SCOPE 53 - Methods to Assess the Effects of Chemicals On Ecosystems

This chapter provides a conceptual overview of the approaches and methods to conduct environmental monitoring and assessments of the effects of chemicals on geographic scales that cover entire regions. Most environmental monitoring is limited to the local environment so that specific pollutants can be linked to the source and controlled. Unfortunately, this type of monitoring provides little information about the effects of chemical releases on entire ecosystems. To better identify "systemic" environmental problems (e.g., acid deposition and loss of diversity) and the ecosystems at greatest risk, regional monitoring and assessment programs are being developed, such as the Environmental Monitoring and Assessment Program (EMAP) conducted by the US Environmental Protection Agency (Messer et al., 1991; Kutz et al., 1992). Regional programs like EMAP use large geographic scale monitoring to identify significant long-term changes in the condition of ecosystems.

In a manner similar to that used in Chapter 3, this chapter discusses important concepts that form the basis of large geographic-scale assessment methods, such as time scales, assessment endpoints, study design, examples of assessment methods, and methods to analyse data. 

4.1 IMPORTANCE OF LARGE GEOGRAPHIC SCALE MONITORING AND ASSESSMENT

The effects of chemicals on ecosystems frequently extend beyond local boundaries. Increasingly, scientists and risk managers are becoming aware of environmental problems that occur on large geographic scales: acid deposition, non-point source pollution, and loss of biodiversity, to name a few. Consequently, environmental managers should focus not only on effects locally, but also on multimedia impacts that occur on larger scales. Ecosystems are interconnected, and disruption in one section may cause repercussions throughout the entire system in ways that could not be anticipated by evaluating chemical impacts on the system on only a local level.

A narrow focus on only the individual or community levels of an ecosystem is insufficient to make decisions and allocate resources for several reasons. First, releases of chemicals to the environment have impacts that are most often significant on large geographic scales. An example is the formation of acid precipitation in one area that results from fossil fuel combustion in distant areas. Second, the ecological resources of interest exist on large scales (i.e., fisheries and forests), and are affected by many complex climatic processes that themselves function on global levels. If environmental managers consider the complexities of the structural components of ecosystems and the functional processes that drive them, they can make more realistic and accurate evaluations of the ecological resources of interest. Such management decisions are often triggered by knowledge of the extent or magnitude of a problem. Recognizing that effects are widespread (i.e., a certain percentage of the ecological resource is degraded) may initiate corrective action to remedy degradation at the regional scale. While knowledge of the effect of chemicals at the organismal level is needed, recognizing that such an effect may be magnified at the population or community level is vital to managing effectively ecosystems of larger geographic scales. 

Regions represent an intermediate hierarchical level between individual ecosystems (estuaries, forests, deserts, and agricultural areas) and the global biosphere. An understanding of regional scale processes is often required to explain or predict the behaviour of individual ecosystems. Regional-scale ecological characteristics also may be the most appropriate scale for addressing ecological effects of global scale processes. Examples of some environmental problems that may require a regional scale ecological assessment are provided below. 

Regions are important units both to assess the effects of chemical releases and to manage biological resources. For example, the airshed is an appropriate unit to assess the success of controls on airborne chemical emissions, whereas the watershed is an appropriate unit for regional water quality. Similarly, forest resources are best managed on a regional scale; therefore, the effects of chemical exposures on these resources are best evaluated on a similar scale. While local scale and site-specific approaches are effective methods to ascertain the influence of some types of stressors and ecological resources, a larger scale perspective permits a more comprehensive view of environmental problems that may impact large geographic areas. 

Programs to assess the adverse impacts of chemical pollutants on ecosystems can also be made considerably more effective if they include regional concerns, values, and characteristics. Making decisions in the overall context of reducing risk over large geographic areas may help to avoid contamination problems that can result from management actions that move pollutants from one medium to another. Employing a regional approach can enhance chemical pollutant control programs in the following ways:

1. Identification of specific chemical pollutants that are major environmental threats within a region, and specific areas within regional systems that are being threatened by the pollutants. Chemical pollutants enter regional environmental systems from numerous point and non-point sources, and their threat is dependent on many factors, including: toxicity, persistence in the environment, and bioavailability; the quantities discharged; and the relative sensitivity of organisms in the area of discharge. Complete assessments of the threats posed by chemical discharges must also consider numerous less obvious, yet equally important, factors, such as land use patterns and socioeconomic systems. Emphasizing a regional approach to environmental assessment should encourage researchers to assemble and integrate the existing data on these other environmental parameters, which are most often described in a regional context.

2. Establishment of regional research priorities, and the allocation of research funds to priority projects. Attempts to develop and implement regional pollution control programs should reveal information gaps that limit the effectiveness of these programs. These informational deficiencies should become research priorities for the region, and research funds should be channelled to these priority topics. 

3. Development of monitoring programs and allocation of funds to program components. Monitoring programs are undertaken to identify changes in environmental systems triggered by biotic or abiotic stressors. A regional approach should include collection and organization of scientific data on the region, accurate interpretation of monitoring data, and the use of these data for resource management. A regional framework to address environmental problems should provide guidance on the development and structure of the monitoring program, parameters to be measured, and the sampling schedule.

4. Design of mitigation and restoration programs and allocation of funds to specific projects. Environmental problems resulting from the discharge of chemicals are frequently addressed by mitigation-restoration projects. A regional approach could identify the highest priority mitigation-restoration sites, thereby assisting risk managers to allocate the limited funds available for mitigation restoration to the most cost-effective projects. 

5. Justification of regulatory programs. Regulatory programs enjoy wider support by legislators, special interest groups, and the general public if rationales for these programs are clearly understood to be rational by all interested parties. The need to regulate an activity can be understood more readily if the regulation of the activity is explained in terms of the positive effects on the region as a whole rather than only on a specific site within the region.

