9.3 A global high-resolution predictability experiment using MPAS

Judt, Falko, National Center for Atmospheric Research/Advanced Study Program

A tremendous increase in computing power has facilitated the advent of global convection-resolving models. Although these models are able to seamlessly predict the weather from local to planetary scales, unresolved issues regarding the predictability of the atmosphere provoke questions about what forecast problems are potentially tractable. To address this issue, we conducted a global high-resolution predictability experiment using the Model for Prediction Across Scales (MPAS). The goal of the experiment was to quantify the intrinsic scale-dependent predictability limits of atmospheric motions. A globally uniform mesh spacing of 4 km allowed for the explicit treatment of organized deep moist convection, alleviating grave limitations of previous predictability studies that either used high-resolution limited-area models or global simulations with coarser grids and cumulus parameterization. For the first time, this experiment was able to shed light on the error growth process from convective to planetary scales. The existence of multiple scale-dependent error growth regimes seems to confirm the radical idea that the global atmosphere has a finite limit of predictability between 2–3 weeks, no matter how small the initial error.