Incorporation of Decadal Trends into the Calibration, Bridging, and Merging (CBaM) Method for Seasonal Prediction of North American Temperature and Precipitation

Seasonal forecast skill of extratropical temperature and precipitation is typically reliant on teleconnection patterns, however, research has shown that dynamical models sometimes fail to represent these teleconnection patterns skillfully, causing reduced model forecast skill. Additionally, multi-model ensembles (MMEs) may not completely represent uncertainty. Recently, Strazzo et al. 2018 applied the Calibration, Bridging, and Merging (CBaM) method to North American Multi-Model Ensemble (NMME) data. Calibration relates a predictand (such as observed regional precipitation) to a predictor (such as forecasted regional precipitation), while bridging relates a climate index, such as forecasted NINO3.4 SST, to observed regional precipitation. These CBaM forecasts achieved higher Brier skill scores and reliability compared to raw NMME forecasts, particularly for 1 month lead temperature, though precipitation showed fewer improvements. Though this methodology was successful, it does not incorporate decadal trend information into the calibration step, which may yield further skill increases over raw model output. Decadal trends are known to be a primary source of skill in seasonal climate forecasts; however, models may not correctly represent trends, and calibration methods do not typically account for trends. We examine the impact of trends on the skill of seasonal forecasts for both terciles and extremes. This presentation aims to show potential for skill increase in seasonal temperature and precipitation predictions with incorporation of decadal trend information.