A Hybrid Ensemble Canonical Correlation Prediction Model of the Winter Siberian High. Issue 4 (16th February 2021)

• Record Type: Journal Article
• Authors: Yang, Hongqing; Fan, Ke
• Is Part Of: Journal of geophysical research. Volume 126:Issue 4(2021)
• Page Start: n/a
• Page End: n/a
• Publication Date: 2021-02-16
• Subjects: CFSv2 -- downscaling ensemble canonical correlation -- hybrid prediction -- winter Siberian High -- year‐to‐year increment; Atmospheric physics -- Periodicals; Geophysics -- Periodicals; 551.5
• DOI: 10.1029/2020JD033780
• Languages: English
• ISSNs: 2169-897X

Abstract

Abstract: As one of the main components of the East Asian winter monsoon, the winter Siberian High (SH) plays an important role in the variability of East Asian climate. However, the Climate Forecast System, version 2 (CFSv2), shows limited prediction skill for the winter SH. To improve the prediction skill, a hybrid ensemble canonical correlation (ECC) prediction model is established for the winter SH and SH intensity index (SHI) basing the year‐to‐year increment method and an efficient downscaling approach. Hence, three preceding predictors from observation/reanalysis [sea‐ice concentration (SIC), snow‐cover extent (SCE), and sea surface temperature (SST)] and one integrated current predictor from CFSv2 [surface air temperature and sea level pressure (SAT&SLP)] are selected based on their fundamental physical roles. Considering these individual predictors, four separate downscaling schemes are constructed. The regional‐mean temporal anomaly correlation coefficient (ACC) of the winter SH for each scheme is 0.54 (SIC‐scheme‐SH), 0.50 (SCE‐scheme‐SH), 0.72 (SST‐scheme‐SH) and 0.42 (SAT&SLP‐scheme‐SH) [four schemes exceed significant at the 1% level]. However, the skill of each scheme differs in regional and temporal distribution. Thus, a hybrid ECC prediction model is proposed by employing multiple linear regression. The regional‐mean temporal ACC of the winter SH between the observed and predicted results increases from −0.12 (CFSv2; not significant at the 10% level) to 0.85 (significant at the 1% level). Besides, the correlation coefficient between the observation and hybrid ECC scheme for winter SHI is 0.90 (significant at the 1% level). Furthermore, the strongest winter SH in 2012 is reproduced well by ECC‐scheme‐SH. Key Points: A hybrid ensemble canonical correlation prediction model is developed for the winter Siberian High with high prediction skill An efficient downscaling method and the year‐to‐year increment approach are applied to the prediction model Three preceding predictors with fundamental mechanisms and a simultaneous integrated predictor with high prediction skill are considered