Answers are grounded in this paper and cite their sources.
The test-field model of turbulence is used to compute the growth of prediction error for inertial-range turbulence in both three and two dimensions. It is found that initial uncertainty in high wavenumbers spreads through the entire inertial range according to a similarity behavior. For the energy inertial range, the time required for error to reach wavenumber k from very high wavenumbers is t=Aε−1/3k−2/3, where ε is the rate of energy transfer per unit mass and A≈10 in three dimensions or A≈2.5 in two dimensions. For the enstrophy inertial range in two dimensions the time for error to propagate from k′ down to k≪k′ (k and k′ both in the inertial range) is t≈4η−1/2{[ln(k′/ k1)]2/3−[ln(k/ k1)]2/3}, where η is the rate of enstrophy transfer and k1 marks the bottom of the enstrophy inertial range. Error growth is also computed for a two-dimensional spectrum that fits the energy spectrum of planetary waves in the atmosphere. An initial state determined with a horizontal resolution feasible with a satellite-based observing system results in significant predictability of large-scale motions for more than a week. It is argued that the test-field model probably underestimates rather than overestimates predictability times.
No takes yet. Share an insight, caveat, or question.
Leith et al. (1972) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: