Design of a Regional Climate Model Ensemble That Incorporates Model Performance and Independence

Abstract:

Due to various commonalities in model design and construction current climate models do not provide independent predictions of climate. Studies indicate that the information contained within the CMIP3 ensemble (25 models) is only equivalent to a set of perhaps five to ten independent models. This suggests that through judicious selection of models one could retain much of the information content of the full ensemble within a smaller sub-ensemble. Given constraints of computational resources we seek to select the most independent Global Climate Models (GCMs) to downscale from, and Regional Climate Models (RCMs) to downscale with, when producing a regional climate projection ensemble. Thus retaining the maximum information content possible.

We propose a method to perform this selection that satisfies the following criteria

1. The chosen models perform adequately for the recent past compared to observations.

2. The chosen models do not exhibit the same strengths and weaknesses in their representation of the climate (i.e. they are independent).

And for the GCMs

3. The chosen models span the plausible future change space.

An application of this method has been performed for the NARCliM project. NARCliM (NSW/ACT Regional Climate Modelling project) is a regional climate modelling project for the Australian area. It will provide a comprehensive dynamically downscaled climate dataset for the CORDEX-AustralAsia region at 50km, and South-East Australia at a resolution of 10km. NARCliM data will be used by the NSW and ACT governments to design their climate change adaptation plans.Using this process an ensemble of 12 simulations (4 GCMs, 3 RCMs) for each period is obtained (Evans et al. 2014). Additionally to the GCM-driven simulations, 3 control run simulations driven by the NCEP/NCAR reanalysis for the entire period of 1950-2009 are also performed in order to validate the RCMs performance in the area. In this talk, we will present the initial evaluation results of the GCM driven simulations and the projected future changes.

Evans, J. P., Ji, F., Lee, C., Smith, P., Argüeso, D., and Fita, L.: A regional climate modelling projection ensemble experiment – NARCliM, Geosci. Model Dev., 7, 621-629, doi:10.5194/gmd-7-621-2013, 2014.