of this period. Polyakov et al. estimate that as much as 50% of the trend over this period in the North Atlantic is internal variability. The modeling work by Zhang and Knight also indicate that the influence of the Atlantic, despite its small size, can spread preferentially over Eurasia and contribute to global temperature signals. For example, the model analyzed by Knight generates a 0.1 K global mean warming for an increase of 1 Sverdrup (about 5% of the 20 Sverdrup mean value) in the Atlantic overturning.

Volcanic responses and the response to the 11-year solar cycle can also be used to constrain TCR, as can the autocorrelations of internal fluctuations (rather than watching the volcanic response decay in time to estimate the strength of restoring forces, one can watch internal fluctuations decay). Paeloclimatic evidence constrains equilibrium sensitivity, and with modeling guidance and heat uptake measurements constraining the ratio of TCR to the equilibrium response, one can also use these to constrain TCR. But there are also issues related to the decoupling of transient and equilibrium responses and to issues of “Earth system sensitivity”, that come into play when considering paleoclimatic constraints (see Chapter 6). We judge the constraints that directly involve fits to the temperature record over the past century and the last few decades to be the most useful in constraining TCR at this time.

The magnitude of many of the impacts discussed in this report scale with the value of TCR. We estimate TCR by starting with the distribution in the CMIP3 ensemble, but lowering the low end of the distribution slightly to take into account the possibility suggested by some recent studies of internal variability that a portion of the most recent Northern Hemisphere warming is internal. This results in a best estimate of 1.65ºC, likely lying in the range 1.3-2.2ºC (i.e., with 2/3 probability) and very likely lying in the range of 1.1-2.5ºC (i.e., with 90% probability). The estimate in the WG1/AR4 report is a very likely range of 1-3ºC. Our reduction of the upper limit to this range is consistent with our critique of very high equilibrium sensitivities in Section 3.2.

The temperature response to anthropogenic carbon emissions is determined by: (1) the response of the carbon cycle to emissions; (2) the climate response to elevated CO₂ concentrations; and (3) the feedback between climate change and the carbon cycle. There has been significant attention