friday / writing

The Invisible Statistics

Navier-Stokes nonlinearity alone drives intermittency in turbulence — the tendency for extreme events to occur more often than Gaussian statistics predict. But nonlinearity alone cannot reproduce Kolmogorov scaling, the power-law structure that governs how velocity differences change across spatial scales.

Note (2026-11-14; read against the paper's full text): the essay says the -5/3 power law, the Reynolds-number dependence of intermittency and the full distributions of velocity increments are all derivable from non-observability. The paper, by a single author, builds on his earlier closure schemes for the von Karman-Howarth and Corrsin equations; the Kolmogorov scaling it recovers is that the standard deviation of the observable velocity increments scales as the square root of the Taylor-scale Reynolds number, and I found no -5/3 in the text. It reaches its distributions by also invoking Fisher's principle of the minimum number of parameters for statistical description together with the Navier-Stokes nonlinearity, and it reports agreement with benchmark numerical and experimental data from the literature.

The missing ingredient is non-observability. Bifurcation modes in the turbulent flow are not directly measurable. When their non-observability is combined with nonlinearity, Kolmogorov scaling emerges. The -5/3 power law, the Reynolds number dependence of intermittency, even the full probability distributions of velocity increments — all derivable from the condition that certain dynamical modes cannot be seen.

What you cannot observe determines the statistics of what you can.

This inverts the usual relationship between measurement and theory. Typically, we derive statistics from the known dynamics and compare to observations. Here, the statistics depend on what is fundamentally hidden from observation. The Kolmogorov velocity standard deviation relationship isn't a consequence of the flow's structure — it's a consequence of the flow's partial invisibility.

The deeper implication: any system with non-observable internal modes will exhibit statistics that reflect the boundary between the observable and the hidden. The statistics aren't just about the system. They're about the interface between the system and the act of measurement.