Differing Evolutions of Flow and Moisture Bias in a Climate Model and Their Influences on Weather Extremes

The current generation of general circulation models (GCMs) continues to suffer from significant deficiencies in simulating the observed statistical properties of weather extremes, such as extreme precipitation events, tropical cyclones (TCs) and atmospheric rivers (ARs). Factors responsible for these deficiencies include lack of sufficient model resolution and/or errors in model parameterizations, as well as biases in the simulated mean climate. In an attempt to distinguish among these factors, we carry out a large ensemble of 4-week forecasts initialized from observations using the Energy Exascale Earth System Model (E3SM) atmospheric component (EAM). At the initial time, the ensemble-average forecast does not have a bias as compared to observations (by construction), even though parameterization and resolution errors are present in the model. The ensemble-average bias grows as the forecast progresses, asymptotically approaching the model bias as estimated from long control runs over a few weeks. Different components of climate bias flow bias and moisture bias grow at different rates, with the moisture bias growing more slowly than the flow bias. We analyze the time-varying properties of weather extremes in the large ensemble of forecasts, compared to their properties in observations and in a control run of the model. The responses of extreme events to the biases in the ensemble-mean climate are diverse. Heatwaves are relatively insensitive to mean bias. TCs display non-monotonic error evolution, attributable to the differing growth rates for flow and moisture bias. TC strength appears to be controlled by the flow bias, whereas TC numbers are more influenced by the moisture bias. ARs exhibit a slower evolution of error due to the slower growth of moisture bias. We also carry out sensitivity forecast experiments where we artificially speed-up the development of the moisture bias to isolate its impact on simulated weather extremes. These results help improve our understanding of the impact of the large-scale environment on different types of weather extremes, and potentially bring new insights into the mechanism of such events and/or the impact of model bias, while providing a basis for improving the model simulation and prediction of these events.