Regional climate models perform poorly against observed data

Have you ever looked out the window and compared what you see with the forecasted weather? Scientists have recently started to do something similar with predicted – and observed – climate change.

We have seen enough of a change in our climate to start comparing models with what’s actually happened so far. In preparation for the Intergovernmental Panel on Climate Change (IPCC) fifth assessment report, scientists have looked back at observed regional temperature and precipitation trends from 1950 to 2011, and compared them with simulated climate models. On the whole, the global predicted trends look fairly similar to the weather over those 60 years. But a closer look at the regional level reveals that the match is not as good. For example, climate models underestimated the warming in Asia, the West Pacific, the East Indian Ocean, and overestimated the warming in the Pacific Ocean and off the coast of the United States.

Even accounting for natural variation in weather and a variety of models (with different approaches, etc.), more than expected of the modelled trends for that period fall outside of changes actually observed. Whatever the cause of this inaccuracy, the implication is that short-term climate projections should not be relied upon as a probabilistic forecast – that is, we shouldn’t use them as we would a weather forecast, which is exactly how short-term climate projections are often used. Future projections can be improved; in the meantime, using models as a forecast should be done with these discrepancies in mind.