OOS 6-1
Downscaling extremes: Abilities of regional downscaling in studying ecological drought
Downscaling is necessary to translate coarsely resolved global climate model (GCM) projections to regional and local changes. However, climatic extremes are often as important to ecosystems as long-term means, and often, means and the extremes are not tightly correlated. But downscaling efforts thus far have focused mostly on temporal means rather than extremes. Also, more attention has been devoted to spatial averages than to spatial variability. To evaluate impacts of projected future climates on ecosystems, it is clear that the extremes as well as the mean climate changes should be considered. The overall deficit of precipitation during drought is a crucial measure, but other phenomena such as heat waves, fire-prone weather, and the spatial variation that occur within a dry spell are also important.
Results/Conclusions
a) Comparison of climate simulations derived from differing downscaling methods to available observations of temperature, precipitation and other ecologically-relevant measures demonstrates how the different downscaling methods yield substantially different distributions of climate extremes in space and time. A set of metrics to evaluate the performance of downscaled output with respect to climatic extremes is needed.
b) Different GCMs and different greenhouse gas emission scenarios also yield quite different changes in extremes, including their frequency, intensity, duration and non-extreme interlude duration. These differences necessitate the use of ensembles of projections by differing models under different emission scenarios in order to provide decision makers with information about the uncertainties they face.
c) Statistical and dynamical downscaling methods both can be useful for ecological applications. Dynamical models provide a full set of atmospheric variables and may yield insights into regional scale changes that depart from historical experience, while statistical models can be trained to reproduce, optimally and economically, climate changes building directly from temporal and spatial variations that have been observed historically. Both approaches also have shortcomings—dynamical models are expensive to run and often require substantial bias adjustments; statistical models are available, usually, for a limited set of variables and are pinned to historical variability and may not capture unexpected or sweeping climate changes.
d) The development of skillful models for regional and local downscaling and the evaluation of those models are limited by the density and quality of observations. Observational gaps are commonly most limiting in complex mountainous terrains that harbor some of the most important ecosystems.