"Using a genetic algorithm to estimate earthquake slip distributions from point surface displacement"

Don't know if that or this will make too much sense but:

The growth of corals around Sumatra are affected by changes in sea level... and for the most part theses are caused by the tectonic motion. Evidence of when the earthquake happened and how big it is is recorded in their growth rings (kind of like trees rings),- that is the point surface displacement

You can calculate how much the earth moved (i.e. the level of slip) to produce a certain amount of vertical displacement. Because the corals record extends over hundreds of years it means it is possible to estimate how far the earth slipped in earthquakes hundreds of years ago.

Other people have done this before to estimate the size earthquake and level of slip on a resolution of hundreds of kilometres. My algorithm does it at a resolution of 20 km.

You can use a brute force approach to this... that is try hundreds of millions of different earthquakes until you find a set of events that fits the data but that takes about a couple of days.

I was able to design a genetic algorithm that treats the parameters describing an earthquake model as a set of genes. I made a load of these earthquake models at random and then 'bred' them to evolve a set of models that fit the data. I was able to use parallel programming techniques to squeeze out a result in about 15 mins.

*****

I got sent the manuscript proofs from the Journal of Geophysical Research on friday evening... it means my technique is now peer reviewed and due to be published in the next few weeks... which should make my thesis defence a bit easier