Multi-Model Seasonal Forecasting at Environment Canada: Past, Present and Future

Wednesday, 17 December 2014: 5:40 PM
William J Merryfield, Canadian Centre for Climate Modelling and Analysis, Environment Canada, Victoria, BC, Canada
Typically, a given operational or modeling center produces seasonal forecasts using a single model version that represents that center’s current state of model development. By contrast, Environment Canada (EC) has employed multi-model ensembles to produce its dynamical seasonal forecasts since their inception in 1995. Successive versions of this forecasting system have drawn upon “models of opportunity” that were developed for other purposes by EC’s climate and meteorological modeling centers. This talk will describe the origins of this single-center/multi-model approach, and its evolution through configurations consisting of two and subsequently four 2-tier models in the 1990s and 2000s, to the current Canadian Seasonal to Interannual Prediction System (CanSIPS). The latter is based on two global climate models, CanCM3 and CanCM4, that employ common ocean, sea ice and land components coupled to two versions of CCCma’s atmospheric model. Lessons learned concerning the efficacy of this multi-model approach will be discussed, focusing on the relative merits of adding more ensemble members versus adding more models, and on methods for combination and calibration. Current efforts aimed at improving the ensemble of prediction models contributing to CanSIPS and the techniques used to initialize them will then be summarized.