INTRODUCTION

1. RATIONALE

A standardized, widely accepted vegetation classification for the United States is required for effective inventory, assessment, and management of the nation's ecosystems. These needs are increasingly apparent as individuals, private organizations, and governments grapple with the escalating alteration and loss of natural vegetation (for examples, see Klopatek et al. 1979, Mack 1986, LaRoe et al. 1995, Mac 1999). Remnants of natural vegetation have become increasingly rare (Noss et al. 1995, Noss and Peters 1995, Barbour and Billings 2000). Some types are now imperiled because of habitat loss or degradation, and others have disappeared entirely from the landscape without ever having been formally documented (Grossman et al. 1994). Losses of vegetation types represent losses in habitat diversity, leading directly to more species being in danger of extinction (Ehrlich 1997, Wilcove et al. 1998, Naeem et al. 1999). Predicted changes in climate, continued atmospheric pollution, ongoing species invasions, and land use changes are likely to cause further unprecedented and rapid alteration in vegetation (Overpeck et al. 1991, Vitousek et al. 1997, Morse et al. 1995). Widespread changes in land use have led to increased social and economic conflicts, resulting in an increasing demand for more robust and timely information about remaining natural and seminatural environments. In addition to these environmental issues, a standardized classification is needed to place basic ecological and biodiversity studies in context. In its application to mapping vegetation, a standardized classification can form the basis for consistently defined and comparable units among different maps. We expect that this standardized classification will play a prominent role in guiding research, resource conservation, and ecosystem management, as well as in planning, restoration activities, and in predicting ecosystem responses to environmental change.

To meet the need for a credible, broadly-accepted vegetation classification, the Ecological Society of America (ESA: the professional organization for ecologists in the United States) joined with cooperating organizations such as the U.S. Geological Survey, U.S. Federal Geographic Data Committee, and NatureServe^{^[1]} to form a Panel on Vegetation Classification. To formalize this partnership, the four participating organizations signed a formal Memorandum of Understanding (MOU)^{^[2]} in August 1998. This MOU defines the working relationship among the signers for the purpose of advancing the National Vegetation Classification.

The objectives of the ESA Vegetation Classification Panel are to: (1) facilitate and support the development, implementation, and use of a standardized vegetation classification for the United States; (2) guide professional ecologists in defining and adopting standards for vegetation sampling and analysis in support of the classification; (3) maintain scientific credibility of the classification through peer review; and (4) promote and facilitate international collaboration in development of vegetation classifications and associated standards. In this document the Panel articulates and explains a set of standards and procedures aimed at achieving the first three of these objectives.

2. BACKGROUND AND PRINCIPLES

The ESA Panel on Vegetation Classification recognizes the Federal Geographic Data Committee’s (FGDC) “National Vegetation Classification Standard”(1997) as the starting point for developing a national vegetation classification. The FGDC classification standard is a physiognomic-floristic hierarchy with higher-level physiognomic units and lower-level floristic units (Figure 1). The FGDC standard, based on the International Classification of Ecological Communities or ICEC (Grossman et al. 1998; now referred to as the International Vegetation Classification, or IVC), introduced the classification hierarchy, documented the component elements of all except the floristic levels, and provided the context for defining those floristic levels. Between 1995 and 1996 the Panel concentrated on assisting the FGDC by reviewing proposed standards for the physiognomic categories (class, subclass, group, subgroup, and formation; Loucks 1996), as well as the specific physiognomic types within these categories.

The guiding principles established by the FGDC for the overall development of the NVC are shown in Box 1 (FGDC 1997, Section 5.3). In particular, the 1997 FGDC standard provided definitions for the floristic units of the classification: the alliance and association. These definitions begin with the premise that a vegetation type represents a group of stands that have similar plant composition and physiognomy, and that types must have diagnostic criteria to enable their recognition. Nonetheless, we recognize that, due to complex biophysical factors as well as chance, vegetation is a continuously varying phenomenon and that species are stochastic in their distribution. As a consequence, floristic vegetation units are not readily defined by precise and absolute criteria. Instead, some examples of vegetation can be seen to be unambiguously members of a particular type, whereas others are intermediate such that their assignment must be defined in terms of relative affinities with alternative types.

Although the 1997 FGDC standard includes the two floristic categories of the NVC hierarchy, Alliance and Association, it provides no list of recognized types, no details about nomenclature, nor methods for defining and describing alliances and associations. With respect to these categories, the document states “The current list of Alliances and Associations for the conterminous United States will be published by The Nature Conservancy in the spring of 1997.” (FGDC 1997, Section 6.0). The list was published in 1998,in cooperation with the Natural Heritage Network (Anderson et al. 1998) and has subsequently been repeatedly refined and improved. Each alliance and association on the list is described in detail in a standardized format (see Grossman et al. 1998, page 48) that contains a compilation of literature and field observations. Collectively, these descriptions constitute a comprehensive summary of our knowledge of the plant communities of the United States. The Panel anticipates that the recognized list of type descriptions will be enhanced and revised in accordance with the FGDC requirement that the alliance and association types must be based on field data conforming to standard methods (FGDC 1997, Sections 5.3 and 7.1) and that the types will be defined so as to meet standard criteria for acceptance. However, the precise standards and criteria were not spelled out by the FGDC. The standards presented here are intended to meet that need.

We have used the FGDC “Guiding Principles” and the definitions for association and alliance to guide the development of standards for defining, naming, and describing floristic units. Our goal for future revisions of the list of alliances and associations and supporting documentation is that they will be based on standards for field observation, type description, peer-review, and data management. Each of these activities is summarized next.

Field plot records. Vegetation associations and alliances should be identified and described through numerical analysis of plot data that have been collected from across the range of the vegetation type and closely related types (irrespective of political and jurisdictional borders). We outline standards for plot data in Chapter 5.

Type description. Proposals for new or revised floristic units must adhere to standards for circumscribing and describing types. Each type description should include sufficient information to determine the distinctive vegetation features of the type and its relation to other types recognized in the classification. Proposals for revision of recognized types must include comparison of the focal types with related types of that level to ensure that they do not duplicate or significantly overlap, but rather enhance, replace, or add to them. We outline standards for type circumscription and description in Chapter 6.

Peer review. Proposals for new and revised types need to be evaluated through a credible, open peer-review process. Standards for the peer-review process are outlined in Chapter 7.

Data management. Plot data used to define and describe an association or alliance must be permanently archived in a publicly accessible data archive, either for revisions to the descriptions of existing type concepts, new descriptions of proposed types, or other uses. Accepted proposals for addition or modification of vegetation types and all supporting documentation must be deposited in the NVC digital public archive. All plant taxa referenced in plot data or community type descriptions must be unambiguously defined by reference to a public database or publication of recognized taxa, or by reference to an authoritative, published circumscription. Unknown taxa should be placed as precisely as possible within the phylogenetic hierarchy of such a database or publication. All three types of data archives (for plant taxa, field plots, and associations and alliances) must be truly archival in the sense that the data will be able to be extracted in their original form and context at some indefinite future time by any reasonably diligent investigator. Data management standards are outlined in Chapter 8.

The standards to be used for collecting field data, describing types, peer review, and data management are enumerated at the end of each of these chapters.

Disclaimers

The NVC is a classification of the full range of existing vegetation, from natural types that include old-growth forest stands and seminatural vegetation (including grazed rangelands, old agricultural lands undergoing natural succession, and stands dominated by naturalized exotics) to planted or cultivated vegetation, such as row crops, orchards, and forest plantations. Various uses and applications may require distinctions with respect to naturalness (see Grossman et al. 1998 Appendix E). Descriptions of types should aid users of the classification in differentiating among natural, seminatural, and planted types.

Consistent with the FGDC principles, the standards described here for floristic units relate to vegetation classification and are not intended as standards for mapping units. Nevertheless, types defined using these standards can be mapped and they can be used as the basis for map various other types of units as well, subject to limitations of scale and mapping technology. The criteria used to aggregate or differentiate within these vegetation types and to form mapping units will depend upon the purpose of the particular mapping project and the resources devoted to it (e.g., Damman 1979, Pearlstine et al. 1998). For example, in using the NVC Alliance class as a target for vegetation mapping by the Gap Analysis Program, not all alliance types can be resolved. In such cases alliance types are aggregated into map units of “compositional groups” or “ecological complexes” (see Pearlstine et al. 1998). Although not part of the NVC standard, such aggregates represent units of vegetation that meet the needs of the mapping activity and have an explicit relationship to established NVC units.

Although vegetation varies more-or-less continuously in time and space, classification partitions that continuum into discrete units for practical reasons. These include, for example, facilitating communication and information-gathering about ecological resources, documenting the diversity of ecological communities, and providing a framework for addressing scientific inquiries into the patterns of vegetation. Alternative classification approaches, particularly those that aggregate alliances and associations differently from the NVC and IVC (which use vegetation physiognomy as the major criteria for aggregating alliances) are available and may be more practical for some particular uses. For example, hierarchical levels of vegetation classifications have been defined based purely on floristic criteria (Westhoff and van der Maarel 1973), on ecosystem processes (Bailey 1996), or on potential natural vegetation (Daubenmire 1968). Each of these approaches meets different needs and the NVC associations that are defined using these standards can nest to varying degrees under any of these hierarchy types. In providing standards for implementation of the floristic levels of the U.S. National Vegetation Classification, we in no way mean to imply that this is the only valid classification approach.

