Q: In planetary motion it has been shown that orbits do not diverge exponentially until 30 Myr. The Lyapunov Characteristic Exponent (LCE) is 1/5 per Myr. [5]

giving a predictablity time of 5 Myr (or perhaps 10 Myr). How does one get 30?

Q: How can numerical weather predictions (at least for large-scale behavior) "get it right" for several days when the LCE

must be very large, due to small-scale turbulence? [6]

Q: How can turbulent wind gusts be predicted, as has been recently shown [7], when, as above, the LCE

must be very large?

A more correct version of the "rule-of-thumb" is::

where the capital and small delta's are the tolerance and initial error, respectively.

But allowing for an unusually large tolerance does not account for the entirity of the issue. For example, in the solar system case, allowing

the tolerance to be 1000 times the initial error gives the prediction time as 10 Myr (as opposed to the rule-of-thumb number of 5 Myr).

You cannot account for the "real" prediction time by increasing the tolerance.