10.6.3.2 Evaluation

The evaluation of downscaling techniques is essential but problematic. It requires that the validity of the downscaling functions under future climates be demonstrated, and that the predictors represent the climate change signal. It is not possible to achieve this rigorously as the empirical knowledge available is insufficient. The analysis of historical developments, e.g., by comparing downscaling models between recent and historical periods (Jacobeit et al., 1998), as well as simulations with GCMs can provide support for these assumptions. However, the success of a statistical downscaling technique for representing present day conditions does not necessarily imply that it would give skilful results under changed climate conditions, and may need independent confirmation from climate model simulations (Charles et al., 1999b).

The classical validation approach is to specify the downscaling technique from a segment of available observational evidence and then assess the performance of the empirical model by comparing its predictions with independent observed values. This approach is particularly valuable when the observational record is long and documents significant changes (greater than 50 years in some cases; Hanssen-Bauer and Førland (1998, 2000)). An example is the analysis of absolute pressure tendencies in the North Atlantic (Kaas et al., 1996). As another example, Wilks (1999b) developed a downscaling function on dry years and found it functioned well in wet years.

An alternative approach is to use a series of comparisons between models and transfer functions (e.g., González-Rouco et al., 1999, 2000). For instance, empirically derived links were shown to be incorporated in a GCM (Busuioc et al., 1999) and a RCM (Charles et al., 1999b). Then a climatic change due to doubling of CO₂ was estimated through the empirical link and compared with the result of the dynamical models. In both cases, the dynamical response was found consistent for the winter season, indicating the validity of the empirical approach, although less robust results were noted in the other seasons.

10.6.3.3 Choice of predictors

There is little systematic work explicitly evaluating the relative skill of different atmospheric predictors (Winkler et al., 1997). This is despite the availability of disparate studies that evaluate a broad range of predictors, predictands and techniques (see Appendix 10.4). Useful summaries of downscaling techniques and the predictors used are also presented in Rummukainen (1997), Wilby (1998), and Wilby and Wigley (2000).

The choice of the predictor variables is of utmost importance. For example, Hewitson and Crane (1996) and Hewitson (1999) have demonstrated how the down-scaled projection of future change in mean precipitation and extreme events may alter significantly depending on whether or not humidity is included as a predictor. The downscaled results can also depend on whether absolute or relative humidity is used as a predictor (Charles et al., 1999b). The implication here is that while a predictor may or may not appear as the most significant when developing the downscaling function under present climates, the changes in that predictor under a future climate may be critical for determining the climate change. Some estimation procedures, for example stepwise regression, are not able to recognise this and exclude variables that may be vital for climate change.

A similar issue exists with respect to downscaling temperature. Werner and von Storch (1993), Hanssen-Bauer and Førland (2000) and Mietus (1999) noted that low-frequency changes in local temperature during the 20th century could only partly be related to changes in circulation. Schubert (1998) makes a vital point in noting that changes of local temperature under doubled atmospheric CO₂ may be dominated by changes in the radiative properties of the atmosphere rather than circulation changes. These can be accounted for by incorporating the large-scale temperature field from the GCM as a surrogate indicator of the changed radiative properties of the atmosphere (Dehn and Buma, 1999) or by using several large-scale predictors, such as gridded temperature and circulation fields (e.g., Gyalistras et al., 1998; Huth, 1999).

With the recent availability of global reanalyses (Kalnay et al., 1996; Gibson et al., 1997), the number of candidate predictor fields has been greatly enhanced (Solman and Nuñez, 1999). Prior to this the empirical evidence about the co-variability of regional/local predictands and large-scale predictors was limited mostly to gridded near surface temperature and/or air pressure. These “new” data sets allow significant improvements in the design of empirical downscaling techniques, in particular by incorporating knowledge about detailed meteorological processes. Taking advantage of these new data sets have allowed systematic evaluation of a broad range of possible predictors for daily precipitation. It has been found that indicators of mid-tropospheric circulation and humidity to be the most critical predictors, with surface flow and humidity information being important under orographic rainfall.

IPCC Homepage