Meta-analysis of fault slip rates across the central Apennines for seismic hazard assessment

Recent seismic hazard models are increasingly relying on fault slip rates as the fundamental quantity for translating the activity of a fault source model into earthquake rates. In this conversion, modelers are often tasked with selecting different estimates and alternative methods to assess the fault slip rates and related uncertainties and incorporate them into the seismic hazard analysis. In the central Apennines, several techniques, such as paleoseismic trenching, mapping of offset geomorphic markers, and dating of scarp profiles have been used to determine slip rates of normal faults. Recently geodetic data have also been used to determine the first estimates of the slip rate (loading) rate on active faults.

Combining measurements obtained with different methods remains challenging because non tectonic processes can introduce noise or spurious signals that are elusive to quantify, and these influence slip rate estimates. After careful and planned data collection, we argue that a rigorous meta-analysis is required to quantify erratic fluctuations and method-related variances. In this case, throw rates are overdispersed with respect to nominal uncertainties in throw and age; therefore, they are commonly affected by unmodeled noise processes to be rigorously quantified for seismic hazard assessment.

Geodetic data can provide slip-rate estimates with a model of elastically unloading seismogenic faults within a viscously deforming lithosphere. However, short-term transients can also infect geodetic data in the central Apennines. Such transients can be isolated and subtracted by time series or included as noise in the long-term covariance matrix; otherwise, the resulting spatial distribution of deformation rates locally fits short-term transients. In some cases, strain rate peaks represent the currently unclear signal of tectonic processes like crustal visco-elasto-plastic deformation and aseismic slip or indicate missing faults in the adopted database. In the central Apennines, we have proved that reasonable estimates of long-term fault slip rates can be extracted even at signal-to-noise ratios of order unity using a more sophisticated modeling approach, including the stress orientations. For well-sampled faults, the slip rate estimates fit the corresponding geological estimates, leading us to conclude that they can be considered for seismic hazard models in regions such as the Apennines.

We remark that geodetic and geological data can be used together to highlight (and possibly model) both the likely occurrence of short-term transients in GPS time series and the existence of non-tectonic processes contributing to the progressive surface exposure of active faults. Given the current understanding of temporal and spatial fault throw rate variability in the central Apennines, producing complex input models for seismic hazard assessment is still not feasible. A base model with a uniform throw rate along the trace (tapering to zero at unconnected fault tips) and merging information from offset features of different ages to constrain a single time-independent rate is still the most reasonable.