Thread by @TheFracDoctor

How do we determine whether earthquakes are natural or induced by industrial activities? A thread,
providing a layman's perspective on our recent paper in SRL: www1.gly.bris.ac.uk/~gljpv/PDFS/Ve…

Obviously making such assessments is a hugely sensitive topic right now - the Newdigate sequence in Surrey being a prime example:

https://twitter.com/seismo_steve/status/1117702442534211585

There's normally nothing inherent in an individual earthquake that tells us if it's natural or induced. The shaking we record on our seismometers at the surface look just the same

So we can't just look at an earthquake and know that it's induced or natural. Instead, we rely on proxy factors and correlations between the earthquakes and the putative industrial cause.

Davis and Frohlich first defined some of the key criteria in their seminal paper in 1993: scits.stanford.edu/sites/default/…
We've adapted and expanded on these questions in our paper

For example - how different are the earthquakes in question to past natural earthquakes in the area. If the earthquakes are suddenly occurring much more frequently, or with much larger magnitudes, then this could be a sign that they're induced.

This can be a challenging question to answer, because we need have a good record of past earthquakes in a region. This varies a lot from place to place, and it depends on the magnitudes of the earthquake's we're interested in.

For instance, we might be sure we know about all the magnitude 5 and 6 earthquakes to have occurred in a region, since these will be widely noticed. But do we know about all the magnitude 3 earthquakes to have occurred over 100s of years?

Next question - is there temporal co-incidence between events and the industrial activity? If the earthquakes begin before the start of the activity in question, then that's a pretty good sign that events are not induced.

And temporal correlation? Coincidence just implies the events happened while the industrial activity was ongoing. Correlation is something more: correlation means that the rates of earthquakes tracked the rates of industrial activity.

For example, injection rates go up, rates of earthquakes go up. Injection rates go down, rates of earthquakes go down. Injection stops, earthquakes stop. And so on. Good temporal correlation can be a really strong indicator that events are induced.

Next question - spatial coincidence. Do the events occur in the same place as the industrial activity, or within a close enough distance. There's still a lot of debate over what distances industrial activities might be able to induce seismicity.

And it depends on the type of activity. Wastewater disposal, where millions of cubic metres of fluid are injected over decades, have been reported to have induced earthquakes at 10 - 20 km distance.

However, they typically induce earthquakes within 5 km of the site, even if seismicity then extends further away (e.g.: www1.gly.bris.ac.uk/~gljpv/PDFS/Ve…)

Events caused by hydraulic fracturing tend to be within 1 km of the well. Events associated with subsidence in conventional oilfields tend to occur within or at the edge of the field being produced.

The other challenge with answering this question is the accuracy of the earthquake location.

If an event is only located with seismometers of a national monitoring network, then the nearest stations might be 10s or 100s of km away, and the resulting location may have several km of uncertainty

Therefore we may not know if an event is right by a well, or several km away from it. This can be addressed by installing local monitoring networks, which provide much better location accuracy

We might also expect to see events occurring at similar depths to the industrial activity

However, in many cases the bulk of the seismicity occurs on faults that extend from the depth of activity into deeper layers (again, examples here: www1.gly.bris.ac.uk/~gljpv/PDFS/Ve…)

so an understanding of the pre-existing structure is vital here. Also, as with the event location, the event depth may be very poorly constrained, so may not know if an earthquake was at the same depth as the industrial activity or not.

Next question - plausible mechanism to have caused the event: Has the industrial activity in question actually affected the subsurface in a way that is likely to cause earthquakes?

We know that small rises in pore pressure can be sufficient to cause earthquakes during injection, but larger pore pressure drops are required to cause subsidence and then earthquakes from production of conventional reservoirs

This can be a tricky question to answer, because we need to have an accurate model of the subsurface, and the impact of the industrial activity on it

As you may have gathered, our ability to answer these different questions will vary - for a given site, some questions may be easily answered, some may be more tricky

and some may give very clear evidence as to events being natural or induced (e.g. if the events begin before the industrial activity), while other answers may be more ambiguous

In the original Davis and Frohlich work, these questions were all given binary answers - yes or no, and all given equal weighting in the final assessment

In our new paper, we recognise that these questions can be answered on a sliding scale, with different questions having different weights depending on how important they are, and how strong the evidence is to answer them

As a result, we don't just chart the result of our assessment - induced or natural. We also chart the "Evidence Strength Ratio", which describes the strength of evidence used to make the assessment

The Evidence Strength Ratio might increase with time as more monitoring stations are deployed, and more information is gathered