Assessing the Skill and Reliability of NCAR-CESM SMYLE Predictions of the North Atlantic Subtropical High in Multi-Year Timescales

Abstract

The North Atlantic Subtropical High (NASH), also known as the Bermuda High, is a semi-permanent high-pressure system located over the North Atlantic Ocean that influences summertime climate across the southeastern United States (SEUS). Variations in the strength and position of the NASH western ridge affect regional circulation and are associated with changes in precipitation, heat, drought, and coastal inundation. Since low-frequency sea surface temperature variability may contribute to variations in NASH position, initialized climate predictions could provide a source of multi-year (1-2 years) predictability for the NASH and its regional impacts. This study evaluates the ability of the NCAR Community Earth System Model Seasonal-to-Multiyear Large Ensemble (SMYLE) hindcasts to predict the NASH western ridge and associated SEUS precipitation at lead times extending to two years. SMYLE western-ridge positions are evaluated against ERA5 reanalysis, and precipitation forecasts are verified against CHIRPS observations. Systematic biases in the simulated NASH are addressed before evaluating western-ridge position and quadrant prediction. Precipitation prediction is assessed using probabilistic verification metrics, including the Brier Score, Brier Skill Score, and receiver operating characteristic area under the curve (ROC AUC). Precipitation skill is additionally conditioned on whether individual ensemble members correctly predict the observed NASH quadrant to determine whether improved representation of the NASH is associated with improved precipitation prediction. SMYLE exhibits a systematic bias in the NASH western ridge position, particularly in longitude, and removing the mean bias improves agreement with ERA5. Considerable case-to-case variability in western-ridge position remains after bias correction, and quadrant prediction varies with initialization, lead time, season, and observed NASH quadrant. SEUS precipitation forecasts exhibit limited skill relative to climatology, although SMYLE retains some ability to discriminate between above- and below-average precipitation conditions. Conditioning the forecasts on correct NASH quadrant prediction generally improves precipitation prediction relative to members that incorrectly predict the NASH, but these improvements do not consistently result in greater skill than climatology or the full ensemble. Circulation composites further show that correct quadrant classification does not necessarily correspond to an exact representation of the NASH western ridge or regional precipitation. In particular, the direction of longitudinal western-ridge errors is associated with systematic precipitation differences, with eastward ridge errors generally corresponding to wetter SMYLE conditions and westward errors to drier conditions. These results demonstrate that accurate prediction of the NASH provides some useful information for SEUS precipitation prediction, but NASH quadrant prediction alone is insufficient to determine precipitation skill in this region. The results highlight both the potential value of the NASH as a source of multi-year regional climate predictability and the importance of representing the position of the western ridge when translating circulation predictability into regional precipitation forecasts.