Environmental Engineering Reference
In-Depth Information
and identified by a highly skilled team (Dines & Murray-Bligh 1997 ). This
process provides a measure of this component (2 above) of uncertainty and
enables quality scores to be corrected for the bias caused by sorting and
identification errors (missed taxa tend to be more sensitive increasing both
metrics used, Ntaxa and ASPT) as well as identifying areas where further
training is required (Dines & Murray-Bligh 1997 ; Haase et al. 2006 ). Sampling
uncertainty (1 above) was quantified by a programme of nested, spatially
replicated sampling across a range of river types and qualities (Clarke et al.
2002 ). Long-term 'natural' variation (3 above) has recently been quantified
using temporally (and spatially) replicated datasets, taking care to separate
sampling and spatial variation from within-year variation and from between-
year variation (Davy-Bowker et al. 2007c ). The importance of natural temporal
variation varies, dependent on the time period that the assessment is intended
to represent and the time interval from reference to test sample: temporal
variation is particularly important when using fossil remains to define refer-
ence condition. The extent of temporal variation is likely to vary between river
types, with some types having more long-term temporal variation than others.
The variation from all these sources has been combined within RIVPACS to
provide simulations of the probable quality classes for any measured value, and
an output is given of the probability of the test site being in those quality
classes (Clarke 1997 ; Davy-Bowker et al. 2007c ). It should be remembered that, if
a site has a score that corresponds exactly with a class boundary, it could be
either class and, thus, there is at least a 50% probability of misclassifying it.
The WFD includes several different biological quality elements for which
there may be several metrics, each of which has associated uncertainty.
A methodology has been developed to assess the effect of the various sources
of variation and errors in the observed and the reference condition values of
one or more metrics on the overall uncertainty in assignment of water bodies
to ecological status class. The software package STARBUGS (STAR Bioassessment
Uncertainty Guidance Software (Clarke 2004 ); see Software at http://www.
eu-star.at or www.ceh.ac.uk/products/software/water.html ) uses the various esti-
mates of the components of uncertainty to generate many random simulations
of the potential metric values for a site, from which predefined metric-based
classification rules and class boundaries are used repeatedly on each simula-
tion to build up estimates of the probabilities that a particular water body
belongs to each of the WFD ecological status classes.
Whilst uncertainty may seem to be an embarrassment to biomonitoring
techniques that we should be able to do without, it should be remembered
that it affects all assessments of quality, whether they be physical, chemical or
biological, as it is not possible to sample the whole water body all the time.
A full understanding of the factors influencing uncertainty, however, can help
us design tools and sampling strategies that reduce uncertainty in the most
Search WWH ::




Custom Search