An Iterative Selection Method: A Summer Eastern Asian Metric for CMIP3 Climate Models as an Example

Abstract

 Global climate metrics are often used for model evaluation and selection of general circulation models. However, most future projection studies have a certain target region or phenomenon, while global climate metrics do not necessarily reflect model performance in reproducing local climate and/or phenomena. In this study, we introduce an iterative selection method of variables and regions to derive a regional climate metric that is also related to model performance in reproducing local phenomena, utilizing a collection of various phenomena metrics. Using the 20th century experiment outputs of the Climate Model Intercomparison Project phase 3 (CMIP3) models, we demonstrate the derivation of a prototype summer Eastern Asian metric. Three locally important phenomena, the seasonal progress of the jet stream, the seasonal progress of the Baiu, and the inter-annual variation of the Pacific-Japan pattern are chosen for the prototype metric. The variables iteratively selected for the new metric were 200-hPa zonal wind, 850-hPa meridional wind, and precipitation. The selected region included Siberia and the western North Pacific. The prototype metric correlates with the three phenomena metrics significantly better than the global climate metric. The resulting metric is strongly influenced by the chosen set of phenomena to be considered, and can be modified by altering the choice accordingly to the purpose of study.