A0253
Title: Functional data analysis-based sensitivity and domain-selective testing for integrated assessment models
Authors: Matteo Fontana - Politecnico di Milano (Italy) [presenting]
Massimo Tavoni - Politecnico di Milano (Italy)
Simone Vantini - Politecnico di Milano (Italy)
Abstract: Climate change is among the biggest threats to the survival of humankind. A lot of research efforts are being undertaken to study its evolution and the impact of policy measures aimed at mitigating its effects. The tool of choice for these tasks are Integrated Assessment Models (IAM), complex pieces of software that are used to predict socioeconomic variables for the next century. These models are black boxes de facto, due to the complex and nonlinear relationships between input and output variables. Moreover, they are often very computationally intensive, so it is virtually impossible to characterize the model response via standard Monte-Carlo based methods. To analyse models with these features, the most common choice is the use of data-parsimonious Global Sensitivity Analysis (GSA) techniques, that are moreover able to decouple additive and interaction effects. The main drawback of these techniques is that can deal only with univariate responses, disregarding the time dimension. The aim is to include the time dimension in the sensitivity analysis of IAMs, by modelling the time-variant outputs as smooth functions over the time domain, and then extend current GSA techniques to functional data. To assess the impact of model uncertainty on the significance of sensitivity indices, we frame the GSA as a Functional ANOVA problem, and then use a domain-selective permutational-based inferential technique for linear models to perform significance testing.
