Skip to content

Initial data inspection

Before aligning anything, a visual check of the imported seismograms catches obvious problems: garbled waveforms, excessive noise, flat traces, data gaps. Catching these early avoids tuning parameters around unusable data.

This pass is for the obvious cases only. Autoselect handles borderline traces algorithmically, across the whole array, not one trace at a time.

Two views help.

By event

A record section: every seismogram for one event, sorted by epicentral distance, with absolute time on the x-axis. Coherent arrivals show as a roughly linear moveout across the traces.

aimbat plot seismograms <ID>

plot seismograms <ID>

In the Project tab, press Enter on an event row and choose View seismograms.

from sqlmodel import Session
from aimbat.db import engine
from aimbat.models import AimbatEvent
from aimbat.plot import plot_seismograms, show_figure

with Session(engine) as session:
    event = session.get(AimbatEvent, event_id)
    fig, _ = plot_seismograms(event, return_fig=True)

show_figure(fig)

Building the figure reads the event from the database, so it happens inside the session; showing it blocks until the window is closed, so it happens outside.

With many traces, only a subset shows at first. Scrolling pans through the rest; scrolling with Shift held pans the time axis.

Look for:

  • traces that are flat, clipped, or incoherent with the array
  • stations with excessive noise relative to signal
  • arrivals much earlier or later than moveout predicts
  • unusually large or small amplitudes after normalisation, often a gain problem in the original file

By station

Every seismogram recorded at one station across all events in the project, aligned on the initial pick (t0) and stacked vertically in time order. Useful for telling whether a station is consistently problematic or the issue is isolated to one event.

aimbat station plotseis <STATION_ID>

station plotseis <STATION_ID>

In the Project tab, press Enter on a station row and choose View seismograms.

from sqlmodel import Session
from aimbat.db import engine
from aimbat.models import AimbatStation
from aimbat.plot import plot_seismograms, show_figure

with Session(engine) as session:
    station = session.get(AimbatStation, station_id)
    fig, _ = plot_seismograms(station, return_fig=True)

show_figure(fig)

Building the figure reads the station from the database, so it happens inside the session; showing it blocks until the window is closed, so it happens outside.

Same scroll behaviour: scrolling pans traces, shift+scroll pans time.

How the data are prepared

Both plots apply the same in-memory preprocessing before display. The files are never modified.

  1. Detrend. Remove the mean and linear trend.
  2. Bandpass filter (optional). Applied only if bandpass_apply is enabled in the event parameters, using that event's bandpass_fmin / bandpass_fmax. Off by default, so inspection shows the waveforms as imported. Data pre-filtered before import can be inspected with it left off.
  3. Resample. To a common 10 Hz for consistent display.
  4. Normalise. Each trace to unit amplitude, so shapes are comparable regardless of gain.

Because the filter follows the current event parameters, the plots change with it. Inspecting both filtered and unfiltered helps separate noise from signal.

Excluding bad data

Excluding a seismogram from processing means deleting it: see Removing data. The file on disk is unaffected.

Deletion is for the obvious cases above; autoselect handles the rest via min_cc.