Online climate change projections report 3.2.2 Process uncertainties

The first task is to define the set of Earth system processes likely to contribute significant uncertainty in 21st century climate (see Box 2.1). These would clearly include surface and atmospheric physical processes (for example water vapour, cloud, surface albedo and soil moisture feedbacks continue to be recognised as key determinants of global and/or regional climate change (Bony et al. 2006; Soden and Held, 2006)). However, other components are also likely to be important. Changes in ocean heat transport have potential to influence both global and regional changes (Raper et al. 2002; Boer and Yu, 2003), while imperfect knowledge of the radiative forcing due to sulphate aerosols (Anderson et al. 2003) is recognised as a significant source of uncertainty, both in determining recent observed climate change and in predicting future changes (Andreae et al. 2005). Uncertainties in the fraction of man-made carbon dioxide emissions likely to remain in the atmosphere (due in particular to terrestrial feedbacks) have also emerged as an important source of divergence in future projections by different models, particularly in changes expected during the second half of the 21st century (Cox et al. 2000; Friedlingstein et al. 2006). We therefore designed our ensemble experiments to sample uncertainties in the atmosphere, ocean, sulphur cycle and terrestrial carbon cycle modules available in the family of HadCM3 components. This covers the major known sources of uncertainty in climate change out to a century or so ahead. Inevitably, however, limitations of computational resource, modelling capability and current understanding imply that some additional drivers of climate change have to be omitted, or included without sampling of the associated uncertainty. For example, our carbon cycle simulations account for feedbacks associated with ocean as well as terrestrial carbon uptake, however uncertainties in processes affecting oceanic uptake are not sampled (see Section 3.2.5). Our simulations do not include forcing from carbonaceous aerosols (e.g. Jones et al. 2005), non-aerosol atmospheric chemistry (e.g. Johnson et al. 2001) or methane cycle feedbacks (Christensen et al. 2004; Archer and Buffett, 2005). The sampling of sulphur cycle feedbacks omits the second indirect effect arising from the effects of reduced cloud droplet size on precipitation efficiency, and hence cloud persistence, as this process is not included in HadCM3, or indeed in most current climate models (see Table 10.1 of Meehl et al. 2007)

Designing ensemble climate projections given finite computing resources

The standard approach to modelling time-dependent climate changes involves simulations which run from pre-industrial conditions up to the end of the period of interest (say from 1860–2100), specifying observed time-dependent changes in external forcing agents (typically man-made changes in greenhouse gases and aerosol precursors, and natural variations arising from solar variability and volcanic eruptions) up to present day, switching to some future scenario of man-made forcings to 2100. The ideal method of sampling modelling uncertainties would be to run a very large ensemble of such transient climate change simulations, in which all the relevant Earth system modules (atmosphere, ocean, sulphur and carbon cycle) are coupled together dynamically, and in which different ensemble members sample multiple perturbations to uncertain parameters in all modules simultaneously, in such a way as to ensure comprehensive coverage of the entire parameter space of each module. Such an experiment would ensure that non-linear interactions between all uncertain processes in all modules were thoroughly sampled. Unfortunately, such an experiment is well beyond the available computing resources, so compromises have to be made based on expert judgement of the relative importance of different sources of uncertainty. Figure 3.1 gives a schematic summary of the major components of our strategy for sampling modelling uncertainties, through the combination of a number of ensemble climate projection experiments. These experiments use several model configurations derived from HadCM3 to sample uncertainties in climate change during the 21st century, and are described below in Sections 3.2.3–3.2.6, and 3.2.11.

 

 

        Figure 3.1: Elements of our methodology to sample modelling uncertainties using perturbed physics ensembles (PPEs) based on configurations of the HadCM3 climate model. Blue boxes denote ensemble simulations using various model configurations derived from HadCM3. Yellow boxes denote statistical tools required to generate alternative estimates of climate change which combine the sources of uncertainty sampled in the various ensemble experiments. Boxes A and B are described in Section 3.2.3. Boxes C, D and E are explained in sections 3.2.4, 3.2.5 and 3.2.11 respectively. Boxes F and G represent our timescaling procedure for deriving very large ensembles of realisations of time-dependent climate change from smaller ensembles of climate model simulations, covered in sections 3.2.4 and 3.2.6. Box H denotes our downscaling procedure (see Section 3.2.11) for the generation of probabilistic projections at the 25 km resolution required for UKCP09, derived from information at larger scales obtained from global climate model simulations.          

Previous  Next

 

   

• Last updated: Sunday, 11 March 2012