Assessments of Wetland Functions: What They Are and What They Are Not THOMAS HRUBY Washington State Department of Ecology SEA Program P.O. Box 47600 Olympia, Washington 98504, USA ABSTRACT / Many methods have been developed over the last two decades to provide information about wetland functions, but there has been little discussion of the models and algorithms used. Methods for generating information about wetlands were analyzed to understand their similarities, differences, and the type of information provided. Methods can first be grouped by the type of information they provide— classifications, characterizations, ratings, assessments, and evaluations. Methods that characterize, rate, or assess wetlands may generate information using one of two conceptual

Over a decade ago Jon Kusler (1986) commented that methods for analyzing wetland functions developed up to then were usually species specific, short term, and narrow in their scope. There was a need for rapid, comprehensive approaches that assessed a range of wetland functions and that could be used in a regulatory or planning context, not a scientific or research one. The need is for comprehensive information that will support sound decisions about compensatory mitigation, acquisition, restoration, and assessment of anthropogenic impacts (National Research Council 1995). Many methods have been developed in the last decade to fill this need. Some are regional in scope, some national, and some try to meet specific local planning or regulatory needs. Many are modifications of a preceding method. Currently the US Army Corps of Engineers is leading a nationwide effort to develop a series of regional function assessment methods based on a hydrogeomorphic classification of wetlands (the HGM approach to assessment; Smith and Bartoldus 1995, Brinson 1995, 1996). This nationwide effort is significant. By 1996, at least $5 million had been spent on the project (C. Rhodes, US EPA, personal communication). Although the last decade has seen many new methods developed to assess wetland functions, there has been little discussion of the assumptions on which the

KEY WORDS: Wetlands; Functions; Assessment; Models; Methods

Environmental Management Vol. 23, No. 1, pp. 75–85

approaches—logic and mechanistic. Most methods that generate a numeric assessment of performance or value of wetland functions rely on the mechanistic approach to constructing models. Rapid assessment methods based on mechanistic models, however, do not assess the rates or dynamics of ecological processes occurring in wetlands. Rather, they provide a clear and concise way of organizing our current, and often subjective, knowledge about wetland functions. This is one limitation of current methods that is often misunderstood both by wetland managers and the scientific community. The advantages and limitations of the assumptions and the computational elements inherent in these approaches are discussed to provide wetland managers and regulators a better understanding of the information they are using.

methods are based, the rationale for choosing the computational models, or the type of information generated by these methods. Furthermore, few of the rapid methods have undergone a validation, where the results of a method are tested against actual measurements of levels of functions (World Wildlife Fund 1992). As a result, wetland scientists often greet new assessment methods with skepticism and question their scientific validity (Brinson and others 1997a). Rapid assessment methods that generate numeric results are especially suspect because they often lack adequate data (Adamus 1986). Environmental managers, on the other hand, are confused about the type of information the methods provide. A recent meeting of the Association of State Wetland Managers in Annapolis, Maryland, USA, in March 1997, and a recent letters by Brinson and others (1997a,b) and Hruby (1997), have highlighted the misconceptions that some wetland managers may have about the type of information provided by assessment methods. They have trouble differentiating among: (1) the levels at which wetlands perform functions and the values given to those functions; (2) the meaning of indices generated by different methods; and (3) the potential that a wetland has to perform a function versus the actual level of performance. Understanding and correcting misconceptions is important because many regulatory and management

r 1999 Springer-Verlag New York Inc.

76

T. Hruby

decisions are based on the results of an assessment method. Wrong decisions may be made, or wetlands may be not be adequately protected, if inappropriate methods are used (Kusler and Niering 1998). This paper examines existing assessment methods and proposed methods (e.g., the US Army Corps of Engineers HGM approach to developing methods) to identify some of their basic assumptions and computational elements. My hope is to provide a better understanding of the information that wetland assessment methods do and do not provide. The first section describes the types of information current rapid assessment methods provide. The second section describes the two types of computational approaches commonly used. The third section describes the structure and elements of mechanistic assessment models that are the basis of most current rapid approaches to wetland assessment.

Information Provided by Wetland Assessment Methods Methods for organizing our knowledge about wetlands have been called classifications, categorizations, characterizations, ratings, assessments, and evaluations. These groupings are meant to indicate the type of information a method provides. Unfortunately, the scientific community has been sloppy in the use of these terms to the extent of mistitling many of the analytical tools developed [e.g., the wetland evaluation technique (WET) of Adamus and others (1987) is actually a rating, not an evaluation]. Table 1 attempts to clarify this problem by categorizing some of the commonly used methods according to the actual information provided rather than by its title. The groupings are based on the following definitions from Webster’s Seventh New Collegiate Dictionary (Anonymous 1963). Classification/categorization—a systematic grouping into categories according to established criteria or shared characteristics. The two most common classifications are those of Cowardin and others (1979), based on shared characteristics of vegetation and water regime, and the hydrogeomorphic classification (Brinson 1993) based on shared characteristics of geomorphic setting and water regime. The criteria used for grouping are generally not linked to specific functions, and thus classifications are not true methods for assessing functions. They can, however, provide a basis on which to develop assessment methods (Brinson 1995). Characterization—a grouping by a distinguishing trait, quality, or property. For example, the Oregon method (Roth and others 1993) characterizes wetlands by the

properties of ‘‘provides’’ a specific function; ‘‘has the potential to provide’’ a function; or ‘‘does not provide’’ a function. These are three distinct attributes that give some information about whether a wetland performs a function, but no information is generated about levels of performance. Rating—classification based on a grade. Ratings usually group wetlands using the qualitative grades of high, medium, or low on a variety of scales such as the performance of a function or its value. The wetland evaluation technique (WET) (Adamus and others 1987) is probably the most widely used rating method. Assessment—an estimate or determination of importance or value. This is the first level at which numbers are generated to represent an estimate of performance, value, or functional value of a function. All commonly used numeric methods fall into this category (see Table 1). All methods to date provide only an assessment that is relative to some predetermined standard. They do not provide an assessment of actual levels of performance or value. In my judgment, the term ‘‘assessment’’ is one of the most commonly misused words in the lexicon of wetland scientists. Almost any method developed is now called an assessment, regardless of whether it might actually be a categorization, a rating, or an true assessment. Evaluation—a determination or fixing of value. The fixing of value for any item is based on having a generally acceptable currency. Up to now the only currency used has been monetary, and evaluations of wetland functions have most often tried to generate dollar values based on different types of economic models such as the travel cost method, random utility model, hedonic techniques, contingent valuation method (Titre and Henderson 1989, Lipton and others 1995), or willingness-to-pay method (Farber and Costanza 1987).

Types of Computations Used The ways in which data are analyzed within a method are called ‘‘models’’ or ‘‘algorithms’’ because most rely on equations or other mathematical rules for achieving a result. I will be using the term ‘‘models’’ in this discussion to represent the individual equations, and ‘‘methods’’ to represent a collection of models. Generally, a method has a separate model for each wetland function analyzed. There are two types of computational approaches commonly used—logic and mechanistic. A model using a logic approach has a qualitative, verbal description that produces a result. In a logic model, the conditions found in a wetland (variables) are combined by logic

Wetland Functions

Table 1.

77

Categorization of some recent methods used to analyze wetlands and some of their characteristics Method (names abbreviated)

Type of model

Range of scores

Reference

Cowardin

logic

not applicable

Hydrogeomorphic Hydrologic

logic logic

not applicable not applicable

Characterizations: grouping by distinguishing attributes

Washington State rating system

logic/mechanistic

Categories 1–4

Oregon method

logic

present/absent

Ratings: classification based on position on a scale

WET NC-CREW

logic logic

H, M, L H, M, L

Minnesota method

logic

H, M, L

EPA synoptic method

mechanistic

H, M, L

Reppert Connecticut method New Hampshire method Hollands

mechanistic mechanistic mechanistic mechanistic

0–1 0–1 0–1 0–1

Ontario method HEP’s

mechanistic mechanistic

0–1 0–1

HAT (for birds) LarsonGolet Eval. planned wetlands

mechanistic mechanistic mechanistic

0–1 0–1 0–1

IVAa HGM Riverine Guidebooka

mechanistic mechanistic

0–100 0–1

Reppert and others 1979 Amman and others 1986 Amman and Stone 1991 Hollands and Magee 1985 Glooschenko 1983 US Fish and Wildlife Service 1980 Cable and others 1989 Larson 1976 Bartoldus and others 1994 Hruby and others 1995 Brinson and others 1995

Willingness-to-pay

Willingness-to-pay

$ value

Farber and Costanza 1987

Classifications: systematic grouping based on shared characteristics

Assessments: estimate of importance or value

Evaluations: fixing of value aThese

Cowardin and others 1979 Brinson 1993 Gilvear and McInnes 1994 Washington State Department of Ecology 1993 Roth and others 1993 Adamus and others 1987 North Carolina Division of Coastal Management 1994 US Army Corps of Engineers St. Paul 1988 Leibowitz and others 1992

are not actual methods but rather provide guidance for developing regionally specific methods.

statements such as ‘‘and,’’ ‘‘or,’’ ‘‘if . . . then’’ to establish a characterization or rating. Logic models have also been called rule-based models (Starfield and others 1989) and descriptive models (Terrell and others 1982). Probably the best known method using logic models is the WET (Adamus and others 1987). Wetland methods based on a mathematical aggregation of numeric data can be called mechanistic because they follow the mechanistic approach to model development described by the US Fish and Wildlife Service (1981) for habitat evaluation procedure (HEP) models. In mechanistic models, environmental characteristics found in a wetland are treated as variables in an equation. Different conditions of these variables are

assigned numbers and combined mathematically to generate an index or score. Examples of wetland methods using mechanistic models are Reppert and others (1979), the Connecticut method (Amman and others 1986), the indicator value assessment (IVA) (Hruby and others 1995), and methods developed using the HGM approach (Brinson 1995, 1996). Almost all wetland assessment methods that generate a number use the mechanistic approach to model development (see Table 1).

Properties of Assessment Models Rapid wetland assessment methods based on logic or mechanistic approaches do not assess the rates or

78

T. Hruby

dynamics of ecological processes occurring in wetlands. Rather, they provide a clear and concise way of organizing our current, and often subjective, knowledge about wetland functions. This is a limitation of current methods that is often misunderstood both by wetland managers and by the scientific community. The misunderstanding is fostered by the fact that many functions are defined as ecological processes that are usually expressed as rates. For example, the function, ‘‘organic carbon export’’ in Brinson and others (1995) is defined as ‘‘export of dissolved and particulate organic carbon from a wetland.’’ When an assessment method provides a number for this function, it is easy to assume that this represents the grams of carbon exported per year, especially when the score is defined as a ‘‘level of performance’’ (Smith and others 1995). A wetland with a score of 0.5 would then be expected to export half the carbon exported by a wetland scoring a 1. Unfortunately, this would be a misinterpretation of the results. Scores are only a numeric representation of a qualitative assessment. A result of 0.5 from an assessment method means that the wetland is judged to be performing a function at a moderate rate (or level of sustainability) relative to those considered to be performing at the highest levels and those performing at the lowest levels. The assessment methods available, and under development, are modeling a process of judgment used by experts to assess how well wetlands perform functions or how sustainable the functions might be. They are not mathematical representations of the actual environmental processes taking place. Measuring the rates or dynamics of environmental processes requires intensive sampling because the processes are highly variable in both space and time. Such procedures, however, are not possible if the method is to be rapid. Rapid, for most wetland managers and environmental consultants, means that a result can be obtained with one site visit and the entire process of data collection and analysis should take less than one day for a single site. Models assessing wetland functions are constructed as a set of hypotheses about relationships between environmental conditions and the performance or sustainability of a function. The hypotheses are developed in three stages during the process of model construction. First, variables are chosen that represent key aspects of performance, value, or sustainability. Second, the relationship between each variable and performance is translated into a verbal and/or numeric hypothesis x value of variable 5 y level of performance or sustainability. In mechanistic models, numeric values are then assigned to different conditions of a variable.

Finally, the scores assigned variables are aggregated by way of an equation to yield a single numeric or qualitative description of a wetland function. For example, a model for the function ‘‘removing sediments’’ might be phrased as follows: ‘‘The performance of a wetland in removing sediments from incoming surface waters is based on its ability to reduce water velocities. This environmental condition then becomes the variable in an equation. The equation for ‘‘removing sediments’’ would be: performance 5 reduction in water velocity. It will probably not be possible to measure how much a wetland reduces water velocities to estimate sediment removal. Such estimates would require measuring changes in current velocities over the entire wetland over at least one year. As a result, model developers might chose several indicators that suggest the wetland provides relatively high amounts of velocity reduction. The verbal description of the model might then be rephrased as ‘‘the potential that a wetland has to remove sediments from incoming surface waters is based on its ability to reduce water velocities as determined by its physical structures (plant stems, etc.) near the ground surface, the size and shape of its outlets, and the depth of water in the wetland.’’ Dense erect vegetation near the ground is an indicator that a wetland has the potential for effectively removing sediments. The erect vegetation will reduce the velocity of floodwaters and allow the sediments to fall out of the water column. The indicator, however, is not correlated with any specific rates of sedimentation because these depend on the amount of sediment coming into the wetland and on the actual velocity of water coming into the wetland. A constricted outlet also acts to reduce water velocities, as do areas of deep water (Fennessey and others 1994). These other two characteristics are also indicators of velocity reduction. The equation for removing sediments could then be rewritten as: performance 5 physical structure 1 outlets 1 depth of water. In a logic model the level of performance would be described using conditional phrases such as ‘‘the wetland rates high for removing sediments if it has a constricted outlet and a area of deep water and erect vegetation over more than 80% of its area.’’ In a mechanistic model, scores would be assigned to different types of physical structures (e.g., dense persistent erect vegetation with more than 100 stems/m2 might be given a score of 1, erect vegetation with 50–100 stems/m2 a score of 0.5, etc.). Different outlet characteristics and different depths of water would also be assigned scaled scores in this manner. A score for the function is then calculated by summing the scores of the three variables.

Wetland Functions

Model builders often have to develop subjective hypotheses on how variables combine to determine performance, because specific data on the actual relationships are lacking. Variables are chosen and scaled based on the judgment of the team or the authors developing the model. Verification and validation is usually done after the release of a method and then is rarely documented (Brooks 1997). It is easier to understand the information provided by logic and mechanistic models if they are treated as environmental decision-making models (also known as ‘‘multiple criteria assessment’’ models). The basic difference between ratings and assessments is that the decision criteria in assessments are given a score and combined to generate a numeric index. Decision-making models represent ‘‘the acquisition and merging of subjective, expert knowledge. Often several persons with varying backgrounds are to be taken into the analysis, e.g., engineers, ecologists, economists, managers, and politicians’’ (Varis and others 1994). Each variable in a model represents a decision criterion used to establish a level of performance, sustainability, or value, rather than an independent variable that estimates the rate of an environmental process. Variables in decision-making models and in wetland assessment methods can be both qualitative descriptions of environmental conditions or some quantitative measure of a condition. In order to combine both types of variables in one equation, they first need to be converted to a scaled number. The computational techniques for combining the two types of variables and the statistical procedures for testing significance under this scenario are well developed (Voogd 1982, Munda and others 1994, Varis and others 1994).

Structure and Elements of Mechanistic Models As mentioned above, mechanistic models are simple numeric representations of hypothesized relationships between environmental characteristics of a wetland, its surrounding landscape, and the performance, value, or sustainability of a function. This section describes the components of mechanistic models (variables and indicators) and some of the conventions used in computations. Variables in Mechanistic Models Variables represent the environmental characteristics that are judged to be important in the performance of a function. These are the characteristics a wetland expert may consider in judging how well a wetland

79

performs a function. The measurements of these specific environmental conditions, or a qualitative description of their condition, are then assigned a score that is used in the equation of the model. Scores for variables are assigned in one of two ways: the first is strictly based on the judgment of those developing models; the second is to assign scores based on conditions found in a number of specific sites that reflect the range of variability within a region (reference sites). The use of data from reference wetland sites to scale the scores for variables is one of the significant features of the HGM approach. Use of Indicators as Surrogates for Variables When it is not feasible to use a variable because its condition cannot be rapidly assessed, it may be possible to assign a score based on an indicator of that variable. Indicators are easily observed characteristics that are correlated with a quantitative or qualitative measure of an environmental variable. Most indicators are fixed characteristics that describe the structure of the ecosystem or its physical, chemical, and geologic properties (Brinson 1995). Such indicators are time-independent conditions (on the scale of most environmental processes), and thus cannot reflect actual rates of performance. Rather they reflect the potential or probability that functions are performed at a certain level. Model scores based on indicators, therefore, do not reflect the levels at which a function may actually be performed. Instead, they estimate the potential or probability that a function is being performed. In the example described above, the reduction of water velocities by the wetland cannot be measured directly. The model is based on structural characteristics that ‘‘indicate’’ a reduction in velocity is taking place. Scores are assigned different indicators based on how well the characteristics are expected to reduce velocities. Thus, a very constricted outlet would have a higher score in the model than an unconstricted outlet because it is expected to reduce velocities more. A wetland with a constricted outlet has a higher potential to reduce velocities than one with an unconstricted outlet under the same conditions of incoming water velocities and sediment loading. As with variables, indicators can be both qualitative and quantitative. Examples of qualitative indicators are ‘‘wetland has a mixed canopy of deciduous and evergreen trees’’ or ‘‘wetland has a surface water outlet.’’ Quantitative indicators might be ‘‘wetland has 150 trees/hectare’’ or ‘‘wetland has water depths between 20 and 40 cm during the winter.’’

80

T. Hruby

Normalizing Scores In mechanistic models, the equation is usually normalized to 1 or 100 for each function. Normalizing is important because each function may have a different number of variables or different mathematical relationships, with correspondingly different maximum scores. In many mechanistic models, equations are normalized by using the total number of variables in the denominator. This approach seems to have its roots in the US Fish and Wildlife Service’s habitat evaluation procedures (HEPs). The HGM approach, which also traces its roots to HEP (Brinson 1996), has also incorporated this mathematical element. If a model equation contains three variables, the equation is written as:

Thus, a wetland with either the best organic soils or the best clay soils for adsorption would, at most, score a 0.5 for its performance because the denominator is 2. The problem is exacerbated if functions are defined as a group of environmental processes such as ‘‘removal of compounds.’’ This function represents a variety of processes including the removal of nitrogen, phosphorus, potassium, metals, and pesticides (Brinson and others 1995). A wetland could not score high for ‘‘removal of compounds’’ unless it scores high for all processes modeled. One approach being used by developers of HGM methods to resolve this problem is to incorporate logic submodels in scaling a variable (R. Brooks personal communication). In the example given above, the equation would be written with one variable:

score (index) for function 5 (V1 1 V2 1 V3)/3 Each variable is scaled between 0 and 1, and thus, the maximum score for a function is also 1. An assumption inherent in this computational approach is that there is only one unique set of conditions (variables) that results in a high score for a function. A function can score 1 only if all three variables also score 1. If a wetland can perform the same function at high levels by way of different environmental conditions (variables), it cannot score 1. This inherent assumption from the computational methods used has a significant impact on the interpretation of model results. I will try to describe the issue using a simplified example. Assume that the wetland process (function) being modeled is the removal of one compound, phosphorus (P), by adsorption to wetland soils. Wetlands can effectively remove P by two separate adsorption processes: binding to clay particles or binding to organic matter (Mitsch and Gosselink 1993). Wetlands that have either a high organic content in their soils or a high clay content can be effective at removing P through adsorption. If the model contains only one variable for the adsorption process (index 5 V1/1), model developers would have to chose either the percentage of organic matter or the percentage of clay in the soil as the variable. If the percentage of organic content is used as a variable, then wetlands with a high organic content would score 1 for V1, and 1 for the overall index of function. Wetlands that remove P by adsorption to clay, however, would score 0 if their soils have no organic content. If, however, both adsorption processes are modeled using separate variables, then no site could score 1. The equation would have two variables: score for function 5 (V1 1 V2)/2

P uptake 5 Vsorp/1 where the highest value for Vsorp would be determined as: Vsorp 5 1 if wetland has 100% organic soils or 100% clay soils. Another solution, one that avoids developing submodels for each variable, is to normalize by the highest scoring wetland or group of wetlands. This is the approach used in the IVA (Hruby and others 1995) and in the new methods being developed in Washington State (Hruby and others 1997).

Discussion Grouping methods by the type of information they generate and the structure of their models can help us understand how best to use them. Characterizations and rating methods that use logic models provide qualitative results. The choice of qualitative approaches is often intentional because: (1) the methods were designed as screening or planning tools that were meant to be general [e.g., WET (Adamus and others 1987), Oregon Method (Roth and others 1993)], or (2) the authors believed current scientific data inadequate to generate numeric scores (Adamus 1986). Recent efforts in developing analytical methods for wetlands have focused on numeric assessments (Brinson 1995, Hruby and others 1995). The previous discussion of assessment models, however, may lead to certain questions regarding their usefulness for wetland managers and regulators. Specifically, the questions that need to be addressed are: 1.

Is it appropriate to represent untested hypotheses numerically to provide an estimate of performance of a wetland function? In other words, is there any scientific validity to using numeric models for

Wetland Functions

2.

3.

estimating performance of functions given the limitations set by the need for rapid assessment, and our current lack of knowledge? If we agree that numeric models provide usable results, then are mechanistic models the appropriate ones to use? Other options might include developing numeric assessments based on logic, multivariate, or more complex nonlinear approaches. Is the use of structural indicators, which represent nondynamic conditions as surrogates for variables, appropriate in assessing functions that represent dynamic processes?

I will try to address these questions in the following discussion and provide my perspective. My goal is to help users of methods better understand the information they have available and to help developers of new methods better understand how the choices they make during development have an impact on the information the methods provide. Numeric Scoring Is it appropriate to represent untested hypotheses numerically to provide an estimate of performance of a wetland function? Or should we refrain from doing so, as Adamus (1986) suggests, until the models can be validated? At present, the numeric scores or indices derived from assessment models have not been linked to a measured rate at which a wetland performs a function because none of the models have been validated. The answer is yes. We need models that provide numeric assessments as long as we recognize and document their limitations. Today’s wetland managers need to be able to assess changes in wetland functions using some kind of numeric approach (Kusler 1986, Brinson 1995). Analyzing impacts to wetlands, establishing credits and debits in a compensation bank, or determining the amount of compensatory mitigation needed for a 404 permit (US Clean Water Act) require some type of numeric assessment of performance or value. Qualitative judgments of does or does not perform a function are inadequate if one is trying to implement the national policy of no net loss of wetland functions. The urgency of making decisions about wetlands often necessitates the use of qualitative information that has been transformed into some numeric results. There is a difference in the accuracy of numeric representations required for scientific experiments (i.e., statistical significance and low autocorrelation between variables) and that required for management [i.e., a numeric

81

scoring of ratings or judgments; see discussion in Romesburg (1981)]. Similar issues have faced the developers of the habitat evaluation procedures (HEPs), but these issues have not precluded the methods from becoming the preeminent tools in wildlife management (Brooks 1997). Mechanistic Models Are mechanistic models the appropriate ones to use? Again the answer is yes. The mechanistic models used in wetland assessment methods provide a means of displaying logical, but scientifically untested, cause-and-effect relationships between variables and functions that are useful in management. As mentioned previously, these methods do not accurately model ecological processes occurring in wetlands. We are, however, documenting the judgments made by wetland scientists in assessing how well a wetland performs a function. With training, we can also achieve a much higher level of consistency among those assessing wetlands. The mechanistic approach used to build the models follows the conventions of ‘‘multiple-criteria decisionmaking models’’ used in environmental decision making (see Munda and others 1994) and should be treated as such. It becomes a moot point, therefore, whether the models are called mechanistic in the tradition of HEP or decision-making models in the tradition of environmental management. The important point is that the modeling approach has a history of successful applications in complex situations that require the combination of judgment, expertise from many disciplines, and both qualitative and quantitative data. The scientific community should recognize that decision-making models have different statistical properties than ecological models. The ususal statistical approaches based on analyses of variance and normally distributed data are not appropriate when mixed qualitative and quantitative data are used or the variables represent subjective judgment. For example, the probability assigned to a judgment is not a frequency, but rather, represents the expert’s degree of belief in a possible proposition or outcome (Cleaves 1995). Model developers, on the other hand, should also recognize that formal analytical procedures exist to assess the validity and variance of their models (see Voogd 1982, Munda and others 1994, Varis and others 1994, Cleaves 1995). They should use these techniques to better justify their choice of variables and the structure of their equations. Using Indicators Instead of Variables Is the use of structural indicators as surrogates for variables appropriate in assessing functions that repre-

82

T. Hruby

sent processes? Again, the answer is yes. A wetland professional analyzing how a wetland functions looks for specific structural elements in making his or her judgment. In most cases no attempt is made to measure performance. For example, a judgment about the effectiveness of a wetland for flood storage may be based on the proximity to a floodplain, the size of a depression, or the presence of outlets. Thus, it is appropriate to use indicators as the variables in models that quantify this judgment. It should be strongly emphasized, however, that decision-making models are often based on the best scientific information available. Much of the judgment used in the models is based on existing literature and expert knowledge. Methods that use indicators, however, can only provide information on the potential or probability that a function is being performed. The structural characteristics of a wetland’s plant community, its water regime, or its soils can only reflect the potential that a function is performed at a high or low rate because these characteristics are time independent. For example, the presence of dense vegetation is a good indicator that a wetland will perform well in filtering and trapping sediments when compared to a wetland without this characteristic. It does not provide any information, however, on the actual rates at which sediments may be trapped by the wetland. The wetland with sparse vegetation may actually be trapping more sediment than the wetland with dense vegetation if there is a source of high sediment loads nearby (e.g., clear-cut logging or development). All wetland rapid assessment methods developed to date use indicators in the models for one or more functions. Users of these methods should, therefore, scrutinize their methods to determine if the model for a function estimates the potential that a function is being performed or actually provides an estimate of performance. Estimates of actual performance need to be based on variables that are actual measurements of rates, counts, or specific processes (e.g., acre feet of storage per storm event or counts of invertebrate taxa per square meter). Functional Capacity, Potential Performance, and Sustainability One issue inherent in developing rapid assessment methods is to understand what aspect of wetland functions is being assessed in the model. The names given to the scores generated by a numeric model are often not defined adequately. The scores generated by assessment methods have been called level of performance, functional capacity index, functional value, suitability index, or site index potential. Furthermore, the same terms

may be used in different methods, but have different meanings. For example the term ‘‘level of performance’’ used in the HGM approach (Smith and others 1995) represents the relative amount of deviation from a sustainable level of performance (M. Brinson and D. Smith, personal communication), while in the methods being developed in Washington State (Hruby and others 1997) it means the level of performance relative to the highest levels found in the wetlands of a specific region and subclass. As mentioned previously, models do not assess the actual rates at which wetlands perform specific functions. They only assess performance relative to those identified as the top performers. Models developed using any of the criteria described below will generate a score for a wetland, often between 0 and 1. The scores, however, are not comparable since they provide different types of information about how a wetland performs a function. The criteria used to identify the highest performing wetlands score (i.e., the reference standards as defined by the HGM approach) determine the type of information about a function that a method provides. Three different criteria have been commonly used in wetland function assessment methods: 1. Authors of a method develop an image of the characteristics found in the highest performing wetland. Wetlands are then compared against this image by assessing its characteristics against the ideal. Methods using this approach include Reppert (1979), WET (Adamus and others 1987), and the Connecticut method (Amman and others 1986). The scores generated in this way provide an assessment of how well a wetland rates against an ideal or hypothetical wetland. The characteristics of the ideal wetland, however, sometimes do not reflect actual conditions in a region or basin since they are not calibrated to local conditions. Models that are not verified to local conditions often generate scores in the middle of the range (0.3–0.7), and do not allow for useful discrimination between sites (Brooks 1997). 2. A set of the highest performing wetlands (reference standard wetlands) for each function is identified within a region. These wetlands are judged to have the highest potential rate of a specific function or the highest habitat suitability for a taxa or group of taxa. The characteristics of these wetlands are used to set the standard against which other wetlands are compared. Methods using this approach include the IVA (Hruby and others 1995) and the ones under development in Washington State (Hruby and others 1997). The scores generated using this approach provide an assessment of the potential performance relative to known wetlands in a region. A score of 0.5 for a function

Wetland Functions

means that the assessed wetland is judged to have a potential level of functioning that is approximately half way between the best and worst wetlands of the region. The numeric score, however, does not provide any information regarding the long-term sustainability of the function at this level or how the function might change in the future. Furthermore, methods developed using this approach need to identify a separate set of wetlands to use as a standard for each function. 3. A reference set of the least altered wetlands is identified within a region. It is then assumed that the characteristics of these wetlands represent the highest level of performance for all functions. The characteristics of the least altered wetlands are used to set the standards against which other wetlands are compared for all functions. Methods using this approach are all based on the HGM approach being developed by the US Army Corps of Engineers (Brinson 1995, 1996). The functional capacity index of HGM methods is a measure of the deviation in the performance of a function from the least altered state (M. Brinson, D. Smith, personal communication). A 1 represents a level of functioning similar to that found in the least altered wetlands, and a low score a level that reflects a large deviation from the standard. The index does not assess whether the actual rate at which a function is performed is higher or lower than the reference standard. Rather, it assesses whether the function is being performed at a sustainable level. For example, a wetland with a man-made weir might store twice the amount of floodwater per hectare as unaltered standard wetlands. Its score for the function of water storage, however, would by definition be no higher than that for the unaltered wetlands. Furthermore, the wetland with a weir might be scored lower because the amount of storage deviates from the unaltered condition. The extra storage provided by the weir might not be considered sustainable under the initial assumption that least altered are the most sustainable.

Conclusions In conclusion, I believe that we are making significant progress in assessing the performance of functions in wetlands, albeit the methods developed are still based on judgment. Both the HGM (Brinson 1995, 1996) and the IVA approaches (Hruby and others 1995) are based on having a structured and formal process for quantifying the judgment of a group of local experts with step by step documentation and peer review of the decisions made. This is an improvement over previous methods that relied mostly on the judgment of the authors and

83

often had limited documentation from existing literature (World Wildlife Fund 1992). We should, however, use the analytical tools developed for modeling environmental decision making in developing wetland assessment methods. Environmental decision making has developed a more rigorous mathematical approach for quantifying judgment that we can adapt to meet the needs of assessing wetland functions. Such an approach might improve the credibility of current efforts. Another dimension that has been lacking in assessment efforts is the validation of the models developed. The common response is that models cannot be validated because the actual measurement of performance may take years and is not financially feasible. The challenge is to find ways to validate models that are more timely and cost effective. A careful approach of development, calibration, verification, and validation is needed for assessments methods, similar to that proposed for HEP models (Brooks 1997).

Literature Cited Adamus, P. R. 1986. Uses and proposed revisions for the Adamus assessment methodology. Pages 73–77 in J. A. Kusler and P. Riexinger (eds.), Proceedings: National wetlands assessment symposium. Association of State Wetlands Managers, Chester, Vermont, Technical Report 1. Adamus, P. R., E. J. Clairain, Jr., R. D. Smith, and R. E. Young. 1987. Wetland evaluation technique (WET), volume II: Methodology. US Army Corps of Engineers, Operational draft. Amman, A. P., and A. L. Stone. 1991. Method for the comparative evaluation of nontidal wetlands in New Hampshire. New Hampshire Department of Environmental Services NHDES-WRD-1991-3. Amman, A. P., R. W. Franzen, and J. L. Johnson. 1986. Method for the evaluation on inland wetlands in Connecticut. Connecticut Department of Environmental Protection DEP Bulletin No. 9. Anonymous. 1963. Webster’s Seventh New Collegiate Dictionary. G & C. Merriam Company, Springfield, Massachusetts, 1221 pp. Bartoldus, C. C., E. W. Garbisch, and M. L. Kraus. 1994. Evaluation for planned wetlands. Environmental Concern Inc., St. Michaels, Maryland. Brinson, M. M. 1993. A hydrogeomorphic classification for wetlands. Technical Report WRP-DE-4. US Army Engineer Waterways Experiment Station, Vicksburg, Mississippi. Brinson, M. M. 1995. The HGM approach explained. National Wetlands Newsletter November–December: 7–13. Brinson, M. M. 1996. Assessing wetland functions using HGM. National Wetlands Newsletter January–February:10–16. Brinson, M. M., F. R. Hauer, L. C. Lee, W. L. Nutter, R. D. Rheinhardt, R. D. Smith, and D. Whigham. 1995. Guidebook for application of hydrogeomorphic assessments to

84

T. Hruby

riverine wetlands. Technical Report WRP-DE-11. US Army Corps of Engineers Waterways Experiment Station, Vicksburg, Mississippi. Brinson, M. M., L. C. Lee, W. Ainslie, R. D. Rheinhardt, G. G. Hollands, R. D. Smith, D. F. Whigham, and W. B. Nutter. 1997a. Common misconceptions of the hydrogeomorphic approach to functional assessment of wetland ecosystems: scientific and technical issues. Wetlands Bulletin 14:16–21. Brinson, M. M., L. C. Lee, W. Ainslie, R. D. Rheinhardt, G. G. Hollands, R. D. Smith, D. F. Whigham, and W. B. Nutter. 1997b. Clearing up misconceptions about HGM. National Wetlands Newsletter July–August:14–18. Brooks, R. P. 1997. Improving habitat suitability index models. Wildlife Society Bulletin 25:163–167. Cable, T. T., V. Brack, Jr., and V. R. Holmes. 1989. Simplified method for wetland habitat assessment. Environmental Management 13:207–213. Cleaves, D. A. 1995. Assessing and communicating uncertainty in decision support systems: Lessons from an ecosystem policy analysis. Artificial Intelligence Applications 9:87–102. Cowardin, L. M., V. Carter, F. C. Golet, and E. T. LaRoe. 1979. Classification of wetlands and deepwater habitats of the United States. US Fish and Wildlife Service FWS/OBS-79/ 31, 103 pp. Farber, S., and R. Costanza. 1987. The economic value of wetland systems. Journal of Environmental Management 24:41–51. Fennessey, M. S., C. C. Brueske, and W. J. Mitch. 1994. Sediment deposition patterns in restored freshwater wetlands using sediment traps. Ecological Engineering 3:409–428. Gilvear, D. J., and R. J. McInnes. 1994. Wetland hydrological vulnerability and the use of classification procedures: A Scottish case study. Journal of Environmental Management 42:403–414. Glooschenko, V. 1983. Development of an evaluation system for wetlands in southern Ontario. Wetlands 3:192–200. Hollands, G. G., and D. W. Magee. 1985. A method for assessing the functions of wetlands. In J. Kusler and P. Riexinger (eds.), Proceedings of the national wetland assessment symposium. Association of Wetland Managers, Berne, New York. Hruby, T. 1997. Continuing the discussion: Scientific and technical issues regarding the hydrogeomorphic approach to function assessment of wetlands. Wetlands Newsletter September:23–24. Hruby, T., W. E. Cesanek, and K. E. Miller. 1995. Estimating relative wetland values for regional planning. Wetlands 15:93–107. Hruby, T., K. Brunner, S. Cooke, K. Dublonica, R. Gersib, T. Granger, L. Reinelt, K. Richter, D. Sheldon, A. Wald, and F. Weinmann. 1997. Draft models to assess riverine and depressional wetlands in the lowlands of western Washington. Washington State Department of Ecology Publication No. 97-93. Kusler, J. A. 1986. Wetland assessment: The regulator’s perspective. Pages 2–3 in J. A. Kusler and P. Riexinger (eds.), Proceedings: National wetlands assessment symposium. Association of State Wetlands Managers, Chester, Vermont, Technical Report 1.

Kusler, J. A. and W. Niering. 1998. Wetland assessment: Have we lost our way? National Wetlands Newsletter 20:8–14. Larson, J. S. 1976. Models for assessment of freshwater wetlands. Publication No. 32. Water Resources Research Center, University of Massachusetts, Amherst, Massachusetts. Leibowitz, S. G., B. Abbruzzese, P. R. Adamus, L. E. Hughes, and J. T. Irish. 1992. A synoptic approach to cumulative impact assessment. US Environmental Protection Agency, EPA/600/R-92/167. Lipton, D. W., K. Wellman, I. C. Sheifer, and R. F. Weiher. 1995. Economic valuation of natural resources—a handbook for coastal resource policymakers. NOAA Coastal Ocean Program Decision Analysis Series No. 5. NOAA Coastal Ocean Office, Silver Springs, Maryland, 131 pp. Mitsch, W. J., and J. G. Gosselink. 1993. Wetlands, 2nd ed. Van Nostrand Reinhold, New York. Munda, G., P. Nijkamp, and P. Rietveld. 1994. Qualitative multipcriteria evaluation for environmental management. Ecological Economics 10:97–112. National Research Council. 1995. Wetlands: Characteristics and boundaries. Water Science and Technology Board, Commission on Geosciences, Environment and Resources. National Academy Press, Washington DC. North Carolina Division of Coastal Management. 1994. NCCREW, A wetland functional assessment procedure for the North Carolina coastal area. A report of the Strategic Plan for Improving Coastal Management in North Carolina, Draft. Reppert, R. T., W. Sigleo, E. Stakhiv, L. Messman, and C. Beyers. 1979. Wetland values: Concepts and methods for wetland evaluation. US Army Corps of Engineers, Institute for Water Resources, Fort Belvoir, Virginia, 109 pp. Romesburg, C. 1981. Wildlife science: Gaining reliable knowledge. Journal of Wildlife Management 45:293–313. Roth, E. M., R. D. Olsen, P. L. Snow, and R. R. Sumner. 1993. In S. G. McCannel (ed.), Oregon freshwater wetland assessment methodology. Oregon Division of State Wetlands. Salem, Oregon. Smith, R. D., and C. C. Bartoldus. 1995. A snap-shot of the corps new wetland assessment procedure. Wetland Journal 6:3–5. Starfield, A. M., B. P. Farm, and R. H. Taylor. 1989. A rule-based ecological model for the management of an estuarine lake. Ecological Modelling 46:107–119. Terrell, J. W., T. E. McMahon, P. D. Inskip, R. F. Raleigh, and K. L. Williamson. 1982. Habitat suitability index models: Appendix A. Guidelines for riverine and lacustrine applications of fish HSI models with the habitat evaluation procedures. US Fish and Wildlife Service, FWS/OBS-82/10.A, 54 pp. Titre, J. P., and J. E. Henderson. 1989. Updated literature review of valuation of wetlands, 1985–present. US Army Engineer Waterways Experiment Station, Vicksburg, Mississippi, EEDP-06-8. US Army Corps of Engineers St. Paul District. 1988. The Minnesota wetland evaluation methodology for the north central United States. US Fish and Wildlife Service. 1980. Habitat evaluation procedures (HEP) manual. 102 ESM. United States Department

Wetland Functions

85

of the Interior Fish and Wildlife Service Division of Ecol. Serv.

Voogd, H. 1982. Multicriteria evaluation with mixed qualitative and quantitative data. Environmental Planning 9:221–236.

U.S. Fish and Wildlife Service. 1981. Standards for the development of habitat suitability index models. 103 ESM. United States Department of the Interior Fish and Wildlife Service Division of Ecol. Serv.

Washington State Department of Ecology. 1993. Washington State wetlands rating system: Western Washington, 2nd ed. Washington State Department of Ecology Publication No. 93-74.

Varis, O., S. Kuikka, and A. Taskinen. 1994. Modeling for water quality decisions: uncertainty and subjectivity in information, in objectives, and in model structure. Ecological Modeling 74:91–101.

World Wildlife Fund. 1992. Statewide wetlands strategies: A guide to protecting and managing the resource. Island Press, Washington, DC.