Predicting any small-scale statistics of high-Reynolds-number turbulence using ensemble simulations

TL;DR

This paper introduces an ensemble simulation approach to predict small-scale turbulence statistics at high Reynolds numbers efficiently, leveraging lower-Reynolds-number simulations and theoretical insights to achieve high accuracy.

Contribution

The authors propose a novel ensemble hypothesis method that predicts high-Reynolds-number turbulence statistics using a mixture of lower-Reynolds-number simulations with variable energy injection.

Findings

Accurately predicts joint QR-PDF and extreme dissipation statistics

Ensemble weights inferred from turbulence scaling exponents

Method reduces computational cost for high-Re turbulence statistics

Abstract

The complex small-scale statistics of turbulence are a result of the combined cascading dynamics through all scales of the flow. Predicting these statistics using fully resolved simulations at the high Reynolds numbers that typically occur in engineering, geophysical, and astrophysical flows will exceed the capabilities of even the largest supercomputers for the foreseeable future. A common observation is that high-Reynolds-number flows are organized in clusters of intense turbulent activity separated by large regions of quiescent flow. We here show that small-scale statistics in high-Reynolds-number turbulence can be predicted based on an ensemble hypothesis, stating that they can be emulated by the statistical mixture of a heterogeneous ensemble of lower-Reynolds-number simulations. These simulations are forced at smaller scales with energy injection rates varying across the ensemble. We show that our method predicts complex gradient statistics from the recent literature, including the joint QR-PDF and extreme dissipation and enstrophy statistics, to unprecedented accuracy. Remarkably, we find that the weight distribution needed for the ensemble method to make predictions can be inferred from the anomalous scaling exponents of turbulence. Thus combining theory with fully resolved simulations, our method can be readily applied to predict a wide range of statistics at high Reynolds number but low cost, while opening up various avenues for further theoretical and numerical exploration.