• About this site|
• Site index|
• Contact us|
• Jobs

A User Guide for the Macroinvertebrate Community Index

Acknowledgements

Part 1: Background to Biotic Indices

Part 2: Guidelines for using the MCI, QMCI and SQMCI

1 Applying MCI indices in different freshwater environments

2 Taxonomy

3 Monitoring and reporting

4 Design of monitoring programmes

5 Detecting trends

6 Reporting the MCI

References

Appendix 1: Calculating the MCI: Excel Macros and User-defined Functions

Appendix 2: Use of MCI, QMCI and SQMCI Throughout New Zealand

4 Design of monitoring programmes

To make sure you achieve what you set out to do, it is essential that monitoring programmes are well-designed. In particular, the methods used for monitoring should be linked to the study or management aims so that the programme will:

• provide useful information for water managers
• be scientifically robust, to satisfy the ecologists and statisticians
• be cost-effective, to enhance the security of ongoing funding.

Data that do not help interpretation, or have no proven value for environmental managers, should not be collected (Likens 1983). The statistical analyses used should be envisaged as part of the study design - not an after-thought once data are collected.

4.1 Sample site selection

Study aims (i.e. data requirements) and budgetary considerations dictate the selection of the sample sites. If an upstream versus downstream (or control versus impact) design is required, then sample site selection can have a major impact on the results. If you require an assessment of differences in water quality (say, as a result of a discharge), then it is essential that the sites are as similar as possible in all other physical features, because habitat differences between sites can lead to differences in MCI values that are unrelated to the health, enrichment or pollution status of the stream ecosystem. It is crucial that the stream substrates to be sampled are similar, but the degree of exposure of a site to sunlight can also be important. One site in the shade for most of the day (where the stones are clean), and the other in full sunlight (where algae proliferate) will invariably have different macroinvertebrate communities. In such cases a false impression of the effects of an impact are highly likely.

The highest MCI values generally are recorded from good stony riffle habitat, with lower values associated with deeper water, slower current velocities, and/or finer bed sediments (in runs and pools). Increasing levels of periphytic algae on the streambed will also reduce MCI values. This is because conditions become less favourable for taxa that require clean conditions (such as mayflies, stoneflies, and most caddisflies) and more favourable for taxa that prefer living in algal mats (such as worms, snails, and chironomids). In soft-bottomed streams, large woody debris appears to be the habitat that is equivalent ecologically to stony riffles, where the most pollution-sensitive taxa exist and the highest MCI values will be recorded (Maxted et al 2003).

It is almost inevitable that there will be differences in substrate, current velocity, water depth, and aspect (i.e. the orientation of the reach in relation to the sun) between sites that may affect monitoring results. However, one approach that largely overcomes the influence of between-site physical habitat differences on biotic index values is to monitor changes over time by sampling at the same place one to four times a year. This approach is particularly useful for SoE monitoring or consent compliance monitoring, where one wants to know whether conditions are being maintained, improved, or are deteriorating. The long-running biomonitoring programme at Kapuni (Taranaki region, North Island) uses this approach, with time-series analysis to examine trends in MCI (and species richness) at monitoring sites (Stark 1993a). Monitoring at Kapuni began in September 1981 and remains ongoing, with sampling four times per year.

Although Stark (1993b) found that water depths, current velocities and substrata commonly encountered in stony riffles appear to have little impact on index values, we recommend limiting sampling in stony streams to water depths of 0.1-0.4 m, current velocities of 0.2-1.2 ms-1, and substratum median rock diameters of 60-140 mm, where possible.

There are a number of different ways of selecting sample sites for SoE monitoring. Perhaps the most intensive approach is the US EPA's EMAP, which involves placing a grid over the study area and selecting sampling sites within each grid systematically from a random start location (Herlihy et al 2000). This probabilistic sampling design invariably results in a very large number of sampling sites in order to satisfy statistical requirements, assumes no prior knowledge about the nature of the sampling sites, and is very expensive.

