Emergent behavior and uncertainty in multimodel climate projections of precipitation trends at small spatial scales

Aspects of model emergent behavior and uncertainty in regional- and small-scale effects of increasing CO2 on seasonal (June-August) precipitation are explored. Nineteen different climate models are studied. New methods of comparing multiple climate models reveal a clearer and more impact- relevant view of precipitation projections for the current century. First, the importance of small spatial scales in multimodel projections is demonstrated. Local trends can be much larger than or even have an opposing sign to the large-scale regional averages used in previous studies. Small-scale effects of increasing CO2 and natural internal variability both play important roles here. These small-scale features make multimodel comparisons difficult for precipitation. New methods that allow information from small spatial scales to be usefully compared across an ensemble of multiple models are presented. The analysis philosophy of this study works with statistical distributions of small-scale variations within climatological regions. A major result of this work is a set of emergent relationships coupling the small- and regional- scale effects of CO2 on precipitation trends. Within each region, a single relationship fits the ensemble of 19 different climate models. Using these relationships, a surprisingly large part of the intermodel variance in small-scale effects of CO2 is explainable simply by the intermodel variance in the regional mean (a form of pattern scaling). Different regions show distinctly different relationships. These relationships imply that regional mean results are still useful, as long as the interregional variation in their relationship with impact- relevant extreme trends is recognized. These relationships are used to present a clear but rich picture of an aspect of model uncertainty, characterized by the intermodel spread in seasonal precipitation trends, including information from small spatial scales.