Estimating the decadal-scale climate predictability limit using optimal local dynamic analogues

Abstract Accurately estimating decadal predictability limits (PLs) is essential for advancing long-term climate predictions and understanding decadal-scale variability. This study combines the optimal local dynamic analog (OLDA) algorithm with the nonlinear local Lyapunov exponent (NLLE) method to estimate decadal PLs of oceanic and atmospheric variables, using long-term reanalysis datasets. Results demonstrate that the OLDA algorithm can enhance identification of analog states and improve PL estimation. The decadal PLs of sea surface temperature (SST) show regional and seasonal differences, with zonal mean values ranging from 8 to 17 years, and higher values in boreal summer and autumn, especially in the Northern Hemisphere and Southern Ocean. Sea level pressure (SLP) decadal PLs range from 8 to 11 years, exhibiting patchy distribution and seasonal variation. The global mean PL of SLP reaches about 10 years in boreal spring and 9 years in other seasons. SLP and SST PL distributions differ across seasons, reflecting the complexity of ocean-atmosphere interactions. Decadal PLs of major climate modes were also estimated, e.g., decadal PL of the SST Inter-Hemispheric Dipole (SSTID) is ~ 17 years, Atlantic Multidecadal Oscillation (AMO) ~ 14 years, Pacific Decadal Oscillation (PDO) ~ 13 years, North Atlantic Oscillation (NAO) ~ 16 years, Northern Hemisphere Annular Mode (NAM) ~ 11 years, and Southern Hemisphere Annular Mode (SAM) ~ 15 years. These modes display distinct predictability patterns and seasonal variations, highlighting their unique roles in regional climate dynamics. These findings enhance our understanding of decadal-scale predictability.