A feature-based perspective on upscale error growth.

<p>Atmospheric predictability is fundamentally limited by the upscale growth of initial small-scale, small-amplitude errors. To better understand this<br>fundamental limitation, it is essential to study upscale growth mechanisms. Upscale error growth is often investigated by spectral analysis. By design,<br>however, spectral analysis is not able to provide detailed spatial information of locally coherent error „features“. The flow dependence of error growth<br>suggests that such local error features play an important role in limitingpredictability at later forecasts times.</p><p>We here take an approach complementary to spectral analysis and apply a feature-based perspective. We have developed an automated algorithm to<br>identify error features in gridded data and track their spatial and temporal evolution. Errors are considered in terms of potential vorticity (PV) and we<br>evaluate a previously derived PV-error tendencies equation to characterize the growth of error features. We examine previously published upscale error<br>growth experiments with the global prediction Model ICON from the German Weather Service, in which the only difference between various realizations is in the seeding of a stochastic convection scheme. Complementing previous work, we here focus on low- to mid-tropospheric errors.</p><p>Spatial composites centered on the centroid of error features indicate that features are primarily generated ahead of an upper-tropospheric trough. The<br>environment surrounding the features at the time of their first detection is characterized by locally enhanced lower to mid tropospheric moisture, latent<br>heat release, and upper tropospheric divergence. Subsequently, this moist-diabatic nature of the error-feature environment becomes gradually less<br>prominent. The development of error features in their early stage is governed by non-conservative processes. Very early after detection, the convective<br>scheme has the strongest contribution to feature amplification. Subsequently, the influence of the convective scheme diminishes. Error features are thus generated by the convective scheme, consistent with the design of the experiments, but further amplification due to that scheme is not diagnosed.<br>Longwave radiation, in contrast, has a persistent and dominant amplifying impact throughout the early stages of the features. PV-error features are<br>found to also exhibit significant moisture errors and the moisture-radiation feedback occurs as a dominant error growth mechanism in our experiments.<br>In addition to the mean perspective of error-feature evolution, the presentation will discuss differences between fast and slowly growing errorfeatures and the interaction of low- to mid-tropospheric errors with errors at tropopause level.</p>