Impact of transient eddies on extratropical seasonal-mean predictability in DEMETER models

Abstract

The impact of transient eddies on extratropical seasonal-mean prediction and predictability was examined using DEMETER seasonal prediction data. Two distinct groups were found among the seven DEMETER models based on the simulated properties of their climatological state: (1) models of a strong jet stream and strong transient activity (strong transient models), which is close to the observed intensity, and (2) models of a weak jet stream and weak transient activity (weak transient models). In addition to climatology, the strong transient models tend to predict strong Pacific North American (PNA) patterns, whereas the weak transient models predict weak PNA patterns. Here we demonstrate that these differences mainly result from differences in the eddy feedback intensity. Due to synoptic eddy feedback, the strong transient models exhibit not only strong signal variance but also strong noise variance compared with those of the weak transient models. Interestingly two groups of models show the potential predictability of deterministic forecast, measured by the signal to noise ratio, which is similar to each other. However, the strong transient models produce the error to spread ratio smaller than that of the weak transient models, implying that the former models produce a more reliable spread for the probabilistic forecast. This study implies that a better representation of transient statistics is needed to improve the extratropical predictability of the dynamical seasonal prediction.