What is this stress triggering about ?

Earthquakes release part of the stress that slowly accumulates as the earth's plates move toward or past each other. An earthquake drops the stress on the fault which slipped, so that earthquake will not recur until the stress rebuilds, typically hundreds to thousands of years hence (these regions are colored blue in all of our illustrations).

But an earthquake also raises the stress elsewhere, at sites off the slipped fault hence (the red regions in our illustrations). All other things being equal, the regions where the stress rises will be the sites of the next earthquakes to occur, both large and small. That's our approach in a nutshell.

We calculate these 'Coulomb' stress changes, and find that aftershocks and subsequent mainshocks tend to occur where the stress rises, and are largely absent where the stress drops. This tendency is strongest immediately after the triggering shock, and fades over the ensuing decades.

Thus, our work explores the 'conversation' between earthquakes on nearby faults, so that learn how one event can promote or inhibit earthquakes on other faults. We also apply these tools to the interaction of earthquakes and volcanic eruptions. This arises because the stresses imparted by an earthquake can squeeze a magma chamber at depth and open conduits to the surface, permitting magma to ascend to the earth's surface.

Earthquake probability

There are so many features of earthquake behavior that we do not understand that the best use of our limited insight is to 'play the odds,' in other words to calculate the probability of future earthquakes and their uncertainties. This permits the hazard of one fault or city to be compared with another, and the threat of earthquakes to be compared to other hazards, such as pollution, storms, or industrial accidents.

Probabilistic seismic hazard assessments typically assume that earthquakes are uncorrelated in space and time (each shock is a dart throw on a map, where one throw has no influence on the next). Occasionally, one assumes that the probability of an earthquake drops after large event, but does not rise elsewhere.

But clustering of earthquakes in space and time is the outstanding feature of seismic catalogs and prehistoric earthquake occurrence, and such clustering is incompatible with such an approach, because it implies that the prospect of an earthquake rises after an event. Stress triggering overcomes this deficiency, and offers a new approach to improve seismic hazard assessments. Our probability calculations thus account for earthquake interaction.

Criticisms and Counter Arguments

First, we can not measure the stress changes in the earth. Instead we model them by treating the earth's crust as a uniformly stiff block of rubber, so how can this possibly represent the behavior of the real earth? We find remarkably good fidelity between the model calculations and observations of seismicity, so we are encouraged that the calculations are sufficient to learn something important.

Second, the stress rise is tiny (as little as 1/4 bar or about 1/8 the pressure you put in your car tires). Reply: So the stress changes can not cause earthquakes of any size, they can only trigger them.

Worse yet, we don't know how close any of these faults are to failure. So how can one in any sense "predict" earthquakes, when we can calculate only the small fluctuation on a much larger driving stress? We borrow from laboratory studies of rock undergoing simulated earthquakes and find that the small stress changes, because they are sudden, cause large changes in the rate and thus likelihood of earthquakes. So even though the perturbation is small, its effect is large.