UNSCENTED KALMAN FILTER BASED INVERSION OF GEODETIC DATA

In seismotectonic and seismic hazard studies, aseismic slip behavior forms a significant body of literature. Transient slow-slip deformations have proceeded some volcanic eruptions, and deep aseismic slip preceded the largest subduction zone thrust earthquakes. Monitoring of slow-slip events plays a significant role in seismotectonic and seismic hazard studies. Owing to the contaminating effects of multiple contributing signals such as fault-slip, local benchmark motion and measurement error, interpretation of displacement time series resulting from aseismic events is not straightforward. In the literature, time dependent inversion algorithm based on the Extended Kalman Filter (EKF) has been introduced as a common inverse method for geodetic data. This method can separate the various signals and allow the space-time evolution of slow-slip transients to be studied in detail. EKF suffers from a number of disadvantages. In this paper, Unscented Kalman filter (UKF), which is able to overcome the mentioned disadvantages by employing Unscented Transform (UT), has been applied. While EKF filter is computationally unstable, the results of UKF-based inversion on a data set for a simulated uniform aseismic slip transient proved to be computationally stable.