Bivariate Analysis of Errors in National Inventory Reporting

The paper aims at analysing errors in the National Inventory Reports, provided annually since 2001 by parties to the UNFCCC. Each said report contains the data on GHG emission from a given year and revisions of past data (back to 1990), recalculated due to improved knowledge and methodology. We consider these revisions the realizations of a non-stationary stochastic process being a sum of two other processes, indexed by time. One of these component processes corresponds to the years, when emission measurements were taken, and involves data gathering errors. The other one corresponds to the years when revisions were made, and involves data processing errors. To disentangle these two we consider the data matrix of realizations up to 2015 and use a bivariate approach to estimate the mean values of the component processes as functions of time. We analyse both types of errors and their changes over time using the mean values estimated. The analysis is conducted for the EU-15 and its individual member countries.