4.2 DEFINING A REGION

Developing a regional ecological monitoring and assessment program requires defining the region of interest, which depends, in part, on the management objectives usually stated as questions. Individual ecosystems are normally defined operationally in terms of areas of relatively uniform physical conditions, biotic community composition, and land use (e.g., freshwater lake, cornfield, or estuary). Regions, in contrast, do not always have obvious natural boundaries, but, instead, are defined by spatiotemporal scale and human interest. A particular point on a map may belong to several different regions defined for different purposes: a 1000 km² watershed, a 10000 km² airshed, and a state or nation of very small or very large size. 

Three relatively natural approaches can define regions: "ecoregion;" watershed (if large) or river basin; and airshed. Ecoregions are defined as areas with relatively uniform ecological characteristics (Omernik, 1987). The ecoregion approach can employ original ecosystem types believed to be present before human development, or it may rely on current land use (e.g., agricultural or urban, suburban, or rural). A watershed-based region, for example, is defined by selecting a particular body of water as a starting point and identifying the land surface area drained by all of the rivers feeding into that body of water. An airshed-based region is defined as an area of restricted air flow, such as an area bounded by mountains (e.g., the Los Angeles Basin in California) or an area within which plumes from individual sources become thoroughly mixed. 

Such approaches are convenient for scientists, but are often impractical for environmental management, because management action is often constrained by political or socioeconomic boundaries. Regions, therefore, are often prescribed by district, state, or national boundaries that cross ecological boundaries or ecoregions. Suter (1993) has suggested that multiple definitions of a region should be employed for many regional assessment problems. Actually, a useful approach may be to first define regions by ecologically relevant boundaries, and then overlay this with political or socioeconomic boundaries. Discussion of three major approaches or paradigms used to define regions (landscape mosaic, watershed, and airshed) follow. 

4.2.1 THE LANDSCAPE MOSAIC PARADIGM

Figure 4.1 depicts an idealized view of a region divided into different ecosystems and land-use types. This view of a region may be called the "landscape mosaic." A simplistic approach to assess the quality or condition of such a landscape is to evaluate the condition of each unit independently, and then to aggregate the results by summing the economic returns and other benefits from each unit; this is also accomplished by calculating an environmental quality index. This approach is severely limited in its ability to aggregate, because it cannot account for interactions between landscape units, and it ignores organisms such as birds and large mammals that require relatively large habitat sizes for survival or can move between landscape units. The frequency and severity of forest fires, for example, are influenced by the size and spatial distribution of vulnerable stands; the abundance of deer in managed forests is enhanced by the presence of clear-cut patches or forest edges containing palatable second-growth vegetation; and the persistence of spotted owls depends on the existence of large stands of undisturbed old-growth forest.

The new subdiscipline of ecological science called "landscape ecology" is devoted to studies of the ecological characteristics of habitat mosaics. Among the research topics pursued by landscape ecologists are the development of matrices that characterize landscape patterns (Turner, 1987), the development and testing of spatial simulation models that relate local-scale ecological processes to regional scale patterns (Dale and Gardner 1987; Costanza et al., 1990), and the analysis of the influence of spatial distributions of ecosystem types on the persistence of populations (Pulliam, 1988) or the severity of pest infestations (Graham et al., 1991).

Figure 4.1 A representation of a landscape mosaic

4.2.2 THE WATERSHED PARADIGM

A region can also be characterized as a hydrologic network of connected streams, lakes, estuaries, and oceans. The ecological characteristics of any particular segment of this network are influenced by events occurring in upstream segments as well as those occurring within the segment. The watershed approach to regional assessment has been practised for several decades in water quality management. Influences of multiple point sources of pollutants on downstream water quality can be calculated using available hydrologic simulation models (Thomann, 1972). Problems of major current interest involve interactions between terrestrial landscapes and watersheds. "Non-point-source pollution," including soils and chemicals contained in runoff from agricultural fields, feedlots, suburban lawns, and industrialized areas is a major source of many of the most persistent and ecologically harmful contaminants. A regional approach to protect or restore an estuarine ecosystem may include managing both the terrestrial and the aquatic components of upstream watersheds. 

4.2.3 THE AIRSHED PARADIGM

Characterizing air quality only in terms of the behaviour of pollutant plumes and mixing zones within an airshed is inadequate to assess influences of atmospheric pollutants on regions. The airshed paradigm characterizes a region as a large volume of air that receives inputs from sources within the region and from other airsheds overlying other regions. Thus, an airshed is not a closed system, and an understanding of the rates, directions, and effects of interchanges with other discrete airsheds is essential to evaluate the regional effects of airborne substances. For example, combustion of fossil fuels in one region may release sulphur dioxide to a local airshed; the sulphur dioxide can then combine with atmospheric water vapour, to form sulphuric acid which lowers the pH of any precipitation. Commonly, although sulphur dioxide is released within a discrete local airshed, often it is carried on the wind to another airshed where it combines with water vapour forming acidic precipitation that falls in a region some distance from the original release. 

4.3 SELECTING A TIME SCALE FOR REGIONAL MONITORING

Having identified spatial boundaries for a region, a time scale of interest must be specified. The scales of time and space are related yet different; and which scale is most relevant to the study objectives must be determined. The analyst must judge whether varying time scales or varying spatial scales provides the data needed to diagnose the problem. Spatial scales are important when evaluating the extent and magnitude of contamination or its effects on biota. Temporal scales are useful to discern contaminant-related trends or changes in populations or media with time. For instance, the same individuals may be tagged and repeatedly sampled with time to determine bioaccumulation of the substance. Tracking reproductive effects of chemicals is possible by evaluating population parameters (e.g., age and sex ratios or natality) of one population with time. In addition, changes in relative proportions of contaminant-sensitive and contaminant-tolerant species of one population will evaluate the effects of the chemical on population dynamics.

The temporal sampling approach employed by the USEPA's EMAP evaluates both local conditions and long-term trends in ecosystems by sampling on a rotating four-year sequential cycle (Overton et al., 1990; Messer et al., 1991). Sites are marked on a grid, and during the first year of sampling, one-fourth of the sites are sampled, with another (different) fourth of the sites sampled during the second, third, and fourth years. Starting with the fifth year, the sampling cycle is repeated with the first fourth of the site being revisited. This sampling design is well suited to detect persistent, gradual changes in local populations and to detect long-term changes that slightly affect the entire region.

Figure 4.2 illustrates the source of problems that can arise from the evaluation of an environmental problem on a limited, perhaps inappropriate, scale. When limited to those events that vary on a large scale (i.e., variability on a cycle of centuries), the assessment cannot discriminate events that vary on a much smaller scale (i.e., days); the converse is also true. 

Figure 4.2 Illustration of the relevance of data at different time scales

For studies of regional effects of environmental pollutants, the relevant time scales are defined by those of pollutant releases and ecological responses. For assessments of regional effects of acid deposition on surface waters, for example, the US National Acidic Precipitation Assessment Program (NAPAP) chose a time scale of 30 years to ascertain reductions of atmospheric sulphur inputs and for the chemical composition of affected streams and lakes to reach equilibrium with the reduced loadings (NAPAP, 1991). When responses to anthropogenic stress are confounded by responses to natural climatic variability, time scales for assessments should include sufficient time to characterize climatic variability. Assessments of responses of forest ecosystems in the eastern United States to air pollution have been greatly complicated because the time scale of variation in temperature and precipitation in the eastern United States is roughly the same as that used to measure and predict changes in pollutant levels. The available time-series of climate records and forest condition data are too short to characterize this variation, and to separate its influence from the influence of pollutant deposition. 

Clearly, the management time scale is also relevant to define time scales for regional assessments. Regulations must be reviewed according to prescribed schedules, and technologies for pollutant control cannot be implemented instantaneously. As with spatial scales, the natural time scales must first be defined for a regional assessment, and then overlaid on the appropriate management time scales. 

4.4 REGIONAL RISK ASSESSMENTS

Suter (1993) described six general situations in which a regional-level ecological risk assessment may be the most appropriate approach. These are presented in Table 4.1, along with examples of each situation. A fuller discussion of these types of situations that may call for regional ecological risk assessments is provided in other chapters of this document. 

Several issues associated with spatial and temporal scales must be considered when planning a regional ecological risk assessment. These include: (1) the spatial and temporal scales associated with various levels of biological organization at which effects may be assessed; (2) the relationship of landscape scales to the distributions of populations in an area, and the way in which they use the available resources; and (3) the spatial and temporal scales of potential chemical exposures. These issues are discussed here only briefly, but are presented in greater detail in a later chapter of this volume. 

The effects of chemicals may be assessed at the biological organization levels of individual, population, community, ecosystem, region, or some combination of these levels. The spatial scale of a regional response to chemical exposure usually overlaps with, but extends beyond, that of individually exposed ecosystems (hundreds of kilometres to thousands of kilometres). The temporal scale of regional dynamics also overlaps that of the component ecosystems, but may be longer than for any one ecosystem in the region (tens of years to hundreds of years). Spatial and temporal scales associated with ecological effects at the various levels of biological organization are discussed by Sheehan (1984a, 1984b, 1992) and Suter (1993).

A second scale consideration for regional ecological risk assessments is the relationship between the scale of landscape patterns (including vegetation structure) and the distribution of animal species, and the scale of their use of resources (Holling, 1992). For example, at the regional scale, landscape responses to chemical stressors are often assessed in terms of changes in vegetation and vegetative structure, and little data are provided about the secondary effects of such changes in vegetation on animal populations. An understanding of the relationship between animal population dynamics and the structure of the landscape is essential to predict changes in animal populations that may result from changes in the vegetative structure of their habitat.

Table 4.1 Appropriate situations for a regional-level ecological risk assessment (Suter, 1993)

The third consideration is the spatial and temporal scale of the chemical exposure itself. If the chemicals are widely distributed or have the potential to be, a regional approach is recommended. Broad-scale applications of pesticides, such as aerial spraying to control grasshoppers in the prairie regions of the United States and Canada (Sheehan et al., 1987) and the spruce budworm in the forests of New Brunswick (Mitchell and Roberts, 1984), cover hundreds to thousands of kilometres, and may occur intermittently over years or decades. At the extreme, widespread aerial transport of photochemical oxidants (Skelly, 1980) and acids (Nash et al., 1992) has contaminated large geographical regions (thousands to tens of thousands of kilometres). Similarly, multiple inputs of chemicals into river systems occurs over hundreds to thousands of kilometres of the watershed. Cumulative exposure to airborne and waterborne chemicals also takes place over long time periods (tens to hundreds of years).

4.4.1 ASCRIBING VALUES ON REGIONAL SCALES

As with any assessment, a clear statement of values and endpoints should be the first step in its planning. Regions possess the characteristics of their component ecological, socioeconomic, and geographic systems as well as unique characteristics associated with the interactions between these systems. The relationships among systems usually differentiate the region from a single ecosystem; therefore, the values associated with a region include the values associated with the individual component systems and the values associated with uniquely regional characteristics.

4.4.1.1 Values inherent in local systems within a region

Both the broad generic and more specific cultural values attributed to ecological systems are equally applicable to regions; therefore, the tables of market, nonmarket, cultural, and ecological values-benefits identified in Chapter 2 (Tables 2.1 and 2.2) are equally useful to analyse regional issues of chemical contamination. In fact, many of these values are best recognized at a regional scale (e.g., forage and timber production). A general goal of management at the regional scale is the conservation of all natural resources (including biological resources) for sustainable use (IUCN, 1980). Biological resources include: genetic diversity at the population-species level, biodiversity at the ecosystem-regional level of organization, and biological productivity (both primary and secondary) at the regional level. 

4.4.1.2 Values unique to regions

The values unique to a region are the result of interactions between ecosystems and socioeconomic systems. Such interactions regulate ecological processes such as the cycling of water and nutrients vital to support sustainable resources. The International Union for the Conservation of Nature (IUCN, 1980) has ranked the conservation of fundamental ecological processes as a first priority objective. The basic components of this objective are:

1. The integrity of the hydrologic cycle is critical to the continued availability of water resources.

2. Disruption of the hydrologic cycles at the watershed-regional scale can result in dramatic changes in the quality and availability of water, and, in fact, has severe effects on human societies. 

3. The cycling of nutrients, such as carbon, phosphorous, and nitrogen, is another regionally important ecological process that must continue uninterrupted to ensure the fertility of arable land on which agricultural production depends.

4. Climatic processes, that essentially control primary productivity, are an additional example of vital regional-level values resulting from interactions within the region. Changes in dependable climatic patterns severely disrupt agriculture and change the inherent liveability of certain areas. 

5. Shifting rainfall patterns can turn a previously productive area into a nonproductive area.

Assessments of the impacts of chemicals on ecosystems should be planned with the consideration of the importance of these fundamental ecological processes that function on large geographic scales.

Table 4.2 Endpoints relevant to regional scale assessments

4.4.2 REGIONAL SCALE ASSESSMENT ENDPOINTS

After considering the target values, the endpoints should be chosen as the focus of any effects studies or monitoring. Although a number of local scale endpoints will also be important at the regional scale, for the ecological consequences of pollution, regional problems cannot be equated to the sum of local problems. In fact, the properties and characteristics of complex systems viewed on the regional scale are specific and qualitatively different from the properties of biological systems viewed from lower levels of organization.

Endpoints relevant to regional scale assessment can be classified into three general categories: traditional ecological endpoints; endpoints specific to the region; and anthropocentric endpoints (important to human cultures). Examples of these assessment endpoints that are important on regional scales are provided in Table 4.2. Certainly many of these examples are relevant to more than one assessment endpoint; for example, acid precipitation is an important regional-level endpoint and an anthropocentric endpoint. 

4.5 DESIGN AND SAMPLING CONSIDERATIONS

Because regions are extensive in both time and space, subsamples are needed to represent the whole population of possible sample units. Site-specific point-based analyses of chemical constituents, for example, can be made at only a limited number of locations throughout the region and on a limited number of occasions to track the changes in those constituents over time.

Reliance on representative subsamples of the entire unit is used commonly in scientific studies. Dose-response tests are performed with only a limited number of organisms that represent an equally limited number of taxa. Even remote sensing techniques depend on reflected and emitted electromagnetic radiation, that is a sample of total radiation, both in terms of the patches of ground from which it comes and the parts of the spectrum over which it is measured. The approach to select sampling sites to be truly representative, therefore, is critical to the interpretation of the data collected for a regional risk assessment.

This section briefly discusses the study design considerations unique to regional scale ecological risk assessments. Chapter 2 provides a more detailed discussion of sampling design considerations that generally apply to most ecological risk assessments, and many excellent references are available that provide greater guidance on sampling design consideration (Green, 1979).

4.5.1 STATISTICAL INFERENCE

In a regional-ecological risk assessment, the desire to make statistical inferences about some defined region can be satisfied only when information that is representative of that region is available. In practice, two ways exist to ensure representativeness when selecting sampling sites. The sampling sites must be selected at random from all possible sites, or they must be selected by some systematic process that eliminates subjectivity. Systematic sampling is often easier in regional risk assessments; however, derivation of completely valid estimates of the precision of a characterization based on systematic sampling is impossible. Most forms of statistical analysis require an element of randomness in the sampling design; consequently, some form of random sampling is preferred to measure physical, chemical, or biological variables.  

4.5.2 NUMBER OF SAMPLES

The number of samples required to characterize adequately the condition of, and changes in, an ecosystem on a regional scale depends on two major factors. The first is the natural variability of the characteristic of the ecosystem being measured. An initial task in developing a sampling design for a regional risk assessment, preliminary sampling (i.e., pilot studies), can help determine the inherent variation in characteristics to be measured. Alternatively, determination of variation from data collected for earlier studies in the same region may be possible.

The second factor determining the number of samples required is the desired precision of estimates, which can be expressed as the standard error of the mean of the sample determination. In practice, the number of samples collected for an assessment will be a compromise between the desired precision of the information and the cost of the research. Excellent references are available that offer more detail on determining sampling parameters (Green, 1979; Gilbert, 1987).

A most critical influence on the number of samples collected is the cost of monitoring and analysis. An approach to balance these costs against data uncertainty is the data quality objectives approach (DQO) developed by the US EPA (USEPA, 1991). This approach helps to ensure that the type, amount, and quality of data collected are adequate to meet project objectives in a cost-effective manner. The DQO process results in a quantitative statement of the level of uncertainty in the results that a data user is willing to accept. 

The USEPA Environmental Monitoring and Assessment Program (EMAP) is a national ecological status and trends monitoring program that developed its sampling design based on the following DQOs. An example of the DQO for trend detection is that "EMAP should be able to detect a 2 percent change per year in a condition indicator over a decade with a of 0.2 and b of 0.3 (power of 0.7) for a resource class (e.g., wetlands or forests) on a regional scale." An example of EMAP's DQO for status assessment is that "for each indicator of condition and resource class on a regional scale, estimate the preparation of the resource in subnominal condition within 10 percent (absolute) with 90 percent confidence based on four years of sampling" (Kirkland, 1994).

4.5.3 STRATIFICATION

The precision of regional estimates can be improved by dividing the region of interest into multiple homogeneous units that are less variable than the region as a whole. In characterizing media chemistry, for example, the soil of a region can be stratified according to land use (including forestry, tillage, horticulture, rough grazing, etc.) and soil type, among other factors. Separate random samples from each stratum can be combined in various ways to obtain a more precise characterization of media chemistry than could be obtained by selecting sampling sites in a completely random fashion. Aquatic media may be stratified by salinity and temperature characteristics. For example, the EMAP has been conducting research on the indicators it uses to assess estuarine condition. These samples are stratified on the basis of salinity (marine, polyhaline, mesohaline, oligohaline-tidal, freshwater) to aid in assessing the reliability of using these indicators to reliably distinguish between polluted and unpolluted environments (Holland, 1990).

Linear features such as roads, hedgerows, and streams must be recognized as separate strata in regional assessments. Such linear features often have particular ecological mechanisms and adaptations that enhance biodiversity, heterogeneity, or productivity, and they usually need to be sampled separately. Ecotones, the interfaces between different habitats, also fall in this category of linear systems that should be regarded as separate strata when attempting to characterize a region.

4.5.4 MEASUREMENT OF CHANGE

Regional risk assessments often require measuring spatial or temporal changes in regional scale processes or ecological characteristics. Measuring the status of an ecosystem and detecting trends over time are competing interests: status is generally best assessed by sampling as much of the resource as possible at a given time, while trends are generally best detected by repeating measurements at the same locations or of the same individuals at regular time intervals (Overton et al., 1990). Revisiting sampling sites at too frequent intervals, however, may provide insufficient time for recovery from measurement stress. To detect subtle trends in diffuse subpopulations, alteration-degradation of the sampling location or individuals must be minimized. Another useful approach is collecting measurements from different individuals within the same population over time, while still allowing for assessments of the regional population (Overton et al., 1990). A difficulty of this approach is ensuring that repeat samples are truly from the same population and account for the natural variability within a population.

The points of measurement or the individuals (e.g., trees and animals) must be marked to ensure that repeat measurements are taken at the same place or from the same individuals. However, care must be taken to ensure that the marking does not, itself, introduce a bias. Unfortunately, marks of any kind in the environment often increase the chance of vandalism, leading either to loss of the marking or to a change in the environmental process being measured. The efficiency and precision of measurements of change in ecological systems or processes can be enhanced greatly by sampling with partial replacement.

4.5.5 IMPORTANCE OF PRELIMINARY SAMPLING

Developing an effective sampling design for a regional assessment requires preliminary knowledge of the region and its features to identify the appropriate size of the sampling units, the number of samples required to obtain the desired precision of estimates, and the possible presence of large-scale spatial patterns that necessitate stratification. Because these problems cannot be remedied by computation after the data are collected, preliminary or pilot studies to investigate such issues are highly recommended. Even though pilot studies are time consuming, and may not generate data that are directly useful in the final assessment, pilot studies cannot be circumvented without jeopardizing the validity of the ultimate assessment.

4.6 METHODS TO MONITOR AND ASSESS REGIONAL ENVIRONMENTS

The appropriate choice of methods to collect and analyse data at the landscape or regional level depends on the objectives of the assessment. For the purpose of discussion, these issues are assumed to have been determined, and an experimental design has been assumed to have been selected. Hypothetically, the issue to be addressed is the use of herbicides to control shrubs on extensive rangelands, and management decisions may be needed to assess the effect of this activity on a management unit scale. The spatial scale for management may be on the order of thousands of square kilometres. The experimental design will include the types of data to be collected, the size of the primary sampling unit, the numbers of samples to be collected, and the means to aggregate those samples at the regional level. 

For this illustration, the effects of the herbicide on the soil and plants and on the region as a whole must be well understood. The effects can be considered at the molecular and cellular levels; the extent and distribution of the effects can be considered; and ecosystem and regional effects can be assessed. Clearly, different techniques are required at each level. Concentrations of chemicals must be measured in soil, soil-water, and plant tissue. The effects of the herbicide at the community level may need to be assessed. Methods for spatial characterization, as well as regional aggregation of physical, ecological, and socioeconomic effects, must be selected. Finally, methods to monitor trends in the effects of the chemicals and to monitor the effectiveness of management strategies regarding the use of the chemical must be identified and evaluated. 

Even on the regional scale, the significance and importance of point (or samplebased) measurements and organism-level or mechanistic-level investigations cannot be underestimated. The organism is the smallest unit that interacts directly with the environment (Suter, 1993), and the organism is directly exposed to chemicals that may be toxic. The reproducing population (or, perhaps, the "canopy level" in forestry studies; Mooney et al., 1991) is the smallest ecological unit that persists on a human time scale; hence, it is the lowest level that can be managed and protected. The aggregation of plant and animal populations into communities and their interaction at the ecosystem level form a basic unit of the landscape or region. The reader is referred to Chapter 3 of this report as well as to chapters on the effects of chemicals on forests, arid ecosystems, soils, aquatic systems, and others for more detailed information on these and other techniques. In the following, methods to conduct regional-level environmental monitoring and assessment are discussed, emphasizing ecosystem or landscape-level methods. Methods applied to the organism, population, and community levels are presented in Chapter 3.

4.6.1 REMOTE SENSING TECHNIQUES

Remote sensing techniques, perhaps most useful at the ecosystem and landscape levels, include systems that acquire data at more detailed levels, and then are "scaled up" or extrapolated in various ways. Aggregations of the French Système Probatoire de l'Observation de la Terre (SPOT) and Landsat (Multi-Spectral Scanner (MSS) and Thematic Mapping (TM)) pixels allow scientists and managers to classify or otherwise characterize land-use and land-cover information at the regional level. The relatively high temporal resolution of both systems allows researchers to monitor changes in the attributes of regions over time. Multidate imagery techniques, which allow quantitative assessments of change, are discussed in a subsequent chapter of this report. Developing techniques in landscape ecology allow assessment of fractal dimension, patch, connectivity, and other parameters from which landscapes can be assessed. METEOSAT and the National Oceanic and Atmospheric Administration's AVHRR (Advanced Very High Resolution Radiometer) may allow assessments of very large areas over short time intervals, and enable researchers to scale up information derived from higher resolution remote sensing, field, and laboratory measurements. AVHRR-derived vegetation indices allow for an assessment of biomass and net primary productivity at regional and landscape levels and, especially, for monitoring changes in those variables over time. Mooney et al. (1991) suggested that remotely sensed measurements of APAR might serve as an index for the combined effects of multiple stresses. Although extrapolations to higher levels of organizations can be a source of uncertainty, using remote sensing to assess stress at the regional level can be an effective tool that may lead to new insights about regional and global stressors and effects. 

Several researchers have investigated relationships between ecosystem variables and indices derived from remote sensing data. Peterson et al. (1987) found strong relationships (R2 = 0.91) between Landsat TM near infrared/red ratios (Normalized Difference Vegetation Index, NDVI) and LAI (Leaf Area Index) (Peterson et al., 1987) in closed-canopy, pure conifer forests in west central Oregon. Figure 4.2 (Peterson et al. 1987) illustrates relationships between LAI and NDVI and between NDVI and NPP (Net Primary Production) in conifer forests in Montana.

4.6.2 GEOGRAPHIC INFORMATION SYSTEMS

Analysing the effects of chemicals on the environment at regional levels typically requires compiling and integrating data from disparate sources. Most of these data sources are geographically referenced; that is, the information they contain is specifically referenced to geographic locations. Geographically referenced data sets can be integrated and manipulated using a geographic information system (GIS). A GIS allows spatial data to be compiled and integrated with additional data layers or non-spatial attributes. A GIS also allows the analyst to simulate impacts in the attributes of a given data layer or indicator in response to a hypothesized change in the environment that might stress that indicator in different ways. As a modelling tool, GISs have been used to estimate potential environmental impacts over large, spatially varying ecoregions. GIS techniques have allowed analysts to identify spatial and temporal trends based on ecosystem and landscape level data, as well as the projected impacts from current and future development. Consequently, GIS is an especially powerful tool for landscape and regional assessments. Excellent references exist on the concepts and applications of GISs (Ripple, 1987; Koeln et al., 1994).

4.7 CHARACTERIZING RISK AT THE REGIONAL SCALE

4.7.1 STATISTICAL METHODS

A statistically-based design is important in any regional assessment of risk. This design will ensure that the data acquired during the program are of sufficient quantity and quality to answer the questions posed by the risk managers and regulators who initiated the study. The use of a sound statistical design and methods also provides a firm basis to deal with uncertainty in the risk analysis. In particular, statistical theory defines the conditions under which the scientist is able to make valid inferences about some defined region or landscape. Those conditions include the formal definition of the region in both time and space, and the examination of "fair" samples from locations, organisms, soils, water, and air from that "population." The statistician's definition of "fair," however, is strict, in that the samples must be derived objectively from the population and, in practice, such samples must be taken either systematically or at random. Inferences based on samples or case studies that are chosen subjectively by the scientists concerned with the risk analysis are not acceptable as a basis for valid inferences about the population (Jeffers, 1988b).

At an analytical level, statistical methodology provides a range of techniques for determining the level of uncertainty in any measurement or combination of measurements. Those techniques depend on replication and random sampling in the original measurements and greatly enhance the control of extraneous variation by the use of stratification, covariance, and ancillary measurements (Green, 1979). The dangers of extrapolation beyond the range of the original data remain a constant potential source of error, as does the accumulation of errors in the aggregation of data over space and time. However, the failure of basic assumptions in traditional statistical methods has become less important with the introduction of new computer-based techniques (Lunn and McNeil, 1991). 

4.7.2 MODELLING

Models, defined here as the formal statement of relationships in physical or mathematical terms, play an integral role in regional environmental monitoring and assessment. A mathematical model may be used to assess the transport and fate of contaminants as well as effects at all ecological levels from organismal to regional. Models used in regional ecological risk assessments may be deterministic (not incorporating random events) or may be stochastic or probabilistic (incorporating random events or variables). 

All wholly deterministic modelling formulations depend on estimation of the model parameters as a special case of statistical inference (i.e. from data obtained as fair samples from some defined population). All models used to synthesize data have specific assumptions that need to be checked before the model is used for making decisions. Indeed, verifying and validating models against new data sets are essential components of the systems analysis in which the model building is embedded. The sensitivity of models to small changes in their basic parameters, a process known as sensitivity analysis, should be conducted whenever they are used.

Ideally, modelling should be embedded in a formal process of systems analysis that enables comparison of alternative models and facilitates validation and refinement of models (Jeffers, 1978). The range of alternative models may include qualitative or semi-quantitative representations of the functional relationships between the landscape elements. More usually, it will include deterministic mathematical models of the basic ecological processes, expressed as differential or difference equations. The effects of airborne and waterborne pollutants on agricultural and forest ecosystems have all been expressed in this way, as have the consequences of alternative energy strategies on political and continental regions. 

Increasingly, however, risk analysis and other disciplines employ stochastic (hence, non-deterministic) models in which probability distributions represent the uncertainty in the relationships between several, perhaps many, variables. Starfield and Bleloch (1986) and Tijms (1986) described the use of such models for a variety of applications, including wildlife conservation and management. Most stand simulation models used to evaluate the effects of pollutant levels, pest infestations, or hydrological processes on forest, agricultural, or wetland landscapes are of this type. Markov models, which depend on the probabilities of transitions to and from a limited number of states, are widely used to simulate ecosystem succession (Jeffers, 1988a), and play an important role in creating regional scale models for ecological risk assessment, particularly when combined with methods of pattern analysis (Howard, 1991). The methods of linear, non-linear, and dynamic programming that emerged from the operations research applications of World War II have long been adopted by environmental managers in the search for optimal control strategies, especially in the management of regional scale systems for multiple goals. An extension of game theory to meta-game theory provides an appropriate model for describing conflict and finding stable resolutions to conflict, and has been used widely to model and resolve conflicts at the regional scale, for example, conflicts between polluting industries and national or regional control agencies. A recent advance in modelling of systems of all sizes is the application of network and topological models (Barnsley, 1988; Higashi and Burns, 1991).

With increasing development of expert systems, namely computer systems aimed at offering the same advice with justification that would be given by a human expert, the role of the "rule bases" (on which such systems depend) is increasing in interest. Such "rule bases" typically consist of sets of logical statements, ideally in common language, to help the manager choose the appropriate option(s) to manage a system. Research on the development of rule-based systems as a basis for environmental decision support systems is currently proceeding (Guariso and Werthner, 1989). Appropriate computer-based methods to extract a "rule base" from data have also been developed and are currently being refined.

Regional scale models are already being used to synthesize regional-level data and make inferences for regional-scale management of environmental risks. The best-known examples are the models used to quantify impacts of acidic deposition on terrestrial and aquatic ecosystems in Europe and North America. Hordjik (1991) described the use of a regional scale acid deposition model in designing sulphur emission control strategies for western Europe. Wright et al. (1991) and Baker et al. (1990) described regional-scale models used to assess effects of acidic deposition on lakes and streams. Although these models have not been used extensively for management purposes, models of the effects of climate change on regional ecosystem characteristics are being developed and tested (Agren et al., 1991).

4.7.3 SOURCES OF UNCERTAINTY

Variation is an inherent property of living organisms and their environment. The genetic processes of sexual reproduction result in a range of phenotypes that expressed similar morphological characteristics or reactions to stimuli or environmental conditions. Chemical and physical processes within the environment also result in heterogeneity in both time and space; consequently, every measurement of the environment is subject to variability that needs to be determined and, if possible, controlled to improve the ability to perform assessments.

This variability is further compounded by the fact that the accuracy with which the physical, chemical, or biological properties of the environment can be measured is at least partly determined by the cost and technical sophistication of the instrumentation available and the technical abilities of the scientist or technician performing the measurements. Errors are possible even with the most technically advanced instrumentation.

In any investigation of complex systems, hierarchy theory emphasizes the importance of working at the coherent level in the hierarchy (Allen and Starr, 1982). The perceived variability in environmental systems may be increased by attempts to link ecosystem processes to forcing functions, such as temperature, precipitation, and moisture content, at too high a level in the hierarchy. Similarly, attempts to characterize ecosystems on the basis of processes at too low a level in the hierarchy may replicate variability that is apparently unrelated to ecosystem function, for example, attempting to determine the response of forests to airborne pollutants on the basis of measurements of photosynthesis in individual leaves. Choosing the proper level in the hierarchy to characterize risk is a significant step in reducing uncertainty in any assessment.

Regional scale risk assessment does not end with the analysis and determination of uncertainty. The information derived from the assessment and uncertainty analysis must be presented to decision makers and risk managers in a clear and effective manner, because decisions are often made on those parts of the assessment that contain the highest degree of uncertainty.

4.8 RECOMMENDATIONS TO ENHANCE REGIONAL ENVIRONMENTAL MONITORING AND ASSESSMENTS

Encourage additional support for research on:

1. ecotoxicological and ecological research at the landscape level;

2. concentrations of organics and trace elements in the environment to establish baseline values and identify long-term trends;

3. interactions between landscape components;

4. sensitivity of various regional landscapes to chemicals;

5. steps in the assessment-management process that have the weakest foundations.

Support development of improved methods to:

6. evaluate exposures and effects at regional levels;

7. determine ecological and socioeconomic values at a regional scale;

8. develop methods to scale from point base to region;

9. develop better methods to organize and present regional data;

10. develop better methods to organize, structure, and present regional ecosystem data for multiuser access.

Encourage and enhance:

11. increased use of synthesis (modelling) in addition to analysis at all spatial and temporal scales in risk assessment;

12. incorporation of sensitivity analysis of the models used in a risk analysis;

13. improved statistical education for those who conduct regional risk assessments and for those who use them;

14. education in effective risk communication methods for environmental professionals.

4.9 REFERENCES

Agren, G.I., McMurtrie, R.E., Parton, W.J., Pastor, J., and Shugart, H.H. (1991) State-of the-art models of production-decomposition linkages in conifer and grassland ecosystems. Ecol. Appl. 1, 118-138.

Allen, T.F.H., and Staff, T.B. (1982) Hierarchy: Perspectives for Ecological Complexity. University of Chicago Press, Chicago.

Baker, J.P., Freda, J., Christensen, S.W., and Sale, M.J. (1990) Biological effects of changes in surface water acid-base chemistry. In: Acidic Deposition: State of Science and Technology. NAPAP Report 13. National Acid Precipitation Assessment Program, Washington, D.C.

Barnsley, M. (1988) Fractals Everywhere. Academic Press, London.

Costanza, R., Sklar, F.H., and White, M.L. (1990) Modeling coastal landscape dynamics. Bioscience 40, 91-107.

Dale, V.H., and Gardner, R.H. (1987) Assessing regional impacts of growth declines using a forest succession model. J. Environ. Man. 24, 83-93.

Dobson, S. (1985) Methods for measuring effects of chemicals on terrestrial animals as indicators of ecological hazard. In: Vouk, V.B., Butler, G.C., Hoel, D.G., and Peakall, D.B. (Eds.) Methods for Estimating Risk of Chemical Injury: Human and Non-Human Biota and Ecosystems, pp. 537-548. SCOPE 26. John Wiley & Sons, New York.

Gilbert, R.O. (1987) Statistical Methods for Environmental Pollution Monitoring. Van Nostrand Reinhold, New York.

Graham, R.L., Hunsaker, C.T., O'Neill, R.V., and Jackson, B.L. (1991) Ecological risk assessment at the regional scale. Ecol. Appl. 1(2), 196-206.

Green, R.H. (1979) Sampling Design and Statistical Methods for Environmental Biologists. John Wiley & Sons, Chichester.

Guariso, G, and Werthner, H. (1989) Environmental Decision Support Systems. Ellis Horwood Ltd, Chichester.

Higashi, M., and Burns, T.P. (1991) Theoretical Studies of Ecosystems: The Network Perspective. Cambridge University Press, Cambridge.

Holland, A.F. (Ed.) (1990) Near Coastal Program Plan for 1990: Estuaries. US Environmental Protection Agency, Environmental Monitoring and Assessment Program, Washington, D.C.

Holling, C.S. (1992) Cross-scale morphology, geometry and dynamics of ecosystems. Ecol. Monogr. 62(4), 447-502.

Hordjik, L. (1991) Use of the RAINS model in acid rain negotiations in Europe. Environ. Sci. Technol. 25, 596-603.

Howard, P.J.A. (1991) An Introduction to Environmental Pattern Analysis. Parthenon Publishing, Casterton.

