Perturbed Parameter Ensembles of Idealized Experiments to Identify Plausible and Diverse Variants of a Model for Climate Change Projections

We use an efficient, idealized set-up to construct five sets of perturbed parameter ensemble (PPE) experiments based on HadGEM3-GA4.0 to explore the uncertainty in model’s performance and its response to climate change. The PPEs, constructed by simultaneously perturbing 22 poorly constrained parameters in five different scheme, include NWP and AMIP-style experiments to understand model performance on weather and climate timescales respectively, and three idealized experiments to determine the effect of parameter perturbations on forcing and feedback components that drive model responses to climate change. We demonstrate that there is a strong relationship between global model errors at the weather and climate timescales for a variety of key variables. We also focus on model performance over two regions: North America and the Indian subcontinent. A seasonal dry bias over central and southern US and a wet bias over the Rockies, and a dry bias over the Indian subcontinent during the monsoon season are some of the biases that are common to current generations of global climate models (CMIP5). We demonstrate that some of these errors in climate simulations develop within the first five days of the TAMIP integration, the design of which facilitates linking these systematic errors to model parameters and processes. We also investigate to what extent these model biases are structural in nature and if there exist model variants (i.e., parameter combinations) that improve model performance significantly. In terms of model response, the effect of perturbing parameters is to produce wide ranges in most forcing and feedback components that are similar in magnitude to those seen in the CMIP5 MME. Overall, we show that the PPE approach in conjunction with seamless assessment across spatial and temporal scales can be used to gain a process-level understanding of model errors and can guide identification of a set of credible and diverse variants of a climate model for future projections.