Use of the MCI and QMCI is well established in New Zealand, possibly because, together with the SQMCI, they are the only comprehensive biotic indices based on tolerances of New Zealand macroinvertebrate taxa for assessing the health of stony streams. All regional councils that undertake SoE monitoring use the MCI and/or SQMCI/QMCI for reporting results. In this appendix we discuss the major findings from the peer-reviewed scientific literature concerning these indices.
A chapter in the "Stream invertebrate book" examines the first 50 years' use of macroinvertebrates in biological monitoring in New Zealand. Boothroyd and Stark (2000) define the term "biomonitoring" and provide a useful discussion on the theory and practice of biomonitoring, including why macroinvertebrate biomonitoring is preferred over other kinds. Some of the historical events that have influenced biological monitoring in New Zealand are detailed. Various methods used for assessing the biological condition of rivers and streams are discussed, and the use of the macroinvertebrate monitoring tool currently favoured in New Zealand, the MCI, is critically examined. Comments on the design of monitoring programmes and data analyses are provided.
Although developed originally in Taranaki for stony streams (Taranaki Catchment Commission 1984; Stark 1985), both the MCI and QMCI were found to be moderately strongly correlated with indicators of enrichment when applied to run/riffle samples from 88 rivers throughout New Zealand (Quinn and Hickey 1990). Quinn and Hickey suggested that the MCI may be a more sensitive index of water enrichment than the QMCI, because it has higher correlations with indicators of enrichment (such as total Kjeldahl nitrogen, periphyton chlorophyll 'a' and ash-free dry weight) for 88 New Zealand rivers. However, they suggested that the extra effort required (mainly in sample processing) to obtain QMCI values may be warranted where water quality changes are expected over relatively short river reaches (e.g. above and below wastewater discharges). In such situations, drift of macroinvertebrates from upstream may introduce taxa (normally in low densities) to polluted downstream sites, where they may not survive in the long term, thereby misrepresenting the "true" character of the site. Quinn and Hickey concluded that these indices were more useful indicators of water quality than species diversity, species richness and the EPT index (i.e. the number of ephemeropteran, plecopteran and trichopteran taxa). This was the first validation of these indices nationwide.
In Southland streams, Quinn et al (1992) found that the QMCI was reduced significantly by intensive grazing and channelisation, and Scott et al (1994) found that increasing intensity of pastoral land development decreased the QMCI. At 29 Northland stream sites, Collier (1995) found significant positive correlations between MCI and shade and the percentage area of native forest in the catchment, but significant negative correlations with an index of periphyton biomass and the percentage of the riparian zone in pasture.
Maloney (1995) found that spraying the herbicide triclopyr to control willows and lupins in the Ahuriri River (South Island) did not alter macroinvertebrate community composition or affect MCI and QMCI values. Quinn et al (1997a) found that pasture streams had significantly lower QMCI than native forest and pine forest catchments at Whatawhata, near Hamilton. Storey and Cowley (1997) found that native forest remnants part-way down second-order streams in pastoral farmland could cause a change from a more enrichment-tolerant macroinvertebrate fauna to one more characteristic of clean water, and that this improvement was detected by the MCI.
Collier et al (1998) evaluated the performance of biotic indices (including the MCI and QMCI) calculated from samples collected from macrophyte, sand, silt, bedrock, and wood substrates at 20 Waikato lowland stream sites. They found that %EPT and MCI were robust under contrasting sampling intensities, and, together with the QMCI, were sensitive to factors relating to water quality and catchment land use, suggesting that these indices are likely to be useful for biomonitoring in lowland stream environments. The MCI was considered particularly suitable for use with rapid bioassessment protocols in lowland streams.
Hickey and Clements (1998) criticised the QMCI because it did not detect the impacts of heavy metal pollution on macroinvertebrate communities in streams on the Coromandel Peninsula. They noted that this was because the QMCI had "incorrect tolerance scores for some taxa to heavy metals". However, this is hardly surprising since the QMCI was developed to detect organic pollution and nutrient enrichment, not metal toxicity.
Hall et al (2001) examined macroinvertebrate communities in streams and rivers dominated by native bush, agricultural or urban land uses within the Water of Leith stream catchment near Dunedin. Both the MCI and QMCI decreased progressively from native bush through agricultural to urban land use, with the QMCI exhibiting the stronger relationship. These results indicate that not only did the representation of pollution tolerant taxa increase along this gradient, but that pollution-tolerant taxa also increased in dominance.
Duggan et al (2002) examined the influence of sample size (100-, 200- and 300-fixed count) on the accuracy and variability of six invertebrate metrics (taxa richness, EPTtaxa, %EPT abundance, % dominant taxon, MCI and QMCI). They found that the MCI provided the most consistent results in terms of having low within-site variability and distinguishing differences in stream impacts between sites (although this was affected by seasonal interactions). Although their study reinforced the use of a range of metrics, they suggested that "better performing metrics such as the MCI, could be given higher weighting than some other metrics when interpreting results." Within the range of sample sizes tested, they found that richness measures (including the MCI) were sensitive to sample size, and that one should be cautious when comparing results from different studies. However, within any particular study the effect of sample size on the interpretation of MCI was not considered likely to be significant.
