trations that would be observed in the absence of feedbacks is estimated on the basis of radiative transfer calculations to be about 1°C, and the water vapor feedback (calculated under the assumption of constant relative humidity) nearly doubles this response (e.g., Held and Soden 2000). Numerical experiments conducted with a variety of climate models that incorporate the full suite of climate feedbacks yield a range of climate sensitivities. The least sensitive models exhibit sensitivities roughly comparable to what would be obtained if only the water vapor feedback were included (about 2°C for a carbon dioxide doubling), whereas the most sensitive models estimate a sensitivity five times as large as radiative transfer calculations (Goosse et al. 2005, Webb et al. 2006, Winton 2006). The midrange models estimate a climate sensitivity of around 3°C for a doubling of carbon dioxide.
The sensitivity estimates derived from the models are checked by comparing observed and simulated responses to various known external forcings. For example, model simulations that consider surface temperature reconstructions for the past 700 years combined with instrumental data estimate climate sensitivity to be between 1.5 and 6.2°C (Hegerl et al. 2006).
The attribution of the large-scale warming of the late 20th century to human influences is supported in part by evidence that the warmth of the most recent one or two decades stands out above the background or natural variability of the last 2,000 years. To place this paleoclimatic evidence in context, it is necessary to consider the other evidence on which the attribution is based.
Based on evidence summarized in Chapter 2, it is known that global mean surface temperature has risen by about 0.6°C during the past century and that most of this warming took place during the period 1920–1940 and during the last 30 years. The troposphere is warming at a rate compatible with the warming of the Earth’s surface (CCSP and SGCR 2006). The spatial pattern of the observed temperature trends resembles the “fingerprint” of greenhouse warming in climate models, with cooling in the stratosphere and an uptake of heat by the oceans (e.g., Meehl et al. 2004, Hansen et al. 2005, Barnett et al. 2005). The warming is also reflected in a host of other indicators: For example, glaciers are retreating, permafrost is melting, snowcover is decreasing, Arctic sea ice is thinning, rivers and lakes are melting earlier and freezing later, bird migration and nesting dates are changing, flowers are blooming earlier, and the ranges of many insect and plant species are spreading to higher latitudes and higher elevations (e.g., ACIA 2004, Parmesan and Yohe 2003, Root et al. 2003, Berteaux et al. 2004, Bradshaw and Holzapfel 2006).
It is also well established that atmospheric concentrations of greenhouse gases have been increasing due to human activities. In recent decades the increases have been documented on the basis of direct measurements at a network of stations. Increases in concentrations of carbon dioxide, methane, and nitrous oxide starting in the 19th century, following many millennia of nearly constant concentrations, are clearly discernible in air bubbles trapped in ice cores recovered from the Greenland and Antarctic ice sheets (Petit et al. 1999, Siegenthaler et al. 2005a, Spahni et al. 2005). The attribution of these increases to human activities rests on both isotopic evidence and the fact that they are consistent with inventories of emissions of these gases from the burning of fossil fuels and other human activities, taking into account the storage in the oceans and the land biosphere. Based on station and ice core measurements, the combined