GCMs are coupled land-atmosphere-ocean numerical models that are used to estimate climate variables such as temperature and precipitation, and are set up for various greenhouse gas and aerosol emission scenarios. GCMs use horizontal grid cells that can have dimensions in the range of 150-500 km. Therefore, GCMs are not capable of accurately estimating climate variables at the groundwater basin scale.
Consequently, spatial downscaling methods have been developed to derive finer resolution data from coarse resolution GCM results. These methods can be grouped into two major categories; the first of these is dynamical downscaling, where higher resolution regional climate models (RCMs) or limited area models (LAMs) are set up using boundary conditions derived from the coarser resolution GCM. The second group of downscaling approaches consists of statistical methods, which rely on the fundamental concept that regional climates are largely a function of the large-scale atmospheric state. This relationship may be expressed in the form of transfer functions, which cover a range of methods: multiple regression, non-linear regression, artificial neural networks, principal component analysis, and canonical correlation analysis. Stochastic weather generators and weather classification schemes are more sophisticated statistical methods.
Both approaches of downscaling have their advantages and limitations. However, comprehensive comparison studies suggest that dynamical downscaling methods provide little advantage over statistical methods. Given the data intensive structure of dynamical methods, it may be preferable to select a statistical downscaling method in groundwater impact studies. The main objective of this type of downscaling is to establish a quantitative relationship between large scale atmospheric variables (predictors) and local surface variables (predictands), which is not always straightforward due to the often weak correlation between the predictors and predictands.
Was this article helpful?