The LA Story - October 24, 2015
If you live in Los Angeles, and you don't live in a cave, you've probably seen this story about an earthquake forecast for the Los Angeles area. The controversial piece of the story is basically that the 99.9% probability computed by the methods on this web site is higher than the quoted probability of 85% probability computed by the US Geological Survey. Both of these probabilities are for a circular region of radius 100 km around La Habra, California, over the time period of 3 years from about June 2015 to June 2018. The 99.9% probability was computed for a paper published in Earth and Space Science, a peer reviewed journal of the American Geophysical Union. The paper appeared online this past week.
Although these two probability calculations seem similar, they are based on very different methods. The USGS probability is based on a pretty massive model of interacting earthquake faults combined with "expert opinion". Their model solves for about 250,000 potential ruptures constrained by about 37,000 data (as per their method), thus the model is underconstrained.
More importantly, their probability of 85% is a time-independent probability, rather than being the type of time-dependent probability used on this web site. Lucy Jones of the US Geological Survey has explained the differences this way:
"Jones said the NASA-led study uses a method that predicts higher numbers of strong earthquakes in the future because fewer than expected have struck in the past. That’s very different from the USGS model, which says the probability of an earthquake doesn’t vary from one year to the next."
To emphasize this latter point, we use the hazard viewer to compute two 1-year forecasts for the same circular region, using the Forecast Timeseries tool on the Hazard Viewer. The top figure is for an M>5 earthquake within the region, and the bottom figure is for a M>6 earthquake within the region. It is clear from looking at the figures that the time-averaged probability for M> 5 events over the past 7 years is about 55%, even though the current probability for the next year is about 62%. For M>6 earthquakes, the current forward-looking 1 year probability is 4.7%, lower than the average value of 5.5%.
The same general conclusions will be true of the 3-year probabilities, meaning that the time-averaged probability for 3-years is probably closer to 90% than the current value of 99.9%.
The fact that earthquake probabilities are time-dependent is a concept taught in all introductory geology classes at the university level, so it should come as no surprise that time-dependent values of probability can differ from time-independent values.
Figure above is for M>5 earthquakes within a 100 km circle at right, for the 3 years from June 2015 to June 2018.