A posteriori adjustment of near-term climate predictions: Accounting for the drift dependence on the initial conditions

Climate predictions initialized from an observationally based state (OBS) drift toward the state of the unconstrained model, which makes the use of a posteriori correction methods essential to disentangle the climate signal of interest from the model bias. We propose that applying a linear regression of the predictions and corresponding OBS on the OBS initial conditions (IC), and substituting the latter for the former, offers an effective method for bias correction. The impact of this new method is examined on monthly means of large-scale sea surface temperature indices and the Northern Hemisphere sea ice extent in EC-Earth2.3 predictions. This postprocessing adjustment through a linear regression on the averaged OBS over the first forecast month as a temporarily smoothed proxy for OBS IC shows a reduction of model error with respect to the two established bias correction methods. Improvements are seen for at least two seasons and for some variables up to 5 years.