3. A BRIEF HISTORICAL BACKGROUND

"Vegetation classification attempts to identify discrete, repeatable classes of relatively homogeneous vegetation communities or associations about which reliable statements can be made. Classification assumes either that natural vegetation groupings (communities) do occur, or that it is reasonable to separate a continuum of variation in vegetation composition and/or structure into a series of arbitrary classes.” (Kimmins 1997).

As we reflected on the history of vegetation classification in the United States and elsewhere and on the opportunities that now lie before us, we became convinced that a clear set of standards for defining floristic units would advance the discipline of vegetation science and make a strong contribution to conservation and resource management. Because our goal is to develop standards informed by the rich historical debate surrounding vegetation classification, we begin this document where the ESA Vegetation Panel began its work: by reviewing the historical basis for some of the fundamental concepts that shape the floristic levels of the US National Vegetation Classification.

3.1. DESCRIBING AND CLASSIFYING VEGETATION

For over a century vegetation scientists have studied plant communities to identify their compositional variation, distribution, dynamics, and environmental relationships. They have used a multiplicity of methods including intuition, knowledge of physiological and population ecology (autecology), synthetic tables, and mathematical analyses to organize and interpret these patterns and relationships. Perhaps Shimwell (1971) expressed the situation best when, after reviewing the large and diverse literature on vegetation classification, he prefaced his book on the subject with the Latin maxim quot homines tot sententiae, "so many men, so many opinions." What follows is not a comprehensive review of vegetation classification; that has been done elsewhere (e.g., Whittaker 1962, 1973, Shimwell 1971, Mueller-Dombois and Ellenberg 1974). Instead, we focus on those elements most significant to the National Vegetation Classification enterprise and particularly those most relevant to the floristic levels.

Vegetation classification is a powerful tool employed for several purposes, including: (1) efficient communication, (2) data reduction and synthesis, (3) interpretation, and (4) land management and planning. Classifications provide one way of summarizing our knowledge of vegetation patterns.

Although different individuals conceptualize vegetation patterns differently, all classifications require the identification of a set of discrete vegetation classes. Several additional ideas are central to the conceptual basis for classification (following Mueller-Dombois and Ellenberg 1974, p. 153):

1. Given similar habitat conditions, similar combinations of species recur from stand to stand, though similarity declines with geographic distance.

2. No two stands (or sampling units) are exactly alike, owing to chance events of dispersal, disturbance, extinction, and history.

3. Species assemblages change more or less continuously with geographic or environmental distance.

4. Stand composition varies with the spatial and temporal scale of analysis.

These fundamental concepts are widely shared, and articulating them helps us understand the inherent limitations of any classification scheme. With these fundamentals in mind, we can better review the primary ways in which vegetation scientists and resource managers have characterized vegetation pattern to meet their needs.

Physiognomic characterization

Physiognomy, narrowly defined, refers to the general external appearance of vegetation based on growth form (gross morphology) of the dominant plants. Structure relates to the spacing and height of plants forming the matrix of the vegetation cover (Fosberg 1961). Often physiognomy is used to encompass both definitions, particularly when distinguishing “physiognomic” classifications from “floristic” ones. The basic unit of many physiognomic classifications is the formation, a "community type defined by dominance of a given growth form in the uppermost stratum of the community, or by a combination of dominant growth forms" (Whittaker 1962). This is the approach used the physiognomic portion of the NVC.

Physiognomic patterns often apply across broad scales as they typically correlate with or are driven by climatic factors, whereas floristic similarities are more regionally constrained as they reflect species composition, which in turn is strongly influenced by geographic discontinuities and idiosyncratic historical factors. Consequently, physiognomic classifications have more often been used in continental or global mapping applications, and floristic classifications in regional applications. A variety of classifications based on physiognomy (e.g., Fosberg 1961) preceded the development of the widely recognized international classification published by the United Nations Educational, Scientific, and Cultural Organization (UNESCO 1973, Mueller-Dombois and Ellenberg 1974). The UNESCO classification was intended to provide a framework for preparing vegetation maps at a scale of about 1:1 million or coarser, appropriate for worldwide comparison of ecological habitats as indicated by equivalent categories of plant growth forms.

Physiognomic classifications have, however, been used for natural resource inventory, management, and planning. Such classifications are based on measurements of vegetation attributes that may change during stand development and disturbance and which have management implications for wildlife habitat, watershed integrity, and range utilization. Criteria for physiognomic classification commonly include (a) plant growth forms that dominate the vegetation (e.g., forb, grass, shrub, tree), (b) plant density or cover, (c) size of the dominant plants, and (d) vertical layering (e.g., single stratum, multistrata). Physiognomic types have been used in numerous regional wildlife habitat studies (e.g., Thomas 1979, Barbour et al. 1998, Barbour et al. 2000), and they have also been used in conjunction with stand age and structure to assess old-growth status (Tyrrell et al. 1998).

Physiognomic classifications alone typically provide a generalization of floristic patterns. However, because they lack specificity at local or regional extents they are often used in conjunction with, or integrated into, thematically higher-resolution classifications that rely on floristics, that is, the taxonomic identity of plants. An exception to this is in certain kinds of floristically rich and complex or poorly understood vegetation, such as tropical rain forests, where physiognomic classification of vegetation remain the most common approach (Adam 1994, Pignatti et al. 1994).

Floristic characterization

Floristic characterization uses the composition of taxa to describe stands of vegetation. These characterizations are usually based on records of formal field observations (“plots”), which are fundamental to the definition, identification, and description of vegetation types. Methods range from describing only the dominant species to listing and recording the abundance of all species present in the stand (total floristic composition). Differences in these characterization methods have an important bearing on the definition and description of the alliances and associations, and are discussed next.

Dominance

One traditional way to classify vegetation is on the basis of dominant plant species of the uppermost stratum. “Dominance types” are typically based on the dominant taxonomic entity (or group of dominants) as assessed by some measure of importance such as biomass, density, height, or canopy cover (Kimmins 1997). Such classes represent the lower levels in several published classification hierarchies (e.g., Cowardin et al. 1979, Brown et al. 1980).

Determining dominance is relatively easy and requiring only a modest floristic knowledge. However, because dominant species often have geographically and ecologically broad ranges, there can be substantial floristic and ecologic variation within any one dominance type. The dominance approach has been used widely in aerial photo interpretation and mapping inventories because of its ease of interpretation and application. With the advances in remotely-sensed image acquisition and interpretation (spaceborne as well as airborne), there has been a significant increase in the level of effort in classifying and mapping dominant vegetation types across large areas (e.g., Scott and Jennings 1998, Lins and Kleckner 1996).

The term “cover type” is almost synonymous with “dominance type.” Cover types are typically based on the dominant species in the uppermost stratum of existing vegetation. In forests cover types may be variously assessed by a plurality of tree basal area or canopy cover. Similarly, rangeland cover types are typically based on those species that constitute a plurality of canopy cover (Shiftlet 1994). Although their limitations have been clearly articulated (e.g., Whittaker 1973), dominance types remain broadly used because they provide a simple, efficient approach for inventory, mapping, and modeling purposes.

Total floristic composition

Total community floristic composition has been widely used for systematic community classification. Two of the major approaches used in the United States are those of Braun-Blanquet (1928; also referred to as the “Zürich-Montpellier School”, see Westhoff and van der Maarel 1973, Kent and Coker 1992), and Daubenmire (1952, 1968); see Layser (1974) and Kimmins (1997) for a comparison of the two approaches). Both approaches use an “association” concept derived from the definition of Flahault and Schröter (1910), which states that an association is “a plant community type of definite floristic composition, uniform habitat conditions, and uniform physiognomy” (Flahault and Schröter 1910; see Daubenmire 1968 and Moravec 1993).

Braun-Blanquet (1928) defined the association as "a plant community characterized by definite floristic and sociological (organizational) features” which shows, by the presence of diagnostic species “a certain independence.” Diagnostic species are those whose relative constancy or abundance distinguish one association from another (Whittaker 1962). Identification of character species, those species that are particularly restricted to a single type, was considered essential to the definition of an association, whereas differential species (those species that delimit one association from another association only; not to be confused with the character species which distinguish one particular association from all other associations), defined lower taxa, such as subassociations (Moravec 1993). Patterns of diagnostic species are assessed using relevés (i.e., plots). A relevé is a record of vegetation composition that includes a comprehensive list of plants in a relatively small, environmentally uniform habitat (Mueller-Dombois and Ellenberg 1974), together with assessment of species cover. The Braun-Blanquet approach combines plant associations with common diagnostic species in a hierarchical classification with progressively broader floristic units called alliances, orders, and classes (see Pignatti et al. 1994). The association concept has been progressively narrowed as more associations have been defined, each with fewer diagnostic or character species (Mueller-Dombois and Ellenberg 1974). Today many associations are defined using only differential species (Weber et al. 2000). Classifications based on the Braun-Blanquet approach continue to be widely employed outside North America (especially in Europe, South Africa and Japan; see Mucina et al. 1993, Mucina 1997, 2001, Rodwell et al. 2002, but also see Borhidi [1996] as a milestone vegetation treatment from the Western hemisphere), and are occasionally applied in the U.S. (e.g., Komárková 1979, Cooper 1986, Peinado et al. 1994, Nakamura and Grandtner 1994, Nakamura et al. 1994, Walker et al. 1994, Peinado et al. 1998,Rivas-Martinez et al. 1999).

