Forecasting Temperature Extremes with the North American Multi-Model Ensemble

This study examines the forecast skill of 2 m temperature extremes in the monthly mean (t2m), maximum (tmax), and minimum (tmin) using the North American Multi-Model Ensemble (NMME), an ensemble of state-of-the-art coupled global climate models. The NMME currently provides real-time guidance for NOAA's operational short-term climate forecasts, and includes a database of retrospective forecasts (1982-2010), used for bias correction, calibration, and skill studies. Seven models from the NMME contribute to this study: NCEP-CFSv2, Environment Canada's CanCM3 and CanCM4, GFDL's CM2.1 and FLOR, NASA-GEOS5, and NCAR-CCSM4. A new tmax/tmin dataset for North America, spanning 1982-2014, was constructed for this study using station observations from three sources, interpolated to a grid matching the NMME. The definition of “extreme” is considered, as well as different forecast construction methods. This study assesses forecasts for extremes in general, and also examines warm-season positive tmax extremes and cold-season negative tmin extremes.