Using small events to constrain the physical mechanism governing slow slip

In conventional models, faults can slip in one of two ways: steadily at rates near plate rate or episodically in earthquakes. However, in the past two decades researchers have discovered another behaviour: slow slip events. In slow slip events, large portions of the plate interface slip accelerate periodically but then stall long before reaching seismic slip rates. The puzzle is why: what physical mechanism allows for slow yet episodic slip?

Episodic slow slip is important to understand because slow slip events occur just below the locked zone, which hosts hazardous megathrust earthquakes. Constraints on slow slip can help us constrain how the earthquake-generating region is loaded to failure. Slow slip may even help us understand how large earthquakes start, as the physics governing the beginning of slow slip is likely the same physics that governs the beginning of large earthquakes.

Slow slip is also important because it is abundant. M>6 events have been observed on most well-instrumented subduction zones, accommodating large fractions of the plate interface slip at depth. If we are to understand the seismic cycle and the stress state at subduction ones, we need to understand slow slip.

A number of physical mechanisms have been proposed to explain slow slip. And a number of these models have reproduced the slow yet episodic slip, usually by allowing for complications in the fault rheology or spatial variations in the fault properties. However, these models have not appeared capable of reproducing all the properties of observed slow slip events. Specifically, the models are designed to be slow and relatively stable, so they struggle to reproduce the complexity that is increasingly observed within a slow slip event. It thus seems that we need a new or modified model. The challenge is that there are many plausible models to consider. So in this project, I propose to evaluate several models and to identify several model properties that are essential to reproducing observed slow slip.

To better evaluate these models, I propose to compare them with existing observations of large slow slip events as well as new analysis of data from the more numerous small slow slip events in central Cascadia. Several tens of small slow slip events occur annually in Cascadia. They are accompanied by seismic tremor: low-amplitude, sustained seismic vibrations thought to be composed of numerous tiny earthquakes driven by the slow slip event. We will track the tremor locations, so that we can see how the small slow slip events grow in space. In addition, we will estimate the magnitude of aseismic slip in the small events, using the unique network of high-precision PBO borehole strainmeters in Cascadia. We will use these new analyses to evaluate several numerical models of slow slip and constrain essential model characteristics.

With the proposed research, we seek to understand how slow slip events work. Slow slip events are an important part of the seismic cycle at subduction zones, which includes hazardous M 8-9 megathrust earthquakes. Slow slip events often occur just down-dip of the zone that generates major earthquakes. That means that each slow slip event increases the stress in the seismogenic zone. That stress has the potential to trigger large earthquakes. It may even be that large earthquakes begin as slow slip events, and that some slow slip events eventually grow and become seismogenic, rupturing up to the trench. In this case, we should be examining properties of slow slip events to see if we can decipher some characteristics that determine the eventual behaviour.

On the other hand, we should also accept that slow slip events may not be useful for predicting large earthquakes. For instance, in Cascadia the stress at the lower limit of the seismogenic zone likely does accumulate mostly during slow slip events, so it seems most likely that the next earthquake will start during a slow slip event. However, there are several hundred slow slip events between megathrust earthquakes, each one lasting several weeks, so the knowledge of the small increase in stress may not be useful in mitigating the time-dependent risks associated with earthquakes at subduction zones---unless there happens to be some property of slow slip events that reveals the current stress regime.

The time-independent risks associated with the slow slip region may be a more useful line of investigation. An important question is whether megathrust earthquakes can rupture into the slow slip region. The slow slip region is located farther inland that the locked zone. That means it is closer to and sometimes directly under large cities. If the slow slip region ruptures in an earthquake, the shaking in these highly populated area could be much larger than if only the locked region regions. Whether or not the slow slip region ruptures in current models depends on which model you use---especially on whether the model is inherently stable and on what stresses exist on the plate interface. If we are to usefully predict whether or not the slow slip region will rupture seismically, we need to know which model we should be using.

To summarise, we currently do not know whether slow slip events pose or tell us about the hazard introduced by earthquakes. Since we often observe only the largest slow slip events, it is not practical to apply hazard analysis techniques in an empirical, statistical sense. It would seem that if we are to effectively use slow slip in hazard analysis, we need to understand what causes it.

Given the current state of slow slip understanding, the knowledge gained here will be initially most useful to the regional geologic surveys---in this case, the Canadian Geological Survey and the United States Geological Survey. As we make progress in understanding what slow slip tells us about earthquake hazard, the information will likely be first applied in estimating the risk in well-instrumented regions, such as the US, Japan, New Zealand, and Costa Rica. Scientists working on earthquake hazard quantification---from the geological surveys and academic institutions---routinely attend conferences and workshops. The most effective communication is likely through the normal academic channels: formal and informal discussions and published papers.