Fidelity of the Simulation of Non-stationary Climate Extremes in Low- and High- Resolution Climate Models

Abstract:

The largest societal impacts of climate occur from climate extremes. It is thus important to evaluate the performance of weather and climate models, used for climate predictions and future projections, in simulating climate extremes. Here, we analyze a suite of atmosphere-only and fully-coupled, low- and high-resolution versions of the US Department of Energy (DOE) baseline Accelerated Climate Model for Energy (ACME v0) to assess their fidelity in simulating the statistics of extremes of surface temperature and precipitation. We generate linear models of extremes using the Generalized Extreme Value (GEV) theory for both observations and model output that capture both the stationary and non-stationary component of the variability of extremes. The non-stationary terms capture the dependence of extremes on large-scale natural modes of variability like the El Nino, North Atlantic Oscillation (NAO), Madden-Julian Oscillation (MJO), etc. We test the null hypothesis that the GEV model parameters derived from observations and the climate simulation are equal. Preliminary results suggest that while the high-resolution simulations capture the stationary component of extremes better over areas with varied topography, there is little gain in simulating the impact of low-frequency variability on climate extremes in simulations of the past few decades with the atmosphere-only model.