Next generation analysis of earthquakes and earthquake catalogs with InSAR

Recent advances in the acquisition and availability of satellite InSAR observations provide new research avenues for researchers to pursue. I use these observations to develop a geodetic earthquake slip distribution catalog to develop and test earthquake scaling relationships using geodetic observations. I also use machine learning algorithms to solve big data challenges in the geodetic community.

I use earthquake slip distributions from geodetic observations to determine earthquake scaling relationships. Earthquake scaling relationships describe how earthquake parameters such as fault area and the amount of slip relate to the earthquake magnitude. I show that relationships produced from geodetic data provide different values than traditional seismological data.

I use machine learning algorithms to classify InSAR interferograms containing surface deformation. I train the algorithm using millions of synthetic interferograms and with modeled surface deformation and noise. The machine learning algorithm is capable of correctly classifying surface deformation in 99.74% of synthetic interferograms and 85.22% of real interferograms. In the cases of correct classification, the algorithm identifies the location of the surface deformation within the interferogram.

I use a machine learning algorithm to remove noise from InSAR interferograms. I train the algorithm using hundreds of thousands of synthetic interferograms with modeled surface deformation and noise. The machine learning algorithm is able to accurately remove the noise in the interferograms and reasonably reproduce the surface deformation in the image.

These studies demonstrate the applicability of machine learning algorithms to InSAR analysis and is a first step toward integration with systematic and operational earthquake analysis.