Evaluating the Utility of Dynamical Downscaling in Agricultural Impacts Projections

Screen Shot 2015-03-19 at 3.56.50 PM.png

Michael Glotter, Ph.D. candidate in the University of Chicago Department of Geophysical Sciences, gave a brief presentation of his paper, “Evaluating the Utility of Dynamical Downscaling in Agricultural Impacts Projections”, at RDCEP’s weekly meeting on Thursday, November 14. Glotter presented his findings at the American Geophysical Union’s Fall Meeting in San Francisco in early December.

Over the years, researchers have developed various regional climate models in order to better understand future impacts of climate change on our society’s resources. In his presentation, Glotter focused specifically on the question of whether existing models provide useful and accurate insight into the effects of climate change on projected agricultural outputs.

According to Glotter, the “cascade of uncertainty” creates challenges for climate impacts projections. This is because researchers must adjust their climate models in order to match the scale of their area of inquiry: Impacts assessment models require multiple steps, from a society-level projection to a phenomenon-level projection (e.g. greenhouse gas emissions) to a regional-level projection to a local-level projection, for example. Each individual step involves some measure of uncertainty, and the uncertainty in climate impacts projections is amplified as the analysis moves through the appropriate steps in order to be carried out at the appropriate scale.

In other words, agricultural outputs create a problem in scale. Because certain models only work for certain scales, researchers sometimes have to apply downscaling techniques to these models to derive relevant projections. For instance, researchers can drive regional climate models (RCMs) with output from general circulation models (GCMs), and the goal of such downscaling is to generate more accurate climate impacts projections. But Glotter pointed out that dynamical downscaling techniques are computationally expensive (they can take up to months to complete, depending on the type of model) and that their potential benefits are largely unknown in impacts assessments.

Bias correcting GCMs and RCMs has a much larger impact on improving the accuracy of projections than dynamical downscaling techniques, Glotter argued. GCMs inherently include substantial bias, and RCMs do indeed correct some GCM biases. After applying a bias correction, little difference remains between GCM- and RCM-driven yields, casting a question mark over the potential added value from RCMs.