Bias, drift, anomaly calculations, and the complications of volcanoes for Pacific decadal climate predictions

In initialized hindcasts, the model drift away from the initial state due to model systematic error is one of the greatest challenges facing initialized prediction today. The resulting bias and drift, and differences in trends between initial states and drifted states, introduce difficulties in calculating anomalies to assess skill of initialized predictions. We examine several methods of calculating anomalies, and use DPLE initialized hindcasts of Pacific SSTs associated with the Interdecadal Pacific Oscillation to illustrate issues with anomaly calculations. Previous results for S2D hindcast skill for the Pacific have shown lower values for the tropical eastern Pacific in time series calculations. The complicating effects of the tropical Pacific SST response to volcanic eruptions is noted to contribute to this effect.