11.10 Assessment of Regional Climate Projection Methods
The assessment of methods recognises the challenges posed by the complex interactions that occur at many spatial and temporal scales, involving the general circulation, cross-scale feedbacks and regional-scale forcing.
11.10.1 Methods for Generating Regional Climate Information
Atmosphere-Ocean General Circulation Models constitute the primary tool for capturing the global climate system behaviour. They are used to investigate the processes responsible for maintaining the general circulation and its natural and forced variability (Chapter 8), to assess the role of various forcing factors in observed climate change (Chapter 9) and to provide projections of the response of the system to scenarios of future external forcing (Chapter 10). As AOGCMs seek to represent the whole climate system, clearly they provide information on regional climate and climate change and relevant processes directly. For example, the skill in simulating the climate of the last century when accounting for all known forcings demonstrates the causes of recent climate change (Chapter 9) and this information can be used to constrain the likelihood of future regional climate change (Stott et al., 2006; see also Section 11.10.2). AOGCM projections provide plausible future regional climate scenarios, although methods to establish the reliability of the regional AOGCM scales have yet to mature. The spread within an ensemble of AOGCMs is often used to characterise the uncertainty in projected future climate changes. Some regional responses are consistent across AOGCM simulations, although for other regions the spread remains large (see Sections 11.2 to 11.9).
Because of their significant complexity and the need to provide multi-century integrations, horizontal resolutions of the atmospheric components of the AOGCMs in the MMD range from 400 to 125 km. Generating information below the grid scale of AOGCMs is referred to as downscaling. There are two main approaches, dynamical and statistical. Dynamical downscaling uses high-resolution climate models to represent global or regional sub-domains, and uses either observed or lower-resolution AOGCM data as their boundary conditions. Dynamical downscaling has the potential for capturing mesoscale nonlinear effects and providing coherent information among multiple climate variables. These models are formulated using physical principles and they can credibly reproduce a broad range of climates around the world, which increases confidence in their ability to downscale realistically future climates. The main drawbacks of dynamical models are their computational cost and that in future climates the parametrization schemes they use to represent sub-grid scale processes may be operating outside the range for which they were designed.
Empirical SD methods use cross-scale relationships that have been derived from observed data, and apply these to climate model data. Statistical downscaling methods have the advantage of being computationally inexpensive, able to access finer scales than dynamical methods and applicable to parameters that cannot be directly obtained from the RCM outputs. They require observational data at the desired scale for a long enough period to allow the method to be well trained and validated. The main drawbacks of SD methods are that they assume that the derived cross-scale relationships remain stable when the climate is perturbed, they cannot effectively accommodate regional feedbacks and, in some methods, can lack coherency among multiple climate variables.