Home

Review of Downscaling Methodologies for Africa Climate Applications

Casey Brown; Arthur M. Greene; Paul J. Block; Alessandra Giannini

Title:
Review of Downscaling Methodologies for Africa Climate Applications
Author(s):
Brown, Casey
Greene, Arthur M.
Block, Paul J.
Giannini, Alessandra
Date:
Type:
Technical reports
Department:
International Research Institute for Climate and Society
Permanent URL:
Series:
IRI Technical Report
Part Number:
08-05
Publisher:
International Research Institute for Climate and Society
Publisher Location:
Palisades, N.Y.
Abstract:
Downscaling is the term used to describe the various methods used to translate the climate projections from coarse resolution GCMs to finer resolutions deemed more useful for assessing impacts. Projections of future climate are produced using complex, coupled atmosphere-ocean models (GCMs). The GCMs are most reliable at the continental scale. Due to the inherent uncertainty of the climate system and the inevitable existence of model errors, multi-model ensembling is the recommended approach for characterizing expected climate changes. As downscaling is dependent on the ability of GCMs to successfully project the climate change signal, it is limited to where that signal is clear. Assessments of climate change in Africa indicate some consensus of reduced winter rainfall in southern Africa, increased annual rainfall in east Africa and uncertainty for the rest of Africa. Selection of GCMs that "do better" over Africa, or any region, is difficult and probably not warranted, given the general parity in model skill and the difficulty in identifying which models are more skillful. Ensemble means or medians offer the highest level of projection accuracy. Downscaling approaches are generally categorized as dynamical, using regional climate models, and statistical, using empirical relationships. However, dynamical downscaling often includes statistical modeling in the form of "bias correction." Dynamical downscaling is useful for incorporating topographic features, such as strong orography, and land use and vegetation. It is recommended where those features play a significant role in regional climate. However, computational time and the uncertainties that accompany complex models outweigh the benefits of dynamical downscaling where these features are not significant. The spatial resolution that can be achieved is on the order of tens of kilometers. Statistical downscaling is simpler and more efficient than dynamical downscaling. It is preferred where estimates of specific variables, especially at point locations, are sought for input to sector models (e.g., hydrologic models) or decision making. However, statistical modeling can mask a true understanding of regional climate dynamics and estimates may be overconfident. In summary, downscaling is best understood as an attempt to increase the understanding of climate change influences at the regional scale. In that context, a variety of methodologies should be explored, using all tools possible to increase that understanding. A set of "Best Practices" is recommended for pursuing this effort.
Subject(s):
Climate change
Item views:
392
Metadata:
text | xml

In Partnership with the Center for Digital Research and Scholarship at Columbia University Libraries/Information Services | Terms of Use