Page 149 - Materials Chemistry, Second Edition

3.4 Procurement, Origin and Quality of Data 133

In particular, chemicals are candidates for estimations 161) because of their very
large number (in the European Union more than 100 000 substances are registered
according to EINECS and REACH). Data procured by estimation have to be tagged
in the inventory and have to be discussed during the interpretation phase. ISO
standards 14040 and 14044 contain several strict regulations for the documenta-
tion of data quality and possible effects of estimations on the final results (see
Chapter 5).

3.4.5
Data Quality and Documentation
One danger in the use of electronic data bases is the difficulty to reconstruct
underlying assumptions of the data generation. In other words, a judgement
of their quality, 162) (original data, average values, estimation, etc) and on their
suitability for a specific study is difficult. Even original data may contain measuring
and allocation errors that have to be documented. Therefore, commissioned by
SPOLD, a uniform data format was developed, to facilitate the electronic data
exchange on the Internet. 163) It was conceived as a network of users and providers of
databases for the exchange of data in the same format. It has been pointed out that,
as a start, the SPOLD format is a transfer format, which may not yet necessarily
represent the best structure for databases. Recent versions of SPOLD are being
used in modern data collections (see below).
LCA data records can only with difficulty be evaluated statistically. This was
already shown on a SETAC Workshop on data quality in Life Cycle Assessment in
Wintergreen, Virginia, USA. As previously illustrated, the procurement of suitable
data is a central problem for LCA. Different options for the procurement such as
those discussed in Section 3.4.4, imply highly varying data quality, which usually
cannot be represented by indicating average values and mean deviations. These
representations are practically only suited for original measurements on single
unit processes. Many of the data used in the inventory can be generic data or have
already been weighted, averaged, aggregated, and include allocations and cut-off
rules applied for their procurement with little transparency to the user.
As the reliability of LCA-results depends considerably on the quality of input data,
questions of quality have been frequently discussed in recent years. 164) It cannot yet
be foreseen, which data quality model will be generally accepted (if a one-fits-all
solution is possible at all). It seems to be certain that a transparent description of
data origin and certain quality criteria remain an important issue in data quality
management. This requires a uniform data format, which was first postulated by
the SPOLD workgroup ‘Promoting Sound Practices’. The paper format converted

161) Bretz and Frankhauser, 1996; Geisler, Hofstetter and Hungerb¨ uhler, 2004.
162) Fava et al., 1994.
163) Singhofen et al., 1996; Hindle and de Oude, 1996; Bretz, 1998; http://www.spold.org.
164) Fava et al., 1994; Chevalier and Le T´ eno, 1996; De Smet and Stalmans, 1996; Kennedy,
Montgomery and Quay, 1996; Coulon et al., 1997; Fernandez and Le T´ eno, 1997; Kennedy et al.,
1997; Huijbregts et al., 2001; Ross, Evans and Webber, 2002; Beaufort-Langeveld et al., 2003;
Ciroth, Fleischer and Steinbach, 2004; Sonnemann and Vigon, 2011.