Abstract. We present a new time-stepping algorithm for nonlinear PDEs that exhibit scale separation in time of a highly oscillatory nature. The algorithm combines the parareal method-a parallel-in-time scheme introduced in [24]-with techniques from the Heterogeneous Multiscale Method (HMM) (cf. [13,2]), which make use of the slow asymptotic structure of the equations [32], [27], [28], [36]. By numerically computing a locally asymptotic solution we are able to "factor out" the fast oscillatory part of the solution, and solve for the remaining slow part of the solution using time steps that are orders of magnitude larger than standard time-stepping methods allow. The scheme has two elements. First, we use HMM to numerically advance the asymptotic form of the equations with a large time step ∆T . Second, we refine the solution in parallel on the subintervals [n∆T, (n + 1) ∆T ] using small time steps ∆t and the iterative scheme scheme in [24]; the intermediate solutions on the sub-intervals [n∆T, (n + 1) ∆T ] are ephemeral, and are used only to converge the overall solution at the large time steps n∆T . Using the asymptotic structure allows the computed solutions to be close enough to the actual solution that parallel-in-time methods converge to high accuracy and with significant parallel speedup.We present error bounds, based on the analysis in [17], that demonstrate convergence of the method. A complexity analysis also demonstrates that the parallel speedup increases arbitrarily with greater scale separation. Finally, we demonstrate the accuracy and efficiency of the method on the (onedimensional) rotating shallow water equations, which is a standard test problem for new algorithms in geophysical fluid problems. Compared to exponential integrators such as ETDRK4 and Strang splitting-which solve the stiff oscillatory part exactly-we find that we can use coarse time steps ∆T that are orders of magnitude larger (for a comparable accuracy), yielding an estimated parallel speedup of approximately 100 for physically realistic parameter values. For the (one-dimensional) shallow water equations, we also show that the estimated parallel speedup of this "asymptotic parareal method" is more than a factor of 10 greater than the speedup obtained from the standard parareal method.