Internal Variability in Regional Climate Downscaling at the Seasonal Scale

Abstract

To study the internal variability of the model and its consequences on seasonal statistics, large ensembles of twenty 3-month simulations of the Canadian Regional Climate Model (CRCM), differing only in their initial conditions, were generated over different domain sizes in eastern North America for a summer season. The degree of internal variability was measured as the spread between the individual members of the ensemble during the integration period. Results show that the CRCM internal variability depends strongly on synoptic events, as is seen by the pulsating behavior of the time evolution of variance during the period of integration. The existence of bimodal solutions for the circulation is also noted. The geographical distribution of variance depends on the variables; precipitation shows maximum variance in the southern United States, while 850-hPa geopotential height exhibits maximum variance in the northeast part of the domain. Results suggest that strong precipitation events in the southern United States may act as a triggering mechanism for the 850-hPa geopotential height spread along the storm track, which reaches its maximum toward the northeast of the domain. This study reveals that successive reductions of the domain size induce a general decrease in the internal variability of the model, but an important variation in its geographical distribution and amplitude was detected. The influence of the internal variability at the seasonal scale was evaluated by computing the variance between the individual member seasonal averages of the ensemble. Large values of internal variability for precipitation suggest possible repercussions of internal variability on seasonal statistics.