Scientists are concerned about discrepancies between model temperature predictions and actual measurements of the upper troposphere because they may affect the models' sensitivity to increased carbon dioxide. Although the discrepancies do not necessarily suggest that estimates of surface global warming will be inaccurate, they may indicate inaccuracies in the predictions for atmospheric circulation.

The essential discrepancy is that models tend to overestimate increases in the "static stability" between the upper and lower-middle troposphere. Greater static stability means that the upper troposphere warms swiftly relative to the lower-middle troposphere; the situation would be stable because warm air prefers to be on top. In measurements taken using satellites, however, the upper-troposphere warming is much less than models project. Scientists are not sure whether the discrepancy is due to measurement or model biases.

Stephen Po-Chedley and Qiang Fu at the University of Washington in Seattle considered whether sea-surface temperatures imposed as a boundary condition could solve the problem. Sea-surface temperatures are well coupled with atmospheric temperatures: their variations in time tend to match. By constraining atmospheric models with sea-surface temperatures, therefore, Po-Chedley and Fu could remove the possibility that the temperature discrepancies were due to simple natural variability.

The researchers used atmospheric global-climate models in the latest Community Model Intercomparison Project (CMIP5) – a collaboration to which climate-modelling groups from around the world contribute. However, they found that these models still significantly overpredicted the upper-troposphere temperature between the late 1970s and the mid 2000s, even when the sea-surface temperature boundary conditions were imposed.

Po-Chedley and Fu now say that the discrepancies represent "an important problem". What is more, they still do not know whether it is the models that are to blame, or biases in the satellite data.

"In future research, we would like to try to understand what is leading to the model-observational discrepancy in upper-tropospheric warming," said Po-Chedley. He and Fu plan to impose different sea-surface warming patterns on the models to check they fully understand the sea-surface temperature coupling with the atmosphere. They will also investigate how atmospheric convection can affect the upper- to lower-tropospheric warming ratio.

Finally, added Po-Chedley, they would like to "take a closer look at the satellite data" to see whether it could be observational uncertainty or bias that is causing the discrepancy with the climate models.

The study is published in Environmental Research Letters (ERL).