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

PARALLEL SESSION 4C: CROP PRODUCTION SYSTEMS 8 th Int. Conference on LCA in the
Agri-Food Sector, 1-4 Oct 2012
Table 1. Characteristics of the 12 selected field representative for against the season tomato in Benin
Field
codes
414
Geo
Loc
Irr syst
Field
area (m²)
Duration
(days)
Yield
(kg.ha -1 )
Nitrogen
rate (kg.ha -1 )
Number of
pesticide appli-
cations
Water supplied
(m 3 .ha -1 )
P06 Cot Man 306 118 4902 2684 29 7200 Sand
P07 Cot Man 126 104 12452 2807 4 6360 Sand
P10 Cot Sprink. 196 104 0 905 10 2665 Sand
P12 Cot Man 546 111 7875 755 17 10400 Silt
P28 Pa Man 922 110 8 1328 7 586 Silt
P17 Pa Hose 169 102 127 665 5 4515 Sand
P19 Pa Hose 1915 56 0 1389 12 919 Silt
P38 Pa Hose 895 88 10 72 11 3759 Sand
P39 Pa Hose 760 103 5662 507 17 10581 Sand
P33 GPP Hose 2200 104 4498 2258 10 771 Clay
P37 GPP Hose 576 113 21163 1595 5 5318 Sand
P40 GPP Hose 963 113 1703 60 1 7048 Sand
2.2. Life-cycle assessment methodology
An ISO-compliant LCA was performed to compare the environmental impact of the 12 fields selected.
The functional unit assigned for all fields was one hectare of tomato production in agreement with the farmers’
strategy to give value to an area through the production of a commercial product. The fields were assessed
from-cradle-to-field-gate. Input production and infrastructures were included, while transport and
end-of-life were not taken into account. Data on the production of agricultural inputs was taken from the
Ecoinvent database (v2.2). Field emissions were estimated using the best available methods for such a specific
context. Ammonia (NH3) emissions following the application of mineral fertiliser nitrogen were estimated
using emission factors from the ECETOC report (ECETOC 1994). Emission factors of group I (high
NH3 emission potential due to high temperature and pH) were chosen to be representative as much as possible
for the tropical context with high temperatures. For volatilisation from poultry manure, a 20% emission
factor was selected from the Ecoinvent guidelines (Nemecek and Kägi 2007). The nitrogen content of poultry
manure was set to 3% according to the organic fertilisation guide of the Reunion Island (Chabalier et al.,
2006) taken as representative for a tropical context. The emissions of nitrous oxide (N2O) and nitric oxide
(NOx) were estimated following the Tier 1 methodology of the IPCC guidelines (IPCC 2006) including both
direct and indirect emissions. Due to the lack of a suitable method for horticultural crops in the Tropics, we
estimated nitrate leaching as 30% of total N applied (IPCC, 2006). Emissions of carbon dioxide (CO2), phosphorus,
and pesticides were estimated following methods from Nemecek and Kägi (2007). The SimaPro
(v7.2) software was used to analyse the environmental impact through the 13 impact categories defined by
ReCiPe (Goedkoop et al., 2009). The names and units of the impact categories are given in Table 2.
2.3. Two-step statistical analysis
To describe the diversity of environmental impacts for the studied cropping systems, we used a two-step
statistical treatment. The first step was to transform the impacts categories into non-correlated variables with
a Principal Component Analysis (PCA). The number of variables was reduced into less dimensions representative
for most of the variability observed in the population. Once individual fields were spread on this multidimensional
space, the PCA method allowed identifying one most representative field located closest to the
intersection of the dimensions and, fields responsible for the variability located at the edge of the cloud.
Fields were then grouped into clusters corresponding to cropping system types using an Agglomerative Hierarchical
Clustering (AHC) algorithm, in which the principal components of the PCA were used as input variables.
As PCA allows the calculation of one representative field for the whole population, AHC allows identifying
one representative field (called paragon) for each cluster (or type). These paragons are located near
the centroid of each cluster. The population can therefore be analysed using those typical individuals, avoiding
any aggregation or averaging often leading to discrepancies with the reality and the system logic of the
cropping system. Such a two-step statistical treatment is also called a goal-oriented typology and has been
used in studies from the agronomic discipline to analyse the diversity of cropping systems and farms
(Poussin et al., 2008). Our analysis was strengthened through the testing of the distribution of additional
qualitative or quantitative variables (geographical, irrigation system, soil types or cropping systems components)
using Kuiper’s test against dimensions and clusters. These variables are summarized in Table 1.
Soil
type