Seismic declustering and seismic events pattern recognition in Norway: Preliminary results using the Wolfram Mathematica machine learning tools.

Machine learning techniques are booming in many areas of science and seismology is not the exception. Many attempts have been done in order to use it for earthquake forecasting. For example and considering the broad range of tools and methods that it offers, several studies are focused on delving into the classification and prediction functions which can help to this hard task. Seismicity rates in Norway are the highest in northern Europe. Along the country, a few zones show elevated amounts of activity, including Nordland, the south-western coast and the Oslo area, among others. Some of them even present seismic swarm activity, as is the case of the Nordland area. In all the abovementioned zones, the magnitude of seismic events is low to intermediate, and strictly intraplate. So far, it is catalogued as Poissonian/memory free distributed. However, this conception may be revisited as the machine learning methods have been to our knowledge never applied to study possible pattern recognition. Using the seismic catalogs provided by the Norwegian National Seismic Network (NNSN) through the Norwegian EPOS project webpage, we developed a seismic declustering using the Gardner & Knopoff technique, centering our attention on northern and southern Norway. After the identification of the main events and aftershocks, we used prediction and sequence recognition functions implemented in Wolfram Mathematica® in order to identify temporal patterns and possible spatial locations for future events. Results show a high clustering degree using different seismic parameters (e.g. magnitude, location, date) and predicted events seem to fit well with the available catalogue.