Downscaling

Definition

Downscaling is a technique for applying information derived from larger-scale models or data analyses to smaller-scale models. In the context of global warming research, downscaling is used to link large-scale atmospheric variables with local climates. Global climate models, also known as general circulation models (GCMs), are data-rich, three-dimensional computer models of the climate mapped on a grid. GCMs employ variable atmospheric data to make estimates of future climate change on a global, hemispheric, or continental scale. Their resolution is generally about 150 to 300 kilometers by 150 to 300 kilometers. However, smaller-scale, local models require data at scales of 10 to 100 kilometers. To translate the larger-scale information to a regional scale requires downscaling into higher spatial and temporal resolution grids. With this finer resolution, the effect of sub-grid topographical features such as clouds, mountain ranges, and wetlands can be incorporated. There are two main methods of downscaling: dynamical downscaling and statistical, or empirical, downscaling.

Significance for Climate Change

Downscaling is an important tool for the study of anthropogenic climate change caused by increased carbon dioxide (CO2) emissions. GCMs allow for predictions of large-scale climate change based on assumed patterns of greenhouse gas (GHG) emissions and other impacts. However, this scale can lack accuracy for regional and sub-regional models where differences in climate occur at a scale below GCM resolutions. Downscaling is used to make more accurate predictions at this local-to-regional scale.

Downscaling has been used in atmospheric forecasting for several decades. In dynamical downscaling, limited area models (LAMs) and regional climate models (RCMs) are nested into GCMs using GCM data as boundary conditions. In statistical downscaling, statistical relationships are calculated between factors simulated by the GCM at the large-scale level and variable data measured at the local level, such as rainfall occurrences and surface air temperatures. Dynamical techniques produce more data but are computationally expensive and depend on the accuracy of the GCM grid-point data used to calculate boundary conditions.

Dynamical downscaling has been used, for example, to calculate the number of extreme temperature days per year, as well as changes in their distribution. Such data can then be used to calculate the increased drought risk in dry climate areas, such as prevail in Australia and New Zealand. Statistical downscaling has been used, for example, to chart projected temperature ranges in various regions through the remainder of the twenty-first century, as well as mean monthly temperatures and precipitation. A combination of dynamical and statistical methods will probably obtain the best results in the future, allowing local planners to better adapt strategies and policies for coping with climate changes in their regions.