformation component. On the other hand these measurements are 1-dimensional, since, on each satellite pass, only the displacement component in the radar line-of-sight (LoS) is measured. Furthermore, applicability of the techniques is subject to a much higher degree of variability compared to the GPS case, which is influenced by data availability and by the significance of poorly modelled error sources, mainly related to atmospheric propagation, phase unwrapping, and other sources causing image-wide error trends, which can be mistaken for interseismic deformation (e.g. orbital uncertainties and sensor oscillator drift). Current remedies to these technical problems include exploiting all the available data redundance (Ferretti et al., 2011) and integrating external information from numerical weather models (e.g. Jolivet et al., 2012) and GPS (e.g. Manzo et al., 2012).

To exploit the above mentioned complementarities of SAR and GPS for the retrieval of the interseismic displacement vector, several integration techniques have been proposed (e.g. Wang et al., 2012, Guglielmino et al., 2011). The fundamental problem which cannot be overcome however, is that even in the best case in which ascending and descending SAR acquisitions cover the area of interest, only two independent deformation components are measured. Therefore any integrated measurement approach with the ambition of preserving the high spatial density of SAR measurements, while deriving all three deformation components, ends up performing some kind of interpolation of the GPS north-south measurements. This can be unjustified in areas with low GPS density and/or complex fault structures, where the north-south component may in principle have a high spatial variability.

1.1.2Current and planned external initiatives

Improvement of seismic hazard evaluations is the goal of several organizations in all countries affected by high seismicity rates. While SHA is normally approached as a national matter, several initiatives exist, which are dedicated to spreading the best practices for SHA and risk analysis as well as increasing the exploitation of space geodetic techniques.
The SHARE project is an EC Collaborative Project, ended in 2013, whose main objective is to provide a community-based seismic hazard model for the Euro-Mediterranean region with update mechanisms. The project established new standards in Probabilistic Seismic Hazard Assessment (PSHA) practice by a close cooperation of leading European geologists, seismologists and engineers. SHARE successfully delivered a European wide probabilistic seismic hazard assessment across multiple disciplines spanning from geology to seismology and earthquake engineering. SHARE introduced an innovative weighting scheme that reflects the importance of the input data sets considering their time horizon, thus emphasizing the geologic knowledge for products with longer time horizons and seismological data for shorter ones.  To estimate fault activity rates, however, SHARE did not consider geodetic data, but only seismological and geological data.
The Global Earthquake Model is instead a collaborative initiative including government institutions, research and academia, and private companies (mostly from the insurance sector), to promote best practices and transparent processes for the assessment of earthquake risk and hazard (www.globalquakemodel.org). In GEM the strain rate is considered an important contributing data for SHA, and GEM stimulated an update of the Global Strain Rate Model (GSRM), based on over 5000 GPS site velocities, representing the largest self-consistent set of secular GPS velocities and the best estimate of present-day plate motions and strain accumulation at a global scale, which will be released during earthquake events. Strain rate maps will in the future be an input data of the OpenQuake engine, a software platform for computing earthquake hazard and risk. Currently InSAR data are not included in the production chain.

The INSAR-based GSRM initiative (iGSRM) is an international effort aiming at building a strain rate model constrained by GPS and InSAR for the Earth’s tectonic belts. This project relies mainly on future Sentinel-1 data to generate improved, high accuracy estimates of ground deformation. The objectives are to generate high resolution strain rate and velocity fields for the Alpine-Himalayan belt, to map unknown faults, using also optical EO data, and to assess time-dependent seismic hazard following major earthquakes. The iGSRM initiative should provide strain rate data to the GEM initiative, although it shall not address the issues of improving SH assessment through GPS and InSAR data.

The CEOS Disaster Risk Management (DRM) Observation Strategy is a response to a collection of observation requirements from the user community to enable the delivery of three coordinated pilots to be implemented in 2014-2016 in three thematic areas: floods, seismic hazards and volcanoes. Each of these thematic pilots aims to serve as a showcase for the international DRM community, in particular demonstrating a) the added value and uniqueness of increased CEOS coordination in this area; b) benefits of closer ties to users and ease of access to data; c) potential for increased roles of space agencies in DRM beyond the current Hyogo Framework for Action, for the following 10-year period starting in 2015. INGV actively participated to the thematic panel on Seismic hazard, which defined three main lines of activity for the DRM Observation Strategy:

1. 
   Support the generation of globally self-consistent strain rate estimates and the mapping of active faults at the global scale by providing EO InSAR and optical data and processing capacities to existing initiatives, such as the iGSRM [Wide extent satellite observations]
   
2. 
   Support and continue the Geohazard Supersites and Natural Laboratories (GSNL) for seismic hazards and volcanoes [Satellite observations focused on supersites]
   
3. 
   Develop and demonstrate advanced science products for rapid earthquake response. [Observation of earthquakes with M>5.8]
   

The CEOS agencies will exploit the Observation Strategy by granting data access to existing projects and initiatives. Some agencies will also contribute work relating to system architecture to develop geohazard services for the global user community. One of the goals of the CEOS Strategy is the creation of a new global seismic hazard map, which should be built on a global strain rate model and a global map of active faults. These goals are long-term and will be addressed by specific projects, as the Interferometric Global Strain Rate Model (iGSRM) by Leeds University, the Global Earthquake Model (GEM) initiative, and others. In particular, these goals address the need of a better knowledge of seismic hazard levels in under-developed regions. However, in regions in which the assessment of seismic hazard has reached a good level of detail, the added value of integrating higher resolution strain data in the hazard model needs to be demonstrated.