Duggan et al (2003) also compared the performance of 100-, 200- and 300-fixed count subsampling with coded abundance (R, C, A, VA, VVA) subsampling, and full counts for rapid assessment biomonitoring. They found 1:1 relationships between QMCI and %EPT for assessments made with both rapid assessment methods and full counts. However, they concluded that variability was greater using coded abundance than fixed counts, which they considered could lead to incorrect conclusions "on occasion". These authors used only two data sets in their analyses (from Quinn et al 2002, and Quinn and Hickey 1993), and recommended the 200-fixed count as the preferred rapid assessment sampled processing protocol (which is protocol P2 in Stark et al 2001). We have done calculations using other data sets that do not support the conclusion that fixed count is better than coded abundance, suggesting, perhaps, that the nature of the data may affect the results. The critical issue, however, that Duggan et al (2003) did not consider was the relative time and cost of the different processing protocols. In our experience, coded abundance is usually more cost-effective.
Death et al (2003) examined the effect of exotic forest logging on stream macroinvertebrate communities in Hawke's Bay. They found that the MCI and QMCI reflected the impact of forest harvesting - the main impact was considered to be an increase in fine sediment in the streambed - and that communities had not recovered to pre-harvest condition by 1.5-2.5 years post-harvest. In contrast, recovery from a natural storm event was much more rapid (five months).
Parkyn et al (2003) and Parkyn and Davies-Colley (2003) evaluated the success of stream rehabilitation by comparing agricultural streams where riparian buffers have been restored with unbuffered control reaches upstream or nearby. They found that buffer widths appeared to be closely related to stream health (QMCI), especially those greater than 10 m, although they were unsure of the mechanism for this and suggested that further study was warranted. However, recovery seemed closely related to reducing stream temperatures, suggesting that restoring in-stream communities might take many years and would be achieved only after canopy closure, with long buffer lengths and protection of upstream tributaries.
Wright-Stow and Winterbourn (2003) examined the correspondence between the MCI and QMCI using fixed-count data from 230 stream and river sites in Canterbury. The two indices ranked sites similarly (rs = 0.86), but the MCI placed most sites in the "good" and "fair" pollution classes, whereas most sites were assigned to the "excellent" or "poor" classes by the QMCI. Wright-Stow and Winterbourn concluded that either the MCI was a more conservative index, or that the boundaries between pollution classes are not equivalent. The latter reason was considered more likely, and given the difficulties inherent in defining classes based on continuous distributions and the fact that there is no way of knowing which index gives the "right" answer, Wright-Stow and Winterbourn suggested a return to fuzzy boundaries between classes as proposed initially by Stark (1985). Alternatively, when comparing large numbers of sites (e.g. in SoE monitoring), they suggested that the percentile the site of interest falls within could be stated. A site with an MCI of 130, for example, could be described as being within the top 10% of sites within the region.
Quinn et al (2004) found that riparian buffers mitigate the effects of pine plantation logging in New Zealand streams and that QMCI values were significantly lower following logging if riparian vegetation was removed too.
Maxted et al (2003) studied the effects of sample substrata, sample area, and land use on various metrics, including the MCI and SQMCI. They found that the MCI and SQMCI were sensitive to urban and rural land uses. The differences in metric scores between streams with hard-bottomed and soft-bottomed substrata were used as the basis for developing separate sampling methods for these two stream types (Stark et al 2001).
Riley et al (2003) studied the consequences for stream physico-chemistry and ecology of the agricultural (pastoral) development of tussock grassland. They found no difference in MCI between tussock, grazed tussock, and pasture catchments. However the QMCI was significantly higher in streams in pasture catchments than in tussock or grazed tussock catchments. Despite changes to the physical and chemical nature of the streams due to pastoral development no sensitive taxa had been lost and, in fact, the relative abundances of some sensitive taxa such as the mayfly Deleatidium had increased in pastoral streams. All MCI and QMCI values in the study streams were indicative of good or very good stream health.
Maxted et al (2005) studied the effects of in-line ponds on stream water quality and macroinvertebrate communities in the Auckland region. They found that ponds in soft-bottomed geology in rural catchments caused reductions in the EPT richness and SQMCI indices downstream, but in hard-bottomed geology bushed catchments there was no significant difference.
Doledec et al (2006) examined the performance of structural and functional approaches for assessing land-use effects on stream macroinvertebrate communities along a gradient of increasing agricultural development: ungrazed native tussock (UT), grazed tussock (GT), extensively grazed pasture (PA), and intensive deer and dairy farming (DD). Macroinvertebrate densities, EPTtaxa, MCI and QMCI differed very little among UT, GT, and PA sites, but densities were somewhat higher and the indices significantly lower at DD sites. The MCI was correlated significantly with the percentage of the streambed covered by sediment particles <1 mm (%FINES), dissolved inorganic nitrogen (DIN) and dissolved reactive phosphorus (DRP), whereas the QMCI was correlated significantly with %FINES and DIN. The authors concluded overall that these traditional structural measures (i.e. biotic indices) were just as effective as the species traits approach for differentiating land-use effects on their grassland stream communities. They predicted that the functional species trait approach may be more effective on a larger spatial scale, but this has not yet been tested.