Uncertainties in Regional Simulations of the U.S. Climate in the 20th and 21st Centuries

In support of the development of the 2013 National Climate Assessment report, we analyzed climate model simulation data from the Climate Model Intercomparison Project phase 3 (CMIP3) archive and from the North American Regional Climate Change Assessment Program (NARCCAP) for the continential U.S. Inter-model uncertainty was characterized by a scheme following Tebaldi et al. (2011, Geophysical Research Letters, 38, L23701). Our scheme categorized the future projections at each grid point into one of the following three categories: (1) Category 1: If less than 50% of the models indicate a statistically significant change then the multi-model mean is shown in color. This implies that model results are in general agreement that simulated changes are within historical variations; (2) Category 2: If more than 50% of the models indicate a statistically significant change, and less than 67% of the significant models agree on the sign of the change, then the grid points are masked out, indicating that the models are in disagreement about the direction of change; (3) Category 3: If more than 50% of the models indicate a statistically significant change, and more than 67% of the significant models agree on the sign of the change, then the multi-model mean is shown in color with simple hatching. Model results are in agreement that simulated changes are statistically significant and in a particular direction.

This scheme was applied to a set of 15 CMIP3 models and a set of 8 NARCCAP models. In virtually all cases, grid point multi-model mean changes of temperature variables fell into Category 3. Precipitation variables exhibited much more varied spatial behavior. Mean precipitation is simulated by both the CMIP3 and NARCCAP models to generally increase in the north and decrease in the southwest. Under the A2 scenario, the changes are statistically significant (Category 3) in the far north and southwest by the end of the 21st Century. In the central part of the country, these changes are either not statistically significant (Category 1) or the models are not in agreement on the sign of the changes (Category 2). For extreme metrics of precipitation, most NARCCAP-simulated changes are not statistically significant (Category 1). However, there are areas of statistically significant changes (Category 3), generally in the same regions where mean precipitation changes were statistically significant.

CMIP3 simulations of trends of the climate of the 20th Century were compared with observations, relative to a base period of 1901-1960, and displayed in the context of 21st Century projections for six continental U.S. regions. In most regions, the observed trends are quite close to the multi-model mean. For the southeast, the center of the so-called “warming hole”, the observations for the most recent 20 years are generally within the envelope of the model simulations but on the extreme lower end. The projected multi-model mean changes for both the A2 and B1 scenarios are considerably larger than the range of 20th Century simulations.