Tuesday, December 22, 2015

Extrapolation without physics


I like this article because it goes into the danger of trying to extract signal from noise for a correlation.  People then want to use this correlation to predict future events.  This happens all the time in earthquake studies, but was a new thing for climate.

For earthquakes, people are so jaded by the numerous attempts that they demand physics.  The electromagnetic signals are so noisy that it is easy to extract 'some signal'.  Again, this is like trying to correlate the stock market with an infinite universe of possible things, such as football, Miss Universe mistakes, or global temperature.  You can always find a correlation.

I have great fun applying rock mechanics to earthquakes, and I'm in a very tiny minority.  This is only physics I know that applies to earthquakes, and it doesn't look good for prediction.  All earthquakes start out exactly the same, with that first quartz grain slipping in water.  Then it's a series of falling dominoes.  Will it go on or stop?  To be useful, you need a signal that maps out the future rupture plane.  Is it all at the critical displacement, ready to fall?  To predict earthquakes you need some new physics, such as advance 'micro ruptures' that can map out the zone, and prepare it for the final rupture.  This would be neat, and could be tested in the lab if they ever figure out that water is essential.  They always do those tests with dry rocks.

So, here is the testable hypothesis:  For large failures, there is an 'advance' process that aligns all the quartz grains at the critical displacement, ready for stick-slip.  At this moment, there is a detectable physical property change, such as electrical conduction, or a seismic wave guide.

Like I have said, the problem with 'politicized science' is that nobody wants to dive into a hypothesis that could be proven wrong.  Bad for the career.  You never see the busgus doing it.  

No comments: