Limitations of ENSO-Based Prediction of Korean Winter Temperature

Abstract This study investigates the statistical and dynamical relationship between the El Niño-Southern Oscillation (ENSO) and winter surface air temperature (SAT) in Korea, using station observations and reanalysis data from 1920 to 2023. Historical SAT records are compiled from 7, 14, and 60 stations for 1920–1959, 1960–1972, and 1973–2023, respectively. Despite the statistically significant correlation ( r = 0.28) between the Niño 3.4 index and winter SAT in Korea, ENSO alone explains only a limited amount of interannual variability. Classifying the SAT anomalies according to the ENSO phase (i.e., warm for El Niño and cold for La Niña), the Niño 3.4 index yields binary-classification accuracy of 0.68; however, about half of the correctly classified anomalies fall within ±0.5 standard deviations from the climatological mean. Also, composite circulation patterns based on ENSO phases differ structurally from those associated with actual SAT anomalies. A multiple linear regression analysis reveals that mid- to high-latitude climate variables, such as the East Asian winter monsoon, western North Pacific (WNP) sea surface temperatures (SSTs), and the Arctic Oscillation, exhibit stronger and more stable associations with Korean winter SAT than ENSO. Especially WNP SSTs show the largest standardized regression coefficients (> 5.0) to indicate their dominant role. This study suggests the need for integrated forecasting approaches that consider both the tropical and extratropical influences, rather than relying solely on ENSO signals for improving the accuracy of seasonal climate predictions and supporting adaptive risk management strategies for wintertime extremes in the Korean Peninsula.