Quantifying and Constraining Structural Uncertainty in Future Climate Projections over India

Abstract:

The quantification of uncertainty in future climate change projections is essential for appropriate use by policy makers.In this study, coupled climate model projections from 27 models participating in the CMIP5 are analyzed for projected future changes over India and sub-regions (homogenous climatic zones) within. We partition the uncertainty of future temperature and precipitation in the RCP4.5 and RCP8.5 experiments into Epistemic (structural or model dependent), Aleatoric (chaotic or initial condition dependent), and Reflexive (scenario dependent) components. For this, a subset of 8 models is selected on the basis of availability of more than three realizations from each model. Our results indicate that while there are regions and seasons with large model dependent uncertainty, in many instances the Aleatoric uncertainty component exceeds it. We also address the question whether using appropriate performance metric based weighting for individual models can reduce structural uncertainty. We explore combinations of metrics across variables (and multiple observational datasets where available) and phenomena that matter for the different regional climatic zones to weight models and examine the resulting uncertainty range.