Skip to content

Quality Assessment

Overview

AIMBAT provides a suite of statistical metrics to help you assess the reliability of your arrival-time picks. Metrics come from two sources:

  • ICCS CC — computed whenever an ICCS instance is created (i.e. on any operation that touches the active event). No explicit run step is needed.
  • MCCC metrics — computed only when you explicitly run the MCCC pass.

Both types are captured in snapshots and displayed in the Event and Station views.

ICCS Cross-Correlation (Live)

For every seismogram, AIMBAT records the Pearson cross-correlation coefficient between that seismogram and the current ICCS stack as iccs_cc.

  • When it is computed: automatically whenever AIMBAT loads or processes an event — for example when running ICCS, displaying waveforms, or adjusting the minimum CC threshold. You do not need to run any explicit step.
  • What it tells you: How closely the waveform matches the array stack under the current window and filter settings. It is the basis for the --autoselect threshold set with aimbat pick cc.
  • Interpretation: Values closer to 1.0 indicate high similarity to the stack. Values near 0 or negative suggest misalignment, poor SNR, or a polarity flip.

Because ICCS CC is computed automatically, it is available for every event that has been opened in the current session — including events that are not currently active.

MCCC Metrics

When you run MCCC, the following statistics are calculated for every participating seismogram:

CC Mean (Waveform Similarity)

The CC Mean is the arithmetic mean of all pairwise cross-correlation coefficients involving a specific seismogram.

  • What it tells you: How similar a station's waveform is to the rest of the array. It serves as a proxy for the Signal-to-Noise Ratio (SNR).
  • Interpretation: Values closer to 1.0 indicate very high similarity. Low values (e.g., < 0.6) often suggest noisy sites or instrument issues.

CC Std (Waveform Consistency)

The CC Std is the standard deviation of those same correlation coefficients.

  • What it tells you: Whether the station's waveform matches the entire array consistently, or only a subset of it.
  • Interpretation: A high CC Std indicates that the waveform shape is evolving as it passes through the array, likely due to significant site effects or complex geology (structural boundaries).

Timing Error (Precision)

The Timing Error is the formal standard error of the arrival-time estimate, derived from the covariance matrix of the least-squares inversion.

  • What it tells you: How "stable" the station's timing is relative to the network geometry and the quality of the correlations.
  • Interpretation: This is your primary metric for QC. High error values suggest inconsistent relative delays, often caused by cycle skipping or severe noise.

Global RMSE (Network Fit)

The RMSE (Root-Mean-Square Error) is a single value for the entire event.

  • What it tells you: The overall "tightness" of the mathematical fit for the entire array.
  • Interpretation: A high global RMSE suggests the array may be too large or sparse to be treated as a single coherent arrival (e.g., the wavefront is distorted by a major tectonic boundary).

Quality Control Quick-Reference

Use the combination of ICCS CC, CC Mean, and Timing Error to triage your data:

ICCS CC CC Mean Timing Error Interpretation Recommended Action
High High Low Robust pick. Keep.
High High High Likely Cycle Skip. Manually re-pick or discard.
High Low Low Noisy site, but stable timing. Keep with caution.
Low — — Poor waveform similarity. Review window/filter or discard.
— Low High Poor data quality. Discard seismogram.

Aggregated Statistics

In the Event and Station views, AIMBAT provides aggregated statistics across the seismograms included in the most recent MCCC run. The label shows "Averages across N seismograms" where N is the count of seismograms that have quality records in the active snapshot.

These aggregates include the SEM (Standard Error of the Mean), which quantifies the uncertainty of the average itself.

Scope of Quality Data

ICCS CC

ICCS CC is updated automatically each time AIMBAT processes an event, so it is available for all seismograms in any snapshot created after the event was first opened. It does not require an explicit MCCC run.

Rollback

When you roll back to a snapshot, AIMBAT restores the live quality metrics from the most recent snapshot whose parameter hash matches the restored state and that has MCCC quality data. This means your quality view reflects the metrics that were valid for those parameters, without having to re-run MCCC.

If no snapshot with a matching hash contains MCCC data (e.g. you rolled back to a state that pre-dates any MCCC run), the live MCCC metrics are left unchanged.

MCCC Metrics

MCCC can be run in two mutually exclusive modes, controlled by the --all flag:

  • Selected only (default, --all omitted): quality metrics are computed and stored only for seismograms whose select flag is True.
  • All seismograms (--all): quality metrics are computed and stored for every seismogram, regardless of selection state.

The mode used is determined from the snapshot data itself: if any seismogram that was deselected at snapshot time has MCCC quality records, the run is treated as "all seismograms". In particular:

  • A deselected seismogram will show MCCC quality data only if it has a record in the most recent MCCC snapshot — meaning MCCC was run with --all at that time. If the most recent run used selected-only mode, its MCCC quality view will be empty, even if an older snapshot contains data for it.
  • Running MCCC twice with the same parameters but different --all values produces two snapshots; AIMBAT always reports from the most recent one.