3.5.3.1. Baseline Period

Any climate scenario must adopt a reference baseline period from which to calculate changes in climate. This baseline data set serves to characterize the sensitivity of the exposure unit to present-day climate and usually serves as the base on which data sets that represent climate change are constructed. Among the possible criteria for selecting the baseline period (IPCC, 1994), it should be representative of the present-day or recent average climate in the study region and of a sufficient duration to encompass a range of climatic variations, including several significant weather anomalies (e.g., severe droughts or cool seasons).

A popular climatological baseline period is a 30-year "normal" period, as defined by the WMO. The current WMO normal period is 1961-1990, which provides a standard reference for many impact studies. Note, however, that in some regions, observations during this time period may exhibit anthropogenic climate changes relative to earlier periods.

3.5.3.2. Sources and Characteristics of Data

Sources of baseline data include a wide variety of observed data, reanalysis data (a combination of observed and model-simulated data), control runs of GCM simulations, and time series generated by stochastic weather generators. Different impact assessments require different types and resolutions of baseline climatological data. These can range from globally gridded baseline data sets at a monthly time scale to single-site data at a daily or hourly time scale. The variables most often required are temperature and precipitation, but incident solar radiation, relative humidity, windspeed, and even more exotic variables sometimes may be needed.

Two important issues in the development of baseline data sets are their spatial and temporal resolution and uncertainties related to their accuracy (New, 1999) (see TAR WGI Section 13.3.2 for further details). Evaluation of the differences between baseline data sets recently has become an important step in scenario development because these differences can have an important bearing on the results obtained in an impact assessment (Arnell, 1995; Pan et al., 1996).

Techniques for constructing climate scenarios (i.e., scenario information that is directly usable in impact studies) have evolved very slowly during the past 2 decades. However, in the past few years several new developments in climate modeling and scenario development have expanded the array of techniques for scenario formation. The following subsections discuss some of these issues and present some background illustrative material.

3.5.4.1. Choosing Variables of Interest

In principle, GCM-based scenarios can be constructed for a wide range of variables at time resolutions down to subdaily time steps. In practice, however, not all data are available at the desired temporal and spatial resolutions. Most scenarios are conventionally based on changes in monthly mean climate, although with greater quantities of model output now being saved operationally, daily output and information on certain types of extreme events (e.g., mid-latitude cyclone intensities) can be accessed readily. However, consideration must be given to whether model output regarding a particular phenomenon is deemed "meaningful." For example, although information on changes in the frequency and intensity of El Niño-Southern Oscillation (ENSO) events may be desirable from an impacts point of view, analyses of possible future changes in this oscillation still are very preliminary (see TAR WGI Chapter 9).

3.5.4.2. Selecting GCM Outputs

Many equilibrium and transient climate change experiments have been performed with GCMs (Kattenberg et al., 1996; TAR WGI Chapter 9). Several research centers now serve as repositories of GCM information (see, e.g., Hulme et al., 1995; CSIRO, 1997). The IPCC Data Distribution Centre (IPCC-DDC, 1999) complements these existing sources. Table 3-5 lists GCM experiments that have been used to develop scenarios for impacts studies evaluated in this report.

Four criteria for selecting GCM outputs from such a large sample of experiments are suggested by Smith and Hulme (1998):

1. Vintage: Recent model simulations are likely (though by no means certain) to be more reliable than those of an earlier vintage since they are based on recent knowledge and incorporate more processes and feedbacks.
2. Resolution: In general, increased spatial resolution of models has led to better representation of climate.
3. Validation: Selection of GCMs that simulate the present-day climate most faithfully is preferred, on the premise that these GCMs are more likely (though not guaranteed) to yield a reliable representation of future climate.
4. Representativeness of results: Alternative GCMs can display large differences in estimates of regional climate change, especially for variables such as precipitation. One option is to choose models that show a range of changes in a key variable in the study region.
   

IPCC Homepage