Search

Purpose: This study aims to analyse the variability in the carbon footprint (CF) of organically and conventionally produced Valencian oranges (Spain), including both farming and post-harvest (PH) stages. At the same time, two issues regarding sample representativeness are addressed: how to determine confidence intervals from small samples and how to calculate the aggregated mean CF (and its variability) when the inventory is derived from different sources. Methods: The functional unit was 1 kg of oranges at a European distribution centre. Farming data come from a survey of two samples of organic and conventional farms; PH data come from one PH centre; and data on exportation to the main European markets were obtained from official secondary sources. To assess the variability of the farming subsystem, a bootstrap of the mean CF was performed. The variability of the PH subsystem was assessed through a Monte Carlo simulation and a subsequent subsampling bootstrap. A weighted discrete distribution of the CF of distribution and end-of-life (EoL) was built, which was also bootstrapped. The empirical distribution of the overall CF was obtained by summing all iterations of the three bootstrap procedures of the subsystems. Results and discussion: The CF of the baseline scenarios for conventional and organic production were 0.82 and 0.67 kg CO2 equivalent·kg orange−1, respectively; the difference between their values was due mainly to differences in the farming subsystem. Distribution and EoL was the subsystem contributing the most to the CF (59.3 and 75.7% of the total CF for conventional and organic oranges, respectively), followed by the farming subsystem (34.1 and 19.8% for conventional and organic oranges, respectively). The confidence intervals for the CF of oranges were 0.72–0.92 and 0.61–0.82 kg CO2 equivalent·kg orange−1 for conventional and organic oranges, respectively, and a significant difference was found between them. If organic production were to reach 50% of the total exported production, the CF would be reduced by 5.4–8.4%. Conclusions: The case study and the methods used show that bootstrap techniques can help to test for the existence of significant differences and estimate confidence intervals of the mean CF. Furthermore, these techniques allow several CF sources to be combined so as to estimate the uncertainty in the mean CF estimate. Assessing the variability in the mean CF (or in other environmental impacts) gives a more reliable measure of the mean impact. [ABSTRACT FROM AUTHOR]