Parameter estimation and model selection of gravitational wave signals contaminated by transient detector noise glitches

The number of astrophysical sources detected by Advanced LIGO and Virgo is expected to increase as the detectors approach their design sensitivity. Gravitational wave detectors are also sensitive to transient noise sources created by the environment and the detector, known as ‘glitches’. As the rate of astrophysical sources increases, the probability that a signal will occur at the same time as a glitch also increases. This has occurred previously in the gravitational wave binary neutron star detection GW170817. In the case of GW170817, the glitch in the Livingston detector was easy to identify, and much shorter than the total duration of the signal, making it possible for the glitch to be removed. In this paper, we examine the effect of glitches on the measurement of signal parameters and Bayes factors used for model selection for much more difficult cases, where it may not be possible to determine that the glitch is present or to remove it. We include binary black holes similar to current detections, sine Gaussian bursts, and core-collapse supernovae. We find that the worst effects occur when the glitch is coincident with the signal maximum, and the signal to noise ratio (SNR) of the glitch is larger than the signal SNR. We have shown that for accurate parameter estimation of future gravitational wave signals it will be essential to develop further methods to either remove or reduce the effect of the glitches.