In our view, the EMAP approach is not an efficient use of resources when you already have some information about the streams and rivers within the region. A stratified sampling design with sites classified into categories (e.g. based on stream type, source of flow, and land use) will prove much more cost-effective. The River Environment Classification system could provide a more efficient basis for site selection, with sites replicated within the dominant classes within the region (Snelder and Biggs 2002; Snelder et al 2004). Alternatively, a stratified sampling design based on representative stream/land-use types across the full disturbance gradient could be adopted, as it was for Auckland Regional Council's SoE monitoring programme (which provided a superb data set for developing the MCI-sb and for SOE reporting).

4.2 Sampling frequency

Expanding on the general recommendation given above, sampling frequency is determined by the study objectives, modified by cost considerations (Resh and McElravy 1993). New Zealand stream faunas show less seasonal variation in species presence than is seen in the Northern Hemisphere (Towns 1985), and absence of key species or segments of the fauna is related more to environmental disturbance than life-history patterns. Towns (1985) suggested that one summer and one winter sampling should be sufficient to provide a close approximation of potential species richness at a site. For most biomonitoring programmes, sampling one or two times per year is likely to provide a reasonable balance between cost and ensuring that in-stream conditions are acceptable. If time-series analyses are required, then seasonal sampling and replication may be advisable for the first few years to build up a picture of biotic index variability, or to establish the reference condition if sampling begins before the exercise of a discharge consent. After this, sampling frequency can be reduced to once or twice per year, particularly for long-term or SoE monitoring, where year-to-year trends are of interest.

Sampling macroinvertebrate communities more frequently than seasonally is unlikely to be appropriate for SoE monitoring, due to the cost and because macroinvertebrate communities are a product of their environmental experience over the past weeks and months. However, more frequent sampling might be appropriate for more intensive special investigations.

From a sample processing viewpoint, spring sampling has some advantages because most aquatic insects are present as large, later instar [Instar = a stage in the life of arthropods between two periods of moulting (shedding of the exoskeleton in order to grow).] larvae or pupae, and are more easily identified. Most keys (e.g. Winterbourn et al 2006) are based on final instar larvae, and caddisfly pupae often develop adult characteristics (while retaining larval sclerites), enabling specific identifications to be made. Summer can be the preferred season for SoE monitoring, however, because that is when flows are lowest, water temperatures are highest, and aquatic communities are under most stress. The summer season was selected in the Auckland region because many catchments are small (<1000 ha) and dominated by small streams that dry up in late summer. Sampling during the dry late summer ensures that the streams are perennial and excludes intermittent streams affected by frequent desiccation. However, sample processing can be tedious due to the presence of many early instar larvae, which can be difficult to identify.

So, what is the preferred season for undertaking SoE monitoring? Preliminary investigations into seasonal variability (JD Stark, unpublished) suggest that spring and winter MCI values may be higher, on average, than those recorded in summer or autumn. The difference appears to be small (perhaps 4−5%), but often is statistically significant. However, given that there is approximately ±10% error in estimating the MCI from single hand-net samples (Stark 1998), it could be argued that seasonal variability is not a major cause for concern.

4.3 Sampling methods

Standard protocols for macroinvertebrate sampling in wadeable streams have been available since late 2001 (Stark et al 2001). These protocols were developed in association with regional council personnel and aim to promote standardised methods by building on (rather than overturning) current practice. As a result, the standard protocols have experienced wide acceptance.

4.4 Replication

Replicate samples are collected for one or more of the following reasons:

• to obtain sample mean values of indices such as the MCI and a measure of variance (e.g. standard deviation, standard error)
• to meet requirements of statistical significance tests for detecting differences between sites or times
• to increase the sampling effort and collect more taxa from the site.

In general, increasing replication improves the precision of estimates of biological indices, taxa richness (expressed as number of taxa per sample), overall macroinvertebrate densities, or densities of individual taxa. Note, however, that precise estimates of the densities of rare taxa may require collecting hundreds of replicates (see Elliot 1977).

