Using a self-organizing map (SOM) we identify nine large-scale meteorological patterns (LSMPs) that span the set of circulation patterns on atmospheric river (AR) days in western North America. By limiting ourselves to this set of days, we can better explore the variations among the circulation patterns that often lead to west coast extreme precipitation. The SOM is defined from reanalysis geopotential heights, winds, and column water vapor, and tends to separate AR days by latitude as well as orientation of landfall. The relative frequency of each of the LSMPs is then compared between the reanalysis and each of 25 models from the Coupled Model Intercomparison Project - phase 5 (CMIP5). Some models better match the relative frequency of each pattern historically. We explore the extent to which the inter-model spread can be tied to biases in the placement of the subtropical jet and eddy-driven jet variability. We also investigate the amount of variance in pattern frequency that can be produced due to internal variance alone using the Community Earth System Model large ensemble (CESM LE). Finally we examine projected jet shifts under climate change in relation to the change in the relative frequency of each of the patterns between the historical period and end-of-century. Many models show an increase in the percent of all AR days associated with the most North-impacting anti-cyclonic AR pattern.