Assessing the influence of background state and model bias on weather extremes using initialized ensembles in a climate model

Abstract:

Extreme weather events with great socioeconomic impact, such as tropical cyclones (TCs) and atmospheric rivers (ARs), have attracted increasing attention among the weather and climate researchers. The genesis and evolution of TCs and ARs are significantly impacted by the large-scale background environment flow. The current generation of general circulation models (GCMs) still suffers from significant deficiencies in simulating the observed statistical properties of these extreme events, and a major cause of this could be errors in the simulated mean climate. In an ensemble of GCM runs initialized from observations, the ensemble-average climate will slowly drift from the observed climate to the simulated climate over a period of a few weeks.

We analyze the time varying properties of weather extremes in a large ensemble of initialized runs carried out using the Energy Exascale Earth System Model (E3SM) atmospheric component (EAM), as compared to their properties in a control run of the model and in observations. This allows us to distinguish between errors associated with model physics or resolution, which directly affect the extreme events, as opposed to the errors in the model mean state, which have an indirect effect on the extreme events.