Error quantification for numerical relativity surrogate models

The past decade of gravitational wave astronomy has seen over 200 mergers of black holes and neutron stars. The detection and parameter estimation of such events is largely made possible using accurate waveform models. With ground-based detectors reaching their design sensitivities and future space-based detectors requiring waveforms of unprecedented accuracies, understanding and mitigating errors in existing waveform models has become increasingly relevant. Moreover, tests of general relativity require precision since any deviations from the theory could be attributed to errors in waveform models. In this work, we use the numerical relativity surrogate model NRSur3dq8, trained on waveforms of aligned-spin black hole binary mergers, to build a model that predicts errors in the phase and amplitude of their corresponding gravitational waveforms. The data-driven nature of surrogate modeling and the complexity of the underlying non-linear models introduce several different sources of error, and accordingly, we explore different approaches to building the error models. We also discuss the applications of surrogate waveform error quantification to parameter estimation studies.