An empirical characterisation of signal versus noise in CO₂ data

Abstract

Uncertainties in the interpretations of CO₂ data arise from errors in the observations and model relations. The space–time variations of CO₂ on the global scale are analysed in terms of the singular-value decomposition, in order to obtain a characterisation of observational error that matches the requirements of global-scale estimation of fluxes. It is found that for monthly-mean data, a first-order moving average model of error is a far better representation than earlier assumptions of independent white noise.