Initial Data Inspection

Before running any alignment, it is worth visually inspecting the imported seismograms to catch obvious problems: garbled waveforms, stations with excessive noise, flat traces, or data gaps. Catching these early avoids wasting time tuning parameters around fundamentally unusable data.

AIMBAT provides two complementary views for this purpose.

By event — record section

Plots all seismograms for an event as a record section: waveforms sorted by epicentral distance, with absolute time on the x-axis. This gives an immediate overview of the array — coherent arrivals should appear as a roughly linear moveout across the traces.

aimbat plot seismograms <ID>

plot seismograms <ID>

When there are many traces, only a subset is shown initially. Scroll the mouse wheel to pan through the remaining traces; hold Shift and scroll to pan along the time axis.

What to look for:

• Traces that are flat, clipped, or visually incoherent with the rest of the array
• Stations with excessive noise relative to the signal
• Traces where the arrival appears to arrive much earlier or later than expected from moveout
• Unusually large or small amplitudes after normalisation (can indicate a gain issue in the original file)

By station — across events

Plots all seismograms recorded at a single station across every event in the project, aligned on the initial pick (t0). The x-axis shows time relative to the pick; traces are stacked vertically in chronological order. This view is useful for checking whether a station is consistently problematic across multiple events, or whether an issue is isolated to one.

aimbat station plotseis <STATION_ID>

station plotseis <STATION_ID>

The same scroll behaviour applies: scroll to pan through traces, shift+scroll to pan the time axis.

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

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

How the data is prepared

Both plots apply the same preprocessing before displaying:

1. Detrend — removes the mean and linear trend
2. Bandpass filter (optional) — applied if bandpass_apply is enabled in the event parameters; uses the bandpass_fmin / bandpass_fmax values set for that event. Filtering is off by default, so the initial inspection shows the raw waveforms as imported. Users who pre-filter their data before import can leave it disabled and work directly with the filtered waveforms they already have
3. Resample — resampled to a common 10 Hz sample rate for consistent display
4. Normalise — each trace is normalised to unit amplitude so waveforms are visually comparable regardless of original gain

The original files are never modified. These steps are applied in memory for display only.

Because the bandpass filter uses the current event parameters, the inspection plots will look different depending on whether a filter is applied. It can be useful to inspect both with and without filtering to distinguish noise from signal.

What to do with bad data

If you identify a seismogram that should not be included in processing, see Removing data. Deleting a seismogram from the project does not affect the underlying file.

For borderline cases — noisy but potentially usable traces — it is better to leave them in and rely on ICCS autoselect to exclude them based on cross-correlation quality rather than deleting them outright.