Julien Tang, Shamik Ghosh, Jacques Delabrouille, John C. Groh, Oliver Jeong, Reijo Keskitalo, Theodore Kisner

CMB observations from ground-based observatories are limited in sensitivity by the fluctuating emission from the Earth's atmosphere, mostly due to water vapor inhomogeneities. Even in atmospheric windows, this spurious signal remains the major source of contamination of the data. Traditional mitigation techniques include low-frequency filtering, or for polarization measurements specifically, modulation with a rotating half-wave plate. The first method filters out a significant fraction of the target cosmological signal while the second leaves residuals due to imperfections, temperature to polarization leakage, or polarized atmospheric emission. In this work, we present a new data analysis framework, based on estimation of atmosphere emission templates using a multi-frequency focal plane. The core novelty of this setup is to have detectors dedicated to atmosphere monitoring, allowing for removing atmospheric contamination with time-domain component separation techniques. We introduce a multipole-dependent atmospheric decontamination factor $A^\mathrm{atm}_\ell$ to quantify the relative reduction of the atmospheric contamination angular power spectrum achievable with this approach. We demonstrate that this decontamination factor scales as the inverse of the survey time and that, using that criterion, our component separation pipeline can outperform a classical filter-bin map-making pipeline by a factor of 4000 for $30\le \ell \le 300$.