Daubenmire (1952) purposely looked for and sampled the least disturbed and oldest plant communities ("near-climax") that he could find across a full range of environments as a basis to define "climax associations". This was based upon the premise that a classification "based upon climax types of vegetation best expresses the potential biotic productivity of a given combination of environmental factors" (Daubenmire (1953). Stands were grouped by traditional synecological synthesis tables for study of community floristics and evaluation of diagnostic species. Daubenmire (1968) narrowed the definition of association to represent a type of climax phytocoenosis and suggested the word "associes" could be used to indicate plant communities in earlier recognizable stages of succession. Later, many authors preferred to use a different term—"community type"—for seral and disclimax plant communities to avoid confusion between climax and seral types. In contrast to earlier definitions of "climax" Daubenmire and Daubenmire (1968) noted that their use of the term was relative to the longevity of seral, shade-intolerant tree species and that the "climax" condition was generally achievable in 300 to 500 years.

Although the Daubenmire and Braun-Blanquet methods have strong underlying similarities (see Layser 1974) the original approach of Daubenmire (1952) was to define climax associations as floristically stable reference points for interpreting vegetation dynamics and site attributes. Conversely, the Braun-Blanquet association was intended as a systematic unit of classification, irrespective of successional status. Thus, under the Braun-Blanquet approach, old fields, pastures, and forests were all described using the association concept, with no preconceptions as to how such types relate to a climax association or successional sequence. Another fundamental difference between the Braun-Blanquet and Daubenmire approaches is apparent in forest vegetation, where the latter assigns primary weighting to diagnostic members of the predominant growth form (tree species), particularly those expected to dominate in late-successional states, and only secondary weighting to diagnostic members of the undergrowth vegetation. Another difference is that the Daubenmire approach makes an explicit effort to use the late-successional natural vegetation to predict the climax vegetation. Because the two methodologies rely on similar vegetation data and analysis, the units defined for late-successional vegetation under these two methods may appear similar. However, if one considers trees and undergrowth vegetation equally in terms of total floristic composition, different types of associations could be defined for the same area, as illustrated recently by Spribille (2001).

Daubenmire’s “habitat types” represent parts of the land surface capable of supporting the same kind of climax plant association (Daubenmire 1952,1968). During the 1960s and 70s, with an emerging emphasis on natural resource management, Daubenmire’s approach of using climax associations as a conceptual framework for a site classification gained preeminence in the western United States. Financial support was provided, particularly by the US Forest Service, for developing plant association and habitat type taxonomies on a systematic basis over large areas of the American West. With millions of hectares to cover, methods were optimized for efficiency (Franklin et al. 1971). In addition, sampling was no longer restricted to “climax” or "near-climax" stands; rather, vegetation was sampled with relevés from "late-successional" (maturing) stands across the full range of environmental conditions (Pfister and Arno 1980). The term "series" was introduced by Daubenmire and Daubenmire (1968) for grouping forest associations having a common climax overstory dominant species. Associations, nested within series, were defined by diagnostic species (identified from a synthesis of field samples) in the forest understory. By the 1980s, more than 100 monographs had been published on habitat types of forestlands and rangelands in the western United States (Wellner 1989), and accompanying keys were provided to identify the habitat types and to infer their potential climax association (also called potential natural vegetation type). However, it should be noted that all these efforts first classified late-successional existing vegetation associations as the starting point for inferring potential vegetation and habitat type interpretations.

Physiognomic-floristic characterizations

Descriptions of vegetation need not rely solely on either floristics or physiognomy. A classification that combines physiognomic and floristic criteria allows flexibility for characterizing a given area by both its physiognomy and composition. Driscoll et al. (1984) proposed a multi-agency ecological land classification system for the United States that consists of a combination of the physiognomic units of UNESCO (1973) and the floristic "late-successional" associations or habitat types. Subsequently, The Nature Conservancy developed a combined physiognomic-floristic classification of existing vegetation titled the International Classification of Ecological Communities (see Grossman et al. 1998) using modified physiognomic units of UNESCO for the upper levels and the floristic alliance and association units for the lower levels (see Figure 1). Units at all levels of the classification were developed across the United States, based on a synthesis of existing information and ecological expertise (Anderson et al. 1998). The Conservancy’s definition of the association was based on Flahault and Schröter’s (1910) association concept of an existing vegetation type with uniform floristic composition, habitat conditions, and physiognomy. Both the Driscoll et al. (1984) and the TNC classifications use a formation concept that incorporates some elements of climate and geography into the physiognomic units, and integrates them with floristic units based on variations of the association concept.

More strictly floristic classifications, such as those of the Braun-Blanquet school, occasionally find it convenient to organize vegetation classes by formations (Rodwell et al. 2002). Westhoff and van der Maarel (1973) note that since the “floristic-sociological characters of an association are supposed to reflect all other characters a floristic-sociologically uniform association might be expected to be structurally uniform as well.” Though not always true (Westhoff 1967), there is often sufficient structural or physiognomic uniformity to make such an integration meaningful. Indeed, it may be possible to conceive of a “phytosociological formation,” in which the definitions of the formation units are informed by the floristic units they contain (Westhoff and van der Maarel 1973, Rodwell et al. 2002).

Floristic classifications and community concepts

Continuum concepts and vegetation classification

Curtis (1959) and Whittaker (1956; also see McIntosh 1967) explicitly recognized that vegetation varies continuously along environmental, successional, and geographic gradients. In addition, these workers embraced the observation of Gleason (1926) that species respond individualistically to these gradients and that chance plays an important role in the composition of vegetation (but see Nicolson and McIntosh 2002 for an important recent view of Gleason’s individualistic concept). The necessary consequence is that in many cases there are not clear and unambiguous boundaries between vegetation types, and that vegetation composition is not consistently predictable. Any decision as to how to divide the continuously varying and somewhat unpredictable phenomenon of vegetation into community types is of necessity somewhat arbitrary with multiple acceptable solutions.

A common approach to capturing vegetation pattern across landscapes is to describe change in floristic composition relative to gradients in geographic or environmental factors such as climate and soils. The set of techniques used to relate vegetation to known physical gradients is referred to as direct gradient analysis. In contrast, techniques for ordering vegetation along compositional gradients deduced from stand similarity and independently of knowledge of the physical environment are referred to as indirect gradient analysis (Gauch 1982, Kent and Coker 1992). Gradients observed using indirect methods can be divided to form a classification, or these gradients can be used to identify key variables driving compositional variation, and these in turn can be used to create an optimal direct gradient representation. Gradient analysis need not lead to classification, yet many researchers have "classified" or summarized vegetation into types based on gradient patterns (e.g., Whittaker 1956, Curtis 1959, Peet 1981, Faber-Langendoen and Maycock 1987, Smith 1995).

Many natural resource professionals and conservationists have used gradient analysis to develop local classifications. Practitioners have also used a “natural community” type concept to develop widely differing kinds of regional classifications, defining units by various combinations of criteria, including vegetation physiognomy, current species composition, soil moisture, substrate, soil chemistry, or topographic position, depending on the local situation (e.g., Nelson 1985, Reschke 1990, Schafale and Weakley 1990, Minnesota NHP 1993). This approach has been used with great success for conservation and inventory at the local and state level, but the utility declines with increasing spatial scale.

Ecological land classifications

There are a number of classification systems that include vegetation as one of several criteria for classifying ecological systems (e.g., McNab and Avers 1994, Avers et al. 1994). Vegetation physiognomy is often used at broad scales to help delineate biogeographic or bioclimatic regions (e.g., Loveland et al. 1999), whereas floristic information is often used at finer scales to define ecological types and delineate ecological land units (e.g. Bailey et al. 1994, Cleland et al. 1994). The habitat-type approach (see above) relies primarily on species occurrence criteria and potential vegetation to define habitat types. Ecological land classification approaches typically use potential natural vegetation as one of several key elements to define ecosystem or ecological land units (Lapin and Barnes 1995, Bailey 1996). These classifications have often been used to guide forest management.

The site classification approach does not provide direct information on existing, or actual vegetation, and care must be taken not to confuse this distinct goal with the study of existing vegetation. Instead, once the ecological unit is defined, existing vegetation information may be used to characterize the current condition of the unit (Bailey 1996). As Cleland et al. (1997:182) state, “Ecological unit maps may be coupled with inventories of existing vegetation, air quality, aquatic systems, wildlife, and human elements to characterize...ecosystems.” Thus, vegetation classifications can play an important role in other classification approaches.

Existing vegetation and potential natural vegetation

Ecologists have developed classifications of both existing vegetation and potential natural vegetation. These should always be kept distinct in considerations of vegetation classifications as they support different, but possibly complementary, objectives and applications. By existing vegetation we simply mean the vegetation found at a given location at the time of observation. By potential natural vegetation we mean “the vegetation that would become established if successional sequences were completed without interference by man or natural disturbance under the present climatic and edaphic conditions” (Tüxen 1956, in Mueller-Dombois and Ellenberg 1974).

Classifying existing vegetation requires fewer assumptions about vegetation dynamics than classifying potential natural vegetation. Emphasis is placed on the current conditions of the stand. Classifications that emphasize potential natural vegetation require the classifier to predict the composition of mature stages of vegetation based on knowledge of the existing vegetation, species autecologies and habitat relationships, and disturbance regimes. For this reason, sampling to identify potential vegetation types is often directed at stands thought to represent mature or late seral vegetation. The 1997 FGDC vegetation standard pertains to existing vegetation and does not address issues related to the study of potential natural vegetation. This document has been written specifically in support of the FGDC standard and is intended solely to support study of existing vegetation.

