28.12.2012 Views

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

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

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

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

PARALLEL SESSION 2C: QUANTIFICATION AND REDUCTION OF UNCERTAINTY 8 th Int. Conference on <strong>LCA</strong> <strong>in</strong> the<br />

Agri-<strong>Food</strong> Sector, 1-4 Oct <strong>2012</strong><br />

2. Discussion<br />

Limited data availability rema<strong>in</strong>s a generic problem <strong>in</strong> LCIs, ma<strong>in</strong>ly due to the great resources needed to<br />

collect primary data. Where datas<strong>et</strong>s are available, however, they often describe limited geographical areas<br />

where practices <strong>in</strong> e.g. agriculture differ depend<strong>in</strong>g upon micro-climate and soil characteristics. The <strong>in</strong>consistencies<br />

<strong>in</strong> the <strong>in</strong>ventories presented above may therefore be the result of six alternative soybean processes<br />

rather any nationwide averages. In search for such more general processes, the m<strong>et</strong>hodology presented here<br />

enables secondary foreground data to be critically evaluated. When averag<strong>in</strong>g unit process data <strong>in</strong>to, for example,<br />

country-wide averages, different economic and environmental flows may also be used complementary<br />

to produce b<strong>et</strong>ter characterised processes. This may be values for water use reported <strong>in</strong> some studies,<br />

while <strong>in</strong>ventories of pesticide use only are available <strong>in</strong> other studies. Moreover, the proposed approach enables<br />

unit process data to move beyond po<strong>in</strong>t values and m<strong>in</strong>imizes the use of unrepresentative param<strong>et</strong>ers. It<br />

also helps to identify areas of great uncerta<strong>in</strong>ty and <strong>in</strong> the process allows for <strong>in</strong>herent variability to be calculated.<br />

With many <strong>in</strong>centives to <strong>in</strong>clude quantified uncerta<strong>in</strong>ties <strong>in</strong>to <strong>LCA</strong> results, an <strong>in</strong>itial step will be to def<strong>in</strong>e<br />

basic uncerta<strong>in</strong>ty param<strong>et</strong>ers <strong>in</strong> the LCI phase, allow<strong>in</strong>g for different k<strong>in</strong>ds of uncerta<strong>in</strong>ty tests/m<strong>et</strong>hods to be<br />

applied (e.g. not only Monte Carlo analysis). Includ<strong>in</strong>g most sources of uncerta<strong>in</strong>ty <strong>in</strong> these param<strong>et</strong>ers is<br />

crucial, as they will d<strong>et</strong>erm<strong>in</strong>e the application and outcome of any later implemented test/m<strong>et</strong>hodology. With<br />

a wide range of sources and types of uncerta<strong>in</strong>ties, we here<strong>in</strong> highlight the dist<strong>in</strong>ction b<strong>et</strong>ween <strong>in</strong>herent standard<br />

deviations and representativeness. While the latter easily can be estimated, the former are not always<br />

available for unit processes. Even when available, <strong>in</strong>herent uncerta<strong>in</strong>ties are often neglected, as <strong>in</strong> the above<br />

example of soybean yield averages over several years. When calculated us<strong>in</strong>g the approach proposed above,<br />

<strong>in</strong>herent standard deviations often exceed those of representativeness. For example, the geom<strong>et</strong>ric standard<br />

deviation amongst the reported uses of potassium <strong>in</strong> the studies discussed above, amounts to 1.43 compared<br />

to an estimated geom<strong>et</strong>ric standard deviation of 1.1-1.27 for representativeness. The <strong>in</strong>herent geom<strong>et</strong>ric standard<br />

deviations of nitrogen showed even greater divergence, but these are somewhat distorted by the reported<br />

null values. Temporal <strong>in</strong>herent variability is also relevant for food production systems, where the<br />

yield average over five years, reported by FAO, exhibit a geom<strong>et</strong>ric standard deviation of 1.11. This is<br />

largely the result of a close l<strong>in</strong>k b<strong>et</strong>ween food production systems and stochastic natural events. Agriculture<br />

practices also exhibit large model uncerta<strong>in</strong>ties when calculat<strong>in</strong>g field emissions, which add further <strong>in</strong>herent<br />

uncerta<strong>in</strong>ty to this field of <strong>LCA</strong> research.<br />

Interconnected (co-variation) param<strong>et</strong>ers (e.g. the amount of fertiliser used and result<strong>in</strong>g N2O emissions)<br />

were here not accounted for, result<strong>in</strong>g <strong>in</strong> an overestimation of the uncerta<strong>in</strong>ty. In the meantime, <strong>in</strong>herent<br />

uncerta<strong>in</strong>ties are not always considered <strong>in</strong> the background database, thereby underestimat<strong>in</strong>g the degree of<br />

overall uncerta<strong>in</strong>ties. Neither were the temporal correlations embedded <strong>in</strong> the database representative of present<br />

time, as they are benchmarked at the time of release of the database. With some NUSAP categories be<strong>in</strong>g<br />

quantifiable (e.g. less than 3 years), others rema<strong>in</strong> open for <strong>in</strong>terpr<strong>et</strong>ation which may also result <strong>in</strong> <strong>in</strong>terpr<strong>et</strong>ation<br />

errors.<br />

5. Conclusion<br />

Inventory data rema<strong>in</strong>s a major source of uncerta<strong>in</strong>ty for <strong>LCA</strong> results, where LCIs describ<strong>in</strong>g the same<br />

production system often experience large differences at the unit process data level. Many of the values used<br />

are also cross-references which som<strong>et</strong>imes date back over ten years. Scrut<strong>in</strong>y is therefore needed when develop<strong>in</strong>g<br />

process datas<strong>et</strong>s, as is b<strong>et</strong>ter report<strong>in</strong>g of the orig<strong>in</strong>s of data, standard deviations around means and<br />

identification of known sources of uncerta<strong>in</strong>ty. Us<strong>in</strong>g the standardised decision tree for data sourc<strong>in</strong>g and<br />

weight<strong>in</strong>g means amongst studies can therefore result <strong>in</strong> more representative and rigid results.<br />

Quantitative uncerta<strong>in</strong>ties need to be <strong>in</strong>cluded <strong>in</strong> <strong>LCA</strong>s <strong>in</strong> order to statistically validate conclusions and<br />

strengthen the credibility of <strong>LCA</strong> results. The here<strong>in</strong> proposed approach presents one way for entail<strong>in</strong>g uncerta<strong>in</strong>ty<br />

param<strong>et</strong>ers that complement each other (<strong>in</strong>herent and representativeness), rather than replac<strong>in</strong>g each<br />

other. A standardised vocabulary is also needed for further uncerta<strong>in</strong>ty discussions with<strong>in</strong> the field of <strong>LCA</strong>.<br />

Here we have e.g. shown that a notable dist<strong>in</strong>ction is needed b<strong>et</strong>ween <strong>in</strong>herent uncerta<strong>in</strong>ty and variability,<br />

and representativeness. <strong>Food</strong> production systems are especially sensitive to <strong>in</strong>herent deviations, especially as<br />

yields which are subject to annual fluctuations often represent the functional unit.<br />

6. References<br />

Cederberg, C., 1998. Life Cycle Assessment of Milk Production – a Comparison of Conventional and Organic Farm<strong>in</strong>g. SIK Report<br />

643. Gothenburg.<br />

196

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!