Online Briefing report 5.14 Limitations
The procedure used in UKCP09 to convert the ensembles of climate model simulations into probabilistic estimates of future climate necessitates a number of expert choices and assumptions, with the result that the probabilities we specify are themselves uncertain. We do know that our probabilistic estimates are robust to reasonable variations within these assumptions, and this is covered in some detail in Annex 2 of the UKCP09 report Climate change projections. Although it is important that prospective users understand the limitations and caveats, it is also worth emphasising that (a) current models are capable of simulating many aspects of global and regional climate with considerable skill; and (b) they do capture, albeit imperfectly, all the major physical and biogeochemical processes known to be likely to exert a significant influence on global and regional climate over the next 100 yr or so.
As our understanding of the climate system and our ability to represent it in models gets better, as statistical methods to convert model results into probabilistic projections are developed further, and as computing power increases, it is likely that uncertainties will become smaller, although natural variability will always provide an irreducible level in the long term. The consequence of these expected improvements is that the shape of a given is likely to change in the future. Users need to understand clearly that, if they choose to adapt to a climate change corresponding to a specific probability level, this is likely to change in future projections — and the changes are likely to be greater at the extremes of probability levels (that is, 10 and 90%). If our understanding of climate processes, and model representations of them, does not change substantially in future, then we foresee a general reduction in uncertainties (except that due to long-term natural variability) because of improvements in our ability to represent processes currently modelled, and we would hence expect the shape of the PDF to change, with a reduction in its width. However, we do not know in what way this reduction in width will occur; in particular it may not be in a symmetrical manner. Although we cannot say what the next generation of PDFs will look like, it is likely that the spread of plausible changes they would indicate would be encompassed by the corresponding PDFs shown in UKCP09. Thus, in the absence of any major change in model projections, users who are incorporating the probabilities given in UKCP09 into their decision-making are likely to find that their decisions are robust to changes in the next generation of projections.
On the other hand, there is also the potential for uncertainties to become greater if processes not yet included, or included imperfectly, in the models turn out to exert a substantial influence on climate change. Less than a decade ago, for example, carbon cycle feedbacks were not included in models, yet these are now known not only to change the projections substantially but also to add significantly to the uncertainty in them — which is why they are included in UKCP09. Further such effects, for example, methane feedbacks from land and oceans or the dynamics of ice sheets, may be shown to be important in due course. Uncertainties could also widen if future (improved) models reveal that a process which is represented in the current generation of models, but with a common bias, turns out to exhibit a larger response to man-made forcing than current models suggest (see Box 4). However, the consistency between model simulations and observations of change over the last century provides some reassurance that any unknown processes are unlikely to change projections fundamentally, at least for the next few decades.
An obvious follow-up question is: should decisions be made now, based on UKCP09 projections, or should they be delayed in the hope that better projections will be available in a few years time? How rapidly will climate projections change in future? Although modellers have improved many aspects of their models over the past decade or so, the current range of changes over the UK (Figure 4) is not significantly narrower than that shown in UKCIP02. In practice, the prospects for better projections will depend on which aspects of future climate users are most interested in. The width of the PDFs in UKCP09 are substantial even for the next few decades, due mainly to natural variability, and grow larger through the century due to uncertainties in climate feedbacks. It may be possible to reduce short-term uncertainties with higher resolution models which may simulate better (for example) the North Atlantic storm track, and by starting model experiments with the recently-observed state of the ocean. However, this may not improve projections of (say) changes in surface temperature a hundred years ahead; at these lead times improved projections would come from more faithful representations of climate feedbacks and the in models. Dialogue between decision makers and climate scientists, on the potential for emerging research to update projections, will be essential. We reiterate the key point made earlier that the UKCP09 methodology is designed to capture known uncertainties in the climate system built into the current generation of climate models, and is the most comprehensive approach to do so to date. The UKCP09 projections can make a useful contribution to assessing risks posed by future climate; they are appropriate for informing decisions on adaptation to long-term climate change which need to be taken on the basis of current knowledge, and the uncertainty quantified in them is likely to be a conservative estimate.
In the UKCP09 report Climate change projections and the online UKCP09 User Guidance, we discuss in depth the limitations of the projections and their sensitivity to some of the assumptions. Decision makers use projections of change in many factors; not just climate itself but also demography, economics, technologies, etc.; uncertainties in these are generally poorly quantified. The great strength of the UKCP09 projections is that they give decision makers the first quantification of the major known uncertainties in climate change over the UK; that these are large reflects the current situation in climate modelling.