The Seasonal-to-Multiyear Large Ensemble (SMYLE) Prediction System Using the Community Earth System Model Version 2
We introduce a new initialized prediction system using CESM2 that targets the heretofore neglected timescales between seasonal and decadal prediction. The SMYLE system encompasses a collection of 20-member ensemble coupled hindcasts initialized quarterly (February, May, August, and November) between 1970-2019 and integrated for 24 months, and also the code, data, and analysis infrastructure needed to replicate the experiments and skill verification. SMYLE exhibits skill for El Niño–Southern Oscillation that is very competitive with other prominent seasonal prediction systems (e.g., the NMME multi-model mean), and it permits a deeper exploration of predictability limits than most seasonal systems given its broad temporal sampling and extended-length hindcasts. The broad overview of prediction skill in this paper reveals varying degrees of potential for useful multiyear predictions of seasonal anomalies in the atmosphere, ocean, land, and sea ice. Overall, SMYLE demonstrates that a relatively simple prediction system design (no CESM-specific data assimilation was used to initialize component models) can yield impressive seasonal to interannual skill, making it a powerful tool for studying predictability mechanisms and for testing prediction system innovations.
SMYLE will likely serve as a template for multi-model efforts (including international efforts coordinated through the World Climate Research Programme) to probe the Year 2 prediction space. Community interest appears to be high (metrics for the SMYLE pre-print include 1,000 views and 238 downloads). The full availability of SMYLE datasets and code infrastructure facilitates community engagement with and extensions of the experiment. Multiple university scientists (including graduate students and postdocs) are already using SMYLE in their research.
A novel seasonal-to-multiyear prediction system using CESM2 has been generated and assessed, and the results are sufficiently promising that SMYLE is expected to catalyze the growth of a new subfield within the domain of initialized Earth system prediction for climate timescales.