March 1 2006
Japan Agency for Marine-Earth Science and Technology

Improved estimation of the degree of temperature rise due to global warming
- An improved estimate of climate sensitivity -


James Annan and Julia Hargreaves, and colleagues in the Global Warming Research Program of Frontier Research Center for Global Change (FRCGC), Japan Agency for Marine-Earth Science and Technology (JAMSTEC) have made an improved estimate of climate sensitivity (Note 1), showing for the first time that this value is less than 4.5C at the 95% confidence level (Note 2) (fig 1).

Two recent papers have contributed to this work. The first, using the MIROC3.2 GCM and the Earth Simulator, demonstrated the use of past climate data in estimating climate sensitivity, and limited climate sensitivity to less than 7C. Other recent research by other researchers had proposed that climate sensitivity could be as high as 10C.
In the second paper, the researchers used climate data from a number of historical periods and showed that, by correctly combining information from multiple sources using Bayesian methods (Note 3), climate sensitivity can be even more tightly constrained to be very likely (95% confidence) less than 4.5C, with a lower limit of 1.8C. A likely range of 2.2-3.9C was also established at the 70% confidence level. In the future, increasingly confident and useful climate forecasts are expected.
The first paper was published in Scientific Online Letters on the Atmosphere in December 2005 and the second paper is shortly to be published in Geophysical Research Letters.


In 1979 it was first stated that climate sensitivity was likely to lie in the range 1.5-4.5C. This range was a very subjective estimate based on the limited climate modelling performed at that time. Subsequent IPCC Assessment Reports have supported this estimate without being able to narrow its range, and the Third Assessment Report (TAR) quantified "likely" to mean a confidence in those bounds only at the 66-90% level. Furthermore, recent research based on observational data had indicated that much higher values for climate sensitivities could not be reasonably ruled out, suggesting that this subjective range might be too optimistic. Values for climate sensitivity of up to 10C have been proposed, which would have disastrous consequences for the future of the planet.
This failure to decrease the uncertainty on the estimate of climate sensitivity, as well as the possibility of the true value being very high, has caused considerable alarm in the global warming debate at both the scientific and political level.


In this new research, it has been shown that by combining observational evidence from a wide range of sources, a greatly improved observationally-based estimate can be formed. For the first time, an upper bound of 4.5C (at the 95% confidence level) has been established from an analysis of observational data.
The greatest achievement of the experiment is to rule out the extremely high estimates of climate sensitivity that have previously been proposed. The work also describes the methods that can be used to combine information so that the full benefit can be gained from available data and models.
The Earth Simulator enabled ensembles of models to be run, greatly opening up the possibility for exploration of the problem using state of the art climate models. The value of climate sensitivity has now been demonstrated to lie within a reasonably narrow range, and this range is highly consistent with the results achieved from the state of the art climate models run at climate research centres throughout the world. This results should increase confidence in the model results and enable researchers to move forward in forecasting the impacts of climate change in greater detail.

Future plans:

The result increases the confidence that we can have in climate models and it is now hoped that researchers can use this information to start making accurate prediction of climate change on regional scales and over the time scales of interest to policy makers. It is also likely that the methods described will lead to a more precise estimate of climate sensitivity in the near future.


J. D. Annan, J. C. Hargreaves, R. Ohgaito, A. Abe-Ouchi, S. Emori. Efficiently constraining climate sensitivity with paleoclimate simulations. SOLA Vol 1 pages 181-184.
J. D. Annan and J. C. Hargreaves. Using multiple observationally-based constraints to estimate climate sensitivity. In press, GRL
Notes 1: Climate sensitivity is defined as the globally-averaged surface temperature change that would result in the long-term from a doubling of the atmospheric CO2 concentration. It describes the most fundamental response of the climate system to anthropogenic forcing, and so determining its value is of great importance in predicting future climate change.

Note 2: Confidence level is the probability value associated with a confidence interval.

Note 3: Bayes' Theorem is a fundamental result in probability theory which explains how multiple sources of information can be combined to generate an improved estimate. Observations of recent global warming, short-term cooling after major volcanic eruptions, cooling at the Last Glacial Maximum and other periods in the historical record, and the seasonal variation in climate, all provide some information which helps to determine the value of climate sensitivity.


Japan Agency for Marine-Earth Science and Technology
Mr. Katsuhiko Masuda, Director
Frontier Research Promotion Department,
Frontier Research Center for Global Change
Tel:045-778-5670 Fax:045-778-5497

Mr. Kenichi Takahashi, Manager
Public Relations Division, Administration Department
Tel:046-867-9050 Fax: 046-867-9055