3.2. A NATIONAL VEGETATION CLASSIFICATION FOR THE UNITED STATES

Agency and scientific consensus on classification

Vegetation classification, especially the concept of a unified, nationwide classification, received little support in the U.S. academic community prior to the 1990s. Most academic ecologists viewed classification as having little to contribute towards a general conceptual synthesis of broad applicability and were little interested in products of largely local or regional applicability. This view also stemmed in part from the diversity of approaches to interpreting and understanding the nature of vegetation patterns, as reviewed in the previous section (Nicolson and McIntosh 2002). As a consequence, little attention was paid to creating a unified national vegetation classification.^{^[3]}

Individual federal and state agencies in the U.S. charged with resource inventory or land management often required vegetation inventories or maps of public lands, both of which depend on classification for definition of units. Prior to the 1990s most of these projects were generally limited in scope and geography and tended to use divergent methods and categories (see Ellis et al. 1977) such that their various products did not fit together as components of a larger scheme. Instead, the disparate, disconnected activities resulted in development of incompatible sets of information and duplication of effort (National Science and Technology Council 1997). Nevertheless, the importance of broadly applicable systems for coordination of efforts had already become apparent during the 1970s and 80s, and some useful and geographically broad classifications were produced, including the habitat type classification of western forests by the U.S. Forest Service (Wellner 1989) and the Cowardin classification of U.S. wetlands (Cowardin et al. 1979). The Society of American Foresters has historically used a practical dominance-based approach for classifying forest types in North America (Eyre 1980). In addition, in the early 1980s, five federal agencies collaborated to develop an ecological land classification framework integrating vegetation, soils, water, and landform (Driscoll et al. 1984).

In the late 1970s, The Nature Conservancy (TNC) initiated a network of state Natural Heritage Programs (NHPs), many of which are now part of state government agencies. The general goal of these programs was inventory and protection of the full range of natural communities and rare species present within the individual states. Because inventory requires a list of the communities being inventoried, the various programs proceeded to develop their own state-specific community classification systems. As TNC started to draw on the work of the NHPs to develop national-level priorities for community preservation and protection, it quickly recognized the need to integrate the disparate state-level vegetation classifications into a consistent national classification.

In the late 1980s, the U.S. Fish and Wildlife Service initiated a research project to identify gaps in biodiversity conservation (Scott et al. 1993), which evolved into what is today the U.S. Geological Survey’s National Gap Analysis Program (GAP; Jennings 2000). This program classifies and maps existing natural and semi-natural vegetation types of the United States on a state and regional basis as a means of assessing the conservation status of species and their habitats. Because a common, widely used, floristically based classification was critical to this work GAP supported TNC’s effort to develop a nationwide classification (Jennings 1993). Collaboration between GAP and TNC led to a systematic compilation of alliance-level information from state natural heritage programs and from the existing literature on vegetation (e.g., Bourgeron and Engelking 1994, Sneddon et al. 1994, Drake and Faber-Langendoen 1997, Weakley et al. 1997, Reid et al. 1999). With support from TNC and an array of federal programs, Grossman et al. (1998) and Anderson et al. (1998) produced the first draft of what became the U.S. National Vegetation Classification (USNVC, referred to here as the NVC). The NVC was initially populated with a compilation of described natural vegetation types taken from as many credible sources as could be found and drawn from the experience vegetation ecologists with extensive regional expertise. Although the majority of the types described were not linked to specific plot data, they were often based upon studies that used plot data or on the knowledge of regional and state ecologists (Weakley et al. 1998, Faber-Langendoen 2001).

The Federal Geographic Data Committee and the ESA Vegetation Panel

In the early 1990’s the US federal government formally recognized the need for a standard nationwide vegetation classification. In 1990 the government published the revised Office of Management and Budget Circular No. A-16 (Darman 1990)^{^[4]}, which introduced spatial information standards. This circular described the development of a National Spatial Data Infrastructure (NSDI) to reduce duplication of information, reduce the expense of developing new geographically based data, and make more data available through coordination and standardization of federal geographic data. The circular established the Federal Geographic Data Committee (FGDC) to promote development of database systems, information standards, exchange formats, and guidelines, and to encourage broad public access.

Interagency commitment to coordination under Circular A-16 was strengthened and urgency was mandated in 1994 under Executive Order 12906 (Federal Register 1994), which instructed the FGDC to involve state, local, and tribal governments in standards development and to use the expertise of academia, the private sector, and professional societies in implementing the order. Circular A-16 was revised in 2002 to incorporate the mandates of Executive Order 12906. Under these mandates, the FGDC established a Vegetation Subcommittee to develop standards for classifying and describing vegetation. The subcommittee includes representatives from federal agencies and other organizations. After reviewing various classification options, FGDC proposed to adopt a modified version of the TNC classification. During the review period, ecologists from the National Biological Survey,^{^[5]} TNC, and academia discussed the need to involve the Ecological Society of America (ESA) to provide peer review as well as a forum for discussion and debate among professional ecologists with respect to the evolving NVC (Barbour 1994, Barbour et al. 2000, Peet 1994, Loucks 1995). The FGDC Vegetation Subcommittee invited ESA to participate in the review of the physiognomic standards as well as development of the standards for the floristic levels. This document is a direct product of the collaboration of ESA, FGDC, USGS, and NatureServe to provide formal standards for vegetation classification within the United States.

STANDARDS FOR ESTABLISHMENT AND REVISION OF FLORISTIC UNITS OF VEGETATION

The following chapters present formal standards for those seeking to propose or modify associations and alliances represented within the US National Vegetation Classification. It is our intent that these standards and procedures will facilitate continued rapid development, wide acceptance, and scientific maturation of the NVC.

4. THE ASSOCIATION AND ALLIANCE CONCEPTS

The historical record of vegetation classification as well as recent developments shows a continuing convergence of the basic concepts that underlie recognition of associations and alliances. Ecologists have long recognized the need to communicate the context of ecological and biological phenomena and to understand interactions within biotic communities. These needs have led to frequent use of “community type” or “vegetation type” as a unit of vegetation. Vegetation types can be understood as segments along gradients of vegetation composition, with more-or-less continuous variation within and among types along biophysical gradients. Despite the range of analytical tools and approaches that are now used to assess vegetation patterns, the basic and practical needs for classifying vegetation have led to substantial convergence in approaches in conceptualizing types of vegetation.

4.1. ASSOCIATION

The association is the most basic unit of vegetation recognized in the NVC. The earliest definition (Flahault and Schröter 1910) is “a plant community of definite floristic composition, uniform habitat conditions, and uniform physiognomy”. Gabriel and Talbot (1984) also include a definition of the association as “a recurring plant community of characteristic composition and structure.” Curtis (1959) defined the plant community, a segment along a continuum, as a “studyable grouping of organisms which grow together in the same general place and have mutual interactions.” Some commonalities are evident in the words used in the three definitions including four central ideas: characteristic composition, physiognomy and structure, habitat, and a recurring distribution across a landscape or region.

As the various association concepts merged into common use, our conceptualization of vegetation also shifted so as to accept more or less continuous variation. As noted in Section 3, Mueller-Dombois and Ellenberg (1974) recognized that “species assemblages change more or less continuously, if one samples a geographically widespread community throughout its range.” Their phrasing highlights an important element, the variability within an association that occurs across its range. In addition, the early recognition of Gleason (1926) that chance plays a major role in the local expression of vegetation has become an important part of our understanding of vegetation composition. Many classifications, including this one, have been framed around some characteristic range in composition, physiognomy, and habitat rather than the “definite” composition and habitat of the original association definition of Flahault and Schröter (1910).

Three other points should be considered:

1. “Habitat" refers to the combination of environmental or site conditions and ecological processes (such as disturbances) that influence the community. Temporal variation (e.g., recurrent fire in temperate grasslands; extreme weather) is included as part of an overall characteristic habitat, as long as it does not fundamentally change species presence.

2. Characteristic physiognomy and habitat conditions may include fine-scale patterned heterogeneity (e.g., hummock/hollow microtopography in bogs, shrub/herb structure in semidesert steppe).

3. Unlike strictly floristic applications of the association (and alliance) concept, the definition for the NVC standard retains an emphasis on both floristic and physiognomic criteria as implied by membership of floristic types in higher order physiognomic units of the classification.

Accordingly, establishment of a plant association implies application of a standard set of methods for describing a complex ecological reality, yet a practical, meaningful classification must accept a degree of variation within the association. As a synthesis of the above considerations, we adopt the following definition of association as the basic unit of vegetation:

A vegetation classification unit consistent with a defined range of species composition, diagnostic species, habitat conditions, and physiognomy.

In the context of this definition, diagnostic species refers to any species or group of species whose relative constancy or abundance can be used to differentiate one type from another. Guidelines have been proposed for the minimum number of diagnostic species required to define an association” (e.g., Schaminée et al. 1993). Obviously, the more diagnostic taxa that are used to define an association and the stronger their constancy and fidelity, the better the case for recognizing the unit. Moravec (1993) stated that associations may be differentiated by (1) character species, i.e., species that are limited to a particular type, (2) a combination of species sharing similar behavior (ecological or sociological species groups), (3) dominant species, or (4) the absence of species (groups) characterizing a similar type.

