A1 - Multiscale analysis of the evolution of forecast uncertainty
This project will investigate the upscale growth and amplification of forecast uncertainty in the mid-latitudes by combining dedicated numerical experiments, development of theory, and quantitative process-based analysis. Numerical ensemble experiments with the ICON model will be performed. Ensembles will be generated with varying magnitude of initial condition uncertainty and by introducing uncertainty in the representation of convection, using both stochastic parameterization and convection-permitting simulations.
The processes that govern the projection of the convective-scale uncertainty on larger scales will be examined in detail. We will employ a quantitative diagnostic framework developed during Phase 1, which is based on a potential-vorticity (PV) budget. Evaluating PV tendencies due to parameterized diabatic processes, in combination with advective PV tendencies, has proven useful to gain insight into the processes underlying amplification of forecast errors (Baumgart et al. 2018; Baumgart et al. 2019).
Differences and commonalities in the processes that govern the evolution of forecast uncertainty will be quantified within different weather situations. The uncertainty associated with convection will be investigated in detail by developing diagnostics to quantify the convective heat source, the redistribution of moist static energy, and the convective momentum transport for both parameterized and resolved convection.
A new theoretical framework for understanding upscale error growth will be developed. Single-scale asymptotics (Klein 2010) and numerical experiments (Craig and Selz 2018) suggest that convectively-forced motions on the mesoscale can be represented by the weak temperature gradient (WTG) approximation. We will derive a multiscale asymptotic model that will describe interactions between the WTG motions and the quasigeostrophic, synoptic-scale motions. The scale-interaction terms resulting from this theory will be evaluated to test the accuracy of the approximate theory.
Finally, a synthesis is sought by comparing the scale-interaction terms with the tendencies obtained in the PV framework. From this analysis we will identify the weather situations where upscale error growth is most likely to create an intrinsic limit on predictability, as well as the physical mechanisms that must be represented in an ensemble forecast system to accurately describe the error growth process.