A Preseismic Strain Anomaly Detected Before M6 Earthquakes in the South Iceland Seismic Zone From GPS Station Velocities

Abstract

The South Iceland Seismic Zone (SISZ) accommodates E‐W translation due to oblique spreading between the North American and the Eurasian plates in South Iceland with ruptures on N‐S faults. Strain is released in earthquake sequences that last days to years, at average intervals of 80–100 years. Two M6.5 earthquakes struck in the SISZ in June 2000, and two M6 earthquakes in May 2008. These events released only half of the strain accumulated since the last earthquake sequence in 1896–1912. GPS station velocities are estimated from annual campaigns and continuous measurements in the SISZ during 2001–2015. The GPS station velocities are used to calculate strain rates from a new method using the vertical derivatives of horizontal stress. Our new strain rates are obtained by integrating vertical derivatives of horizontal stress rates, rather than differentiating interpolated GPS velocities, allowing finer resolution of strain. Estimating the strain rates for eight time intervals, we find high strain rates in the SISZ (0.5–1 microstrain per year) with significant temporal and spatial variation. The strain rates in the SISZ reveal contraction in the Hengill area and dilatation over a large area in eastern SISZ, indicating inflation of the magma system beneath Hekla volcano. A prominent strain anomaly is evident in the epicentral area prior to the May 2008 earthquakes. We suggest that this signal is primarily caused by plate motion and an increase in the rate of contraction in the Hengill area, located ∼10 km west of the epicentral area.

Slow Slip Event Detection in Cascadia Using Vertical Derivatives of Horizontal Stress Rates

“…The a priori probability function has contributions from minimizing , that is, a constraint on the VDoHS rates to be no larger than the minimum necessary to fit the observations, as well as (i.e., a constraint on the short‐wavelength variations in the VDoHS rates, and the same constraint on horizontal derivatives of the vertical velocity). The probability function that is minimized is described in detail in section of Haines et al (), and see section S3 in the supporting information of Árnadóttir et al (). The damping factor coefficients for these three components of the a priori probability functions are determined by constructing the marginal probability function for the observations from their joint probabilities with the VDoHS rates, and then finding the damping factors that maximize the marginal probability of the observations.…”

“…In each horizontal equation, the first two terms are dependent on the horizontal velocities, u x and u y , and their horizontal derivatives as well as the stress‐strain parameters, while the remaining force terms and are the VDoHS rates. Table outlines step‐by‐step the approach we use to obtain VDoHS rates, which is also described further in the following section, and in greater detail in Haines et al (), Dimitrova et al (), and Árnadóttir et al ().…”

New Zealand‐Wide Geodetic Strain Rates Using a Physics‐Based Approach

“…The previous methods lack the constraints provided by including the physics of elastic behavior near the Earth surface, and to compensate have often been applied with strong damping, which results in highly smoothed strain rates, typically giving mismatches to the GPS data of ~2 standard deviations or larger. We outline the fundamentals of the method below, although we refer the reader to Haines et al (), Dimitrova et al (), and Árnadóttir et al () for more detail on application of the method to GPS velocity fields. An analytical approximation for the Haines et al () method has been developed by Sandwell and Wessel () to create gridded 2‐D velocity fields in the gpsgridder algorithm in GMT.…”

“…We use a Bayesian inversion for the coefficients F i that best fit the GPS velocity variation from the elastostatic solution . A detailed discussion of the Bayesian inversion, which is based on the principal of maximum entropy (e.g., Jaynes, ; Akaike & Krishnaiah, ) can be found in Dimitrova et al (, see “Inversion Process” section), and (Árnadóttir et al, ; see supporting information section S1). The a priori probability function has contributions from minimizing , which constrains the VDoHS rates to be no larger than the minimum necessary to fit the observations; we also minimize , to avoid short wavelength variations in the VDoHS rates (both of these constraints impose some degree of regularization of the results).…”

“…Because we integrate the VDoHS rates to obtain strain rates, there is no amplification of spatial domain noise, unlike other methods that rely on taking derivatives of displacement fields. The VDoHS method has been implemented in Iceland (Árnadóttir et al, ) and the North Island of New Zealand (Dimitrova et al, ) to produce a high‐resolution view of tectonic and volcanic deformation in those locations. It has also been adapted for use with GPS time series to resolve transient slow slip events and locking processes at the Cascadia subduction zone (Haines et al, ).…”