Most climate scenarios for impact assessments were developed by using outputs from AOGCMs (Hulme and Brown, 1998). Often these results are scaled toward the desired emission levels with simple climate models (Hulme et al., 1995; Harvey et al., 1997; see Section 3.8.3). In this simple approach, most interactions are neglected.
The only models that can be used to develop more consistent scenarios that incorporate most of the important interactions are IAMs (see Section 188.8.131.52). IAMs have been developed with different levels of complexity, from extremely simple to highly complex (Harvey et al., 1997). Different interactions are included, although no single model provides a fully comprehensive treatment. The models are most commonly used for emission scenario development and mitigation policy assessment (Schimel et al., 1997a; Alcamo et al., 1998a; Pepper et al., 1998). All simulate a causal chain (e.g., human activities, emissions, climate change, sea-level rise, and other impacts). Emissions, climate change, impact, and mitigation scenarios derived from these models have been published (Schimel et al., 1997a; Leemans et al., 1998; Pepper et al., 1998) and collected in several databases (Alcamo et al., 1995; Nakicenovic et al., 2000). Unfortunately, it is not always clear which interactions are explicitly included in individual IAMs. This reduces the comparability of individual IAM-derived scenarios and thus their utility.
Depending on assumed interactions during scenario development, a wide range
of estimates of climate change and its impacts is possible (see Table
3-7). However, within this range certain responses are more likely than
others. To define appropriate and realistic levels of interactions, expert judgment
and sensitivity experiments with models could be very valuable (van der Sluijs,
1997). Innovative, objective, and systematic approaches have to be developed
to evaluate underlying scenario assumptions and to validate the scenario results.
This is still an immature area of scenario development.
One difficulty faced by authors in attempting to summarize and synthesize the
results of impact studies for previous IPCC assessments (i.e., IPCC, 1996b,
1998) has been a lack of consistency in projections. Different climate projections
have been adopted in different studies, in different regions (or within the
same region), and in different sectors. Moreover, even where the same climate
projections are assumed, they might not be applied in the same way in different
impact studies. Finally, some studies also are inconsistent in their methods
of projecting changes in climate alongside concurrent changes in related socioeconomic
and environmental conditions.
For example, GHG concentrations often are transformed into CO2-equivalent concentrations to determine radiative forcing levels and climate change. The GCM community often presents climate change simulations as "doubled CO2" anomalies. Depending on the scenario, however, 5-40% of the forcing is caused by non-CO2 GHGs (30% in 1990). The doubled-CO2 scenarios often are interpreted as CO2 only (e.g., Cramer et al., 1997); others add an explicit distinction between CO2 and non-CO2 gases (e.g., Downing et al., 1999). In determining the impacts of direct CO2 effects and climate change, this can easily lead to inconsistencies. Similar discrepancies exist for other types of interactions.
Finally, it is a significant challenge to integrate climate or sea-level rise scenarios, with a time horizon of decades to hundreds of years, with nonclimatic scenarios of social, economic, and technological systems that can change rapidly over a time scale of years. For instance, it is difficult to devise credible socioeconomic scenarios that extend beyond the lifetime of current infrastructure and institutions. Moreover, social/economic actors who need to be involved in the scenario development process (e.g., business, governments) often find long time horizons difficult to contemplate. Box 3-2 illustrates a recent example of an attempt to harmonize climate change, sea level, atmospheric composition, and socioeconomic scenarios in a multi-sectoral global impact assessment.
Figure 3-2: Scaled outputs of mean December-February (left) and June-August (right) temperature and precipitation change by the 2050s relative to 1961-1990 over land grid boxes representing Central North America (top), Southern Africa (middle), and Southern Asia (bottom) from eight simulations with five AOGCMs (experiments b, c, e, h, and a four-member ensemble from t; see Table 3-5). Simulations assume forcing by greenhouse gases but not aerosols, and are standardized according to the climate sensitivity of each AOGCM. Lines connect four points for each simulation, all in the same order from the origin: B1-low, B2-mid, A1-mid, A2-high. Each point represents the standardized regional changes in climate from the AOGCM, linearly scaled according to the global warming estimated with a simple climate model for one of four preliminary SRES marker emissions scenarios (B1, B2, A1, and A2) and a value of the climate sensitivity (low = 1.5°C; mid = 2.5°C, and high = 4.5°C). Also plotted are ±1 and ±2 standard deviation ellipses from the 1400-year HadCM2 and 1000-year GFDL unforced simulations, which are used to indicate natural multi-decadal variability and are orientated according to the correlation between modeled 30-year mean temperature and precipitation. Results from two other AOGCMs did not extend to the 2050s (Carter et al., 2000).
Other reports in this collection