The proposed research will take an alternative approach to differential evaluation of model credibility by focusing on process-level evaluation rather than on simple metrics. We hypothesize that a consistent set of process-oriented model analyses can be developed and applied in different climate regimes, and that this suite of model analyses will help define credible model members whose future simulated climates will have value for regional climate change assessment. We are focusing on three regions in North America (Southwest, Great Plains, Northeast) with the following objectives: (1) Establish a framework for determining the differential credibility of climate simulations using a process-based methodology for three specified regions; (2) Develop process-level time-series analysis to follow identified mechanistic errors in the evolution (from current period into the future) of warm season precipitation in the regions; (3) Based on the process-level analysis, differentiate the credibility of the models using collective expert evaluation (CEE); (4) Translate the process-level information into quantitative metrics; (5) Compare these metrics with the baseline metrics of ENSEMBLES; (6) Compare credibility rankings based on our process-level collective expert evaluation and developed metrics with rankings based on the ENSEMBLES metrics and diagnose causes of differences. (7) Apply the developed framework to Coupled Model Intercomparison Project (CMIP5) high-resolution decadal predictions.
This research will employ simulation data sets produced through the North American Regional Climate Change Program (NARCCAP) and the global coupled model (50 km) decadal predictions in progress as part of CMIP5 for the present day and near future periods. The process-level investigations will first be conducted for the coupled global models (AOGCMs), atmospheric global models (time-slices), and the regional climate models (RCMs) involved in NARCCAP. Time series of process-level errors will be examined in the reanalysis-driven and AOGCM-driven present day simulations and followed into the future simulations. The importance of the errors as the models respond to the new forcing in the future will be evaluated. In this way, our analysis will focus on the effect of the more consequential errors on the model future response. Our purpose is to perform qualitative process-level collective expert evaluations for each region, which may then be transformed into quantitative indicators.
The ultimate goal of this analysis is to provide meaningful differential weighting of the models using a process-level approach that results in more robust estimates of future climate change. The process-level analyses have value in that, by enhancing our understanding of the evolution of processes under greenhouse gas forcing, uncertainties may be reduced in the sense that better understanding of important mechanisms at work will result. The proposed research directly addresses MAPP FY2011 Priority 1b: ’to evaluate uncertainties in the longterm prediction and projection of twenty-first century climate over North America leveraging NARCCAP and new CMIP5 projections’.