I found this article to be very interesting-

http://www.bbc.com/news/science-environment-32472310

In summary-

Larent Bollinger and his colleagues anticipated a major earthquake to happen in the exact location that last Saturday’s event occurred. Bollinger’s team dug trenches across the main earthquake fault where the fault met the surface and “used fragments of charcoal buried within the fault to carbon-date when the fault had last moved.” They discovered that this segment of the fault had not erupted for a very long time, 1344 to be exact. They also discovered that the 1344 event was preceded by large event on a neighboring fault in 1255. They theorized that the movement from 1255 event caused strain to be transferred westward along the fault, which was finally released in 1344 (89 years later).

In 1934, a large earthquake, taking over 17,000 lives, ruptured the segment of the fault where the 1255 took place. The Nepal event that just recently happened, some 81 years later, seems to follow this historic pattern.

Another article that goes more in depth with regards to the historical seismicity of Nepal.

http://www.lebret-irfed.org/spip.php?article787

Gotta love the title- “Weather Underground.”

This article makes for an interesting read about the recent increase in unusually large earthquakes in Oklahoma caused by fluid injection wells. The article goes on to discuss the transformation of the oil state due to an accidental discovery of oil back in 1859. It also talks about how the lack of transparency of oil companies is irritating some of the citizens of OK.

Here’s a snippet:

“At least eight bills have been proposed that aim to make it difficult for communities to set their own rules for oil drilling.”

“The first oil discovered in Oklahoma was found accidentally, in 1859, in a well drilled to find salt, near present-day Salina; the oil was sold as fuel for lamps. As related in “Oklahoma Oil: Past, Present, and Future,” by Dan Boyd, the next find came in 1889, near Chelsea, where a well produced half a barrel of oil per day; it was used to treat cattle for ticks. Then, in 1897, a well drilled near Bartlesville became a major oil producer, and many others followed. Within ten years, Oklahoma was producing more oil than anywhere else in the world. Not coincidentally, in 1907, Oklahoma went from being a territory to being the forty-sixth state. The state constitution includes a legal definition of kerosene”

http://www.newyorker.com/magazine/2015/04/13/weather-underground

I found these articles/videos to be helpful in understanding how the different inversion techniques work that Chuck had talked about in class. Hopefully, you will find them useful as well.

Genetic Algorithm

http://www.ai-junkie.com/ga/intro/gat2.html

Simulated Annealing

https://www.youtube.com/watch?v=tdsTfZMqAxw

https://www.youtube.com/watch?v=iaq_Fpr4KZc

http://www.sciencedirect.com/science/article/pii/S0031920199001570

Monte Carlo Inversion

http://onlinelibrary.wiley.com/store/10.1029/2000RG000089/asset/rog1559.pdf;jsessionid=7287C1CE73229D2541651D48A7BE6A84.f01t04v=1&t=i8eihb7a&s=679c56c98ee431899290e0571863ae83469d7cea

http://www.ipgp.fr/~tarantola/Files/Professional/Papers_PDF/MonteCarlo_latex.pdf

A brief summary of each method

Genetic Algorithms (GA’s) mimic mother nature in a sense; GA’s use the same combination of selection, recombination, and mutation to evolve a solution to a problem.  GA’s begin with a randomly generated population. Each member is evaluated and assigned a “fitness” score. The fittest members of the population are kept while the rest are discarded. New members are created by combining aspects of the selected, fittest members. Add a little bit of randomness (mutation) and repeat until the solution converges. One great thing about GA’s – they can avoid being trapped in a local minima unlike other optimization methods. Anyone who can offer an explanation as to why gets two high fives!

“The Simulated Annealing method works by approximating a global optimum solution for a solution space of interest. During each cycle (iteration), the algorithm tries to find a better solution than its current solution. It looks at immediate neighbors and determines what’s better instead of randomly searching through the solution space. One problem with this one- because the technique is greedy, the solution can frequently get trapped in local minima (premature convergence).

“Monte Carlo sampling is useful when the space of feasible solutions is to be explored, and measures of resolution and uncertainty of solution are needed.” With a Monte Carlo inversion, you would first need to identify a mathematical model of the process you want to explore. Next, define the parameters for each factor in your model and create random data according to those parameters. Lastly, simulate and analyze the output of your process. SA’s, GA’s, and Neighborhood Algorithms are Monte Carlo techniques. Monte Carlo methods are favored over linearization techniques for two reasons: “First, they are numerically more stable in the optimization/parameter search stage. Second, they are more reliable in estimating the uncertainty of the solution by means of model covariance and resolution matrices.”

http://www.ipgp.fr/~tarantola/Files/Professional/Papers_PDF/MonteCarlo_latex.pdf