Main steps toward scenarios of climate variability and extremes
In order to develop climate change information from all tools suggested previously (GCMs or downscaling methods), two main approaches are generally used for impacts studies:
The most popular and simple way to develop climate change information is to use the delta method and apply the GCM signal or anomalies calculated in the future to the observed climate. In this case, the scenario of future climate is simply obtained by adjusting the baseline observations by the .change fields. or the difference (or ratio) between period-averaged results for the GCM or RCM experiments or other downscaling simulations (usually 30 year periods) and the corresponding averages for the GCM or RCM simulated baseline period (e.g. 1961-1990).
Differences are usually applied for temperature changes (e.g. 2040-2069 minus 1961-1990) while ratios are commonly used for precipitation change (e.g. 2040-2069 divided by 1961-1990), though differences may be preferred in some cases. One of the limitations of this method is that it assumes no changes in variability are expected in the future, as this variability is inherited from the current climate regime. This also assumes that any biases in the simulation of present-day climate are the same as in the simulation of future climate. This inconsistency may pose severe constraints on the usefulness of GCM information, especially in regions characterised by complex physiographic settings or where regional or local forcings play a key role in the climate regime.
If the biases of the models are not stable (i.e. not stationary), the error does not cancel out the difference between the future and current simulated periods. In that case, the signal should incorporate a various part of the biases, sometimes higher than the signal itself, especially where the heterogeneous land/ocean surface conditions are not well simulated by the models as the feedbacks are strong, and not homogeneous and linear with time. In that respect, suitable regionalisation or downscaling techniques may be able to overcome some of these biases, but not systematically, as added values or new insight that have been gained through the use of higher resolution scenarios must be first demonstrated. However, if this is the case, downscaling methods may be used to improve GCM results over land/sea discontinuities, the seasonal sea-ice margin, and in regions of complex topography and non-homogeneous land surface types;
Beyond analogue or synthetic scenarios (see box 3.1 in Barrow et al., 2004), the use of full temporal series of the simulated variables from downscaling methods may be necessary, especially if the impacts model requires daily or hourly climate information which are not well simulated by coarse scale GCMs. In this case, some potential changes in variability and extremes are better taken into account in the impacts model. However, some rigorous assessment of the capacity of the downscaling model to simulate the high frequency variability of the climate regime at the scale of interest is needed, especially for the highest quartiles (extreme values) which have the greatest impact.
This might involve the comparison of event spectra, for example. Statistical and dynamical downscaling are potential sources of realistic scenarios at the scales required by this community, but research is still in its early stages and it is likely to be some time before comprehensive scenarios which directly provide changes in extremes are available for use in VIA studies.
In most recent impacts studies, it has been common practice to simply take the scenario information from the GCM grid box within which the study site or region is located. Although most impacts studies have followed this approach, lack of confidence in regional estimates of climate change from GCMs has led to the suggestion that the minimum effective spatial resolution should be defined by at least four and possibly more GCM grid boxes (von Storch et al., 1993). It must also be remembered that all scenarios are calculated with respect to a particular reference period, currently 1961 1990, and so the scenarios should be applied only to observed data representing that period. It is incorrect to apply a scenario calculated with respect to 1961 1990 to observed data representing, for example, the 1951 1980 or 1971-2000 periods (see for further explanations and details, the following reports: IPCC-TGCIA, 1999; IPCC, 2001; Barrow et al., 2004).
Finally, as mentioned in Chapter 10 of the IPCC Third Assessment Report (Giorgi et al., 2001), the combined use of GCMs and different downscaling techniques may provide the most suitable approach for the construction of climate change scenarios for impacts and adaptation studies. The comparison of results from different approaches applied to the same problem can increase the confidence in the results and help the evaluation and the understanding of the behaviour of these different methodologies. A suite of techniques allows for a more comprehensive picture of all possible outcomes. This requires a co-ordinated effort to intercompare downscaling methods and GCMs in order to improve our capacity to build a coherent picture of regional climate change that is useful for impacts and adaptation research in Canada (e.g. Barrow et al., 2004).
In summary, as suggested in Barrow et al. (2004), the construction of scenarios of climate variability and extremes is one of the major challenges facing the scenarios research community. Extremes - by definition - are of very low temporal frequency, and also very often occur at very high spatial resolution. This is the most difficult combination of parameters to attempt to model. Climate change scenarios which are currently available are at spatial and temporal scales which are too coarse to provide meaningful information about future extreme events to the VIA community, although GCM output can be used to obtain more qualitative information about these changes. Many of the extremes which are of importance to the VIA community, such as extreme precipitation which can overwhelm municipal stormwater and drainage infrastructure, require much higher temporal and spatial resolution climate information than is currently available from the scenarios community. The CCDS will help to partially overcome this limitation by facilitating the access of higher resolution climate change information (both temporal and spatial) using state-of-the-art research. The content of the CCDS web site is developed to facilitate the comparison and the combination of climate information from GCMs and from different downscaling techniques.