Based on multiple proxies from the Southern Hemisphere, an austral summer (December-January-February: DJF) Antarctic Oscillation Index (AAO) since 1500 A.D. was reconstructed with a focus on interannual to interdecadal variability (<50 a). By applying a multivariate regression method, the observational AAO-proxy relations were calibrated and cross-validated for the period of 1957 89. The regressions were employed to compute the DJF-AAO index for 1500 1956. To verify the results, the authors checked the explained variance (r 2 ), the reduction of error (RE), and the standard error (SE). Cross-validation was performed by applying a leave-one-out validation method. Over the entire reconstruction period, the mean values of r 2 , RE, and SE are 59.9%, 0.47, and 0.67, respectively. These statistics indicate that the DJF-AAO reconstruction is relatively skillful and reliable for the last ～460 years. The reconstructed AAO variations on the interannual and interdecadal timescales compare favorably with those of several shorter sea level pressure (SLP)-based AAO indices. The leading periods of the DJF-AAO index over the last 500 years are ～2.4, ～2.6, ～6.3, ～24.1, and ～37.6 years, all of which are significant at the 95% level as estimated by power spectral analysis.
