ESSAY: Model interpretation of climate signals: an application to the Asian monsoon climate - The Global Climate System: Patterns, Processes, and Teleconnections - page 298

present, there are four basic approaches: statistical, nested models, time-
sliced method, and variable-grid GCM. In the statistical approach, cross-
scale relationships, known as ''transfer functions'' are first derived from
large-scale observational and local-scale data, and checked for consistency
with the synoptic-scale forcings of the GCM. For a given climate scenario
provided by the low-resolution GCM (typically with horizontal resolution of
250-500 km), the transfer functions are used to generate the statistics from
global GCM outputs to regional scales. This approach is limited by the
amount of available global and regional data required for robust statistics,
and the possible inconsistency between model and observational data. It has
the clear advantage of computational ease. In the nested regional model
approach, the large-scale forcing functions derived from the GCM are used
as lateral boundary conditions to drive a regional climate model (with typical
resolution of 20-50 km) applied to a specific geographic region usually of
continental to sub-continental scale, and for a chosen time period of interest.
This time period may be related to the occurrence of a devasting drought or
flood in a certain region, and one wants to see what are the causes, and if they
are related to the underlying large-scale climate forcings or to local feedback
processes. Multiple nesting grids, with increasing resolution are sometimes
used to zoom in on a subregion to resolve even smaller scale features. The
nested regional models may have numerical instabilities at the lateral bound-
aries, so appropriate buffer zones have to be designed (Giorgi andMearns 1991 ).
Alternatively, to avoid the lateral boundary problems, a time-sliced
approach is used, by re-running the atmospheric component of the coupled
GCM, at a higher spatial resolution and for a shorter time period, using the
large-scale lower boundary forcings, such as sea surface temperature, from
the coupled model. This approach has the disadvantage of ''wasting''
valuable computational resources outside the region of interest. More
recently, a new strategy has been developed to use GCMs with variable
resolution, or so-called stretched or ''telescoping grids,'' in which the GCM
can zoom in on a specific region, with high resolution to resolve local
features, while keeping the computations elsewhere at the coarse resolution
(Fox-Rabinovitz et al. 2001 ). This approach can achieve considerable
savings in computation resources, while achieving the desired higher spa-
tial resolution in the region of interest. The variable-grid GCM requires the
redesign of the model numerics, as well as the physical parameterizations
to maintain dynamical consistency between the regions with high and low
resolutions.
Depending on the space-time resolution, outputs from the regional models
can be used for climate assessments, and for water resource management. For
applications to river-basin catchment scales, further downscaling may be
needed (Lattenmaier et al. 1999 ). For that purpose, outputs from the regional