Principal uncertainty patterns in precipitation among CMIP5 models: Dominant modes of intermodel disagreement in precipitation climatologies and projected change patterns

Projections of modeled precipitation change in global warming scenarios demonstrate marked intermodel disagreement, especially at regional scales. While these differences are often considered within a geographically local domain, they are in part caused by intermodel uncertainty inherited from the large scale. It is therefore important to identify the major aspects of model disagreement at larger scales in order to better understand differences at the regional level. One way to do this is to pinpoint the major modes of intermodel disagreement through objective analysis of modeled precipitation change patterns, as well as the disagreement in precipitation climatologies in historical and radiative forcing scenarios. For brevity, these modes are labeled Principal Uncertainty Patterns (PUPs). For the Coupled Model Intercomparison Project phase 5 (CMIP5) models, a PUP analysis is applied to projected changes in precipitation, upper-level wind, and temperature fields, for both coupled model runs (36 models) and atmosphere-only simulations (30 models). This analysis is also applied to the simulated historical and future climatologies for the same ensembles. We take a global approach first, and then focus on specific regions: Africa, the tropical and subtropical Americas, and the mid-latitude Pacific storm tracks. For Africa, the leading June-July-August (JJA) PUP is associated with disagreement in the amplitude of positive end-of-century precipitation changes in the monsoon region. For the tropical Americas, intermodel uncertainty in the amount of end-of-century drying is the dominant PUP for JJA. The two leading December-January-February (DJF) PUPs in the storm tracks region appear to represent (1) an amplitude mode that shows the eastward extension of mid-latitude Pacific storm tracks trailing into the North American coast, and (2) a gradient mode associated with the meridional shift of these storm tracks. Relationships of precipitation uncertainties to model disagreement in upper-level winds and temperature fields are also examined, as well as comparison to observations available through the obs4MIPs project. Modes of joint variability between projected precipitation change and quantities from the historical base period in models provide a way of objectively identifying patterns of intermodel disagreement on future climate change that are related to intermodel disagreement within the current climatology.