Dear Friends,

 

The latest issue of EOS, Earth and Space Science News magazine published by the American Geophysical Union, has an interesting discussion about the use of probabilistic models in earthquake prediction

 

https://eos.org/opinions/reality-check-seismic-hazard-models-you-can-trust

 

https://eos.org/project-updates/seismic-hazard-assessment-honing-the-debate-testing-the-models

 

These models assume the exact values of the parameters of the corresponding probability distributions, and since there is often not enough data to determine these values (e.g., in regions with few earthquakes), some crude estimates are used. As a result, the quality of the predictions is often not good, major earthquakes sometimes occur in areas where the current probabilistic models predicted  probability close to 0.

 

The first cited article shows that in some situations when we do not know exact values of the probabilities, deterministic models lead to better predictions. However, the conclusion that we should completely switch to deterministic models is not welcome by many other geophysicists who correctly point out that in other situations, when we have enough data to determine the corresponding probabilities more accurately, probabilistic models are helpful.

 

As of now, the discussion is mostly probabilistic vs. deterministic models, but in my opinion, the fact that we sometimes have a good knowledge about probabilities and in some cases, don’t, seems to indicate that this is a potential area of application of imprecise probability techniques – techniques that cover both situations with known probabilities and with unknown ones as particular cases and allows to combine these two different types of data.

 

Googling for “"imprecise probability" for earthquake prediction” finds some papers that applied imprecise probability to earthquake prediction, such as:

 

http://ac.erikquaeghebeur.name/content/AvC-2011-ISIPTA-poster.pdf

 

http://ac.erikquaeghebeur.name/content/AvC-2011-ISIPTA-abstract.pdf

 

https://books.google.com/books?id=fb7DAQAAQBAJ&pg=PA31&lpg=PA31&dq=%22imprecise+probability%22+for+earthquake+prediction&source=bl&ots=2VLt95v7cG&sig=bnRnTB_WzgtX1zDi7X56cxihQ5g&hl=en&sa=X&ved=0CB4Q6AEwAGoVChMInKXIw9nixgIVyRuSCh181gTq#v=onepage&q=%22imprecise%20probability%22%20for%20earthquake%20prediction&f=false

 

This may be a good opportunity to promote our ideas for the earthquake geophysics community – maybe some authors can write a letter to the editor to EOS or comment on the web – EOS papers allow for web comments.