pipette
ESEspañol

Prediction Accuracy Can Select the Scaling Law of Required Precision

Song-Ju Kim

Preprint

In the authors' words

How much precision must be allocated, and to which uncertainty channels, to meet a prediction requirement? We formulate a minimum one-shot specification cost over initial state variables and persistent forcing parameters. The central result is that prediction accuracy can select the horizon-scaling law of this resource by activating different physical uncertainty channels. Within any common uniformly scale-separable regime with the same active channels, changing a fixed tolerance changes the required precision by only bits. An exactly solvable multiscale toral system shows that one fixed dynamics and target can nevertheless switch from to when finer accuracy activates an expanding channel. A forced Lorenz--84 calculation then realizes the same channel-switching mechanism numerically over finite horizons in a single two-scale diagnostic field: a sampled post-transient eddy envelope supports a coarse branch controlled by the phase rate of a large-scale forcing, whereas finer accuracy activates exponentially sensitive atmospheric degrees of freedom and produces an approximately linear increase of the linearized one-shot cost over the resolved window. Thus accuracy can determine not only how much predictive precision is required, but which physical channel controls its growth with horizon.

Main resultLimitation the authors admit

Appeared: Friday, October 2. arXiv. Preprint, not yet peer-reviewed.

Authors' comment: 14 pages, 6 figures