Assessment of ionospheric activity tolerances for Epoch of Reionisation science with the Murchison Widefield Array

Structure imprinted in foreground extragalactic point sources by ionospheric refraction has the potential to contaminate Epoch of Reionisation (EoR) power spectra of the 21~cm emission line of neutral hydrogen. The alteration of the spatial and spectral structure of foreground measurements due to total electron content (TEC) gradients in the ionosphere create a departure from the expected sky signal. We present a general framework for understanding the signatures of ionospheric behaviour in the two-dimensional (2D) neutral hydrogen power spectrum measured by a low-frequency radio interferometer. Two primary classes of ionospheric behaviour are considered, corresponding to dominant modes observed in Murchison Widefield Array (MWA) EoR data; namely, anisotropic structured wave behaviour, and isotropic turbulence. Analytic predictions for power spectrum bias due to this contamination are computed, and compared with simulations. We then apply the ionospheric metric described in Jordan et al. (2017) to study the impact of ionospheric structure on MWA data, by dividing MWA EoR datasets into classes with good and poor ionospheric conditions, using sets of matched 30-minute observations from 2014 September. The results are compared with the analytic and simulated predictions, demonstrating the observed bias in the power spectrum when the ionosphere is active (displays coherent structures or isotropic turbulence). The analysis demonstrates that unless ionospheric activity can be quantified and corrected, active data should not be included in EoR analysis in order to avoid systematic biases in cosmological power spectra. When data are corrected with a model formed from the calibration information, bias reduces below the expected 21~cm signal level. Data are considered `quiet' when the median measured source position offsets are less than 10-15~arcseconds.