Despite the use of diagnostic species in vegetation classification, it should be recognized that diagnostic species can never precisely define lines between two similar associations or alliances. In addition to the fact that vegetation varies continuously, species are stochastic in their distributions (given the vagarities of, for example, dispersal, reproduction, and establishment) and chance events influence their occurrence at any given site. Various tools are needed to help assist users of the classification in recognizing the limits of a given vegetation type, whether in the form of more class-like concepts that provide relatively unambiguous criteria (e.g., vegetation keys) or in the form of type clusters that provide a core concept of the type, but leave the intermediate categories to be assessed through measures of similarity, or other specified criteria that are more probabilistic in nature. Good practice requires the identification of types based on overall species composition, diagnostic species, and other criteria that minimize ambiguity.

There is no consensus on some fixed amount of variability that is acceptable within an association or alliance. Mueller-Dombois and Ellenberg (1974) suggest, as a rule of thumb, that stands with a Jaccard presence/absence index (of similarity to the most typical plot) between 25% and 50% could be part of the same association and that stands with greater levels of similarity may better define subassociations. The subject of “stopping rules” in classification is a complex one, and a variety of criteria are often applied, including physiognomic and habitat considerations. In addition, the nature of the vegetation itself strongly influences decisions about where to draw conceptual boundaries between vegetation types. Important considerations may include species richness, variability, degree of anthropogenic alteration, and the homogeneity of the vegetation. No simple rule can be applied to all cases.

4.2 ALLIANCE

The vegetation alliance is a unit of vegetation determined by the floristic characteristics shared among its constituent associations, and is constrained by the physiognomic characteristics of the higher levels of classification within which the alliance is included. Its makeup is broader in concept than the association (i.e., more floristically and structurally variable), yet it has discernable and specifiable floristic characteristics. We define alliances as:

A group of associations with a defined range of species composition, habitat conditions, and physiognomy, and which contains one or more of a set of diagnostic species, typically at least one of which is found in the upper most or dominant stratum of the vegetation.

This definition includes both floristic and physiognomic criteria, in keeping with the integrated physiognomic-floristic hierarchy of the NVC. It also builds directly from the association concept.

The vegetation alliance concept presented here differs somewhat from the concept used in the more floristically-based Braun-Blanquet approach (Braun-Blanquet 1964, Westhoff and van der Maarel 1973). For example, using the Braun-Blanquet criteria, the Dicrano-Pinion alliance, which typically contains evergreen tree physiognomy, can include common juniper (Juniperus communis) shrublands (Rodwell 1991). The Vaccinio-Piceion (or Piceion Excelsae) alliance, with typically evergreen physiognomy, can include broadleaved deciduous birch (Betula pubescens) woodlands (Ellenberg 1988, Rodwell 1991). Nonetheless, alliances of the Braun-Blanquet system typically contain broadly uniform physiognomic and habitat characteristics comparable to the concepts and standards put forth here. Specht et al. (1974) used a similar approach to define alliances for Australia.

In comparison with the association, the alliance is more compositionally and structurally variable, more geographically widespread, and occupies a broader set of habitat conditions. Alliances that are defined narrowly based on specialized local habitats, locally distinctive species, or differ primarily in the relative dominance of major species, are to be avoided.

Many forest alliances are roughly equivalent to the "cover types" developed by the Society of American Foresters (SAF) to describe North American forests (Mueller-Dombois and Ellenberg 1974, Eyre 1980). In cases where the cover type is based solely on differences in the co-dominance of major species (e.g. Bald Cypress cover type, Water Tupelo cover type, and Bald Cypress-Water Tupelo cover type), the alliance may be broader than the cover types, or recombine them in different ways based on floristic and ecologic relationships. In cases where the dominant tree species extend over large geographic areas and varied environmental, floristic or physiognomic conditions, the alliance may represent a finer level of classification than the SAF cover type. In these situations, diagnostic species may include multiple dominant or co-dominant tree and understory species that together help define the physiognomic, floristic, and environmental features of an alliance type. For example, the broad ranging Jack Pine forest cover type (Eyre 1980, No. 1) may include at least two alliances, a more closed, mesic jack pine forest type and a more xeric, bedrock woodland type.

The alliance is similar in concept to the "series" of Daubenmire, a group of habitat types that share the same dominant species under apparent climax conditions (Pfister and Arno 1980). The series concept emphasizes the composition of the tree regeneration layer more than tree overstory composition in order to reveal the potential homogeneity of late-seral or climax canopy conditions based on the current tree population structure. Alliances differ from the series concept in that alliances, like associations, are based on existing vegetation, regardless of successional status. For example, a shrub type that dominates after a fire would be classified as distinct from both the forest type that was burned and the possible forest type that may eventually reestablish on the site.

4.3 STANDARDS FOR FLORISTIC UNITS

1. The NVC definitions for the floristic units of vegetation are:

a. Association: A recurring plant community with a characteristic range in species composition, specific diagnostic species, and a defined range in habitat conditions and physiognomy or structure.

b. Alliance: a grouping of associations with a characteristic physiognomy and habitat and which shares one or more diagnostic species typically are found in the uppermost or dominant stratum of the vegetation.

2. Diagnostic species exhibit patterns of relative fidelity, constancy or abundance that differentiate one type from another.

3. Diagnostic criteria used to define the association and alliance should be clearly stated, and the range of variability in composition, habitat, and physiognomy and structure should be clearly described, including similarity with other related types

4. Associations and alliances are categories of existing vegetation (i.e., , the plant species present and the vegetation structure found at a given location at the time of observation).

5. Associations and alliances recognized within the NVC must be defined so as to nest within categories of the recognized physiognomic hierarchy (e.g. in FGDC 1997, Association, Alliance, Formation, Subgroup, Group, Subclass, Class; see Figure 1).

5. VEGETATION FIELD PLOTS

A basic premise underlying these standards is that the alliance and association units are to be described and recognized through use of plot data. A second basic premise is that adherence to common standards for recording field plots is of critical importance for the development and consistent application of a scientifically credible NVC. Without data collected in compliance with such standards, recognition, description, and comparison of vegetation types could well be inaccurate, inconsistent and less than fully repeatable. The types of information that need to be collected in the field are discussed below and are listed in Appendix 1.

5.1. MAJOR TYPES OF REQUIRED DATA

The purpose of field plots is to record the vegetation and its environmental context. In addition, later interpretation of information collected in the plot requires metadata. Data recorded for field plots for the NVC fall into these three main categories.

1. Vegetation data: Floristic composition and physiognomy that can be used to classify vegetation constitute the key component of plot data. Floristic data consist of a list of the taxa observed, often recorded by the vertical strata they occur in, and usually associated with some measure of importance such as the relative amount of ground covered by them. Vegetation structure is typically assessed in terms of overall cover by vertical strata and the physiognomic attributes of the taxa associated with those strata.

2. Site data: Vegetation is best interpreted in the context of habitat, geographic location, and stand history information. This includes

a. abiotic factors such as soils, parent material, elevation, slope, aspect, topographic position, and climate,

b. stand history and disturbance regime, and

c. geographic location

3. Metadata: Data that describe the methods used to obtain vegetation and environmental data, or that are critical for subsequent uses of plot data. Examples of required metadata are the method and precision used to determine plot location, field methods, the nomenclatural (taxonomic) source or standard for identifying and naming plant species, the field personnel (including contact information and institutional affiliation) and the sampling date. Optional metadata include interpretations and reidentifications of plant taxa and the assignment of the plot to a particular type or types within the NVC.

Not all studies that use vegetation plot data are focused on classification. Investigators may have a variety of objectives when collecting plot data including, for example, documentation of ecological patterns and processes, assessment of vegetation structure, assessment of long-term change and human impacts, determination of targets for restoration, and validation of remote-sensed data. This chapter describes the plot information needed to support the development of associations and alliances of the NVC. It is not intended to serve as a definitive guide to recording and describing vegetation; discussion of these issues can be found in other references (e.g., Mueller-Dombois and Ellenberg 1974, Kent and Coker 1992, Jongman et al. 1995). In particular, this chapter is intended to alert investigators to the major issues that must be considered when collecting vegetation plot data for the purpose of developing or supporting a vegetation classification and to inform them as to the critical data that must be collected for plots to be used in the context of the NVC. If plot data are to be used to support development, refinement and identification of NVC types, investigators need to collect the core data described next.

5.2. STAND SELECTION AND PLOT DESIGN

Plot Selection

Vegetation surveys typically focus on detecting the range of vegetation variation in a region, or on a range wide assessment of one or more vegetation types. To achieve adequate representation of the vegetation in a focal area or type, plot selection is usually preceded by reconnaissance (ground or aerial) to assess the major patterns of variation in vegetation (or its underlying environmental gradients) and to develop a method for stand and plot selection. For example, major environmental factors may be used to create an “abiotic grid within which to select plots (e.g., stratified sampling of Peet 1980, or the gradsect technique of Austin and Heyligers 1991). The selection method is a critical step because it determines how well the plots will represent the area under study.

Selection of stands (contiguous areas of vegetation that are reasonably uniform in physiognomy, floristic composition, and environment) may be made by either preferential (subjective) or representative (objective) means, or some combination of these (sensu Podani 2000). With preferential methods, stands are selected based on the investigator’s previous experience, and stands that are “degraded”, “atypical”, or redundant may be rejected. A stand selected for plot records is considered typical of the vegetation of which it is a part, and each plot recorded is expected to yield a more or less typical description in terms of both floristic composition and physiognomy (Werger 1973). The same is true of representative selection, except that this approach also involves selecting stands with some degree of objectivity so as facilitate characterization of the full universe of vegetation within which the study is being conducted. The selection of representative stands may be via a simple random, stratified random (including the environmental grid or gradsect approach noted above), systematic, or semi-systematic method (Podani 2000). Either preferential or representative methods will yield plots suitable for the NVC, but representative sampling will typically lead to a less biased set of plots. In contrast, the representative method may miss or under sample rare and unusual types. Consequently, it is often necessary to supplement representative sampling with plots from rare or unusual types encountered in the course of field work. When working in highly modified landscapes, preferential selection is often the only way to assure that reasonably natural vegetation is adequately observed and sufficiently understood to be compared to other vegetation. Stratification of a landscape into a priori units within which plots are randomly located represents a hybrid approach and is often the preferred method.

