Supplementary Material to Assessment of Uncertainties in Greenhouse Gas Emission Profiles

Supplementary material to “Assessment of uncertainties in greenhouse gas emission profiles of livestock sectors in Africa, Latin America and Europe”

Biqing Zhu, Johannes Kros, Jan Peter Lesschen, Igor G. Staritsky and Wim de Vries

Appendix S1 Assessing the CV of statistical data from the temporal variation in yearly data and actual trends in time

Appendix S2 Full list of model input and parameters and their uncertainties and their uncertainties

Appendix S1 Assessing the CV of statistical data from the temporal variation in yearly data and actual trends in time

The pdfs of the MIPs based on statistical data, which were all derived from FAOSTAT data for each continent, were calculated from the temporal variation in yearly data for the past 10 to 15 years (depending on the available data). Because those MIPs are all activity data, a physical minimum of 0 and a physical maximum of infinity was used. Since the mean values of statistical data are usually large (e.g. annual crop production of a country) and the variances are usually very small, a normal distribution was applied for all the MIPs in this group. However, since this variation not only represents uncertainty, but also the “real” year-to-year variation and/or trends, the standard deviations (SD) were derived from the measurement noise in observations, while accounting for linear trends in data collected from the same country for a given MIP. By using the root mean square error (RMSE) as a proxy for the standard deviation of the measurement noise (Janssen & Heuberger, 1995), the SD was calculated form the quadratic sum of the mean error of an observation (Oi) and the predicted (Pi) value based on a linear trend over N years (Janssen & Heuberger, 1995):

SDi≈RMSE=i(Pi-Oi)2N

This means that the CV can be approximated as the RMSE divided by the mean of the observations, also denoted as the normalized RMSE (NRMSE), i.e.:

CVi≈NRMSE=iPi-Oi2N