Exploring regional patterns of uncertainty over precipitation change among CMIP5 models using empirical mode techniques, with a focus on the midlatitude Pacific storm track region

Thursday, 18 December 2014
Baird Langenbrunner1, J David Neelin1, Benjamin R Lintner2 and Bruce T Anderson3, (1)University of California Los Angeles, Los Angeles, CA, United States, (2)Rutgers, New Brunswick, NJ, United States, (3)Boston University, Boston, MA, United States
At the regional level, precipitation (P) change can be one of the most uncertain aspects in climate change assessments, and general circulation models (GCMs) show marked disagreement over P projections at these scales. This intermodel uncertainty owes itself to several factors, including internal model variability, differences in model physics and parameterizations, and intermodel disagreement over end-of-century changes to circulation and moisture transport at larger scales. Here, we use empirical mode techniques, including empirical orthogonal functions (EOFs), rotated EOFs, and singular value decomposition, to diagnose spatial patterns of intermodel disagreement over end-of-century P changes in the Coupled Model Intercomparison Project Phase 5 (CMIP5) archive. We label these Principal Uncertainty Patterns (PUPs) as a general term for the above empirical techniques applied to a model ensemble. Globally, the intermodel uncertainty is dominated by disagreement over changes in the deep tropics. To understand more regional signals, we focus on the midlatitude Pacific wintertime storm track domain, given its relevance to concern over regional P changes in western North America. The first PUP for the storm track domain highlights disagreement over the extension of the eastward edge of storm track precipitation. The second PUP is a more complex pattern involving disagreement over zonally asymmetric changes occurring to the Hadley cell and associated meridional displacement of the storm track. We consider the sensitivity of these results to methodological details, such as use of normalized and pattern-scaled P changes, regular versus rotated EOFs, and domain size. We further consider results based on multivariate PUPs of precipitation and other fields. For example, a survey of coupled disagreement between changes to P and surface air temperature or upper- and lower-level winds sheds light on larger-scale circulation and moisture transport uncertainties feeding into regional disagreement. Finally, PUPs of intermodel disagreement over the historical climatology show little relationship to model spread in end-of-century P changes. This latter result implies that intermodel uncertainty over P climatologies is not a good predictor for disagreement over end-of-century P projections.