Climate models at different spatial scales and levels of complexity provide the major source of information for constructing scenarios. GCMs and a hierarchy of simple models produce information at the global scale. These are discussed further below and assessed in detail in Chapters 8 and 9. At the regional scale there are several methods for obtaining sub-GCM grid scale information. These are detailed in Chapter 10 and summarised in Section 13.4.
The most common method of developing climate scenarios for quantitative impact assessments is to use results from GCM experiments. GCMs are the most advanced tools currently available for simulating the response of the global climate system to changing atmospheric composition.
All of the earliest GCM-based scenarios developed for impact assessment in the 1980s were based on equilibrium-response experiments (e.g., Emanuel et al., 1985; Rosenzweig, 1985; Gleick, 1986; Parry et al., 1988). However, most of these scenarios contained no explicit information about the time of realisation of changes, although time-dependency was introduced in some studies using pattern-scaling techniques (e.g., Santer et al., 1990; see Section 13.5).
The evolving (transient) pattern of climate response to gradual changes in atmospheric composition was introduced into climate scenarios using outputs from coupled AOGCMs from the early 1990s onwards. Recent AOGCM simulations (see Chapter 9, Table 9.1) begin by modelling historical forcing by greenhouse gases and aerosols from the late 19th or early 20th century onwards. Climate scenarios based on these simulations are being increasingly adopted in impact studies (e.g., Neilson et al., 1997; Downing et al., 2000) along with scenarios based on ensemble simulations (e.g., papers in Parry and Livermore, 1999) and scenarios accounting for multi-decadal natural climatic variability from long AOGCM control simulations (e.g., Hulme et al., 1999a).
There are several limitations that restrict the usefulness of AOGCM outputs for impact assessment: (i) the large resources required to undertake GCM simulations and store their outputs, which have restricted the range of experiments that can be conducted (e.g., the range of radiative forcings assumed); (ii) their coarse spatial resolution compared to the scale of many impact assessments (see Section 13.4); (iii) the difficulty of distinguishing an anthropogenic signal from the noise of natural internal model variability (see Section 13.5); and (iv) the difference in climate sensitivity between models.
Other reports in this collection