10.6 Empirical/Statistical and Statistical/Dynamical Methods
10.6.1 Introduction
As with the dynamical downscaling of RCMs, the methods described in this section
rely on the concept that regional climates are largely a function of the large-scale
atmospheric state. In empirical downscaling the cross-scale relationship is
expressed as a stochastic and/or deterministic function between a set of large-scale
atmospheric variables (predictors) and local/regional climate variables (predictands).
Predictor and predictand can be the same variables on different spatial scales
(e.g., Bürger, 1997; Wilks, 1999b; Widmann and Bretherton, 2000), but more
commonly are different.
When using downscaling for assessing regional climate change, three implicit
assumptions are made:
- The predictors are variables of relevance to the local climate variable
being derived, and are realistically modelled by the GCM. Tropospheric quantities
such as temperature or geopotential height are more skilfully represented
than derived variables such as precipitation at the regional or grid scale
(e.g., Osborn and Hulme, 1997; Trigo and Palutikof, 1999). Furthermore, there
is no theoretical level of spatial aggregation at which GCMs can be considered
skilful, though there is evidence that this is several grid lengths (Widmann
and Bretherton, 2000).
- The transfer function is valid under altered climatic conditions (see Section
10.6.2.2). This cannot be proven in advance, as it would require the observational
record to span all possible future realisations of the predictors. However,
it could be evaluated with nested AOGCM/RCM simulations of present and future
climate, using the simulation of present climate to determine the downscaling
function and testing the function against the future time slice.
- The predictors fully represent the climate change signal. Most downscaling
approaches to date have relied entirely on circulation-based predictors and,
therefore, can only capture this component of the climate change. More recently
other important predictors, e.g., atmospheric humidity, have been considered
(e.g., Charles et al., 1999b; Hewitson, 1999).
A diverse range of downscaling methods has been developed, but, in principle,
these models are based on three techniques:
- Weather generators, which are random number generators of realistic looking
sequences of local climate variables, and may be conditioned upon the large-scale
atmospheric state (Section 10.6.2.1);
- Transfer functions, where a direct quantitative relationship is derived
through, for example, regression (Section 10.6.2.2);
- Weather typing schemes based on the more traditional synoptic climatology
concept (including analogues and phase space partitioning) and which relate
a particular atmospheric state to a set of local climate variables (Section
10.6.2.3).
Each of these approaches has relative strengths and weaknesses in representing
the range of temporal variance of the local climate predictand. Consequently,
the above approaches are often used in conjunction with one another in order
to compensate for the relative deficiencies in one method.
Most downscaling applications have dealt with temperature and precipitation.
However, a diverse array of studies exists in which other variables have been
investigated. Appendix 10.4 provides a non-exhaustive
list of past studies indicating predictands, geographical domain, and technique
category. In light of the diversity in the literature, we concentrate on references
to applications since 1995 and based on recent global climate change projections.
