SCEC Award Number 18109 View PDF
Proposal Category Individual Proposal (Data Gathering and Products)
Proposal Title Testing the Quality of Geodetic Fault Slip Rates by Explicit Consideration of Spatial Correlations
Investigator(s)
Name Organization
Michael Floyd Massachusetts Institute of Technology
Other Participants Gareth Funning
SCEC Priorities 1a, 1b, 5c SCEC Groups Geodesy, CXM
Report Due Date 03/15/2019 Date Report Submitted 05/10/2019
Project Abstract
 Although potentially minor in comparison to epistemic uncertainties, the inclusion of spatial correlations between data when solving for model parameters such as fault slip rate is important to consider. This is also particularly important as projects such as the Community Geodetic Model progress to produce dense geodetic solutions including both GNSS and InSAR velocities. With such a density of points, assumption of independence and lack of accounting for spatial correlations (a) inherent in the data and (b) as a consequence of physical phenomena that generate and perturb the observed velocities will over-average and under-estimate the formal uncertainties associated geophysical parameter estimates, which may lead to meaningful discrepancies with other methods or put too much confidence in . Just as temporal correlations between data points in a GNSS time series is known to increase the magnitude of velocity uncertainties by a several (e.g. 3–5) times, so accounting for spatial correlations between geodetic data points is likely to result in the uncertainties associated with geophysical parameters such as fault slip rates to increase by several times. Correlations between model parameters such as creep rate and locking depth are also likely to introduce a factor of up to three when estimating moment deficit accumulation rate; however, as the parameters are negatively correlated, this may still be a useful approach to estimating such a quantity, bearing in mind these limitations.
Intellectual Merit This project addresses the impact of neglected quantities in the data covariance matrix and overlooked correlations between estimated geophysical model parameters. Commonly, geodetic data is assumed to be independent where, in reality, it is likely to be correlated both due to the system by which it is observed (e.g. GNSS, InSAR, etc.) and by the correlations associated with physical phenomena, such as a fault's deformation field. This is even more important where deformation fields overlap or otherwise interact, creating more correlations within the model by which we choose to describe the phenomena.
Broader Impacts The impacts of this work effect the confidence with which we pass geophysical quantites to probabilistic seismic hazard assessments, such as UCERF. Common and fundamental discrepencies between geologic and geodetic estimates of meaningful geophysical parameters such as fault slip rates is also addressed, as the full bounds of uncertainties are important for both these methodologies.
Exemplary Figure Figure 3: Estimates of slip rate and locking depth (top row) for the 100 uniform creep rate (left column) and 100 variable creep rate (right column) experiments. For the uniform creep rate experiment, the input slip rate is 25 mm/yr and locking depth is 15 km. Equivalent comparison of input versus estimated moment accumulation rate (per unit length along strike) is shown in the bottom row for both experiments.