Item

Abstract

We improved near-realtime processing of seismic data by creating an automated workflow for seismic events detection, location and spectral parameter estimation in a form of a Python package. For detection task we take advantage of EQTransformer - deep neural network picker supported by GaMMA phase association. For source location we use both NonLinLoc (picking/travel-time based algorithm) and BackTrackBB software (partial waveform stacking algorithm). The source parameters including seismic moment, source size, static stress drop, apparent stress and radiation efficiency are calculated using spectral fitting method, where the inverse problem is solved using Markov-Chain Monte-Carlo method. Package performance was tested on the data from the LUMINEOS local surface network composed of 24 sensors located in Legnica-Głogów Copper District (LGCD), Poland. The region is one of the most seismically active in Europe, with several thousand seismic events induced every year (M>0.4) by copper ore mining and maximum local magnitudes reaching M4.2. We used our workflow to analyze the seismicity framing strong mining rockburst in the LGCD on November 29, 2016. This tragic M3.4 event caused 8 fatalities and massive tunnels collapse. We aim to recognize plausible spatio-temporal changes of the source parameters revealing the source contribution to the damaging potential of this induced earthquake sequence. The obtained results can be used not only for scientists but also mining engineers, who are responsible for safety in the mine.