ABSTRACT
The climate change projections presented in Chapter 2.1 address an important issue. Climate modellers, and those interested in climate impacts and adaptation, are aware that despite rapid ongoing increases in computer power and improvements in techniques, computational resources are considerably less than the community would require in order to address climate change. This situation is unlikely to change in the near future. Important decisions are therefore faced when assessing climate change and its impacts. Critically, is the ensemble size adequate for the development of probabilities or will resources, instead, be focused on ensuring adequate spatial resolution to simulate key processes (e.g. orographic flow) and to provide geographically specific information for impacts modellers? Challinor et al (2009a) discuss this issue, and other related topics mentioned below. They also illustrate the accuracy and regional detail in the simulation of precipitation that can be achieved by high-resolution simulations over Africa. Chapter 2.1 takes this thinking a step further by using a parallel suite of simulations at 20km, 60km, 120km and 180km. It also assesses predictability across lead times by comparing 2015–2039 to end-of-century simulations. Interestingly, it has been suggested (e.g. Cox and Stephenson, 2007) that total uncertainty in climate prediction may be at a minimum at 30 to 50 years’ lead time, after which uncertainty in initial conditions has fallen significantly, while uncertainty in greenhouse gas emissions is not yet prohibitively large. Thus, comparisons across lead times are important, and the use of hatching to denote agreement across ensemble members – as in many of the (lower-resolution) results presented in Chapter 2.1 – is a useful visualization for assessing predictability.
