Evaluation of extreme rainfall and temperature over North America in CanRCM4 and CRCM5

Abstract

We assess the ability of two Canadian regional climate models (RCMs), CanRCM4 and CRCM5, to simulate North American climate extremes over the period 1989–2009. Both RCMs use lateral boundary conditions derived from the ERA-Interim reanalysis and share the same dynamical core but use different nesting strategies, land-surface and physics schemes. The annual cycle and spatial patterns of extreme temperature indices are generally well reproduced in both models but the magnitude varies. In central and southern North America, maximum temperature extremes are up to 7 °C warmer in CanRCM4. There is a cool bias in minimum temperature extremes in both RCMs. The shape of the annual cycle of extreme rainfall varies between simulations. There is a wet bias in CRCM5 extreme rainfall on the west coast throughout the year and in winter rainfall elsewhere. In summer both RCMs have precipitation biases in the south-east. These rainfall and temperature biases are likely associated with differences in the physical parameterisation of rainfall. CanRCM4 simulates too little convective rainfall, while over-estimating large-scale rainfall; nevertheless, cloud cover is well simulated. CRCM5 simulates more large-scale rainfall throughout the year on the west coast and in winter in other regions. The spatial extent, intensity and location of atmospheric river (AR) landfall are well reproduced by the RCMs, as is the fraction of winter rainfall from AR days. Spectral nudging improves agreement on landfall latitude between the RCM and the driving model without greatly diminishing the intensity of the rainfall extreme.