Subseasonal to Decadal Prediction Research Framework with NCAR’s CESM1 and CESM2

Interest in earth system prediction on timescales from subseasonal to decadal (S2D) has grown rapidly within the international research community in the last decade. However, many fundamental questions related to prediction and sources of predictability on these time-scales are still unanswered. Secondly, there is very limited understanding of how choices in forecast system design, such as model initialization, ensemble perturbation methods, model top/resolution affect forecast skill. We discuss here the challenges and choices that have been made for S2D prediction based on NCAR’s Community Earth System Model, v1 (CESM1), particularly focusing on initialization of the land, ocean, and atmospheric components for the various timescales, and present prediction skill of this model on subseasonal, seasonal, and decadal timescales. In addition, we discuss ongoing efforts to develop a more seamless S2D prediction framework using CESM2 that synthesizes the experience and expertise developed in previous, largely uncoordinated, prediction projects focused on distinct timescales. The goal is a unified system that can be used by multiple communities to address fundamental science questions related to sources of predictability across S2D timescales, and to improve our understanding of how choices of system design impact predictability.