IUCN (1980) World Conservation Strategy. International Union for the Conservation of Nature, Gland, Switzerland.

Jeffers, J.N.R. (1978) An Introduction to Systems Analysis: With Ecological Applications. Edward Arnold, London.

Jeffers, J.N.R. (1988a) Statistical and mathematical approaches to issues of scales in ecology. In: Rosswall, T., Woodmansee, R.G., and Risser, P.G. (Eds.) Scales and Global Change: Spatial and Temporal Variability in Biospheric and Geospheric Processes. SCOPE 35. John Wiley & Sons, Chichester.

Jeffers, J.N.R. (1988b) Practitioner's Handbook on the Modelling of Dynamic Change in Ecosystems. SCOPE 34. John Wiley & Sons, Chichester.

Kirkland, L.L. (1994) EMAP Quality Management Plan. US Environmental Protection Agency, Washington, D.C.

Koeln, G.T., Cowardin, L.M., and Strong, L.L. (1994) In: Bookout, T.A. (Ed.) Research and Management Techniques for Wildlife and Habitats, 5th edn., pp. 540-566. The Wildlife Society, Bethesda.

Kutz, F.W., Linthurst, R.A., Riordan, C., Slimak, M., and Frederick, R. (1992) Ecological research at EPA: new directions. Environ. Sci. Technol. 26, 860-865.

Lunn, A.D, and McNeil, D.R. (1991) Computer-Interactive Data Analysis. John Wiley & Sons, Chichester.

Meadows, D.H. (1991) The Global Citizen. Island Press, Washington, D.C.

Messer, J.J., Linthurst, R.A., and Overton, W.S. (1991) An EPA program for monitoring ecological status and trends. Environ. Monitor. Assess. 17, 67-78.

Mitchell, M.F., and Roberts, JR. (1984) A case study of the use of fenitrothium in New Brunswick: The evolution of an ordered approach to ecological monitoring. In: Sheehan, P.J., Miller, D.R., Butler, G.C., and Bourdeau, P.H. (Eds.) Effects of Pollutants at the Ecosystem Level. SCOPE 22. John Wiley & Sons, Chichester.

Mooney, H.A., Winner, W.E., and Pell, E.J. (1991) Response of Plants to Multiple Stresses. Academic Press, San Diego.

NAPAP (1991) 1990 Integrated Assessment Report. US National Acidic Precipitation Assessment Program, Washington, D.C.

Nash, B.L., Davis, D.O., and Skelly, J.M. (1992) Forest health along a wet sulphate/pH disposition gradient in north-central Pennsylvania. Environ. Toxicol. Chem. 11, 1092-1104.

Omernik, J. (1987) Ecoregions of the conterminous United States. Ann. Assoc. Am. Geogr. 77(1), 118-125.

Overton, W.S., White, D., and Stevens, D.L. (1990) Design Report for EMAP. Report No. EPA 600/3-91/053. US Environmental Protection Agency, Corvallis.

Peterson, E.B., Chan, Y.-H., Peterson, N.M., Constable, G.A., Caton, R.B., Davis, C.S., Wallace, R.R., and Yarranton, G.A. (1987) Cumulative Effects Assessment in Canada: An Agenda for Action and Research. Cat. No. EN106-7/1987E. Minister of Supply and Services, Canada, Ottawa.

Pulliam, H.R. (1988) Sources, sinks, and population regualtion. Am. Natur. 132, 652-661.

Ripple, W.J. (Ed.) (1987) GIS for Resource Management. American Society for Photogrammetry and Remote Sensing, and American Congress on Surveying and Mapping, Falls Church, Virginia.

Sheehan, P.J. (1984a) Effects on individuals and populations. In: Sheehan, P.J., Miller, D.R., Butler, G.C., and Bourdeau, P.H. (Eds.) Effects of Pollutants at the Ecosystem Level. SCOPE 22. Jobn Wiley & Sons, Chichester.

Sheehan, P.J. (1984b) Effects on community and ecosystem structure and dynamics. In: Sheehan, P.J., Miller, D.R., Butler, G.C., and Bourdeau, P.H. (Eds.) Effects of Pollutants at the Ecosystem Level. SCOPE 22. John Wiley & Sons, Chichester.

Sheehan, P.J. (1992) Ecotoxicological considerations. In: Cote, R.P., and Wells, P.G. (Eds.) Controlling Chemical Hazards. Unwin Hyman, Boston.

Sheehan, P.J., Baril, A., Mineau, P., Smith, K.K., Hartenist, A., and Marshall, W.K. (1987) The Impact of Pesticides on the Ecology of Prairie Nesting Ducks. Technical Report Series No. 19. Environment Canada, Canadian Wildlife Service, Ottawa.

Skelly, J.M. (1980) Photochemical Oxidant Impact on Mediterranean and Temperate Forest Ecosystems: Real and Potential Effects. Effects of Air Pollutants on Mediterranean and Temperate Forest Ecosystems. General Technical Report PSW-43. Pacific Southwest Forest and Range Experimental Station, Berkeley.

Starfield, A.M., and Bleloch, A.L. (1986) Building Models for Conservation and Wildlife. Macmillan, London.

Suter, II, G.W. (1993) Ecological Risk Assessment. Lewis, Boca Raton, Florida.

Thomann, R.V. (1972) Systems Analysis and Water Quality Management. McGraw-Hill, New York.

Tijms, H.C. (1986) Stochastic Modelling and Analysis. John Wiley & Sons, Chichester.

Turner, M.G. (1987) Simulation of landscape changes in Georgia: A comparison of 3 transition models. Landscape Ecol. 1, 29-36.

USEPA (1991) Planning for Data Collection: The Data Quality Objectives Process for Environmental Decisions (Draft, 16 October 1991). US Environmental Protection Agency, Office of Research and Development, Quality Assurance Management Staff, Washington, D.C.

Wright, R.F., Cosby, B.J., and Hornberger, G.M. (1991) A regional model of lake acidification in Norway. Ambio 20, 222-225.