LCA Food 2012 in Saint Malo, France! - Manifestations et colloques ...

PARALLEL SESSION 2C: QUANTIFICATION AND REDUCTION OF UNCERTAINTY 8 th Int. Conference on LCA in the
Agri-Food Sector, 1-4 Oct 2012
Six inventories for Brazilian soybeans highlighted in da Silva et al., (2010), with the use of Schmidt
(2007) to represent Dalgaard (2008), and the addition of FAOSTAT data (http://faostat.fao.org) were included
in this study. Each inventory flow was sourced to its origins in order to avoid double counting of
cross-references and to determine the actual representativeness of each inventory flow. The resulting
weighted means and inherent standard deviations were then calculated amongst the original values (excluding
cross-references and assumptions). Given space limitations, we here only present and discuss the inventories
for fertilisers, fuel and resulting yields.
3. Results
The inventories supporting the six different soybean inventories and FAO data are reported in Table 1.
All studies rely upon a divergent set of inventories describing an identical process. Many of the values are,
moreover, cross-references of previous publications. The amount of diesel used, for example, stems back to
two single primary data points, Ostermayer 2002 in Jungbluth et al., (2007) and Cederberg (1998). Ecoinvent
adopts the average of these two publications (55.25 kg assuming 0.85 kg litre -1 diesel) and refers to an FNP
Agro report from 2000 for fertiliser use (Jungbluth et al., 2007). Schmidt (2007) on the other hand adjusted
Argentinian primary data to a Brazilian scenario. This study was therefore not considered as a primary data
source for Brazilian soybeans. Da Silva et al., (2010) unfortunately failed to report on inventory values used
and could therefore not be included. Most studies use the average yield over five years reported by FAO in
order to eliminate any annual fluctuations.
Table 1: Unit process data for six different articles and FAOSTAT data for one hectare of Brazilian soybean
crops
FAOS Cederberg
Lehuger Cavalett
Ecoinvent Cederberg TAT and Fly- Schmidt et al., and Or-
2.2 1998 2002 gsjö 2004 2007 2009 tega 2010
unit Database Report FAO Report PhD thesis Article Article
Economic inputs
Diesel kg 55.25 51 - 55.25 44.63 - 55.25
Nitrogen, as N kg 3.11 0 3.73 8 0 8 0
Phosphorus, as P kg 26.18 17.46 28.8 31 0 31 33.8
Potassium, as K
Economic outputs
kg 24.9 33.2 51.46 57 20 57 65.4
Soybeans kg 2 544 2 200 2 613 2 500 2 680 2 500 2 830
With all cross-references and non-primary data points excluded, diesel was narrowed down to two primary
data sources and fertilisers to five values. The economic input data were then standardised to one tonne
of soybeans according to reported yields in order to aggregate values into a single weighted mean (Table 2).
Table 2: Reported inputs of diesel and fertilisers in the only studies presenting real novel field data, and their
accompanying standard deviations in brackets as estimated using the NUSAP approach. The calculated
weighted geometric mean and the inherent geometric standard deviation amongst the values are presented in
Eq. 1
the rightmost columns. All values are normalised to the production of one tonne of soybeans in Brazil.
FNP 2000/
Ostermayer
2002
Cederberg
and Flygsjö
2004
Cavalett
and Ortega
2010
Cederberg
Wt. Inherent
1998 FAO 2002
mean STDEVg
Diesel (kg) 23.4 (1.34) 23.2 (1.31)
23.3 1.01
Nitrogen (kg)
Phosphorus
1.2 (1.27) 0 (1.24) 1.4 (1.22) 3.2 (1.27) 0 (1.10) 1.4 3.48
(kg) 10.3 (1.27) 7.9 (1.24) 11.0 (1.22) 12.4 (1.27) 11.9 (1.10) 10.6 1.19
Potassium (kg) 9.8 (1.27) 15.1 (1.24) 19.7 (1.22) 22.8 (1.27) 23.1 (1.10) 17.6 1.43
The inherent variability for yields was calculated using the yields reported by FAO between 2001 and
2005, the same time range adopted by ecoinvent. Calculating the geometric mean over these five years
yielded an average harvest of 2 537 kg ha -1 and an inherent geometric standard deviation of 1.11.
195