1: \begin{abstract}

2: Accurate assessment of uncertainty in the approximation of infinite-time-averaged statistics of statistically stationary ergodic processes (that is, of signals, obtained experimentally or computationally, that meander about some unknown mean) is a topic of significant importance in a host of engineering applications.  Among them, for example, the statistics of many turbulent flows are generally considered as stationary and ergodic after some initial transient is identified and set aside. As taking infinite time averages is not practically feasible, finite-time-averaged approximations of these statistics are generally used.  For problems in which the measured samples are independent and identically distributed (i.i.d.), the expected squared averaging error reduces only like $\sigma/\sqrt{N}$, where $\sigma$ is the standard deviation of the data.  In problems for which the measured samples are not i.i.d., convergence is even slower (eventually, like $Q/\sqrt{N}$ for $Q<\sigma$), and an uncertainty quantification (UQ) method is needed.  The present paper presents a new method to quantify the expected squared averaging error which is \emph{multiscale}, meaning that it is based on an autocorrelation model that is tuned to the data to fit the statistic of interest at a large range of different timescales.  The method is also \emph{asymptotically unbiased}, meaning that the expected squared averaging error asymptotically

3: converges like $Q/\sqrt{N}$ for the same value of $Q$ as the actual system, if it is modelled as a random process with the same mean, variance, and autocorrelation.  The new UQ method is tested on three representative test problems, and shown to be highly effective.

4: \end{abstract}

5: