3.0 General Guidelines for Wetlands Biological Characterization

Impacts on Quality of Inland Wetlands of the United States:
A Survey of Indicators, Techniques, and Applications of Community Level Biomonitoring Data
Excerpts from Report #EPA/600/3-90/073
(now out of print)

Wetlands pose unusual challenges for monitoring programs. Because wetlands, as transitional environments, are located between uplands and deepwater areas, their biota exhibits extreme spatial variability, triggered by very slight changes in elevation. Temporal variability is also great, because the shallowness of any surface water results in its being highly influenced by slight, fleeting changes in precipitation, evaporation, or infiltration. Only a minority of all wetlands in the United States have permanent surface water (Shaw and Fredine 1956), so sampling techniques developed for other surface waters are not always applicable. The extreme spatial and temporal variability often requires that large numbers of samples be collected if the wetland community is to be properly characterized.

Such extensive sampling is made difficult, however, by potentially severe problems of access. Physically, access to many wetlands is hindered by water too shallow for rapid boat access, soil too fluid for rapid foot or vehicular access, and vegetation canopies too dense for easy aerial or airboat access. Access to many wetlands is also seriously hindered by the widespread (and sometimes misguided) public perception that wetlands, in contrast to other waters regulated by the Clean Water Act, are exclusively private land. Landowner awareness of the potential for regulation has led to commonplace denial of requests for access to wetlands during other EPA projects. Proportionally few wetlands are publicly owned, and these are not necessarily representative of the total wetlands population. These factors all combine to potentially increase the costs of an effective wetland monitoring program, and pose significant demands for study design and logistical planning.

Despite these difficulties, the need for more vigorous wetland sampling efforts is compelling. Because most wetlands are located in a topographically low, depositional environment and have long hydraulic detention times, they accumulate contaminants from a wide area. At the same time, undisturbed wetlands are characterized by exceptional biological productivity, suggesting a greater need for more extensive monitoring of wetlands. However, wetlands seldom are monitored, so much remains to be learned about the extent to which contamination and other stressors have altered their condition.

3.1 WHAT TO MONITOR

Monitoring of multiple indicators--having both short and long lifespans, and both localized and broad home ranges--is preferable to monitoring a few because indicators differ in their sensitivity to different types of stress in different types of wetlands, and in their temporal and spatial occurrence. By monitoring both short- and long-lived taxa, the effects both of stressors that occur briefly (e.g., herbicides) and of those that occur over longer time periods (e.g., bioaccumulation of metals) can be detected. By monitoring both resident and wide-ranging/ migrant species, the cumulative landscape-level impacts that may not be detectable on a local scale may become apparent. Ideally, monitoring of a wetland should encompass as long a time period, as many indicators, and as many microhabitats within the wetland as possible, given available resources. However, the need to make choices is inevitable.

Another choice concerns the which level of ecological hierarchy should be measured--e.g., physiology of individuals, demographics of a population, structure of a community, or processes of an entire ecosystem. Conclusions from one level cannot necessarily be extrapolated to another. As noted in Chapter 1, the scope of this report is limited primarily to the community level. A good discussion of factors affecting the choice of an appropriate hierarchical level in wetlands is presented by Farmer and Adams (1989).

Sometimes, the analysis of initial data collections can be used to target particularly sensitive groups or processes and identify optimal numbers of samples. Also, if life histories and ecological relationships are sufficiently well-understood in a particular area, monitoring could be limited to a few taxa known for their sensitivity to a particular stressor or their role as ecological "keystones." Keystone species include those which physically alter the landscape so profoundly that they create or destroy habitat for a much larger group of species over a wide area.

Examples of taxa that are considered to be keystones in particular regions or wetland types include:

• woodpeckers, which excavate cavities required by dozens of species;
• bees and other pollinating or seed-dispersing organisms, which control habitat structure through their major collective effects on vegetation;
• gopher tortoises and other burrowing species that create shelter critical to survival of many other animals;
• beaver, which create wetlands and temporarily destroy forest;
• muskrats, alligators, and some herbivorous birds, which through grazing and physical movement cause locally major increases in open water patchiness of wetlands.

Caution is necessary because it is seldom possible to validly infer trends in all species by monitoring only one or a very few "keystone" or "indicator" species. Thus, changes in community-level metrics usually give a clearer indication of "abnormal" biological stress than does the presence or absence of a single indicator species, regardless of its reputation as a keystone (Browder 1988, Cairns 1974, Couch 1982, Grigal 1972, Hellawell 1984, Karr 1987, Kelly and Harwell 1989, Landres et al. 1988).

In other aquatic systems, stable isotope techniques have been used to help identify keystone species, ecosystem components, or processes. In the case of vascular plants, attempts to identify the most sensitive species have also been made by measuring exposure of a host of species to a particular substance (e.g., a nutrient) and then monitoring the varying degrees to which the substance accumulates in tissue (e.g., Canfield et al. 1983), or alters germination and other physiological processes. Species which accumulate the substance and/or show the greatest physiological response the most are presumed to be likely to be affected if the substance increases.

To identify the most sensitive indicators, greater efforts could be made to comb the literature on experimental toxicology. However, although use of standardized conditions in most toxicity testing allows some degree of comparison among taxa regarding their relative sensitivities, the usefulness of laboratory toxicological data can be limited by the dissimilarity of test conditions and typical wetland conditions (e.g., altered toxicant mobility and toxicity due to increased organic carbon; interactions between hydroperiod effects and chemical toxicity--see Chapter 2.0).

Conceptual models (e.g., Patterson and Whillans 1984) or simulation models (e.g., Summers and McKellar 1981) of wetland ecosystems also could be applied to identify impact networks and thus, taxa that are likely to be most vulnerable to a particular stressor, and/or are potential keystones in ecosystem energy flow (Levins 1973). However, modeling approaches are also limited by lack of data on many wetland species and stressors (e.g., tolerance of wetland organisms to desiccation, burial).

Inevitably, the choice of what to monitor is governed by both policy and scientific considerations. The following criteria (derived from AMS 1987, Hellawell 1984, Kelly and Harwell 1989, Landres et al. 1988, Schaeffer et al. 1988, and Temple, Barker, & Sloane 1989), may apply:

Decision factors related to policy implications:

• Unambiguous - The indicator is socially relevant and easily understood as an indicator of ecological integrity and/or health;
• Evaluative - The indicator is capable of evaluating the effectiveness of regulations, control, or management strategies;
• Cost-effective - The indicator is capable of giving a maximum amount of information for a minimum cost, and thus fiscally attractive;
• Accessible - The indicator is capable of being generated from accessible data sources;
• Anticipatory - The indicator is capable of providing a warning in time to avoid widespread or irreversible damage.

Decision factors related to scientific implications:

• Sensitivity - The indicator is responsive to the range of conditions likely to be encountered;
• Common - The indicator is sufficiently present in wetlands to be captured by reasonable sampling effort;
• Integrative - The indicator is capable of integrating effects over time and space;
• Standardized - The indicator is either broadly used and possessing standard methods, or capable of development of standard methods;
• Reliable - The indicator provides comparable results over a wide range of conditions;
• Predictive - The indicator provides a predictable response to a given stressor or set of stressors;
• Rigorous - The indicator is scientifically accurate, precise, explicit and capable of standard measurement and reporting protocols that are congruent with the data quality objectives.

The relative weights given each of these evaluation factors will vary depending on the programmatic context, i.e., for which of the following potential purposes the indicator is being used:

• Determining simply whether a wetland is changing, and in what direction;
• Assessing how aberrant is the community structure of a particular wetland, e.g., to set priorities for restoration or strategies for mitigation;
• Evaluating the success of management of a wetland, e.g., compliance with permits and mitigation plans;
• Pinpointing the source of degradation of a wetland;
• Evaluating overall program success of wetland quality protection efforts;
• Priority ranking of wetlands;
• Gaining an understanding of fundamental wetland processes and advancing the science.

As we examined the technical literature on the most commonly monitored taxonomic groups, we applied the unweighted criteria to the indicators in a non-systematic, qualitative manner. A resulting summary of the advantages and disadvantages of each taxonomic group is presented as Appendix A. As data become available, a more thorough analysis would consider, more specifically, the differences among taxa with regard to particular stressors in particular wetland types.

To date, there appears to be only one field study (Brooks et al. 1990) that has attempted to compare the relative sensitivity of major phyla (at the level of community structure) for indicating anthropogenic stress in inland wetlands. Experimental studies making such comparisons are also virtually non-existent. Future efforts to develop and compare indicators could focus on studies that circumstantially span a gradient of disturbed and undisturbed (but otherwise as similar as possible) wetlands of all types. They could compare all taxa, metrics and data reduction techniques, which, from a theoretical perspective and studies to date, show promise for use (e.g., Do vegetation similarity measures respond more sensitively to heavy metal pollution than does wetland invertebrate biomass?). Such future efforts to develop and compare metrics could emphasize comparisons under different types of temporal and spatial variability.

Given this situation, an alternative approach is to query wetland experts regarding their personal opinions of taxa and metrics that might be most useful for a given purpose. Some of these opinions have been published (Table 6). However, recommendations can be unintentionally colored by the expert's degree of experience with a particular taxon.

As resources allow, rigorous approaches to indicator evaluation might involve integrated laboratory and field dosing experiments, conducted in parallel with empirical field studies of a series of wetlands that are as similar as possible but are situationally exposed to various levels (i.e., a gradient) of the same stressor. This is proposed in EPA's implementation plan for wetland - water quality research (Adamus 1989).

3.2 TYPES OF MONITORING

Monitoring methods might be classified as qualitative and quantitative. Qualitative methods are generally faster, based largely on visual observation, require little or no sampling equipment, and are usually applied just along the edges of wetlands. Compared to measurement-based quantitative methods, qualitative methods are often less replicable and accurate.

One type of qualitative method used in wetland biological monitoring involves use of ground-level (or low-level) photography. This typically consists of establishing fixed stations at several points around or within a wetland and taking photographs at specified times. Stations may be surveyed in to known benchmarks to assure that they may be subsequently located with accuracy, or objects expected to be immobile over time (e.g., heavy metal stakes) may be included in each picture. Range poles can also be included in pictures to document scale. Photographs are often pieced together to form a panorama, and video cameras are being used increasingly to comprehensively document conditions. Photographs can subsequently be evaluated visually, primarily for major changes in woody vegetation. Time-lapse photography can be used in some settings to monitor wildlife use. Cameras tethered to balloons have also been used in emergent wetlands to record interspersion of open water areas with vegetation, and distribution of submersed macrophytes (e.g., Edwards and Brown 1960).

Table 6. Wetland Monitoring Indicators Suggested by Various Scientists.

Aust et al. (1988):

These authors studied silvicultural impacts to wetlands, and found that the most efficient indices of changes in ecological function (from helicopter logging, skidding, and herbiciding) were soil acidity, redox potential, oxygen concentration, temperature, soil mechanical resistance, sedimentation, and vegetation cover. These require short sampling periods, a minimum of laboratory work, and easily operable and maintainable equipment. Less complex to interpret were sedimentation, net primary productivity, plant N and P uptake, cellulose decomposition, and bird richness, diversity, and abundance. Most responsive to disturbance (i.e., showing significant differences across gradients or between treatments) were total N and P concentrations in soil water, soil acidity, redox potential, saturated hydraulic conductivity, temperature, soil mechanical resistance, sedimentation, net primary productivity, plant N and P uptake, and cellulose decomposition. Most integrative of ecological processes were soil redox potential, net primary productivity, plant N and P uptake, and cellullose decomposition rates.

Brooks et al. (1989) and Brooks and Hughes (1988):

For general monitoring of inland wetlands, the following monitoring parameters were suggested: hydrology, water quality, hydric soils, vegetation (richness, density, productivity, vertical stand structure, horizontal patchiness), macroinvertebrates, fish, amphibians, birds, mammals.

Brown et al. (1989):

They proposed the following (in approximate priority order) be monitored for EMAP (EPA's proposed Environmental Monitoring and Assessment Program for wetlands, in which a probability sample of 3000 wetlands (50-100 of each of about 13 types) nationwide would be visited once every 3-4 years, with perhaps more-frequent airphoto coverage):

1. Regional changes in the acreage, type diversity, and spatial patterns of wetlands.
2. Nutrients
3. Other pollutants in sediments
4. Hydroperiod
5. Vegetation (patterns, abundance, richness, composition)
6. Sediment and organic matter accretion
7. Waterbird abundance and species composition
8. Bioaccumulation
9. Macroinvertebrates (abundance, biomass, composition)
10. Leaf area, percent light transmittance, greenness
11. Microbial community structure
12. Bioassays and biomarker measurement

Florida DER (Schwarz et al. 1987):

For state-required monitoring of wooded and cattail-dominated wetlands receiving treated wastewater, the following parameters are measured: water quality, detention time, vegetation ("importance value" of dominant species), macroinvertebrates (Shannon diversity index), and fish (biomass ratio of rough fish to sport and forage fish).

Kadlec (1988):

Chemical inputs and outputs normalized to flow, vegetation biomass, sediment and organic matter accretion.

US EPA (1983):

For monitoring of wetlands receiving wastewater, the following parameters were listed: hydrology, nutrients, other dissolved substances, trace metals, refractory chemicals, sedimentation, vegetation (species composition, areal distribution, biomass, growth, production), detrital cycling (organic matter accretion), bioaccumulation, macroinvertebrates, fish (productivity, biomass, spawning success, bioassays, incidence of disease), wildlife communities (habitat structure, species richness, density, indicator species, incidence of disease).

US EPA (Sherman et al. 1989):

For comparison of multiple sets of constructed wetlands with reference wetlands in Florida, New England, and the Pacific Northwest, the following were measured: water depth, depth to water table, ambient nutrient concentrations, sediment chemistry, soil oxidation, morphometry and bank slope, vegetation (species composition, cover, natives vs. exotics).
Table 6. Wetland Monitoring Indicators Suggested by Various Scientists.

Qualitative methods such as these can be used to develop maps of vegetation within wetlands, e.g., Farney and Bookhout (1982), Meeks (1969), and Morgan and Philipp (1986). Use of cover maps, aerial photos, and ground photos can be used to identify broad changes in plant composition, as well as providing permanent records. Suitable, existing, low-altitude color photographs often can be obtained from state offices of the Agricultural Stabilization and Conservation Service, from the U.S. Forest Service (forest pest management monitoring programs), and (near roads) from state highway departments, as well as other sources. Remote sensing has also been used under ideal circumstances to estimate soil saturation, primary productivity, and sedimentation (Heilman 1982).

A second type of qualitative monitoring involves making visual, ground-level estimates simply of presence/absence of indicator species and physical conditions (e.g., Terrell and Perfetti 1989) and, in the case of vegetation, of percent cover. Vast numbers of such unpublished "species lists" are available from university botanical visits to wetlands, consultant reports, and other sources. While preferable to no data at all, these represent the "data rich - information poor" syndrome. However, some investigators go beyond a simple listing of species and visually estimate abundance in relative terms, e.g., rare, common, and this allows improved interpretation of data. Examples are reports by Dunn and Sharitz (1987), Ehrenfeld 1983, Kadlec and Hammer (1980), Nilsson and Keddy (1988), Taylor and Erman 1979, Wilcox (1986).

A third type of qualitative monitoring approach involves the use of "wetland evaluation" methods. Many such methods are available (e.g., see reviews by Adamus 1989, Kusler and Riexinger 1986, Lonard et al. 1981), but differ little in terms of their time requirements. Perhaps the most widely used are:

• Habitat Evaluation Procedures (HEP) of the U.S. Fish and Wildlife Service.
• Wetland Evaluation Technique (WET) developed by EPA and the Corps of Engineers (Adamus et al. 1987, Adamus et al. 1990).

Although these methods may benefit from or require a limited number of field measurements, they are predominantly qualitative. They do not directly measure biological communities, but rather, assume biological community structure or wetland function using information on habitat structure (Schroeder 1987). Most are applicable at the individual-site level (e.g., WET), while others (e.g., the "Synoptic Approach"-- Abbruzzese et al. in press) operate at regional scales and require more cursory data inputs.

Quantitative methods are the focus of this report. Although many reports and books describe protocols for biological sampling of lakes and flowing waters, few have attempted to comprehensively describe or evaluate sampling modifications appropriate for the highly variable, transitional environments of wetlands. Some relevant information can be found in the following:

Brooks 1989, Brooks and Hughes 1988, Erwin 1989, Fredrickson and Reid 1988a,b, Harris et al. 1984, Horner and Raedeke 1989, Murkin and Murkin 1989, Platts et al. 1987, USEPA 1986, Welcomme 1979, Woods 1985.

Other parts of EPA's Wetlands Research Program have developed protocols for wetland sampling. For example, at the EPA-Corvallis Laboratory, the Wetlands Team has developed protocols for monitoring created or restored (mitigation) wetlands (Sherman et al. 1989). The EPA-Duluth Laboratory has developed protocols for biological field-sampling of wetlands impacted by a variety of stressors. EPA is refining these and developing other protocols for support of its nationwide Environmental Monitoring and Assessment Program (EMAP). This report is not intended to substitute for these other protocols, but rather, includes them in discussions of a full range of methods available.

3.3 STUDY DESIGN

For detecting wetland ecological change and estimating its causes, a statistically powerful approach would involve sampling both before and after the expected change, in both exposed wetlands and in similar, unexposed wetlands (or in a large, random sample of similar wetlands with unknown exposure history). Selection of unexposed, or "reference" wetlands is discussed briefly in section 2.1. Statistical issues associated with wetland biomonitoring are discussed in greater detail by Simenstad et al. (1989).

Because of the high variability of wetland environments, sample collections should be replicated, both within and among wetlands, and within and among sampling times. One simple option for estimating the minimum effective number of samples or hours of effort involves plotting a curve. The "x" axis of the curve would describe the number of samples collected and the "y" axis would describe the community metric being measured (e.g., cumulative number of species), or its cumulative percent error or variance. Assuming a reasonably large number of samples have been initially collected, the point where the curve levels off might be considered to represent the minimum effective sampling effort. Statistical protocols are also available for estimating requisite number of samples in wetlands, given a desired detection level and initial information on sample variability (e.g., Downing and Anderson 1985, Eberhardt 1978, Jackson and Resh 1988, Resh and Price 1984).

There are several options for placement of sampling stations. In previous wetland studies, stations most often have been situated in one of the following ways:

• randomly;
• along transects (usually perpendicular to wetland gradient or flow and extending to the deepest part of the wetland, and sometimes intentionally aligned to intersect all habitat or topographic "types" within the wetland);
• at ecotones (spatial boundaries between major vegetation types, and open water and vegetation);
• in proportion to occurrence of habitat types (or hydroperiod classes) present within the wetland;
• at locations subjectively felt by the investigator to "represent" the wetland.

Seasonal timing of sampling is also important, and can be scheduled to coincide with (a) times at which organisms of concern are most likely to be at maximum numbers, (b) times when these organisms are most physiologically sensitive to a particular stressor, and (c) times at which concentration of, or organism exposure to, the stressor is greatest. From this, it is apparent that cost-effective wetland biomonitoring requires knowledge of (a) life history aspects of wetland organisms, (b) physiology and relative sensitivities to stressors of the component organisms, and (c) dynamics of physical and chemical factors that largely determine stressor availability. Most biological surveys of wetlands have been conducted during the growing season, and relatively little is known of exposure or community structure and function during stressful conditions of ice cover, severe anoxia, or drought. Time-of-day is also an important consideration, particularly when monitoring vertebrates. Unless diurnal behavior patterns are well-understood, or there is sufficient labor available to sample wetlands simultaneously, it may be desirable to alternate the order in which wetlands are visited, to avoid temporal bias.

The optimal seasonal timing from a biological perspective may not coincide with the best timing from a perspective of physical human access. Physical access into wetlands is notoriously difficult, and the more accessible edges of a wetland do not represent the biological conditions in a wetland generally. Although interior parts of wetlands may be more accessible during ice cover or drought, seldom are these the most biologically appropriate times for sampling. Previous investigators have used hip boots, canoes, inflatable rafts, airboats, helicopters, snowshoes (in summer, for distributing weight in peat bogs to prevent sinking), and scuba gear for dealing with problems posed by the semi-fluid substrate of many wetlands. For vegetation, remote sensors can be used for general coverage estimates. Low-altitude video can provide digital data directly, facilitating spatial analysis (pers. comm., M. Scott, U.S. Fish and Wildlife Service, Fort Collins, CO). Biomass of submersed aquatic macrophytes was measured electronically by Canfield et al. 1983, Duarte 1987, and Thomas et al. 1990.

3.4 DATA ANALYSIS AND INTERPRETATION

After addressing the question, "What should we measure?" the next logical question is "How do we express the data?" Thus, in developing and applying wetland biocriteria, the selection and interpretation of appropriate metrics is at least as important as the selection of appropriate taxa and sampling techniques. Questions related to wetland metric selection, such as the following, must inevitably be addressed if "data" are to be converted to "information:"

• Is abundance, biomass, or species richness a more sensitive indicator of wetland biological change?
• When are "guilds" an appropriate way to compile data?
• Do similarity indices and ordination procedures indicate stress from contaminants better than they show stress from hydroperiod alteration?
• When metrics describing ecosystem structure (such as the above) show that a wetland has changed, what can be inferred about the wetland's change in function?

Providing a detailed description of all possible techniques for analysis and interpretation of wetland data was considered beyond the scope of this report. Similarly, the validity and sensitivity of various metrics and procedures, as applied to the specific taxa and stressors described in later chapters, is not evaluated by this report. Such an evaluation, perhaps using the evaluation factors listed in section 3.1, would be extremely important in developing biocriteria for wetlands, but is not currently feasible due to lack of sufficient comparative data. Some of the more commonly used metrics and analysis procedures are shown in Tables 2, 3, and 4. Review and comparisons of performance of various indices in non-wetland ecosystems are given, for example, in Green and Vascotto 1978, Huhta 1979, Krebs 1989, Magurran 1988, Matthews et al. 1982, Polovino et al. 1983, Wolda 1981, Washington 1984, and others. For further information on statistical analysis of wetland community data the following references (among hundreds) might be consulted: Gauch 1982, Green and Vascotto 1978, Hill 1979, Isom 1986, Jongman et al. 1987, Ludwig and Reynolds 1988, Pielou 1984, and Wiegleb 1981.

Community-level metrics can also vary greatly in their sensitivity for detecting environmental stress. To optimize detection of ecologically degraded condition, it is usually best to use several metrics or procedures in combination (Schindler 1987), as is done by the "Index of Biotic Integrity" that was developed for other surface waters (Karr 1981).

For other surface waters, information compiled by Sheehan (1984) and others suggests that the approximate statistical sensitivity of community-level metrics/procedures to pollution has often been:

cluster/ordination > similarity > richness per unit area or effort> biomass/abundance procedures indices and diversity indices

However, generalizations such as this contain a high degree of uncertainty. This is because of biases potentially arising from unknown (and perhaps inconsistent) dependence on a metric's or procedure's sensitivity to (a) statistical properties of the data set, (b) the particular combination of taxa contained in the data set (and associated life histories varying from sample to sample), (c) taxonomic level-of-identification, (d) wetland or community type, (e) spatial scale of measurement, (f) temporal scale of measurement (e.g., frequency of sampling, time elapsed since the stressor was maximal), (g) sampling equipment, level-of-effort, and techniques used. Thus, when only a few metrics and statistical procedures can be applied, results may be difficult to interpret. Unfortunately, few wetland studies have examined these potential biases. Also, of particular interest would be (a) the correlation of responses of metrics at several ecological levels, e.g., do metrics based on response at the organism level show the same response as those based on data from the population, community, and ecosystem levels? and (b) the correlation of responses of metrics to responses in ecosystem function (processes).

A single number from a metric, if used alone, sometimes provides little useful information. Often more instructive is the particular taxonomic composition that led to a particular summary metric value. Thus, where data on sensitivities and life histories of organisms are available, aggregating species-level monitoring data by functional groups ("guilds", see Table 4) of species can provide for more meaningful data interpretations. It can also reduce the statistical variability in data sets, thus reducing the number of requisite samples (pers. comm., Dr. James Karr, University of Virginia).

Moreover, shifts in taxonomic composition in response to contaminants frequently are likelier to occur than changes in total number of species or biomass (e.g., Ferrington and Crisp 1989). However, predicting which species will become dominant following a wetland disturbance is generally more difficult than predicting that species composition, overall richness, or biomass-abundance will change (Nilsson and Keddy 1988). In wetland macrophyte communities, richness is frequently correlated with biomass (Nilsson and Keddy 1988). This is not true in some communities of wetland fish (Tonn 1985).

All of the above metrics/procedures, except biomass/abundance, commonly employ species-level data. Such data are easily determined for taxa such as birds, but are much more difficult to acquire for microbial communities, which have large numbers of species, and for which comprehensive regional references on taxonomy are virtually non-existent. The need for species-level identifications for the determination of anthropogenic effects has been asserted by some studies and disputed by others; the need may depend on the factors listed above that pertain to metric biases, as well as on costs of making more-detailed identifications vs. costs of collecting a larger number of samples that are only identified at a general taxonomic level.

The utility of some metrics and procedures, as well as their sensitivity, may vary by wetland type. For example, metrics and procedures that depend on species-level data (richness, ordination, similarity indices) may be ineffective in describing the ecological condition of wetlands that characteristically have low species richness (e.g., breeding bird richness in salt marshes, fish richness in montane wetlands).

The metrics and procedures listed in this report represent only our current abilities to quantify wetland community structure. From the emerging discipline of "stress ecology" (e.g., Lugo 1978, Odum 1979, 1985), there may be additional theoretical properties of wetland community structure--such as inertia, elasticity, amplitude, resilience, hysteresis, malleability, and persistence (to use the terms of Sheehan 1984 and Westman 1978)--that have potential for quantification and testing. However, only a very few experimental studies (e.g., Meffe and Sheldon 1990) have quantitatively examined some of these in a regional set of inland wetlands. If conceptual and operational problems associated with these metrics can be overcome, they may hold potential for more sensitively measuring impacts.

After addressing the question of "How do we express the data?" the next logical question is "What represents normal (or desirable) conditions?" Data interpretation is critical to every monitoring program, and (as discussed in Chapter 2) "normal" can be defined either in terms of (a) the condition of a reference wetland, (b) average regional conditions, or (c) ecological conditions necessary for sustaining the ecosystem type and/or a dynamic balance of its important species. In deference to the vital processes of natural succession that prevail in many wetland types, a definition of "normal condition" should encompass not only a mean condition, but the naturally-occurring extremes in structure and function that may be expected over decades of time (i.e., temporal and spatial variability). This report has not sought to go beyond this general consideration and attempt to define nominal (normal) and subnominal (abnormal) wetland conditions. Such an exercise would require an understanding of specific resource management objectives, considerably more data, and significant public involvement.

Finally, if it has been determined that a wetland is "abnormal," it may sometimes be necessary to conclusively determine causality. This typically involves laboratory and field bioassay work, a discussion of which is beyond the scope of this report.

Regardless of which approach is used, caution must be exercised in interpreting community-level data as a potential indication of anthropogenic stress. Absence of a species may be due merely to random events (e.g., Grigal 1985). Sampling metrics, particularly species richness, are often very sensitive to the intensity of sampling, i.e., number of samples, level of effort, size and natural heterogeneity of the wetland sampled. Also, genetic mutation, natural selection, and/or adaptation can result in evolution of tolerant "ecotypes"--local forms of a species that have become tolerant even of normally toxic contaminants. This can alter competitive relationships and ultimately, community structure. Although it is uncertain as to how widespread this phenomenon may be, it can be locally important and has been documented to occur in communities of microbes (Baath 1989), macrophytes (e.g., Christy and Sharitz 1980, McNaughton et al. 1974), aquatic invertebrates (e.g., Krantzberg and Stokes 1989, Kraus and Kraus 1986), and amphibians (e.g., Karns 1984).

Also, the possibility that mobile fish or wildlife are avoiding contaminated areas (even temporarily) should be considered when evaluating community-level vertebrate data. Conversely, wide-ranging biological indicators may not occur even in the "healthiest" wetlands if most other surrounding wetlands have been contaminated or altered.

Finally, wetland function cannot always be assumed to change whenever the structure of the biological community changes. Changes in community composition may be compensatory, such that new species replace the function of original species and overall community biomass and perhaps richness does not change (Cairns and Pratt 1986, Herricks and Cairns 1982). An example of this specifically from wetlands is provided by Cattaneo and Kalff (1986), who conducted invertebrate exclusion experiments in an aquatic bed wetland. For this reason, it may be advisable to develop and employ, whenever possible, measurements of both structure and function.

The following sections of this report summarize information relevant to monitoring specific taxonomic groups, wetland types, and stressors. Again, the purpose is not to be prescriptive, but rather to partially survey techniques used by other investigators and summarize conclusions that are relevant to future monitoring. It is expected that these descriptions will be refined and evolve during the review process and as more data are collected from wetlands. The order of these sections does not necessarily reflect priorities, but rather is based on phylogeny (taxonomic relationships). Despite the manner of organization, by major taxa, it is important to recognize that a massive array of interactions can occur in any wetland among the separate taxonomic groups, and such competitive interactions, as noted in a few cases in the following text, can temper the response of an individual taxon to a particular stressor.