Friday, May 20, 2011

Random Earthquake Thoughts

The following are random thoughts that may be good for someone who is on a Great Thoughts Friday mode.  Large Earthquakes are few and could really be considered as sparse objects but since we are talking about time series, shouldn't some of the tools used in detecting those in high dimensional time series be using rank minimization techniques mentioned this week ? By high dimensional time series, I don't just mean acceleration traces from different geographical locations but rather more complex data from other types of sensor networks.

By reading a little more about the potential connection between earthquakes and ionospheric measurements in Ionospheric Precursors of Earthquakes; Recent Advances in Theory and Practical Applications by Sergey Pulinets, I decided to watch for a day the ionospheric TEC measurements from JPL updated every five minutes




The Ionospheric TEC map are constructed from tomographic measurements performed between GPS satellites and some ground stations. For a day, I noticed something obvious: direct sunlight on a region increases the electron content in the ionosphere over that region albeit not uniformly. Where is the non-uniformity coming from ? I am not sure but if one follows the theme of the paper above, these non-uniformities may be linked to solar and/or geomagnetic activity and gases releases in the cases of earthquakes and therefore, they might be co-located to earth faults.




Both the TEC map and the night map seem to correlate. Now I wonder how these maxima average over time and correlate with the known faults ? or


Earthquake Density Maps for the World from the USGS)


one wonders if any of these maxima can also be correlated with geostationary observation of earth or the activity on the Sun. Over the past day, I didn't see an obvious connection with the last two. 

Let us note that the general public cannot seem to have access to previous ionospheric TEC maps

In a different direction, some folks are also looking at earthquake prediction based on a network analysis of past events. The paper is Earthquake networks based on similar activity patterns by Joel N. Tenenbaum, Shlomo Havlin, H. Eugene Stanley. The abstract reads:
Earthquakes are a complex spatiotemporal phenomenon the underlying mechanism for which is still not fully understood despite decades of research and analysis. We propose and develop a network approach to earthquake events. In this network, a node represents a spatial location while a link between two nodes represents similar activity patterns in the two different locations. The strength of a link is proportional to the strength of the cross-correlation. We apply our network approach to a Japanese earthquake catalog spanning the 14-year period 1985-1998. We find strong links representing large correlations between patterns in locations separated by more than 1000 km. We also find significant similarities in the network structure upon comparing different time periods.

The results are not overly convincing but I wonder if this type of analysis would not be amenable to a diffusion wavelet approach.

Liked this entry ? subscribe to the Nuit Blanche feed, there's more where that came from

No comments:

Printfriendly