Page 122 - Freshwater-Biology-and-Ecology-Handbook

CHAPTER 2 13 13

QUALITY ASSURANCE OF LABORATORY

ANALYSIS AND DATA HANDLING AND

DATA QUALITY STANDARDS

Error is inevitable in ecological survey data, but it can (including the UK River Communities team). STAR/AQEM
be minimised, and also measured by numerical quality samples were collected according to the method described
assurance procedures so that it can be considered in data in Section 15 and national samples were collected by a
analysis and interpretation. Measuring errors ensures that variety of methods according to different national protocols
data is interpreted correctly by enabling real ecological used in 2002. The results showed the considerable range
differences to be distinguished from those caused by error. in quality that is typical of laboratories that have not been
audited before, and therefore their analysts are unaware
Clear, unambiguous and comprehensive instructions are of what they are missing. Experience from the UK audit
an essential step in quality assurance. The Environment showed that most errors stem from not noticing the
Agency has produced a series of Operational Instructions presence of specimens in the sorting tray rather than from
that cover almost every aspect of its work. Those relevant to misidentifications. It also showed that quality tends to vary
invertebrate sampling and analysis are mentioned here. much more between laboratories than within them. As a
result, the UK’s environment protection agencies audited
Unless an error is highlighted, it is often impossible for an each of their laboratories by having 20 random samples
individual to know that they are making mistakes. Active re-analysed from each laboratory each year.
feedback is at the core of quality assurance procedures.

It is impossible for a human to sort samples without error. Table 2.9
Small inconspicuous specimens are easily missed. To Results of the family-level sorting audit undertaken in the
assess sorting error, the environment protection agencies STAR project
in the UK undertook audits in which a number of samples
are re-sorted by experts. For many years, these were METHOD
undertaken annually by the River Communities team that
analysed the RIVPACS reference samples. The audit Partner STAR/AQEM NATIONAL
measured analytical quality against the RIVPACS baseline. Mean Range Mean Range
This enabled analytical quality to be adjusted to that used
by the RIVPACS model, so that observed results were A 1.00 0 - 2 0.83* 0 - 3
comparable to the expected results predicted by RIVPACS. B 0.83 0 - 3 0.83 0 - 1
C 4.17 0 - 11 3.50 1 - 7
The first time that most analysts, including experienced
ones, have their samples audited they are often surprised D 1.83 0 - 3 4.83 2 - 11
at the number of errors. Unless someone points out E 1.67 1 - 3 - -
unintended errors to analysts, they may be unaware of
them. This was demonstrated in the STAR project (Table F 3.50 0 - 9 3.17 0 - 6
2.9), in which laboratories across Europe were audited G 0.33 0 - 1 0.83 0 - 2

H 4.25 3 - 6 8.25 5 - 12
I 1.00 0 - 2 2.75 2 - 3
J 1.00 0 - 3 8.83* 3 - 18
K 4.33 2 - 7 6.33 1 - 11
L - - 5.33* 2 - 11
M 1.33 0 - 3 1.50 0 - 3
N 2.67 0 - 4 5.67* 1 - 13
O 3.17 0 - 10 3.17 0 - 7

122 | Freshwater Biology and Ecology Handbook
–