Anomaly recording station detection
process
According to Figure S1, the detection process of abnormal recording
stations is as follows:
Step1, acquisition of repeating earthquake event data. We prepare a list
of repeating earthquakes and retrieve the waveform data of the station
to be detected from the NEDBC database (Chai et al., 2020).
Step 2, calculation of the correlation coefficient. We retrieved
relevant event waveforms, and performed waveform pre-processing such as
demean, detrend, band-pass filtering (0.01-0.05 Hz), and normalization.
Then, we calculated the CC and σSF (variance of scale
factor) with a sliding window method. The statistical parameter
hypothesis testing method is used to obtain the confidence interval
thresholds ηCC and ηSF of CC and
σSF, respectively.
Step3, filtering of potentially anomalous stations. After determining
the detection threshold of each channel, if the CC of the channel falls
outside the confidence interval, the station is recorded as a potential
abnormal station. While for cases where the CC values fall in the
confidence area, the variance σSF of the 3-channel scale
factor SF is calculated to determine whether the σSFfalls within the confidence interval. If the σSF falls
within the confidence interval, the station is recorded as a normal
record station, otherwise it will be considered as a candidate abnormal
station.
Step 4, confirmation of abnormal recording station. Combining the
original waveforms and the PSD curve characteristics, we verified the
anomalous stations and classified them into four categories.