Quantification of Uncertainty in High-Resolution Temperature Scenarios for North America

Abstract

A framework for the construction of probabilistic projections of high-resolution monthly temperature over North America using available outputs of opportunity from ensembles of multiple general circulation models (GCMs) and multiple regional climate models (RCMs) is proposed. In this approach, a statistical relationship is first established between RCM output and that from the respective driving GCM and then this relationship is applied to downscale outputs from a larger number of GCM simulations. Those statistically downscaled projections were used to estimate empirical quantiles at high resolution. Uncertainty in the projected temperature was partitioned into four sources including differences in GCMs, internal variability simulated by GCMs, differences in RCMs, and statistical downscaling including internal variability at finer spatial scale. Large spatial variability in projected future temperature changes is found, with increasingly larger changes toward the north in winter temperature and larger changes in the central United States in summer temperature. Under a given emission scenario, downscaling from large scale to small scale is the most important source of uncertainty, though structural errors in GCMs become equally important by the end of the twenty-first century. Different emission scenarios yield different projections of temperature change. This difference increases with time. The difference between the IPCC’s Special Report on Emissions Scenarios (SRES) A2 and B1 in the median values of projected changes in 30-yr mean temperature is small for the coming 30 yr, but can become almost as large as the total variance due to internal variability and modeling errors in both GCM and RCM later in the twenty-first century.