The top priority when processing samples to enable the calculation of biotic indices is to have good taxonomy. After all, the tolerance values that reflect environmental health vary among taxa, so if taxa identifications are not correct, then you may end up with an incorrect assessment. Stark et al (2001) have provided quality control (QC) methods that can be used to ensure samples are processed to a high standard.
Identifying aquatic macroinvertebrates is not easy despite the existence of good keys such as Winterbourn et al (2006), and there are plenty of traps for the inexperienced. The increasing adoption of QC procedures in sample processing since they were provided by Stark et al (2001) has highlighted that accurate taxonomy in sample processing cannot always be taken for granted. There has also been some confusion over exactly how best to undertake the sample processing QC required by the protocols, so we provide further clarification here.
It is important to remember that the overall objective of QC is to ensure data quality, and because QC is an overhead cost we believe that this objective should be achieved as cost-effectively as possible. This entails reconciling any differences between the identifications made by the original sample-processing laboratory and the laboratory chosen to undertake QC (which should be separate agencies).
We recommend the following procedure.
It may seem more objective to require the QC laboratory to undertake QC without having a copy of the data generated by the processing laboratory. It is possible, for example, that an inexperienced person could undertake QC simply by agreeing with the identifications provided on the data sheets, which would not be a QC check at all. However, from our experience, both here and in the United States of America, when QC is undertaken blind (i.e. without the data provided by the sample processors) it is extremely difficult to reconcile any differences in identifications (without someone having another look at the specimens, which adds extra time and cost). Furthermore, given the very small size of some specimens, it is easy for the QC laboratory to overlook taxa that should be in the vial, leaving doubt as to whether they were there and were missed, or whether the sample processors forgot to put specimens into the vial. When the QC is undertaken blind like this, the result can be two slightly different lists of taxa for each sample and an additional step of reconciliation is required. Since a reconciled data set is the aim of QC, we believe that the reconciliation should be part of a one-step QC procedure undertaken by the second laboratory, leaving a third stage only if there is disagreement over the identification of specific taxa. In such cases, these can be provided to an agreed independent expert, as stated in the protocols (Stark et al 2001).
QC is not required on every batch of samples processed, especially when the processing laboratory has a proven track record of excellent performance. Even then, however, QC should be undertaken every now and then to ensure that high-quality work is being maintained.
The value of reference collections (i.e. a set of vials containing clearly labelled identified examples of different macroinvertebrate taxa) in QC should not be overlooked, especially if reporting is to be based on the MCI or MCI-sb (which require only presence-absence data). Although examining a reference collection containing all taxa identified from a particular batch of samples does not evaluate the complete processing performance of the processing laboratory, it is the most cost-effective way to confirm the taxonomy and resolve any disputes over identifications.
The MCI was developed initially in the early 1980s (Stark 1985), shortly after the publication of Winterbourn and Gregson's (1981) landmark first edition of the "Guide to aquatic insects of New Zealand". Although the Guide is a huge aid to better identifications, it does require some experience and training to use reliably. This was one of the reasons why it was decided to develop the MCI based on generic (at best) taxonomy. This appears to have been a wise decision given that the two most recent editions of the Guide have reverted to generic keys (except for the Simuliidae) "because so many described species are unknown as larvae" (Winterbourn et al 2006). Stark (1985) also found that an MCI based on family-level data was not very sensitive and could distinguish only gross pollution from everything else. Generic-level taxonomy was adopted because there was sufficient sensitivity for assessing stream health at this level of taxonomy, and also because it was more cost-effective and practical than species-level taxonomy.
There is no doubt that MCI-level taxonomy has become the norm in New Zealand, and, in general, this has proven suitable for bioassessments and SoE reporting based on the MCI and its variants. The fact that most data sets are identified to the MCI level does mean, however, that we don't have as many data identified to the species level as would have been the case if the MCI had not constrained the identifications. It could well be that an MCI based on species-level identifications (where practical) may perform even better than the existing generic-level indices. However, to develop a species-level MCI, species-level data are required, which, in general, are not being collected.
Wright-Stow and Winterbourn (2003) evaluated the effect of taxonomic resolution on biotic index performance by comparing MCI and QMCI values determined using ordinal-level taxonomy with the conventional MCI and QMCI. Ordinal-level tolerance values were obtained by averaging MCI scores for 10 insect orders and five other higher taxonomic groups. They found that biotic indices based on coarse-level taxonomy ranked the health of streams in Canterbury in a comparable way to the MCI and QMCI.