Within the Working Group California Earthquake Probability framework (WGCEP, 2007), a deformation model was developed to assign a slip rate and an aseismic-slip factor (plus their uncertainties) to each fault section in a fault model. WGCEP developed a preferred deformation model and alternatives that are consistent with the geological slip-rate studies, as well as with geodetic data and the overall Pacific North America plate rate. Actually, the deformation models were derived primarily from geologically- estimated fault slip rates. Following previous working groups on California earthquake probabilities, WGCEP 2007 used expert opinion to select average long-term slip rates for individual fault sections from the wide range of published rates and, in some cases, geodetically constrained slip rates were considered. Finally, in the WGCEP 2007 geodetic data were also used to constrain the strain rates for the crustal shear zones.
Finally, we quote an Italian national initiative relevant to this study:
Nationwide multitemporal SAR processing: Italy is the first country to have applied operationally the multitemporal InSAR techniques. Following a law issued in 2002, a national plan for the measurement of ground deformation using the PS-InSAR technique, has been carried out. Presently, maps of Permanent Scatterers detected over the entire national territory have been generated using ERS and ENVISAT data. The PS location and mean velocities are available on the Geoportale Nazionale, managed by the Ministry of the Environment (http://www.pcn.minambiente.it/viewer/index.php?project=ps_envisat_asc). Although these data are very useful for the analysis of local deformation (landslides, subsidences, sinkholes, etc.), they cannot be used to extract tectonic deformation, since they have not been not processed to retain sufficient accuracy over low spatial frequency deformation gradients.

1.1.3Objectives of this feasibility study

This proposal describes a feasibility study named: "Constraining Seismic Hazard Models with InSAR and GPS (CHARMING)".

The project has two main objectives of broad interest in the context of the international initiatives mentioned in section 1.1.2, and one main objective of Italian national interest.

The first objective concerns regions characterized by poorly understood plate boundary zones, such as that between the Eurasian and African plates, and availability of GPS as well as InSAR data. Given the complementarities of InSAR and GPS mentioned in section 1.1.1, in particular concerning the higher spatial resolution and sensitivity to the vertical motion component of SAR, it is of interest to investigate the marginal contribution of SAR compared to GPS in improving PSHA models. This is quite a different scenario from those addressed by the iGSRM and WGCEP initiatives. The former considers larger fault systems (e.g. Tibet) in areas with sparse GPS networks, increasing the expected benefit of including SAR in strain rate calculations. The latter considers once again larger fault systems (e.g. San Andreas) in areas with very dense GPS networks, where the marginal benefit of including SAR can be expected to be low. Our study addresses contexts in which the relative importance of each technique in the retrieval of the interseismic deformation field is not at all obvious. Furthermore, unlike iGSRM, our objective is not limited to the estimation of the interseismic velocity components (or equivalently of the strain rate tensor elements), but rather it extends to analyzing the impact of these measurements on PSHA models.

A second broad objective is to investigate in more detail some common assumptions. The first of these is that InSAR methods cannot provide sufficiently accurate azimuth measurements for interseismic deformation studies; the second is that the coverage provided by multi-temporal DInSAR techniques published in the last decade cannot be improved significantly; the third assumption is that the vertical interseismic velocity component is not useful for hazard modeling. These aspects are of interest also for all the previously mentioned international initiatives. Furthermore, any finding in the scenarios considered in this study, featuring a dense distribution of small faults, shall be immediately applicable to the scenarios considered in the international initiatives. It should be noted that the opposite is not true, in the sense that methods and assumptions applicable to Tibet or California are not necessarily applicable to the Mediterranean.

Finally, an Italian national objective can be pursued as a fringe benefit of the main objectives discussed so far. In fact, Italy is located in the poorly understood plate boundary zone between Eurasia and Africa. By choosing an area of interest on the Italian territory to investigate the main objectives, we can hope to obtain an improved seismic hazard map of these areas, compared to the state-of-the-art. In fact, even only the incorporation of GPS deformations in PSHA models would represent an improvement for the Italian territory.

The objectives stated above pose several challenges to the current state-of-the art, from the deformation measurement as well as from the modeling point of view.

From the modelling point of view, this study shall devote a significant effort in developing models that can exploit spatially dense deformation data and are consistent with geological slip-rate studies as well as with geodetic rates. A key parameter that should be assigned to each fault source and used in the deformation models is the average aseismic-slip factor. To determine these factors, geodetic strain rates can be compared with geologically- and seismologically–derived strain fields. Thanks to the availability of long time series of geodetic and SAR datasets, this analysis can be conducted for several different time frames to investigate the spatio-temporal variability of aseismic-slip factors. The tectonic and geodynamic information can then be combined with geodetic data to build a deformation-based ERF model for the investigated regions. Unlike the standard PSHA approach (e.g., Cornell, 1968), we propose to estimate the activity rates from a grid of points based on earthquake-independent data, and use the available earthquake catalogues to carry out a “sanity check”, through a retrospective testing of deformation-based ERF models.
From the measurement point of view, there are multiple challenges. The first objective shall be to explore, through the development and testing of an innovative measurement technique, whether interseismic north-south deformations can be derived from InSAR data stacks, with sufficient accuracy and spatial resolution to complement the GPS measurements. The second objective is to understand if state-of-the-art multi-temporal DInSAR techniques allow the derivation of reliable interseismic deformation measurements over large areas, as opposed to the limited-area test-cases considered so far. To this end capabilities under development within ongoing projects (Italian Space Agency MuSA project) shall be exploited, concerning the incorporation of numerical weather model data in the SAR processing chain, whereas resources shall be devoted within CHARMING to explore the recently proposed Intermittent Small BAseline Subset (ISBAS) approach to improve the coverage of interseismic DInSAR measurements, as well as to exploit the available GPS stations within the area of interest to mitigate long-wavelength error terms.