Quantifying and Reducing the Uncertainties in Future Projections of Droughts and Heat Waves for North America that Result from the Diversity of Models in CMIP5

Abstract:

There are many sources of uncertainty regarding the future projections of our climate, including the multiple possible Representative Concentration Pathways (RCPs), the variety of climate models used, and the initial and boundary conditions with which they are run. Moreover, it has been shown that the internal variability of the climate system can sometimes be of the same order of magnitude as the climate change signal or even larger for some variables. Nonetheless, in order to help inform stakeholders in water resources and agriculture in North America when developing adaptation strategies, particularly for extreme events such as droughts and heat waves, it is necessary to study the plausible range of changes that the region might experience during the 21^st century. We aim to understand and reduce the uncertainties associated with this range of possible scenarios by focusing on the diversity of climate models involved in the Coupled Model Intercomparison Project Phase 5 (CMIP5). Data output from various CMIP5 models is compared against near surface climate and land-surface hydrological data from the North American Land Data Assimilation System (NLDAS)-2 to evaluate how well each climate model represents the land-surface processes associated with droughts and heat waves during the overlapping historical period (1979-2005). These processes include the representation of precipitation and radiation and their partitioning at the land surface, land-atmosphere interactions, and the propagation of signals of these extreme events through the land surface. The ability of the CMIP5 models to reproduce these important physical processes for regions of North America is used to inform a multi-model ensemble in which models that represent the processes relevant to droughts and heat waves better are given more importance. Furthermore, the future projections are clustered to identify possible dependencies in behavior across models. The results indicate a wide range in performance for the historical runs with some models hampered by poor interannual variability in summer precipitation and near surface air temperature, whilst others partition too much precipitation into evapotranspiration with implications for drought and heat wave development.