Whether or not sample replication is necessary (or even possible) depends on the study objectives and budget or time constraints. We believe that sample replication is essential when undertaking robust scientific research programmes and intensive AEEs, but that for routine ongoing compliance monitoring or SoE monitoring, sample replication is optional and, in most cases, unnecessary, not only because routine biomonitoring programmes should be as cost-effective as possible (with additional investigations initiated only if routine monitoring detects a problem), but also because single samples provide reasonably precise estimates of biotic index values. Single hand-net samples provide estimates of MCI and MCI-sb that are less than ±12% of average (100) index values (Stark 1998; Stark and Maxted 2007), which we believe is more than adequate for SoE monitoring. Replication to define the variance of index scores from single samples under local conditions is recommended and should cover a range of land-use and habitat quality conditions. Replication need only be done once when sampling methods are established, and then updated periodically when new site types are added to a sampling network. Alternatively, replicate samples collected at a certain percentage of sites (e.g. 5%) would gradually build such a data set over time.

Knowledge of index variability among replicate samples is essential if statistically significant differences between samples are to be detected. The number of replicate samples collected is inevitably a compromise between the cost (mainly of sample processing) and the desired sensitivity of the monitoring programme. If we expect a given site to have an MCI around 100, and hope to detect statistically significant differences in index values within ±10% of this value, then two hand-net samples or five Surber samples are likely to be required (see Table 5 in Stark 1998). If detection of a ±20% change is acceptable, a single hand-net sample and duplicate Surber samples would suffice. Table 5 in Stark (1998) also reveals the cost-effectiveness of the SQMCI compared with the QMCI: only three hand-net samples compared with eight Surber samples are required to achieve a detectable difference of 0.48 index units (which represents <±10% of an average index value of 5.0).

Once replicates have been collected, between-site differences can be assessed using standard statistical procedures such as ANOVA and t-tests. Information in Stark's (1998) Table 5 is useful for determining whether index values from single samples are likely to be significantly different.

4.5 Other factors that can affect monitoring results

When designing monitoring programmes it is important to consider the influence of any factors that may confound the interpretation of monitoring results. We have already considered habitat differences between sample sites in Section 4.1 above. However, flow variability can also have a profound effect on bioassessments.

The influence of flow variability on biotic index values is under investigation as part of FRST-funded research in collaboration with NIWA's Water Allocation Research programme (JD Stark, unpublished). Exploratory data analyses have indicated that increased MCI values may be significantly correlated with flood events, and that extended periods of low flow (or periods without significant freshes) are related to lower MCI values. At these extremes, MCI values will be higher (soon after floods) or lower (during prolonged low flows) than the annual mean MCI. It has been suggested that high flows may elevate MCI values by washing higher-scoring taxa from "better" habitat upstream and/or by flushing away algal proliferations (which generally are inhabited by communities of lower-scoring taxa). Floods five or six times the magnitude of the preceding base flow appear to have significant influences on MCI values. This is consistent with the findings of Biggs and Close (1989) for periphyton communities.

Most councils try to factor out the effects of significant floods by delaying sampling until a set time after the last significant flood - two weeks is common. This gives the stream time to recover from the immediate effects of the flood, so that it will be closer to its "average" state. Most councils do not, however, have a similar rule that avoids sampling under extreme low flow conditions. Variability in flow over time does not affect synoptic surveys or the interpretation of compliance biomonitoring data collected on a single day (because all sites normally will have experienced similar flows), but it can affect comparisons between times or SoE monitoring programmes if it takes several weeks to sample all monitoring sites within the region.

4.6 What other environmental data should be recorded?

This section focuses on SoE monitoring, although our recommendations may apply also to other forms of monitoring (e.g. compliance, AEE, and biodiversity).

The fundamental aim of SoE monitoring is to collect data that will provide robust assessments of stream health or condition. Macroinvertebrate sampling alone can achieve this aim, but it often is desirable to collect additional information to identify the causal environmental factors. Only when these are understood can water managers implement policies or actions designed to improve stream condition. For example, an ecologist might report an MCI of 90 for a stream, which indicates only "fair" condition or "probable moderate pollution". Clearly, there is room for improvement, but this information alone is not sufficient for any action to be taken. Knowing that chironomids dominate the community and that the stream bed is covered in thick periphyton is helpful, but still not enough to suggest remedial action unless we understand what has caused the problem. If we also know that this river is in a rural setting we might suspect that the periphyton proliferation (with its associated macroinvertebrates) might be the result of nutrient enrichment from point or diffuse sources and may have been exacerbated by the removal of riparian shade. Controlling nutrient inputs (by fencing off a riparian strip to prevent stock access and to intercept nutrient run-off, and treating any point-source discharges), and planting shade trees along the stream banks could bring about some improvement in stream condition.

