2.4.2 Artificial experiments

The most significant advance in artificial experiments since the TAR is the development of a new set of commitment runs by AOGCMs. These are climate change projections that assume that the radiative forcing at a particular point in time (often the current forcing) is held constant into the future (Meehl et al., 2007). The projections demonstrate the time-lags in the climate response to changes in radiative forcing (due to the delayed penetration of heat into the oceans), and of sea level to warming. Recent experiments estimate a global mean warming commitment associated with radiative forcing in 2000 of about 0.6°C by 2100 (Meehl et al., 2007). Sea-level rise due to thermal expansion of the oceans responds much more slowly, on a time-scale of millennia; committed sea-level rise is estimated at between 0.3 and 0.8 m above present levels by 2300, assuming concentrations stabilised at A1B levels in 2100 (Meehl et al., 2007). However, these commitment runs are unrealistic because the instantaneous stabilisation of radiative forcing is implausible, implying an unrealistic change in emission rates (see Nakićenović et al., 2007). They are therefore only suitable for setting a lower bound on impacts seen as inevitable (Parry et al., 1998).

2.4.3 Sensitivity analysis

Sensitivity analysis (see Box 2.1) is commonly applied in many model-based CCIAV studies to investigate the behaviour of a system, assuming arbitrary, often regularly spaced, adjustments in important driving variables. It has become a standard technique in assessing sensitivity to climatic variations, enabling the construction of impact response surfaces over multi-variate climate space (e.g., van Minnen et al., 2000; Miller et al., 2003). Response surfaces are increasingly constructed in combination with probabilistic representations of future climate to assess risk of impact (see Section 2.4.8). Sensitivity analysis sampling uncertainties in emissions, natural climate variability, climate change projections, and climate impacts has been used to evaluate the robustness of proposed adaptation measures for water resource management by Dessai (2005). Sensitivity analysis has also been used as a device for studying land-use change, by applying arbitrary adjustments to areas, such as +10% forest, -10% cropland, where these area changes are either spatially explicit (Shackley and Deanwood, 2003) or not (Ott and Uhlenbrook, 2004; van Beek and van Asch, 2004; Vaze et al., 2004).