Fault-model parameters estimation using a feature-voting technique: Dead Sea fault as a case study

Abstract

Estimation of fault-model parameters is a challenge faced by researchers. Traditional statistical tools rely on a velocity field derived from precise coordinates of control-points in GNSS networks, changing over-time. This requires dealing with geodetic datum and deformation-monitoring datum, while handling ill-conditioned state models.This paper describes an algorithm for evaluating full fault-model parameters all at once using a feature-voting technique, while relying on precise GNSS vectors computed with scientific GNSS processing software. The algorithm bypasses the datum issues and deals with numerical instabilities using perturbed data replications. Dealing with gross-errors is an embedded attribute of the technique, and noise is an intrinsic feature needed to cope with the ill-conditioned equation-set problem with respect to the low-frequency nature of the data. Thus, the geodetic datum and deformation monitoring datum becomes irrelevant for this algorithm.Simulations of a locked fault-model, based upon a true solved epoch, derive proper parameter values for a preselected fault-model scenario, including simulations with noise. A case study relies on the northern section of the Dead-Sea fault, on 10 northern sites in Israel's continuously operating permanent stations, based on four true-data epochs of computed vectors in order to evaluate Dead-Sea fault model parameters.Similarity is shown between computed fault-model parameters using the presented feature-voting technique and those estimated by traditional processes; with respect to other researchers' work, at the same area over a similar time period, carried out by a traditional statistical least-squares adjustment procedure, including efforts in gross-error filtering and coping mechanisms with the biases caused from both datum issues.The feature-voting based methodology for analyzing fault-model parameters, using precise GNSS vectors and not precise coordinates; enables simultaneous full fault-model parameter computation while avoiding biases originating in deformation-monitoring datum and other error factors.