Near-term climate forecasts such as decadal climate projections are increasingly used for planning adaptation measures. To ensure the applicability of these forecasts, systematic errors of the forecasting system must be corrected with the help of a statistical model.

The data

The algorithm used does not restrict the number and type of predictors a priori, but uses a systematic model selection strategy based on a boosting procedure to select the most relevant predictors. This provides an improved algorithm that ignores irrelevant predictors and thus reduces model uncertainty and also improves the prediction quality of medium-term climate prediction models.

Publication of the data set: Pasternack A, Grieger J et al (2020) https://doi.org/10.5281/zenodo.3975759