Constructing Intelligent Ensemble Averages with Multiple Datasets

The CMIP5 GCM archive contains the newest simulation data from the most updated and advanced set of global coupled climate models. These models, however, exhibit well-known bias when simulating 20th century climate conditions which is presumably also present in the 21st century future projections. A novel approach for constructing “intelligent” climate projections using a process-based framework has been previously tested on the CMIP5 archive: this approach applies a bias-corrective weight to the trend in key energy budget and hydrological variables for each model, which are then ensemble-averaged to produce future projections. The updated study now evaluates the importance of selecting an observational dataset when producing the bias-corrected trends. Several types of observation data are compared, including satellite data from NASA's Clouds and Earth's Radiant Energy System (CERES) and Tropical Rainfall Measuring Mission (TRMM) experiments, merged satellite and in-situ data from the Global Precipitation Climatology Project (GPCP), and reanalysis datasets (namely Modern Era-Retrospective Analysis for Research and Applications (MERRA) and ECMWF Interim). The performance of each dataset is also indirectly compared through this study as the trustworthiness of the “intelligent” climate projections relies heavily on the quality of observational data. It is found that the quality of projections is potentially improved using the alternative ensemble-averaging approach, but the choice of an appropriate observational dataset is crucial when considering the selected region and climatological variable.