Large deviations of temperature over Northern Hemisphere regions

Abstract:

When applying a large deviation limit law to some data set, one checks usually the convergence of rate function estimates as a function of increasing averaging window length. Depending on the averaging window of convergence and on the properties of the chosen observable, the rate function estimate can be useful to study persistent extreme events. We consider CMIP6 simulations of the MPI-ESM-LR model and study the convergence of large deviation rate functions of near-surface, summer and winter temperature, over several regions of the Northern Hemisphere. We find that, in general, rate functions converge in case of land regions on time scales shorter than one season. This suggests that the rate functions can be used to compute the probability of persistent temperature events. In case of oceanic regions, the rate functions do not converge, probably due to the low-frequency variability of temperature over the oceans. Furthermore, we study the large-scale atmospheric fields related to the regional temperature large deviations.