Arctic Oscillation Modulation of Winter Air-Sea Coupling in the East/Japan Sea: Persistence, Timescales, and Extremes
Abstract
The winter climate of the East/Japan Sea (EJS) is strongly affected by the Arctic Oscillation (AO), yet how AO polarity reshapes the memory, coupling patterns, and predictability of sea-surface temperature anomalies (SSTA) remains poorly quantified. Using 30 winters (1993--2022) of daily OISST and ERA5 fields, we combine multivariate Maximum Covariance Analysis (MCA) with an Ornstein--Uhlenbeck (OU)-like integration of atmospheric principal components (PCs). The leading coupled mode explains 87% (+AO) and 75% (-AO) of squared covariance, with SSTA hot spots in East Korea Bay and along the subpolar front. Zero-lag correlations between the SSTA PC and OU-integrated atmospheric PCs reveal characteristic memory timescales ($\tau$) of $\sim$18--25 days for wind-stress curl (CurlTau), $\sim$15--30 days for near-surface air temperature (ATMP) and zonal winds, and $\sim$30--50 days for sea-level pressure (SLP) and meridional winds -- longer under -AO. Detrended Fluctuation Analysis (DFA) shows SSTA persistence $H \approx 1.3$--$1.4$ and that integrated atmospheric responses acquire ocean-like persistence, validating Hasselmann's stochastic framework for winter EJS. AO-phase contrasts align with a curl$\rightarrow$Ekman pumping$\rightarrow$eddy/SSH$\rightarrow$SST pathway: +AO favors anticyclonic/downwelling responses and warmer SSTA, whereas -AO favors cyclonic/upwelling and cooler SSTA. These diagnostics identify phase-specific predictor windows (e.g., 3-week OU-integrated CurlTau/ATMP; 4--7-week SLP/V-wind under -AO) to initialize subseasonal extremes prediction (marine heatwaves and cold-surge-impacted SST). The approach quantifies memory scales and spatial coupling that were not explicitly resolved by previous composite analyses, offering a tractable foundation for probabilistic forecast models.