Measuring persistence time in stellar atmospheres.

The covariance matrix representation of red noise in time series (e.g. Delisle et al. 2020) requires an estimate of the e-folding time of an observation's predictive power, or persistence time. We have recently adapted the TAUEST computational method, which was developed for paleoclimate time series, to estimate the persistence times of observables that trace stellar activity. Given that astronomical time series are, in most cases, irregularly spaced, one of the main challenges in applying TAUEST is the variation in timesteps between different datasets. We explored how to adjust the TAUEST algorithm for broad timestep distributions and attempted to determine the types of datasets it could be applied to with greatest precision. Our study also examines the contexts in which the debiasing procedure that is part of the geoscience version of TAUEST is appropriate. We will show TAUEST persistence time estimates from radial velocities, activity indicators, and photometry and present our assessment of the algorithm's accuracy in each dataset.