AbstractOn the basis of multivariate linear regression with an adaptive choice of climate indices as predictors, a seasonal forecast with a lead time of 2 months was applied to Korea on a monthly basis, and leave‐one‐out cross‐validation was applied to obtain forecasting skill at the 1% significance level. The monthly ACC (anomaly correlation coefficient) skill was 0.42–0.65 for temperature and 0.35–0.63 for precipitation. COD (coefficient of determination) was 18–42% for temperature and 14–39% for precipitation. The first coupled SLP pattern related to Korean climate is very similar to the correlation pattern between the preceding climate index and SLP at the target month, indicating that preceding climate indices can be dynamically linked to Korean climate. For example, the PNA index at a lead time of 5 months prior to October is closely related to a circulation anomaly with weak negative correlation over the Okhotsk Sea to East Sea and strong positive correlation over a broad band from Lake Baikal to China. This SLP pattern provides conditions that can dynamically induce cold advection from northwestern Asia around Lake Baikal toward the Korean Peninsula, resulting in cooling over Korea. Copyright © 2006 Royal Meteorological Society