Abstract

A simple method is proposed to generate synthetic vector fields as surrogates for turbulent velocity fields. The method is based on the minimal Lagrangian map, by which an initial Gaussian field generated using random-phase Fourier modes is deformed. The deformation is achieved by moving fluid particles of a sequence of low-pass filtered fields at their fixed velocity for some scale-dependent time-interval, interpolating onto a regular grid, and imposing the divergence-free condition. Statistical analysis shows that the resultant non-Gaussian field displays many properties commonly observed in turbulence, ranging from skewed and intermittent velocity gradient and increment probability distributions, preferential alignment of vorticity with intermediate strain rate, and nontrivial vortex stretching statistics. Differences begin to appear only when interrogating the data with measures associated with intense vortex tubes that are conspicuously absent in the synthetic field. To explore the dynamical implications of these observations, the synthetic non-Gaussian fields are used as initial conditions in DNS and LES of decaying isotropic turbulence, and results are compared with initializations using Gaussian fields. The non-Gaussian synthetic fields yield more realistic results with significantly shortened initial adjustment periods. © 2006 American Institute of Physics.