Seismic event detection in suburban Chicago using a single broadband
seismic station
Abstract
On November 4, 2013, residents near a quarry in the western suburbs of
Chicago felt shaking from a rare, small earthquake. The USGS reported a
magnitude of 3.2 and Dr. Robert Herrmann reported a dip-slip source
mechanism from analyzing surface wave amplitudes recorded by USArray
stations. With the goal of detecting potential aftershocks in this
region of low seismicity and possibly gaining more insight into the
source mechanism, a broadband seismic station was installed in the
source region by researchers of Northwestern University. Due to the
station’s suburban setting and proximity to various transportation
arteries, industrial operations, and city infrastructure including a
deep tunnel and reservoir, detecting and discriminating small
earthquakes from urban noise events poses a serious challenge. Average
daily noise levels can be 50 dB above typical noise levels for broadband
seismometers in Illinois in pertinent frequency bands, so aftershock
signals can be buried deep within the noise, rendering typical STA/LTA
detection methods relatively ineffective. A preliminary analysis of
several months of waveform data identified seismic signals from
~1000 events. None of these events occurred on a Sunday
or at night, implying an anthropogenic origin and further illustrating
the challenge. Recorded signals from these events span a wide range of
waveforms, rendering popular detection methods like template matching
less effective than in other settings. We aim to define and engineer a
set of waveform features to aid with seismic event detection using data
from a single broadband station in a noisy, urban environment. To
identify useful spectral parameters, we first computed power spectral
density (PSD) estimates using segments ranging from the hour-scale to
the second-scale. Week-long spectrograms of the PSD estimates revealed
characteristic frequencies that are likely associated with routine
quarry operations. Select features were then tested for their ability to
detect regional and local seismic events for one month of data. We will
present the results of this analysis, including the performance of
several features and discuss their respective benefits and limitations
for seismic event detection in an urban environment.