CR 07:85-95 (1996)  -  DOI: https://doi.org/10.3354/cr007085

Downscaling, or translation across scales, is a term adopted in recent years to describe a set of techniques that relate local- and regional-scale climate variables to the larger scale atmospheric forcing. Conceptually, this is a direct evolution of more traditional techniques in synoptic climatology; however, the downscaling approach was developed specifically to address present needs in global environmental change research, and the need for more detailed temporal and spatial information from Global Climate Models (GCMs). Two general categories exist for downscaling techniques: process based techniques focused on nested models, and empirical techniques using one form or another of transfer function between scales. While in the long term nested models hold the greatest promise for regional-scale analysis, this approach is still in development, requires detailed surface climate data, and is dependent on high end computer availability. Conversely, empirical relationships offer a more immediate solution and significantly lower computing requirements, consequently offering an approach that can be rapidly adopted by a wider community of scientists. In this paper, an application of empirical downscaling of regional precipitation is implemented to demonstrate its effectiveness for evaluating GCM simulations and developing regional climate change scenarios. Gridded analyses of synoptic-scale circulation fields are related to regional precipitation using neural nets. Comparable GCM circulation fields are then used with the derived relationships to investigate control simulation and doubled atmospheric CO₂ simulation synoptic-scale forcing on regional climates.

Downscaling · Artificial Neural Nets · Climate change