Ten years, sixteen fully coupled global models, and hundreds of research papers later, the North American Multi-Model Ensemble (NMME) monthly-to-seasonal prediction system is looking ahead to its second decade. The NMME comprises both real-time, initialized predictions and a substantial research database; both retrospective and real-time forecasts are archived and freely available for research and development. Many US-based and international entities, both private and public, use NMME data for regional or otherwise tailored forecasts. The system’s built-in evolution, with new models gradually replacing older ones, has been demonstrated to gradually improve the skill of 2 m temperature and sea surface temperature, although precipitation prediction remains a difficult problem.
This paper reviews some of the NMME-based contributions to seasonal climate prediction research and applications, progress on scientific understanding of seasonal prediction and multi-model ensembles, and new techniques. Several prediction-oriented aspects are explored, including model representation of observed trends and the under-prediction of below-average temperature. We discuss potential new directions, such as higher-resolution models, hybrid statistical-dynamical techniques, or prediction of environmental hazards such as coastal flooding and the risk of mosquito-borne diseases.