Low‐order system for eddy growth in the storm track

We use a machine‐learning fit to reanalysis data from the North Atlantic storm track to find the best low‐order model to describe the observed interaction between area‐mean baroclinicity and eddy heat flux. We find that the optimal fit model corresponds to a conservative nonlinear oscillator with a predator–prey‐like interaction between the eddy heat flux and baroclinicity, and a linear period of 30 days. The structure of this best‐fit model has several implications for the baroclinic properties of the North Atlantic storm track. (a) The observed mean growth of eddies is accurately described as linear growth with a growth rate numerically equal to the Eady growth rate. (b) The Eady growth rate is a dynamical variable that, on average, reduces when the eddies grow. (c) The observed bound on eddy amplitude is not due to nonlinear saturation of the eddies but to the concurrent erosion of baroclinicity. (d) The restoration phase for baroclinicity after depletion is best described by logistic growth, with a saturation value of twice the mean observed baroclinicity. (e) The sporadic nature of heat flux maxima is a natural consequence of the interaction between eddies and the baroclinicity. (f) The North Atlantic storm track exhibits dissipative control of the climatological baroclinicity and an eddy saturation control of the climatological eddy heat flux.