Inhomogeneous forcing and transient climate sensitivity

Understanding climate sensitivity is critical to projecting climate change in response to a given forcing scenario. Recent analyses1–3 have suggested that transient climate sensitivity is at the low end of the present model range taking into account the reduced warming rates during the past 10–15 years during which forcing has increased markedly4. In contrast, comparisons of modelled feedback processes with observations indicate that the most realistic models have higher sensitivities5,6. Here I analyse results from recent climate modelling intercomparison projects to demonstrate that transient climate sensitivity to historical aerosols and ozone is substantially greater than the transient climate sensitivity toCO2.Thisenhancedsensitivity isprimarily caused by more of the forcing being located at Northern Hemisphere middle to high latitudes where it triggers more rapid land responses and stronger feedbacks. I find that accounting for this enhancement largely reconciles the twosets of results, and I conclude that the lowest end of the range of transient climate response toCO2 inpresentmodels andassessments7 (<1.3 ◦C) is very unlikely. Modelled transient climate responses were in good agreement with the understanding of historical forcing and observed warming trends in most analyses to ~2006 (ref. 8). Recent measurements posed a problem, however: warming rates were slower during the past 10–15 years while positive forcing continued to increase rapidly and new observations led to reduced estimates of offsetting negative aerosol forcing4,9. Although there are uncertainties in the recent forcing trends and at least part of the reduced warming rate could be due to internal variability, analyses in both the scientific literature1,2 and the popular press3 accounting for those factors concluded that climate sensitivity is likely towards the low end of present models’ range. Inferring climate sensitivity from recent observations requires a thorough understanding of both recent forcing and the global mean response to various forcing agents. Although the forcing has been studied in detail, the global mean response has conventionally been assumed to be the same for all forcing agents (in all such analyses, not only the most recent). I examine the response to historical anthropogenic inhomogeneous forcing in the most recent set of state of the art climate model simulations: the Coupled Model Intercomparison Project Phase 5 (CMIP5; ref. 10). Simulations to examine the influence of aerosols and ozone on climate were part of CMIP5 (ref. 10), but were relegated to a low priority and few results are available. Hence to examine the response to aerosols and ozone, I analyse CMIP5 historical simulations of the response to all forcings (histAll), to well-mixed greenhouse gas (WMGHG) forcing (histGHG) and to natural forcing (histNat), using the residual of histAll− (histGHG+ histNat) following two methods.Method 1 assumes that stratospheric water vapour forcing induces a response similar to WMGHGs, so that the residual (with scaled histGHG) represents the response to aerosol + ozone + land-use (LU; representing anthropogenic changes in vegetation cover and land usage). Method 2 assumes that the response to positive stratospheric water vapour forcing offsets the response to negative LU forcing, leaving a residual representing only aerosols+ ozone (Methods). I include the eight models for which forcing due to aerosols and ozone has been documented11 and all of these transient historical climate simulations are available. I evaluate the transient climate response (TCR), defined as the global mean temperature change in response to gradually increasing (1% yr−1) CO2 at the time of its doubling in a given model12 (all values annual averages). For consistency, the response to other forcings, which I refer to more generally as transient climate sensitivity, is given using the same scale (that is, the response per unit forcing times a model’s doubled CO2 forcing). Uncertainty in the TCR for a particular model stems from both the responses and the forcings, with the poorly documented LU forcing contributing the largest fraction in these calculations. All of the available CMIP5 models show greater TCR for historical inhomogeneous forcing than forWMGHG forcing (Fig. 1 and Supplementary Table 1). The TCR for WMGHG is 2.0± 0.3 ◦C (mean and s.d. across model ensembles), whereas it is 2.9 ± 1.0 ◦C for aerosol+ ozone+ LU (Method 1) and 3.0±1.1 ◦C for aerosol+ ozone (Method 2). Thus, the results are robust to the methodology for treating these minor forcing agents (LU and stratospheric water vapour), and seem to be dominated by the response to aerosol and ozone forcing. The TCR for aerosol + ozone is 45 ± 38% (mean of Methods 1 and 2; Supplementary Table 1) greater than the TCR calculated from historical WMGHG simulations (histGHG). In comparison with independent TCR estimates from these same models from the response to 1% per year CO2 increases13, which are 2.0± 0.3 ◦C, the aerosol+ ozone TCR is 53± 46% greater. Analysis of the large-scale forcing and temperature response patterns provides both further confidence that the enhanced TCR is due to aerosol + ozone forcing and insight into the large spread in model results. The models have greater net negative inhomogeneous forcing in the Northern Hemisphere (NH) than in the Southern Hemisphere (SH) by −0.97 ± 0.57Wm−2, and in the NH extratropics (NHext; 28◦–90◦) than in the SHext by −1.07 ± 0.66Wm−2 (Supplementary Table 3). The extratropical asymmetry results from a very strong asymmetry in aerosol forcing (−1.32 ± 0.57Wm−2) that is partially offset by more positive ozone forcing in the NHext than the SHext (0.40 ± 0.11Wm−2; Supplementary Table 2). It has been shown previously that forcing in the NH extratropics causes a greater global mean temperature response than forcing in the tropics14–18. In particular, NHext forcing caused ∼50% greater