Probabilistic methods and computer simulation are used to analyze streamline properties of flows defined by a general class of random, incompressible velocity fields. Such fields are stochastically modeled by a superposition of simple shear-flow layers. The resulting flow is governed by a nonstationary, random Hamiltonian with Hurst exponent H = 0.5 and having the form
where the W k are two-sided continuoustime random walks. The statistical topography of the flow, as characterized by the level sets or streamlines of H0, is analyzed via an associated Brownian walk space from which asymptotic results are determined. The flow consists of a hierarchy of nested, closed streamlines. For a given box of side length 2L centered at the particle's starting position, an upper bound on the particle's probability of exit, or "noncycling" probability, pnc, is shown to have a power law dependence on the box size, pnc(L) ∼ L -α , for L 1 and positive constant α. We also introduce a constant, nonzero mean flow and denote its relative strength with respect to the r.m.s. fluctuations of the random field by ρ. In the case 0 < ρ < 1, the fraction pnc(ρ) of "percolating" or noncycling particles (in an infinitely large box) satisfies the relation
All particles percolate in the case ρ ≥ 1. Computer simulations for various values of N agree well with earlier work on the N = 2 case by Avellaneda, Elliott, and Apelian, thereby confirming and validating both studies. Numerical results also show the power law exponent α to be remarkably robust with respect to changes in topology, including the existence of traps, irregularly spaced modes, and the value of N . All runs yield a common value of α ≈ 0.22. Likewise, the mean length of streamlines exiting boxes of size 2L, λ(L) , scales like L γ with γ ≈ 1.28 for all N . These exponent values contrast with those predicted by Isichenko and Kalda yet consistently satisfy a "sum rule," α + γ = 2 -H, relating α, γ, and H, the Hurst exponent of the flow.