Validation of a Global Geospace Model With a Systems Science Approach Based on Canonical Correlation Analysis

Abstract

A systems science approach based on canonical correlation analysis (CCA) is applied as a new, behavioral way to validate global geospace models. The biggest novelty of the technique is that it validates models at a system level, whereby a side-by-side comparison is performed of CCA applied to a 30-day observational and the corresponding simulation data sets comprising quiet, moderate and active times. The simulation used the Multiscale Atmosphere-Geospace Environment (MAGE) model. It is shown that (a) CCA must be combined with sensitivity analysis to be effective, (b) the MAGE model generally reproduces the observed behavior (more so for quieter time intervals), quantified by the intercorrelations between different variables and (c) the technique identifies the SuperMAG SML index as a quantity for which refinements of the model are needed.