For a variety of reasons, stand selection may be limited to only part of the vegetation present in an area. Many studies focus only on natural vegetation, including naturally disturbed, and various successional stages of vegetation. Others may focus exclusively on late-successional or mature natural vegetation. However, in principle, the NVC applies to existing vegetation, regardless of successional status or cultural influence. Criteria used to select stands should be thoroughly documented in the metadata.

Plot Location

Following reconnaissance and stand selection, a plot or series of plots is located within all or some subset of stands. Each plot should represent one entity of vegetation in the field; that is, a plot should be relatively homogeneous in both vegetation and habitat and large enough to represent the stand's floristic composition. Specifically, plots should be large enough and homogeneous enough that the relative importance of the dominant species observed within the plot is comparable to that of the surrounding stand. Of course, the investigator must recognize that communities are never fully homogeneous. Indeed, the main requirements for homogeneity can be met as long as obvious boundaries and unrepresentative floristic or structural features present in the stand are avoided (Rodwell 1991). Decisions about plot placement and homogeneity must be included in the plot metadata. These initial decisions are important, as both stand selection and plot placement within stands affect data quality.

Vegetation can be homogeneous at one scale and not at another. Some within-plot pattern is inevitable; small gaps occur within forests owing to death of individual dominants, and bryophytes and herbs can reflect substrate heterogeneity such as occurrence of rocks or logs. Moreover, forests examined at a scale of many kilometers can contain homogenous patches of differing age or disturbance history. For the purposes of the NVC the field worker should not seek homogeneity at the scale of either the mosses on a stump or the forests across a landscape, but rather homogeneous stands within which to place plots at some scale between 10 and 100,000 m² ^{(^[6])} reflecting a typical pattern of plants co-occurring under common environmental and historical conditions.

The floristic composition and structure of a plant community will vary not only in space but also in time. Seasonal changes, even during the growing season, can be dramatic in some types of vegetation. Large shifts in floristic composition over one to several years can occur in response to unusual weather conditions or fire. Some forest types (e.g., mixed mesophytic forests) may have a diverse and prominent, but ephemeral, spring flora. Some deserts have striking assemblages of annuals that appear only once every few decades. Although plot records for the NVC are based on the existing vegetation at the time of observation, plots that are known or expected to be missing a substantive portion of the likely flora must be so annotated to enable future analysts to properly interpret the data quality. Repeated inventories may be made over the course of a season to fully document the species in the plot. Practically speaking, these repeat visits (which should be documented as such) can be treated as multiple visits to the same plot and recorded as one plot observation record. Conversely, multiple visits over a series of years should be treated as separate plot observations (Poore 1962).

Plot design

Two fundamentally different approaches are commonly used for recording vegetation: (a) a plot where the information recorded is taken from a single entire plot, or (b) subplots, where the information recorded is taken from a set of smaller plots from within the stand. Both type of plots can provide adequate data for vegetation classification, but each method has its own requirements and advantages. Each of these is discussed next.

Data taken from an entire large plot

This is an efficient, rapid method for collecting floristic and physiognomic data for classification. The plot size is chosen to ensure that the plot is small enough to remain relatively uniform in habitat and vegetation, yet is large enough to include most of the species that occur within the stand. This approach permits statistical assessments of between-stand variation, but not within-stand variation.

Recommended plot size varies, depending on the structure of vegetation (the size of individual plants, spacing, number of vertical layers, etc). Plot sizes have also been based on the need for the plot to adequately represent the vegetation being sampled such that an increase in plot area yields few new species within the stand, and none significant to the vegetation’s physiognomy (see Moravec 1973 for a method of mean similarity coefficients). Plots larger than this are acceptable, but plots that are too small to represent the stand’s composition and structure are not adequate for developing a vegetation classification. For instance, in most temperate hardwood or coniferous forests, plots of between 200 and 1,000 m² are adequate for characterizing both the herb and the tree strata, whereas in many tropical forests, plots between 1,000 and 10,000 m² are required. Grasslands and shrublands may require plots between 100 and 200 m², whereas deserts and other arid-zone vegetation may require large plots, typically between 1,000 and 2,500 m² because the vegetation cover is sparse and species may be widely scattered. These recommended plot sizes typically satisfy minimum area calculations (McAuliffe 1990). Specialized studies of fine-scale variation, such as zonation around small wetlands or small sized bryophyte assemblages may well require plots that are smaller still, perhaps only a few m², though such small plots are to be avoided in community classification studies wherever possible.

We do not specify or recommend any particular plot shape; in fact, plot shape may need to vary depending on stand shape (e.g., riparian stands tend to be linear). Whenever possible, plot size and shape should be kept constant within a study. Issues of efficiency in plot layout most often dictate the plot shape employed by an investigator.

Data taken from a set of smaller subplots

Data may be collected from multiple subplots within a stand as an alternative to observation of a single large plot. This approach yields data that can be used to assess internal variability within a stand and to more precisely estimate the average abundance of each species across the stand. It is often used to measure treatment responses or evaluate other experimental manipulations of vegetation. The approach also may be useful for characterizing average vegetation composition in topographically gentle terrain where boundaries between stands may be diffuse. This method is inappropriate for measures of species number per unit area larger than the subplot, but can be helpful for assessing the relative variation within and among stands.

Investigators using the multiple small plot methods may locate their sample units randomly or systematically within the stand. The observation unit can be a quadrat, line-transect or point-transect, and can be of various sizes, lengths, and shapes. Quadrats for ground layer vegetation typically range from 0.25 to 5.0 m² and anywhere from 10 to 50 quadrats may be placed in the stand, again, either randomly or systematically. Quadrats for trees, where measured separately, typically are on the order of one m²or more. Even though subplots may be collected over a large portion of the stand, the total area over which data are recorded may be smaller than if the investigator used a single large plot (e.g., 50 one m²quadrats dispersed in a temperate forest stand will cover 50 m², whereas a single large plot would typically cover 100-1000 m²).

When deciding between multiple subplots and a single large plot it is important to consider the tradeoff between the greater precision of species abundance obtained with smaller, distributed subplots versus the more complete species list and more realistic assessment of intimate co-occurrence obtained using the single large plot. A major disadvantage of relying solely on subplots to characterize the stand is that it requires a large number of small sample units to adequately characterize the full floristic composition of the stand, a larger number than is generally employed. Yorks and Dabydeen (1998) described how reliance on subplots can result in a failure to assess the importance of many of the species in a plot. Consequently, whenever subplots or transects are used to characterize a stand, we strongly recommend that a list of “additional species present” within a larger part of the stand, such as some fixed area around the subsamples, be included. The classic Whittaker plot contains 25 one m² subplots plus a tally of additional species in the full 1000 m² macroplot, and the California Native Plant Society protocol incorporates a 50 meter point transect supplemented with a list all the additional species in a surrounding 5 x50 m²area (Sawyer and Keeler-Wolf 1995).

Hybrid approaches

Hybrid methods can combine some of the advantages of the two approaches. Sometimes, several somewhat large subplots (e.g., > 200 m² in a forest) are established to assess internal stand variability. The plots are sufficiently large that, should variability between plots be high, the plots could be classified separately as individual plots. A different strategy is for plots of differing sizes to be nested and used for progressively lower vegetation strata, such that plot size decreases as one moves from the tree layer to the shrub and herb strata owing to the generally small size and greater density of plants of lower strata. Although efficient with respect to quantitative measures of abundance, especially for common species, this method risks under representing the floristic richness of the lower strata, which are often more diverse than the upper strata. This problem can be ameliorated by listing all additional species found outside the nested plots but within the largest plot used for the upper layer. Again, the fundamental requirement is that the plot method provide an adequate measure of the species diversity and structural pattern of the vegetation for the purposes of classification.

Because vegetation pattern and its correlation with environmental factors can vary with plot size (see Reed et al. 1993), no one plot size is a priori correct, and it can be desirable to record vegetation across a range of different plot sizes. The widely applied 1000 m² Whittaker (1960) plots and 375 m² Daubenmire (1968) plots contain a series of subplots for herbaceous vegetation. More recently a number of investigators have proposed protocols where multiple plot sizes are nested within a single large plot (e.g., Naveh and Whittaker 1979, Whittaker et al. 1979, Shmida 1984, Stohlgren et al. 1995, Peet et al. 1998). These methods allow documentation of species richness and co-occurrence for a broad range of plot sizes smaller than the overall plot. Typically, they have the added advantage of documenting all vegetation types at several consistent scales of resolution, thereby assuring compatibility with many types of plot data.

5.3. VEGETATION PLOT DATA

As indicated in section 5.1, there are three types of data needed for effective vegetation classification: vegetation data, site data, and metadata. Of these, data on the structure and floristic composition of the vegetation must meet especially strict criteria. Environmental, or habitat, data, such as soil attributes, topographic position, and disturbance history, are also important, but their requirements are not as demanding. It is the quality of the vegetation data that largely determines whether a plot qualifies for use in the NVC.

