Evaluation of simulated responses to climate forcings: a flexible statistical framework using confirmatory factor analysis and structural equation modelling – Part 2: Numerical experiment

• Published in: Advances in statistical climatology, meteorology and oceanography Vol. 8; no. 2; pp. 249 - 271
• Main Authors: Lashgari, Katarina, Moberg, Anders, Brattström, Gudrun
• Format: Journal Article
• Language: English
• Published: Gottingen Copernicus GmbH 2022; Copernicus Publications
• Subjects: Climate; Climate change; Climate models; Datasets; Experiments; Factor analysis; Frameworks; General circulation models; matematisk statistik; Mathematical models; Mathematical Statistics; Modelling; Multivariate statistical analysis; Noise levels; Numerical experiments; Performance evaluation; Precipitation; Sensitivity analysis; Simulation; Statistical analysis; Statistical methods; Statistical models; Structural equation modeling; Temperature; Temperature data; Temperature variations; Variables
• DOI: 10.5194/ascmo-8-249-2022
• ISSN: 2364-3587; 2364-3579

The performance of a new statistical framework, developed for the evaluation of simulated temperature responses to climate forcings against temperature reconstructions derived from climate proxy data for the last millennium, is evaluated in a so-called pseudo-proxy experiment, where the true unobservable temperature is replaced with output data from a selected simulation with a climate model. Being an extension of the statistical model used in many detection and attribution (D&A) studies, the framework under study involves two main types of statistical models, each of which is based on the concept of latent (unobservable) variables: confirmatory factor analysis (CFA) models and structural equation modelling (SEM) models. Within the present pseudo-proxy experiment, each statistical model was fitted to seven continental-scale regional data sets. In addition, their performance for each defined region was compared to the performance of the corresponding statistical model used in D&A studies. The results of this experiment indicated that the SEM specification is the most appropriate one for describing the underlying latent structure of the simulated temperature data in question. The conclusions of the experiment have been confirmed in a cross-validation study, presuming the availability of several simulation data sets within each studied region. Since the experiment is performed only for zero noise level in the pseudo-proxy data, all statistical models, chosen as final regional models, await further investigation to thoroughly test their performance for realistic levels of added noise, similar to what is found in real proxy data for past temperature variations.