Extratropical Prediction Skill of the Subseasonal‐to‐Seasonal (S2S) Prediction Models

The deterministic prediction skill of the 10 operational models participating in the subseasonal‐to‐seasonal (S2S) prediction project is assessed for both the extratropical stratosphere and troposphere. Based on the mean squared skill score of 50‐ and 500‐hPa geopotential height forecasts, the overall prediction skill is on average 16 days in the stratosphere and 9 days in the troposphere. The high‐top models with a fully resolved stratosphere typically have a higher prediction skill than the low‐top models. Among them, the European Centre for Medium‐Range Weather Forecasts model shows the best performance in both hemispheres. The decomposition of model errors reveals that eddy errors are more important than zonal‐mean errors in both the stratosphere and troposphere. While the errors in the stratosphere are dominated by planetary‐scale eddies, those in the troposphere are equally influenced by planetary‐ and synoptic‐scale eddies. This result indicates that subseasonal‐to‐seasonal prediction could be improved by better representing planetary‐scale wave activities in the model. Plain Language Summary How well do the operational models predict weather and climate beyond 2 weeks? To answer this question and to better understand the gap between short‐term weather forecasts and long‐term seasonal forecasts, the international research organizations recently launched the subseasonal‐to‐seasonal (S2S) prediction project with the participation of over 10 modeling centers across the world. As a first step to improve S2S prediction, this study assesses the prediction skill of S2S models. It is found that while most models reliably predict the stratospheric circulation for 2‐ to 4‐week lead times, they do so only for lead times of about 1 to 2 weeks in the troposphere. Both the stratospheric and tropospheric errors are mainly explained by eddy errors. More specifically, the misrepresentation of planetary‐scale waves, with length scales of a few thousand kilometers or larger, is the key factor that determines the prediction limit. This result suggests that S2S prediction could be improved by better representing large‐scale wave activities in the model. Key Points S2S models have extended prediction skills of approximately 16 days in the stratosphere and 9 days in the troposphere Both the stratospheric and tropospheric prediction errors are mainly caused by eddy errors The planetary‐scale eddies explain most errors in the stratosphere and almost half of the errors in th