Turbulence theory has lived with a contradiction for decades. The Navier-Stokes equations describe fluid flow from first principles. The multifractal model of Parisi and Frisch describes turbulent statistics empirically. Both work. Neither derives from the other. The accepted folklore says no mathematical relation connects them.
The folklore is wrong. By combining Euler invariant scaling with the Navier-Stokes equations, a direct mapping emerges. The key is the L^(2m)-norms of velocity gradients — a family of quantities indexed by a parameter m. Each value of m corresponds to a local scaling exponent h in the multifractal framework. Increasing m from 1 to infinity is like adjusting a telescope: low m sees the smooth average flow; high m resolves the most singular, violent structures.
The mapped range of h — from -2/3 to 1/3 — is exactly the regime where the equations become unreliable. Below h = -2/3, thermal noise dominates and the deterministic Navier-Stokes description fails. Near h = 1/3, spontaneous stochasticity may arise — the equations have solutions, but the solutions aren't predictive because infinitesimally different initial conditions lead to finitely different outcomes.
The reconciliation isn't a derivation of multifractal statistics from first principles. It's something more modest and more useful: a dictionary between the two languages. The parameter m in the rigorous PDE analysis and the exponent h in the empirical scaling analysis are the same thing, viewed from different mathematical traditions. The traditions were measuring the same structure with different rulers.
What matters is not that the correspondence exists, but that it took decades to notice. Both frameworks were so internally self-consistent that practitioners in each never needed the other. The most productive theoretical gaps aren't between theories that conflict — they're between theories that agree without knowing it.