Looking for Seasonal Forecasts of Opportunity in the NMME

Modeling centers participating in the North American Multi-Model Ensemble (NMME) experiment began generating real-time monthly forecasts in 2011. Operational centers, including the National Weather Service Climate Prediction Center (CPC), rely on these NMME forecasts to provide valuable guidance for monthly and seasonal climate outlooks. While the NMME is a state-of-the-art tool for seasonal climate prediction, NMME forecasts require post-processing to improve reliability. We present results from two post-processing techniques: probability anomaly correlation (PAC) and calibration, bridging and merging (CBaM). We compute verification statistics for post-processed and uncalibrated NMME forecasts of North American temperature and precipitation over the real-time period of 2012 through the present and compare these forecasts to empirical ENSO-derived forecasts. Results indicate that the NMME, and in particular the post-processed NMME, outperforms the empirical forecasts. Despite these improvements, seasonal forecast skill with the post-processed NMME remains relatively low on average. However, we highlight specific circumstances under which forecasters can have increased confidence in verification of a predicted outcome. Finally, we provide some context for these “forecasts-of-opportunity” and suggest potential future developmental work that may better leverage these forecasts to further improve operational climate prediction.