Origins and estimates of uncertainty in predictions of twenty-first century temperature rise

Abstract

Predictions of temperature rise over the twenty-first century are necessarily uncertain, both because the sensitivity of the climate system to changing atmospheric greenhouse-gas concentrations, as well as the rate of ocean heat uptake, is poorly quantified and because future influences on climate-of anthropogenic as well as natural origin-are difficult to predict. Past observations have been used to help constrain the range of uncertainties in future warming rates, but under the assumption of a particular scenario of future emissions. Here we investigate the relative importance of the uncertainty in climate response to a particular emissions scenario versus the uncertainty caused by the differences between future emissions scenarios for our estimates of future change. We present probabilistic forecasts of global-mean temperatures for four representative scenarios for future emissions, obtained with a comprehensive climate model. We find that, in the absence of policies to mitigate climate change, global-mean temperature rise is insensitive to the differences in the emissions scenarios over the next four decades. We also show that in the future, as the signal of climate change emerges further, the predictions will become better constrained.

Ensemble climate predictions using climate models and observational constraints

“…Warming as high as 6.9 K (relative to 1990s temperatures) cannot be ruled out at the 95% confidence level in the SRES A1FI scenario (figure 6). Stott & Kettleborough (2002) also showed that as the signal of climate strengthens, the uncertainty in future temperature rise is likely to reduce, potentially halving uncertainties in late twenty-first century warming by 2020 when compared with values estimated in 2000 ( figure 6). The range of warming under the B2 scenario is 1.6-3.7 K by the 2090s decade relative to the 1990s decade.…”

“…For example, introducing a new model into the analysis that has a lower sensitivity than other models has only a second-order effect on the result (Stott et al 2006b). At the same time, uncertainty is reduced in a predictable way as the signal strengthens (Stott & Kettleborough 2002). For a STAID forecast, there should be a transfer function that links a forecast quantity to an observable quantity.…”

“…A typical approach described here, and many of the ASK studies (Stott & Kettleborough 2002; Stott et al 2006a,b;Kettleborough et al 2007) is to partition the observed changes into contributions from the response to well-mixed GHGs, the combined effects of all anthropogenic forcings (ANTHRO) and the combined effects of all natural forcings (NAT). Anthropogenic forcings typically included in climate models are the effects of sulphate aerosols (both their direct effect through increased reflection of incoming solar radiation, and in some model simulations through their indirect effects by making clouds brighter or longer lasting) and the effects of tropospheric and stratospheric ozone changes.…”

“…The assumption that the relative roles of GHGs and aerosols should remain constant, as predicted by models was relaxed by Stott & Kettleborough (2002) who calculated separate scaling factors on the response patterns in HadCM3 to GHG and aerosols as well as including the response to natural forcings in the analysis. Probabilistic forecasts of global mean temperatures were obtained for four representative SRES emission scenarios.…”

The climate prediction .net BBC climate change experiment: design of the coupled model ensemble

“…Forest et al 2000) or by generating 'pseudo-ensembles' by rescaling the output of a single atmosphere-ocean general circulation model (AOGCM; Allen et al 2000;Stott & Kettleborough 2002). This is limited, however, to predicting uncertainty ranges on just a few global climate variables.…”