A new paper published in Nature Climate Change (Harrison et al. Implications of evaluation of CMIP5 palaeosimulations for climate projections. Nature Climate Change 5: 735-743. August 2015) suggests that we need to exercise considerable caution about future projections of regional climate changes.

The paper looks at features of the climate that are characteristic of 21^st century projections and then examines whether state-of-the-art climate models predict these features correctly in the geologic past. Lead author, Sandy Harrison says “This is necessary because future climate changes will be much larger than anything we have experienced in last hundred years or so when we have meteorological observations. But there is abundant geologic evidence for large climate changes in the past that can be used to see whether the models are working”.

The paper shows that climate models capture the large-scale patterns of temperature change, including the fact that warming over land is more than twice as much as over the ocean and that the biggest warming will occur in high latitudes. They also reproduced the observed global relationship between precipitation changes and temperature as temperature increases.

However, the paper shows that the models do not capture the scale of regional climate changes. The models predict an increase in monsoon rainfall both in the future projections and in the mid-Holocene, 6000 years ago, in response to enhanced land-sea temperature contrast. Abundant evidence shows that the Sahara desert was vegetated and supported abundant wildlife during the mid-Holocene, but models underestimate the observed change in precipitation in northern Africa by at least 50%.

Modelled changes can also be opposite to what actually occurred.  Models predict drying in the mid-continents in both the future and the mid-Holocene. The mid-Holocene predictions for central Eurasia are wrong – palaeoenvironmental data show that this region was in fact wetter and cooler than today – and this raises serious concerns about the future projections for the region.

“Our evaluations give us confidence that the general trajectory of modelled global warming is correct and that means that model estimates of what we need to do to limit global warming, say to less than two degrees, are likely to be realistic – which is very good news indeed. However, many government agencies want to use the projections for planning purposes at a local scale and here I think we have to exercise considerable caution about what the models say”.