Slow-slip events and small earthquake clustering - implications for the locked region of the shallow Hikurangi subduction zone

Technical Abstract

Slow-slip events have been observed along the Hikurangi subduction margin since the continuous GPS network began to be established along the North Island east coast in 2002. These events appear at the surface as a change in direction and rate of motion at survey points monitored by continuously-operating GPS instruments. The change may last from days to months, after which the previous motion resumes. The signals are inferred to result from slip on a patch of the subduction interface fault. The slip rate between the two sides of the fault is on the order of 50 mm/day or less, which is orders of magnitude slower than the slip rate during a normal earthquake. Hence the name "slow-slip event" or "slow earthquake".

The slow-slip events release some of the strain energy that is being accumulated in the region of the plate interface as a result of the inexorable motion of the Pacific Plate beneath the North Island. The amount of energy that is released slowly is therefore unavailable for release in a future earthquake on this part of the subduction interface. If slow-slip events (SSE) release strain energy on parts of the plate interface that we have previously expected to be involved in great earthquakes, then either the moment or the frequency of such earthquakes would be less than we previously believed.
Slow-slip events appear to be located near the down-dip end of the well-coupled part of the subduction interface, in the transition between stick-slip behaviour above and continuous deformation below. By mapping the locations of the slow-slip events we can potentially map the well-coupled region of the interface. This provides information about the extent of the well-coupled region, at least as it appears at the present time. With proper interpretation, this can assist in estimating the maximum size of a future earthquake.

The occurrence of a slow-slip event causes stress changes at shallower depths on the subduction interface. From the data collected to date by ourselves and others, it appears that these stress changes may influence the occurrence of small earthquakes both close to the slip event and perhaps on a more regional basis. These effects are not yet well understood.

A slow-slip event causes an increase of the stress towards failure of the well-coupled part of the subduction interface just above. It is therefore possible that a slow-slip event could trigger a large earthquake at shallower depth on the plate interface. This phenomenon has not yet been confirmed, but its possibility means that the monitoring of slow-slip events as they occur may be of importance. For example, deep, slow-slip has been suggested by others to have occurred just prior to the 1960 Chilean (Mw 9.5), and 1944 and 1946 Nankai subduction earthquakes (each Mw 8.2). Stress triggering has been well demonstrated on crustal faults in various locations around the world, with the North Anatolian fault being a classic example.

For all these reasons we have undertaken a multifaceted investigation of slow-slip events in New Zealand over the past two years.

We have modelled the large 2004-05 Manawatu slow-slip event, and have detected and modelled three other events that occurred within the project period to define their location, depth, duration and magnitude. Two additional events started near the end of the project period.

The SSEs along the east coast occur at very shallow depths (10-15 km) and tend to be quite short (days to weeks). Those under Manawatu and along the Kapiti Coast are much deeper (30-50 km) and longer (many months). We have attempted to determine what causes the variation in depth. It has been assumed elsewhere that temperature is a primary control on the depth at which slow-slip occurs, but we show that this is not the case for the Hikurangi margin.

We have worked towards mapping the extent of the strongly-coupled zone of the subduction interface. We have used both the locations of SSEs and inferences from microearthquake locations together with results of seismic tomography, and we find that these give reasonably consistent results. A speculative model is proposed in which the permeability of the rocks in the overriding plate may affect the degree of coupling and hence the location of SSEs.

We have developed an automated method of detecting the onset of SSEs soon after they begin. The algorithm is running every day in test mode. It is based on the cross-correlation between the observed GPS time series and a large set of candidate SSEs.

It was suspected prior to this project that the 2003-04 Kapiti Coast SSE had triggered the occurrence of moderate (M 4-5) earthquakes beneath Upper Hutt in 2004 and 2005. We have not recognised similar triggering for any other New Zealand SSE to date. However, it has been discovered in a companion project that very small earthquakes (M -2) do occur in conjunction with the Gisborne SSEs. Also, preliminary results from other projects suggest that the Gisborne earthquake of 20 December 2007, which occurred within the subducting Pacific Plate, may have triggered a slow-slip event on the plate interface above.