We have developed standards for two different types of plot data, depending on whether (a) the plots can be used to develop vegetation types for the NVC classification (“classification plots”), or (b) they provide supplemental information relevant to existing NVC types (such as geographic extent or abundance) but are incomplete in some manner that prevents their use for primary classification analysis (“occurrence plots”). The minimum set of plot attributes that should be collected for each type of field plot (classification and occurrence) are listed in Appendix 1. Additionally, to ensure that as many kinds of classification plot sampling data as possible are available to develop the NVC, Appendix 1 distinguishes between those fields that are minimally required for classification (category 1) from others that are optimal, or consistent with best practices (category 2). For classification plots, the minimal requirements include a select set of records such as location fields, species (taxon) cover assessments, elevation, slope gradient and aspect, plot area, sampling method used, and the persons who collected the plot. Nonetheless, plots that meet only these minimal requirements are much less valuable for classification than those that contain the optimal set of fields that are part of the standard. Occurrence plots have essentially the same minimum requirements as classification plots, but they do not require a complete species list with cover values, nor do they require slope gradient, aspect, plot area, and there are fewer metadata requirements. In what follows we discuss the main features of the plot sampling standards for classification purposes.

Vertical structure and physiognomy of vegetation

Certain data on vegetation structure and physiognomy are needed to relate associations and alliances to the physiognomic and structural categories of the FGDC (1997) hierarchy. Physiognomy and structure have overlapping but different meanings. Fosberg (1961) defined vegetation physiognomy as the external appearance of vegetation. Physiognomy in this sense is the result in part of biomass structure, functional phenomena (such as leaf fall in forests), and gross compositional characteristics (such as luxuriance or relative xeromorphy). Structure relates to the spacing and height of plants forming the matrix of the vegetation cover. To be of value as a classification tool for the NVC, the description of vegetation structure by strata (or layers) must be standardized to permit consistent comparisons among data sets.

A stratum is a layer of vegetation which includes all plant growth forms that occur within it. Plants are assigned to strata based on their predominant position or height in the stand, not by their taxonomy or mature growth form. Consequently, a tree species that has both seedlings and saplings in a plot could be listed in several strata. In describing the vegetation physiognomy of a plot, the purpose is to capture the essential features of the often-complex stand conditions, rather than to describe the layering in the greatest possible detail.^{^[7]}

In terrestrial environments, four basic vegetation strata should be recognized whenever they are present: tree, shrub, herb, and moss (sensu Fosberg 1961; the ground layer of mosses, liverworts, lichens, and algae). In aquatic environments, floating, and submerged strata should be recognized where present. These six strata are needed to convey both the vertical distribution of overall cover and the predominant growth forms so as to place a plot within the NVC hierarchy. Additionally, they may be used to convey the abundance of each species in each stratum so as to provide a more detailed record of vegetation composition by strata (see below).

The six strata are defined as follows:

Tree stratum: includes tall trees (single-stemmed woody plants, generally more than 5 m in height or greater at maturity under optimal growing conditions). Very tall shrubs with tree-like form may also be included here, as may other life forms, such as lianas and epiphytes, and their contribution to the stratum can be further specified using the “life form” field.

Shrub stratum: includes shrubs (multiple-stemmed woody plants, generally less than 5 m in height at maturity under optimal growing conditions) and by shorter trees (saplings). As with the tree stratum, other life forms present in this stratum may also be included (however, herbaceous life forms should be excluded, as their stems often die back annually and do not have as consistent a height as woody life forms). Where dwarf-shrubs (i.e. shrubs < 0.5 m) form a distinct stratum (either as part of a series of strata, as in a forest, or as the top stratum of more open vegetation, such as tundra or xeric shrublands), they should be treated as a low version of the shrub stratum (or short shrub substratum). In many vegetation types, dwarf-shrubs may simply occur as one life form component of the herb stratum (see below).

Herb stratum: (also referred to as field stratum) includes herbs (plants without woody stems and often dying back annually), often in association with low creeping semi-shrubs, dwarf-shrubs, vines, and non-woody brambles (such as raspberries), as well as tree or shrub seedlings.

Moss stratum: (also referred to as nonvascular, byroad, or ground stratum ): is defined entirely by mosses, lichens, liverworts, and alga. Ground-creeping vines, prostrate shrubs and herbs should be treated in the herb stratum. Where herbs are entirely absent, it is still possible to recognize this stratum if other very low woody or semi-woody life forms are present.

Floating aquatic stratum: includes rooted or drifting plants that float on the water surface (e.g., duckweed, water-lily).

Submerged aquatic stratum: includes rooted or drifting plants that by-and-large remain submerged in the water column or on the aquatic bottom (e.g., pondweed). The focus is on the overall strata arrangement of these aquatic plants. Note that emergent plants life forms in a wetland should be placed in the strata listed above (e.g., cattail or sedges would be placed in the herb stratum, whereas the duckweed would be in the floating aquatic stratum).

Epiphytes, vines and lianas are not treated as separate strata; rather they are treated within the strata defined above, but can be distinguished from other life forms in the strata using the life form field.

More finely divided substrata can be used (for example, the tree stratum may be divided into canopy tree and subcanopy tree, and the shrub stratum may be divided into tall shrub and short shrub), but these should always nest within rather than span multiple standard strata.

For each stratum, the total percent cover and the prevailing height of the top and base of the stratum should be recorded. Cover is a meaningful attribute for nearly all plant life forms, which allows their abundances to be evaluated in comparable terms (Daubenmire 1968, Mueller-Dombois and Ellenberg 1974). Percent cover can be defined generically as “the vertical projection of the crown or shoot area to the ground surface expressed as … percent of the reference area” (Mueller-Dombois and Ellenberg 1974). The use of crown or shoot area results in two definitions of cover as follows:

Canopy cover: the percentage of ground covered by a vertical projection of the outermost perimeter of the natural spread of foliage of plants. Small openings within the canopy are included” (SRM 1989).

Foliar cover: the percentage of ground covered by the vertical projection of the aerial portion of plants. Small openings in the canopy and intraspecific overlap are excluded” (SRM 1989). Foliar cover is the vertical projection of shoots; i.e., stems and leaves.

Canopy cover is the preferred method of collecting cover because it better estimates the “area that is directly influenced by the individuals of each species” (Daubenmire 1968) and canopy cover, or canopy closure, is easier than foliar cover to estimate from aerial photos and is more likely to correlate with satellite image analysis. A classification based on canopy cover is better suited for mapping vegetation than one based on foliar cover.

The best practice for recording the canopy cover of strata is to record percent cover as a continuous value; however it may be estimated using categorical values of, for example, 5-10% intervals or another recognized cover scale (but see below). The percent cover of the three most abundant growth forms in the dominant or uppermost strata should also be estimated (see Table 1 for a list of growth forms). For example, in addition to total cover estimates for all trees in a stand dominated by the tree stratum, separate cover estimates of the dominant growth forms (e.g., deciduous broadleaf trees, needleleaf evergreen trees) should be made. These estimates will help place the plot within the physiognomic hierarchy of the NVC. An approximate total cover representing simultaneously all strata and growth forms should be assigned, but should follow the standard rules for cover, such as having a maximum possible value of 100%. Finally, and importantly, for each taxon a total cover summed across all strata should be assigned, again with a maximum value of 100%.^{^[8]}

In describing vegetation structure, the following rules should be followed:

1. Always recognize the standard strata (tree, shrub, herb, moss, floating, submerged), where present. Substrata (e.g., canopy tree and subcanopy tree, tall shrub and short shrub) can be used, but these should always nest within rather than span multiple standard strata.

2. Provide the prevailing height of the top and the base of each stratum.

3. The cover of the stratum is the total vertical projection of the canopy cover of all species collectively on the ground, not the sum of the individual covers of all species in the stratum. The total cover of the stratum will, therefore, never exceed 100% (whereas, adding up the individual cover of species within the stratum could well exceed 100% since species may overlap in their cover).

4. Plants are assigned to strata based on their predominant position or height in the stand, not by their taxonomy or mature growth form. Consequently, a tree species that has both seedlings and saplings in a plot could be listed in several strata.

5. Epiphytes and lianas are handled in different ways by various field protocols. When treated as individual species for cover assessment, they may be treated as a special growth form-strata, independent of the strata mentioned above, or they may be assigned to the standard strata on the basis of the location of their predominant canopy cover. Bryophytes (including liverworts) and lichens growing on the same substrate as vascular plants are treated as part of the nonvascular strata. When assessing total cover of the primary strata, an epiphyte or liana should be included in the primary stratum where it has its predominant canopy cover.

6. The herb stratum (sometimes called field stratum) includes all woody or semiwoody plants or creeping vines where these overlap in height. This is a compromise between strata based strictly on height versus growth form. More specific measures of growth form (forbs, grasses, dwarf-shrubs) composition within this stratum can either be recognized directly in the field or can be estimated after the fact by assigning species within a stratum to a growth form category and calculating an approximate percent cover of the growth form.

7. The moss stratum (sometimes called nonvascular, byroid, or ground stratum) is reserved strictly for cryptogams (mosses, lichens, liverworts, algae and bacteria), even where herbs or woody plants may be reduced to very short heights.

