How Well Do CMIP5/CMIP6 Models Simulate Northeast U.S. Extreme Precipitation and Its Associated Circulation?

Fourteen Coupled Model Intercomparison Project Phase 5 (CMIP5) model historical simulations (1850-2005) are evaluated for their ability to reproduce observed Northeast US precipitation and its associated circulation, with particular emphasis on extreme (top 1%) precipitation. The observed extreme precipitation circulation regimes are identified using k-means clustering of reanalysis 500-hPa geopotential heights, resulting in four patterns. 20 regionally-based metrics are defined (8 related to precipitation, 12 related to extreme circulation patterns) and used to evaluate the success of the individual models in simulating observed precipitation and extreme precipitation, including regional and seasonal variations, and the four associated extreme precipitation circulation patterns.

While no models completely match observations based on the given metrics, most models realistically reproduce aspects of both Northeast US precipitation and the circulation associated with extreme precipitation. In general, higher resolution models score higher for both precipitation and circulation metrics. While most models are shown to produce too much drizzle and a too-fat extreme precipitation tail, the majority of the models capture the seasonality and magnitude of mean daily precipitation and extreme precipitation well. Although the majority of the simulated extreme precipitation circulation patterns can be matched to an observed pattern, there are differences in how well each model replicates the location and magnitude of specific troughs and ridges within the patterns, as well as the seasonality of the patterns.

As Phase 6 (CMIP6) historical runs become available, similar analysis is done to determine how well the CMIP6 models replicate Northeast US precipitation and extreme precipitation, and its associated circulation, and how the CMIP6 results compare to the CMIP5 results.