Objective Selection of Representative CMIP5 GCM Simulations for Regional Climate Change Impacts Research: Cluster Initialization and the ETCCDI Climate Extreme Indices

Research on the regional impacts of climate change often involves the use of domain-specific environmental models, for example hydrological models, crop models, ecosystem models, etc., that are driven using outputs from Global Climate Model (GCM) simulations. As the costs of running the environmental models may not be trivial, only a small subset of available GCM simulations, ideally those that bracket the range of simulated changes in relevant climate variables, are typically used to assess regional climate change impacts. When considering a small number of variables, for example mean annual temperature and precipitation, one can quite readily visualize and manually select GCM scenarios that capture the spread of simulated changes.

Most environmental systems are, however, sensitive to climate conditions, for example extremes, that cannot adequately be described by a small number of climate indicators. For example, the CCl/CLIVAR/JCOMM Expert Team on Climate Change Detection and Indices (ETCCDI) has recommended a set of 27 core climate change indicators for annual temperature and precipitation extremes. Visualizing and selecting representative climate change scenarios in such a high dimensional space is no longer a simple task.

To overcome this difficulty, an automated, objective procedure based on a cluster analysis initialization algorithm is proposed and applied to more than 120 Reference Concentration Pathway (RCP) simulations from the CMIP5 Coupled Model Intercomparison Project archive. Results are demonstrated on climate change simulations for the 2050s over 21 regions of the globe. The number of models required to exceed threshold levels of explained variance in the ETCCDI indices is analyzed, as is the potential for loss of information when choosing scenarios based only on indicators like mean annual temperature and precipitation.