Page 184 - Materials Chemistry, Second Edition

MODELING THE AGRI-FOOD INDUSTRY WITH LIFE CYCLE ASSESSMENT 169

a result of this flow a part of the pesticides can reach groundwater and finally
the waters of rivers and wells. This phenomenon occurs when a fraction of the
pesticide is removed and dissolved in water run-off and adsorbed on particles
of eroded material. The magnitude of this fraction depends on the slope, on
soil type, amount and intensity of rainfall.
From these considerations it follows that the fate of the pollutants associ-
ated with the use of a pesticide depends on site-specific data that are partly in
contradiction with an analysis such as LCA that should be site independent.
Moreover, the determination of the amount of pesticides remaining in the soil
or their transformation products can be carried out with chemical and bio-
logical methods of analysis and with the implementation of predictive math-
ematical models, which have been developed especially in recent decades. In
the agronomic and environmental chemistry literature there are, therefore,
numerous dispersion models of pesticides in various environmental media
and even software that perform the same type of analysis, developed in vari-
ous countries. Some models are relative to the estimation of only a few effects,
such as evaporation and degradation. These models start from some basic data
concerning the characteristics of the pesticide, the characteristics of the soil,
weather site and the characteristics of the plant to be treated. As quite evident,
these data are site-specific and therefore the data obtained for the agricultural
environment of a country and sold in commercial databases very often cannot
be considered representative of other environments.
Whether dispersion models are employed or not, in any case many of the
necessary data for the LCA, needed to follow the fate of pesticides, are dif-
ficult to obtain. In other cases the solution adopted for the emission estimate
is derived from the analysis of the literature. Margni et at. (Margni, et al. 2002)
state that the fraction of the active ingredient entering the soil is assumed to
be eighty five of the total applied quantity, assuming five per cent stays on the
leaves in addition to ten per cent loss into the air while, few substances reach the
groundwater and in most cases the pesticide run-off is less than ten per cent of
the applied dose, based on literature (Audsley, et al. 1997). A more sophisticated
model developed by Hauschild et ah has been adapted to LCA (Hauschild,
2000; Birkved & Hauschild, 2006). It considers all types of dynamic behaviour
during the emission of a pesticide, regarding emissions to air, groundwater
and surface water in terms of elementary flows. Another approach to the prob-
lem is that offered by Ecoinvent in which the entire amount of pesticide used
as input shall be deemed emitted to the soil (Ecoinvent, 2010).
From the above analysis, it is clear that for the LCA implementer that is
not an expert on pesticide chemistry or on mathematical dispersion models
of pesticides, it is difficult to reach a scientific and reliable solution regarding
the estimation of emissions from pesticides. We can conclude that there are
two prevailing solutions: the first one concerns the use of dispersion models
or literature data for the estimation of pesticide emissions to air, groundwater
and surface water while the second one hypothesizes that the whole amount
of pesticide used as an input is emitted to soil. These approaches naturally lead
to different results both in terms of affected impact categories and absolute