18th Conference on Climate Variability and Change

5.3

Climate change detection and attribution in the upper air

Stephen S. Leroy, Harvard Univ., Cambridge, MA; and J. G. Anderson and J. A. Dykema

We perform a diagnostic study of forced runs of the IPCC AR4 coupled climate models with the goal of evaluating climate models according to their predictive capability. Previous studies have shown that climate signal detection and attribution problem, when put into a Bayesian context, is a special case of testing climate models using observations of trends in the climate system. By comparing the output of the SRES A1B forced runs of the IPCC AR4 suite of climate models, we determine what is common among predictions by different climate models, what is different, and how long it will take before climate models can be tested using specific data types.

We find that global change is better described by thermal expansion of the troposphere than by trends in surface air temperature, that the most predictable component of climate change is poleward drift of the jet streams, and that this latter hypothesis should be testable with 99% confidence in 10 to 20 years. Changes in surface air temperature and polar annular modes are less predictable than jet stream position. Radio occultation by GPS, in the era of the Constellation Observing System for Meteorology, Ionosphere and Climate (COSMIC), should be able to detect poleward motion of the jet streams in 10 to 20 years.

Poleward motion of the jet streams should be discernible even in data sets which are not absolutely calibrated. For example, changes in jet stream position appear as gradients in MSU middle-tropospheric brightness temperature rather than trends in a global average mean. We apply optimal signal detection methods with appropriately formulated observation error covariances to detect anthropogenic change in MSU data even when it is described as inaccurate. While this cannot tell us the sensitivity of the climate system (because climate forcing over the last 25 years is not well described), it still offers the most robust method thus far in attributing global change---rather than global warming---to human influence.

extended abstract  Extended Abstract (968K)

wrf recording  Recorded presentation

Session 5, Climate Modeling: Studies of climate change
Wednesday, 1 February 2006, 8:30 AM-5:00 PM, A313

Previous paper  Next paper

Browse or search entire meeting

AMS Home Page