The function allows detection of discrete seismic events using the STA-LTA
picker approach and including several stations of a seismic network. It
will pick events at all input stations and identify events picked by more
than a user defined minimum number of stations, within a given time window.
It further allows rejecting events that are too short or too long, or occur
to short after a previous event. The function can be used to process long
data archives organised in a consistent file structure. This is the parallel
computation version of the routine. For the single CPU version, allowing to
show verbose progress information, use aux_picknetwork().
aux_picknetworkparallel(
start,
stop,
res,
buffer,
station,
component,
dir,
f,
envelope = TRUE,
sta,
lta,
on,
off,
freeze,
dur_min,
dur_max,
n_common,
t_common,
t_pause,
cpu,
...
)data frame, detected events (start time, duration in
seconds)
POSIXct value, start time for event detection analysis.
If a character string (or vector) is provided, the function will try to
convert it to POSIXct.
POSIXct value, start time for event detection analysis.
If a character string (or vector) is provided, the function will try to
convert it to POSIXct.
Numeric value, time resolution of the snippets used to
detect events, in seconds. A useful value would be one hour
(res = 3600).
Numeric vector, of length two. Size of the two-sided
buffer to account for signal pre-processing effects, in seconds. The first
value is used for the left side, the second for the right side. Events that
happen to be detected during the buffer time will be removed from the output
data set.
Character vector, seismic station IDs to be processed.
Character value or vector, seismic component to
be used. Information can be restricted to a single character, denoting the
last character of the station component code, i.e. "Z" would use the
first and last component of c("BHZ", "HHE", "HHZ"). If omitted,
the function will use "BHZ" by default.
Character value, path to directory that contains the
seismic files. See read_data for details and constraints on the
data structure.
Numeric Numeric vector of length two, cutoff frequencies
(Hz) for the bandpass filter to be applied. If omitted, no filtering will
be performed.
Logical value, option to calculate signal envelopes
before detecting events. Default is TRUE.
Numeric value, duration of short-term window. Attention,
this definition differs from the one in pick_stalta!
Numeric value, duration of long-term window. Attention,
this definition differs from the one in pick_stalta!
Numeric value, threshold value for event onset.
Numeric value, threshold value for event end.
Logical value, option to freeze lta value at start of
an event. Useful to avoid self-adjustment of lta for long-duration events.
Numeric value, minimum duration of an event (s). Picks
with a STA-LTA duration below this threshold will be discarded.
Numeric value, maximum duration of an event (s). Picks
with a STA-LTA duration longer than this threshold will be discarded.
Numeric value, minimum number of stations at which
an event has to be registered.
Numeric value, time (s) within which the picks of an
event need to be registered by seismic stations.
Numeric value, time period after the onset of an event
during which no other event is allowed to happen.
Numeric value, fraction of CPUs to use. If omitted,
only one CPU will be used.
Further arguments passed to the function, i.e. keywords for the signal deconvolution part. See details for further information.
Michael Dietze
The data will be imported time slice by time slice. Optionally, a signal deconvolution can be done. The data will be filtered to isolate the frequency range of interest, and tapered to account for edge effects. The taper size is automatically set be the two-sided buffer (see below). Optionally, a signal envelope can be calculated as part of the preprocessing work. For all successfully imported and preprocessed signal time snippets, events will be detected by the STL-LTA approach. Events from all input signals will be checked for a minium and maximum allowed event duration, set as thresholds by the user. All remaining events are checked for the number of stations of the network that did consistently detect an event. That means an event must be found in each signal within a user defined time window, a window that corresponds to the maximum travel time of a seismic signal through the network. The minimum number of stations that detect the event can be set by the user. In addition, events can be rejected if they occur during a user defined ban time after the onset of a previous event, which assumes that no subsequent events are allowed during the implementation of long duration events happening.
The time period to process is defined by start, end and the
length of time snippets res (s). In addition, a left sided and
right sided buffer can and should be added buffer = c(60, 60) (s),
to account for artefacts during signal preprocessing and events that
occur across the time snippets. The start and end time should
be provided in POSIXct format. If it is provided as text string, the
function will try to convert to POSIXct assuming the time zone is UTC.
There must be at least two seismic stations. The seismic components
component must be a vector of the same length as station and
contain the corresponding information for each station.
Deconvolution of the input signals can be done, although usually this is
not essential for just picking events if the filter frequencies are in the
flat response range of the sensors. If wanted, one needs to provide the
additional vectors (all of the same length as station) sensor,
logger and optionally gain (otherwise assumed to be 1).
See signal_deconvolve() for further information on the deconvolution
routine and the supported keywords for sensors and loggers.
STA-LTA picking is usually done on signal envelopes enevelope = TRUE.
However, for array seismology or other specific cases it might be useful to
work with the waveforms directly, instead of with their envelopes.
The STA-LTA routine requires information on the length of the STA window,
the LTA window, the on-threshold to define the start of an event, the
off-threshold to define the end of an event, and the option to freeze the
STA-LTA ratio at the onset of and event. See the documentation of the
function pick_stalta() for further information.
Events that last shorter than dur_min or longer than dur_max
are rejected. This criterion is applied before the test of how many stations
have commonly detected an event. Hence, the minimum and maximum duration
criteria need to be fulfilled by an event for all seismic stations, or at
least as many as defined as minimum number of stations n_common.
A valid event should be detected by more than one station. At least two
stations (n_common = 2) should detect it to confirm it is not just a
local effect like rain drop impacts or animal activity. At least three
stations (n_common = 3) are needed to locate an event. Detection at
more stations increases the validity of an event as proper signal. However,
the likelihood of stations not operating correctly or that some stations
may be too far to detect a smaller event limits the number of stations that
are be set as (n_common).
An event that happens just outside a seismic network will generate seismic
waves that travel through the network and will require a given time for
that. The time is set by the longest inter-station distance and the apparent
seismic wave velocity. For example, an event who's signal propagated with
1000 m/s through a network with a maximum aperture of 1000 m will need up to
1 s. Hence, the parameter t_common should be set to 1, in
this case. This parameter is good to reject events that show a longer travel
time, e.g. because they are pressure waves that travel through the
atmosphere, like helicopter signals or lightning strikes.
picks <- aux_picknetwork(start = "2017-04-09 01:00:00 UTC",
stop = "2017-04-09 02:00:00 UTC",
res = 1800,
buffer = c(90, 90),
station = c("RUEG1","RUEG2"),
component = rep("BHZ", 2),
dir = paste0(path.package("eseis"), "/extdata/"),
f = c(0.1, 0.2),
envelope = TRUE,
sta = 0.5,
lta = 300,
on = 3,
off = 1,
freeze = TRUE,
dur_min = 2,
dur_max = 90,
n_common = 2,
t_common = 5,
t_pause = 30)
print(picks)
Run the code above in your browser using DataLab