Model Selection

Schematic of a climate model. (source: NOAA)

There are over 20 global climate modeling centers around the world that produce more than 60 model versions, all attempting to model the global climate.

In order to provide the best models and data for our smaller subregion of the globe, we perform a model selection procedure to assess which of the many General Circulation Models (GCMs) best replicate historical climate patterns in Alaska and the Arctic.

While it is possible to downscale all available GCM datasets, SNAP provides climate data for the top 5 models for the Alaska and Arctic regions from the Coupled Model Intercomparison Project (CMIP) for both the CMIP3 and CMIP5 global modeling efforts.

We offer downscaled projections for each model, as well as for the average of all 5 models. 

How the climate models were evaluated

SNAP evaluated the performance of 15 GCMs used in the Coupled Model Intercomparison Project 3 (CMIP3) and 22 GCMs used in the Coupled Model Intercomparison Project 5 (CMIP5). For several regions across Alaska and the Arctic, we calculated the degree to which each model’s output aligned with 1958–2001 climate data for precipitation, sea level pressure, and surface air temperature.

Evaluation results

The AR5/CMIP5 climate model evaluation application developed by SNAP summarizes key results of SNAP's general circulation model (GCM) statistical evaluation and model selection.

The evaluation is of historical climate model runs over several geographic domains with an Alaska and Arctic focus. GCM performance is ranked based on minimum error with respect to a European Re-Analysis (ERA-40) baseline data set using several error metrics. This application supplements the paper Downscaling of climate model output for Alaskan stakeholders (Walsh et al. 2018), which includes additional context and details.

Chosen models

The models that performed best over the larger domains tended to be the ones that performed best over Alaska. From this analysis, SNAP selected the five best-performing models from CMIP3 and CMIP5:

CMIP3/AR4Canadian Centre for Climate Modelling and AnalysisGeneral Circulation Model version 3.1 - t47CCCMA-CGCM3.1(T47)
CMIP3/AR4Max Planck Institute for MeteorologyEuropean Centre Hamburg Model 5MPI-ECHAM5/MPI-OM
CMIP3/AR4Geophysical Fluid Dynamics LaboratoryCoupled Climate Model 2.1GFDL-CM2.1
CMIP3/AR4UK Met Office - Hadley CentreCoupled Model 3.0UKMO-HadCM3
CMIP3/AR4Center for Climate System ResearchModel for Interdisciplinary Research on ClimateJAMSTEC-MIROC3.2(medres)
CMIP3/AR45-model averageCalculated as the mean of the above 5 models5modelavg
CMIP5/AR5National Center for Atmospheric ResearchCommunity Earth System Model 4NCAR-CCSM4
CMIP5/AR5NOAA Geophysical Fluid Dynamics LaboratoryCoupled Model 3.0 GFDL-CM3
CMIP5/AR5NASA Goddard Institute for Space StudiesModelE/Russell GISS-E2-R
CMIP5/AR5Institut Pierre-Simon LaplaceIPSL Coupled Model v5A IPSL-CM5A-LR
CMIP5/AR5Meteorological Research InstituteCoupled General Circulation Model v3.0MRI-CGCM3
CMIP5/AR55-model averageCalculated as the mean of the above 5 models5modelavg

For a description of the model selection process, see Walsh JE, Chapman WL, Romanovsky V et al. 2008. Global Climate Model Performance over Alaska and Greenland. Journal of Climate 21 (23): 6156–74.

We acknowledge the World Climate Research Programme’s Working Group on Coupled Modeling, which is responsible for CMIP, and we thank all the climate modeling groups for producing and making available their model output. For CMIP the US Department of Energy's Program for Climate Model Diagnosis and Intercomparison provides coordinating support and leads development of software infrastructure in partnership with the Global Organization for Earth System Science Portals.