A08 - Characterization and prediction of quasi-stationary atmospheric states
Head(s): Dr. Maximilian Engel (FU), Prof. Dr. Péter Koltai (FU), Prof. Dr. Stephan Pfahl (FU)
Participating institution(s): FU Berlin
Research areas: Mathematics, Geosciences (atmospheric, ocean and climate sciences)
Atmospheric dynamics in the mid-latitudes features quasi-stationary atmospheric states (QSAS), also denoted as atmospheric blocking. Despite the existence of a multitude of data-driven approaches for their identification, a unified mathematical characterization of such states and the conditions leading to transitions between them is still lacking. We aim to remedy this by describing and analysing QSAS as persistent dynamical structures both in nonautonomous deterministic and nonstationary non-deterministic dynamical models. This dual view is expected to elucidate different aspects of the cause, the persistence, and the predictability of QSAS.
For deterministic, potentially infinite-dimensional models, we plan to consider trajectorybased and geometric stability theory to assess how the system enters and exits QSAS. Simultaneously, set-oriented and transfer-operator based approaches will find persistent dynamicalstates in time-series generated by non-deterministic models, yielding quantitative stability measures and the basis for comparison with meteorological data analysis (see below). Geographic localization of the atmospheric models to certain windows of longitude and latitude is necessary to isolate the QSAS from the remainder of the atmosphere. This will be achieved by combining recent data-driven operator-learning techniques with regime-estimation for nonstationary processes (the regimes constituting the neglected part of the atmosphere), developed
by the CRC’s former Mercator Fellow Illia Horenko.
Complementing the above Eulerian view, we will analyse QSAS from a Lagrangian, transportoriented perspective, enabled by modified multiplicative ergodic theory and coherent sets. In addition to advancing theory to allow for a suitable formal characterization of QSAS, a complementary, numerical and data-driven approach will be pursued to investigate the applicability of both the Eulerian and the Lagrangian approach to meteorological data, also in collaboration
with project C06. The data will come from simulations of a hierarchy of models on various scales, from conceptual systems that exhibit only the key characteristics of QSAS to comprehensive general circulation models and reanalysis data with realistic QSAS representation. In particular, it will be investigated how the characteristics of unstable modes leading away from QSAS, but also of coherent set representations, can be exploited towards improved data-driven forecasts of transitions between QSAS.