Does coherent signal discrimination improve faster than glitch frequency grows with detector additions?

This is an excellent question that gets at a fundamental tension in multi-detector gravitational-wave astronomy. The preprints directly address whether adding detectors improves signal discrimination faster than it increases the false-alarm burden.

The Core Trade-off

When you add a gravitational-wave detector to a network, you gain signal strength in two ways: more independent measurements of the same wave, and a higher signal-to-noise ratio (SNR) that makes real signals stand out. This is the "upside" — the log Bayes factor (a measure of confidence that a signal is real rather than noise) increases with each new detector [AG-2024.03-419].

But there's a hidden cost: glitches are non-Gaussian noise transients that mimic gravitational-wave signals [AG-2024.03-419]. As you add detectors, you don't just add signal channels; you add independent *glitch channels* too. Since glitches appear to follow roughly Poissonian statistics (though with important exceptions like 1400Ripples and Power Line artifacts that don't) [AG-2025.11-155], a three-detector network has more glitch events than a two-detector one, increasing the false-alarm rate.

Does Signal Discrimination Win?

The answer appears to be yes, but with caveats. [AG-2024.03-419] explicitly studies whether the gain in signal-versus-glitch Bayes factor ($\ln\mathcal{B}_{\mathcal{S},\mathcal{G}}$) from adding Virgo to the Hanford-Livingston pair outpaces the cost of increased glitch frequency. Their empirical measurement using real O3a data with actual background glitch triggers shows that detection efficiency does improve — the three-detector network performs better than two detectors — but this requires careful accounting of the glitch background, which the earlier study had overlooked.

The practical implication is that signal discrimination does improve faster than glitch frequency grows, provided you use coherent multi-detector methods that exploit the spatial and temporal correlations between detectors. Methods like BayesWave and the improved $χ^2$ discriminators [AG-2025.07-142] leverage the fact that a gravitational wave produces correlated signals across all detectors simultaneously, whereas most glitches are instrumental or environmental and appear in only one or two detectors.

Why This Matters

This trade-off will intensify dramatically with third-generation observatories like the Einstein Telescope, which will detect gravitational waves much more frequently [AG-2024.11-444]. However, the Einstein Telescope's triangular geometry creates a "null stream" — a data combination that contains only noise, not signal — which can be used to characterize and subtract glitches more reliably than current methods [AG-2024.11-444].

One more nuance: glitches far from a signal in time-frequency space don't actually bias parameter estimation, so not all glitches need mitigation [AG-2025.06-690]. This further tilts the balance in favor of multi-detector networks, where spatial separation naturally isolates most glitches.