Put simply, an MCI value does not imply any particular remedial action: it is necessary to understand the causes of degradation in order to remediate it. However, there has been considerable research into the effects of agricultural activity on stream community health (e.g. Hickey et al 1989; Quinn et al 1992; Quinn et al 1997a), so in the case where we have a site photograph showing the periphyton on the streambed and the stream in a rural setting, and where we know the MCI, it probably is sufficient to enable a water manager to implement remedial strategies. In other words, the macroinvertebrate sample alone provides an assessment of stream health and knowing the land use in the catchment provides sufficient information, when combined with existing knowledge, for remedial action to be undertaken.

How much additional information should be collected along with macroinvertebrate samples for SoE monitoring? There is a time and cost associated with data collection, so it is prudent to ensure that this expenditure is worthwhile and contributes to the monitoring objectives. We agree with Likens (1983), who notes that just because the information "might be useful one day" is insufficient justification.

When balancing practicalities (especially the ideal of collecting all SoE samples within a few days - see Section 4.7) and cost with information needs, we believe that SoE monitoring should comprise the following essential and optional elements.

Essential elements of SoE monitoring are:

1. the site location (including the map or GPS reference), sampling date, sampling time, and name(s) of personnel undertaking sampling
2. site photographs showing:
   1. an overview of the site in its setting (stream banks, riparian cover, catchment land use)
   2. the stream bed (substrate type, periphyton type and cover)
3. a single hand-net sample per site collected according to protocol C1 (hard-bottomed) or C2 (soft-bottomed) (Stark et al 2001); descriptions of substrata sampled; and proportions of soft-bottomed substrata sampled (e.g. % submerged wood, macrophytes, bank structure)
4. field observations - local land use and sources of pollution
5. catchment variables - land use, proportion of impervious land cover, elevation, distance from the sea, stream order, pollution sources, barriers.

Optional elements are:

1. field physicochemical measurements − electronic instruments can log pH, dissolved oxygen, temperature and conductivity quite efficiently
2. habitat assessments using a formal protocol
3. periphyton percentage cover and/or samples for chlorophyll a or ash-free dry-weight (AFDW) determinations
4. stream substrata percentage composition
5. width, depth, current velocity and discharge
6. river environment classification class, or some other description of the environment/ eco-region.

4.7 Minimising the time taken for a full sample round

The integrity of the macroinvertebrate data is maximised if all SoE monitoring sites are sampled within the space of a few days rather than over several weeks. This gives less time for conditions to change so that the risk of confounding influences, such as freshes and recessions, affecting assessments is minimised. However, few councils seem able to collect samples from 40 to 60 sites within a few days. There are several reasons for this. Often it is because of all the other sampling and assessment that is being undertaken at the same time (as described in the previous section); sometimes the time taken to travel between sites is a constraint.

We accept that it may be impractical for most councils to sample all sites within a few days, but there are several strategies that may help to achieve this. First, all sites could be visited, with the only work undertaken being site photographs and macroinvertebrate sampling. If electronic meters capable of automatically recording selected water quality variables are available, these could be used too. In this way perhaps eight to 10 sites could be visited each day. All sites could then be visited a second time after the macroinvertebrate samples have been collected in order to undertake more time-consuming activities such as habitat assessments. Another strategy is to send out two or more teams to sample different sites on the same days (although that could increase sample variability if different people collect them).

Ultimately, sampling over a period of weeks may not be a problem. Current research aims to determine seasonal and flow-related correction factors that would enable MCI values (among others) to be standardised to improve the validity of between-site comparisons and stream health assessments.