Estimates of future changes in climate are typically derived from complex climate model simulations driven by the expected changes in anthropogenic emissions including greenhouse gases that warm the planet. These simulations of future climate are widely referred to as so-called climate projections. Climate projections are affected by a range of different uncertainties, whereby internal variability is the dominant source of uncertainty in regional near-term climate change estimates for the next 10-20 years.
We have developed a new method to reduce the uncertainty from internal climate variability in estimates of near-term climate change. We use large ensembles of hundreds of climate projections and in each year sub-select those simulations for which the global patterns of ocean temperature anomalies are most similar to observed anomaly patterns. These ocean temperature anomaly patterns include signatures from different modes of climate variability, and the selection aligns the phasing of the simulated and observed climate variations. This method improves the accuracy of near- term climate projections in large parts of the globe, most notably the tropical Pacific, the tropical and sub-polar North Atlantic, the Indian ocean and land regions in Africa, Asia, southwestern Europe, and Australia among others. The skill improvements in the Pacific region are particularly remarkable, as this is an area where state-of-the-art climate models struggle to reproduce the observed climate variations, limiting the usefulness of climate projections there. The new constraint based on the phasing of climate variability therefore provides more accurate climate information in these regions and addresses some of the previous limitations in projecting and predicting future climate.