Orateur
Description
Ground-based observations of the cosmic microwave background (CMB) are severely challenged by atmospheric emission, which outshines the faint astrophysical and CMB signals. In particular, fluctuations in atmospheric water vapour introduce non-stationary, spatially complex correlations in the time-ordered data (TOD), coupling neighbouring detectors and contaminating large angular scale modes in temperature. The presence of instruments temperature-to-polarisation leakage can further pollute polarisation signal. Traditional atmospheric mitigation techniques, such as high-pass filtering in the frequency domain, risk discarding valuable CMB information, especially at low multipoles where the primordial B-mode polarisation signal is expected to peak.
We present a novel time-domain component separation approach that exploits the distinct temporal and spectral signatures of atmospheric emission. By incorporating physical atmospheric parameters such as wind speed and precipitable water vapour (PWV), our method improves atmospheric decontamination while preserving the underlying CMB signal. We validate the approach using simulations that include both a 2D atmospheric model and a full 3D atmospheric cube with realistic temporal evolution, demonstrating accurate reconstruction of sky maps and recovery of large-scale modes. In addition, we show how the atmospheric signal can be leveraged to better characterise the instrumental response and to model the correlated component of the noise covariance matrix, enabling its whitening for subsequent analysis stages.