Floristic composition

Species list

For field plots used to classify vegetation, sampling should be designed to detect and record the complete species assemblage of the plot. As a minimum standard, only one field visit is required. Generally, plots should be recorded only when the vegetation is adequately developed phenologically so that the prevailing cover of each species can be assessed. However, some plant species may not be visible in certain seasons (e.g., spring ephemerals) or may be unreachable (e.g., epiphytes, cliff species), and thus not identifiable. All reasonable efforts should be made to ensure that such species are recorded, and their occurrence at least noted. The phenological aspects of vegetation exhibiting clear seasonal changes in floristic composition must also be noted (e.g., young grasses, whose abundances may be underestimated in late spring). In cases where phenological changes are pronounced (especially among dominants), repeat visits are highly recommended.

At a minimum, data must include a comprehensive list of all vascular plant species visible in the plot at the time of sampling together with an overall assessment of their cover (but see previous section for legacy data). A conscientious effort should be made to thoroughly traverse the plot to compile a complete species list. Nonvascular plants (e.g. bryophytes and lichens) should be listed where they play an important role (e.g., peatlands, rocky talus). We recommend, but do not require, that a list of additional species found in the stand (but not the plot) also be compiled. However, it is important that species within the plot be distinguished from those outside the plot, in order that diversity estimates for the plot (or area) not be inflated.

All plant taxa should be identified to the finest taxonomic resolution possible. For example, variety and subspecies level determination should be made routinely where appropriate. In addition, it is essential that the basis for the name applied for each taxon be identified. Plant names have different meanings in different reference works, and it is imperative that the meaning of each name be conveyed by reference to a standard authoritative work. In lieu of an authoritative work, an investigator may specify Kartesz 1999 or subsequent editions, though this should only be done with great caution so as to avoid inadvertent misidentifications. Kartesz 1999 is the basis for the list maintained by the USDA PLANTS (2003) database as a taxonomic standard. If using USDA PLANTS as an authority, it is imperative that the version and observation date be provided.

Species by strata

It is desirable and considered best practice (although not strictly required) that each species listed in a plot also be assigned to the main strata (tree, shrub, herb, nonvascular, floating, submerged) in which it is found. Not all plants will fit clearly into the strata recognized, but the purpose of listing species by vegetation structure is to document the composition of the most visible strata of the stand (see the above section “Vertical structure and physiognomy of vegetation”).

Cover

For each species found in the plot, an overall measure of cover must be recorded, and additional cover values by strata are recommended. Percent cover has been widely accepted as a useful measure of species importance that can be applied to all species. As discussed above, cover may be defined either as canopy cover or as foliar cover. Canopy cover is the recommended form of cover estimates. Cover values are relatively rapid, reliable, and, for the purposes of vegetation survey and classification, they accurately reflect the variation in abundance of a species from stand to stand (Mueller-Dombois and Ellenberg 1974).

Total cover should be recorded for all species in the plot. It is recommended that in addition to the overall cover value, four separate cover estimates be provided for each species if it is found in the herb layer, shrub layer, and tree layer. Recording abundance of species cover by strata provides a three-dimensional view of the vegetation and facilitates the interpretation of physiognomic and floristic relationships within the FGDC hierarchy. Cover values should be absolute rather than a relative portion of a layer (e.g., if a species forms a monospecific layer with a cover of 50%, the cover for the species is recorded as 50%, not as 100% of the layer). The cover for all species in any single layer (or overall) may be greater than 100%, as the foliage of one species within a layer may overlap with that of another. Cover can easily be converted from absolute to relative cover at a later stage, if that fits the needs of the investigator.

Species importance can also be measured as density (number of individuals), frequency (percentage of quadrats or points having a species present), biomass, basal area, total canopy cover, or some weighted average of two or more importance measures. Such supplemental measures of importance can add to the value of a plot, but are not required.

Cover scales

Use of cover classes instead of actual percent cover can speed up fieldwork considerably. Cover class estimates are acceptable because individual investigators cannot reasonably expect to discern small differences in the field such as between 52 and 53% cover, and in addition percent cover for a species varies over the course of the growing season (especially for herbs) giving high precision estimates a false connotation of accuracy. A practical cover scale needs to be roughly logarithmic, in part because the human eye can discern doublings better than a linear scale (e.g., it is easier to tell the difference between 1 and 2% cover than between 51 and 52%). In addition, many species are relatively sparse across all stands and small differences in their cover may be particularly important for classification. Table 2 provides a comparison of widely used cover-abundance scales. Among these, the Braun-Blanquet (1932) scale, which has been extensively and successfully used for vegetation classification purposes (Mueller-Dombois and Ellenberg 1974, Kent and Coker 1992), has a set of class boundaries at “few” (somewhere between 0 and 1%), 5%, 25%, 50%, and 75%. It provides a common minimal set of cover classes acceptable for classification. Any scale used for collecting species cover data needs to be convertible to this common scale by having boundaries at or near 0-1%, 5%, 25%, 50%, and 75%. By this criterion, the North Carolina (Peet et al 1998) and Krajina (1933) cover class systems are ideal in that they can be unambiguously collapsed to the Braun-Blanquet (1932) standard, and the Daubenmire (1959), Pfister and Arno (1980) and New Zealand (Allen 1992, Hall 1992) scales are for all practical purposes collapsible into the Braun-Blanquet (1932) scale without damage to data integrity. The Domin (1928), Barkman et al (1964), and Western Heritage Task Force (Bourgeron et al., 1991) scales all are somewhat discordant with the Braun-Blanquet (1932) standard and should be avoided.

When recording species cover in a plot, any species noted as being present in the stand, but not found in the plot, should be assigned a unique cover code, so that these species can be identified as not part of the plot itself.

Tree abundance measures

In North America, tree species abundance has often been assessed using individual stem measurements, basal area totals, or density. Nonetheless, cover is a requirement for trees because, by using cover, it is possible to look at the abundance of all species across all strata and to assess relationships between and among the strata. However, it can be difficult to accurately estimate cover of individual tree species in large plots (e.g., > 500 m²). In such cases, basal area and stem density measures can be used to supplement cover data. In addition, these data will allow comparisons with a wide variety of other forest plot data. For these reasons, collection of basal area and density (stem area and stem counts) for tree species is encouraged. For data sets where cover was not collected for the tree stratum, but which are otherwise acceptable for classification, it may be possible to calculate an estimate of tree cover derived from individual stem measurements or from basal area and density. The methods used for this conversion, including appropriate calibration techniques, should be thoroughly documented.

Environmental data

Environmental data provide important measures of the abiotic factors that influence the structure and composition of vegetation (see Appendix 1). For classification purposes, a select set of basic and readily obtainable measures is highly desirable. Physical features of the stand include elevation, slope aspect and slope gradient, topographic position, landform, and geology. Desirable soil and water features include soil moisture, drainage, hydrology, depth of water, and water salinity (where appropriate). The soil surface should also be characterized in terms of percent cover of litter (including dead stems < 10 cm), rock, bare ground, woody debris (dead stems > 10 cm), live woody stems, surface water, and other physical objects. Surface cover estimates should always add to 100% absolute cover. Habitat and stand conditions should be described, including landscape context, homogeneity of the vegetation, phenological expression, stand maturity, successional status, and evidence of disturbance. In many cases recommended constrained vocabularies (see Appendix 2 for recommended constrained vocabularies also used for automated “picklists”.) have been developed for these data fields and are documented at http://www.vegbank.org/. Plot data should conform to these vocabularies so as to facilitate data exchange.

Geographic data

All plot records must include geocoordinates in the form of latitude and longitude in decimal degrees as per the WGS 84 datum (also known as NAD83; see EUROCONTROL and IfEN 1998). Where data were originally collected following some other system (e.g., USGS quadrangles with the NAD27 datum), the original data should also be included should it become necessary to assess conversion accuracy at some future time. These original data should include x and y coordinates, the datum or spheroid size used with the coordinates, and the projection used, if any. Geographic data should include a description of the method used to determine the plot location (e.g., estimated from a USGS 7.5 minute quadrangle, use of a geographic positioning system). An estimate of the accuracy of the plot’s location information should also be included in the form of an estimate that the plot origin has a 95% or greater probability of being within a given number of meters of the reported location. Additionally, it may be useful to provide narrative information for plot relocation.

Metadata

Metadata are needed to understand how the plot data were gathered (see Appendix 1). All field plots must have a project name and project description, the methodology used to select and lay out the plots, cover method and strata method used, and the name and contact information of the lead field investigators.

Legacy data

Legacy data are plot data collected prior to the publication of these standards or without any documented effort to comply with these standards. Given that collection of vegetation plot data has been going on in the United States for over a century, including extensive sampling of some parts of the country, these data may contribute substantially to the improvement of the NVC. Some plots may represent stands (or even types) that no longer exist. Others may contain valuable information on the historic distribution and ecology of a plant community, or may contain important structural data (such as on old-growth features) that may be difficult to obtain today. Legacy data have no special status and must conform to the same rules as other plot data. However, care should be taken in importing legacy data to assure maximum compatibility with current standards. In using legacy data there are some difficult issues that should be addressed in the plot metadata. Problems include: (1) uncertainty about plot location, which is especially common for data that exist only in published form rather than field records; (2) inadequate metadata on stand selection, plot placement, and sampling method; (3) uncertainty about species identity because of changes in nomenclature and lack of voucher specimens; (4) uncertainty about completeness of floristic data; (5) uncertainty about sampling season; and (6) incompatibility of the cover or abundance measures used.

5.4. STANDARDS FOR VEGETATION PLOTS

1. Sampling methods: Methods used to select plot locations, choice of plot sampling technique, and comprehensiveness of vegetation description must be described in metadata.

2. Plot methods: