USGS:Biology Biological Informatics Program Vegetation Characterization Program Standardized National Vegetation Classification System 3.0 Vegetation Classification: Background

3.0 Vegetation Classification: Background

The objective of classification is to group together a set of observational units on the basis of their common attributes (Kent and Coker 1992). The end product of a classification should be a set of groups derived from the units of observation where, typically, units within a group share more attributes with one another than with units in other groups. For vegetation classification, the unit of observation is typically the "stand," defined as a relatively homogeneous area with respect to species composition, structure, and function.

The process of classifying a particular type of vegetation on the landscape requires a clearly defined objective for the classification and a familiarity with the variability across its range. If the objective of a study is to create an independent vegetation classification system, attribute data on species, cover, vegetation age and structure, leaf characteristics, bark characters, dispersal mechanisms and life history traits should be collected and organized. If the objective is to classify ecosystems, data on the key environmental features such as soils, hydrology, landform, etc., need to be collected. The biological and environmental information to be collected, organized, and described must be carefully chosen to meet the objectives of the classification.

3.1.1 Community Units and Continua

Within the Anglo-American ecological tradition, there has been a disinterest in classification per se. Beginning with the viewpoint of Gleason (1917, 1926), extended by others, including Whittaker (1956, 1962) and Curtis (1959), it is held that vegetation units cannot be defined; species comprising a community respond individually to environmental gradients and to each other. Whittaker (1962) referred to this viewpoint as the "individualistic dissent." The question often became polarized between the "continuum concept" and the "community unit concept." The argument is still presented in such a polarized light today, despite efforts to broaden the discussion (Moravec 1992, Roughgarden 1989).

Despite the polarized viewpoints, several features of communities are widely recognized (Mueller-Dombois and Ellenberg, 1974):

• Similar species combinations recur.
• No two stands (or sampling units) are exactly alike.
• Species assemblages change more or less continuously if one samples a geographically widespread community throughout its range.

Thus, recurring species combinations are variously correlated with their environment, and these shift geographically. Austin (1991) considered that vegetation units will be most interpretable within certain landscape regions. In sum, an ordering is possible, but within limits. Vegetation classifications often require a predetermined consistency that does not do justice to the complexity and variability of the units. The same may be said for land classifications. The goal of classification is to determine the relative degree of similarity and dissimilarity among units while recognizing that the communities are distributed on a continuum across the landscape.

3.2 Review of Different Approaches to Classification

Many vegetation classification systems have been developed, but three have gained widespread acceptance: physiognomic classifications, floristic classifications, and site or ecosystem classifications (Howard and Mitchell, 1985). The intent of all three is to provide a systematic ordering of vegetation or ecosystem pattern and to relate these patterns to ecological processes. Following is a brief survey of various classification systems and a description of their strengths and limitations.

3.2.1 Vegetation Classifications

3.2.1.1 Physiognomic Methods

Beginning in the nineteenth century, with the work of plant geographers such as Humboldt, Warming, and Grisebach, vegetation classification focused on the outward appearance or physiognomy of the vegetation. Broadly speaking, physiognomy refers to structure (height and spacing of the vegetation) and life forms of the dominant species (the gross morphology and growth aspect of the plants). In addition, physiognomy refers to characters of seasonality, leaf shape, phenology, duration, etc. These features are easily recognized in the field and can be applied with little knowledge of the flora. In addition, they permit generalizations about the vegetation at a coarse, often worldwide scale. The basic unit of several physiognomic classification systems is the "formation." a "community type defined by dominance of a given growth form in the uppermost stratum (or the uppermost closed stratum) of the community, or by a combination of dominant growth forms" (Whittaker 1962). In practice, formations are defined by varied, conventionally accepted combinations of growth-form dominance and characteristics of the environment. "Cold-deciduous alluvial forests," "evergreen subdesert shrublands," and "alpine meadows" are examples of formations.

The predominance of certain physiognomic types in a region tend to correspond to major climatic zones. Thus, physiognomic categories are often expressions of macroclimate, soils, and vegetation (Holdridge 1947, Walter 1985, Howard and Mitchell 1985). As a result, broad-leaved evergreen trees tend to be found in tropical climates, evergreen needle-leaved trees tend to be found in boreal climates, etc. Physiognomic features provide a fast, efficient way to categorize vegetation, can often be linked to remote sensing signatures, and are useful for initial reconnaissance of areas requiring survey. Physiognomic classification systems generally emphasize a divisive (or "top-down") approach, subdividing coarse vegetation patterns into units suitable for small-scale assessment. In addition, physiognomy reflects the effects of disturbance and management (such as grazing or fire), though in a relatively coarse way.

In the twentieth century, the physiognomic traditions of Warming and others were expanded in several directions (as described in detail by Whittaker 1962 and Shimwell 1971). In Europe, Brockman-Jerosch and Rubel (1912) and Rubel (1930) emphasized physiognomy together with species dominance. Their methods were expanded by Fosberg (1961), Ellenberg and Mueller-Dombois (1967) and the United Nations Educational, Scientific and Cultural Organization (UNESCO 1973). In the United States, Clements (1916, 1928), and later Brau n (1947, 1950) identified broad-scale regional formations, described by major dominants sharing the same physiognomy. More appropriately called "vegetational regions," these units described what were thought to be the "climatic climax types," areas of vegetation that were typical, mature phases of the vegetation. Other recent descriptions of vegetation in the United States that emphasize physiognomic units can be found in Vankat (1979), Barbour and Billings (1988), and Barbour and Christensen (1993). In Great Britain, the work of Moss (1913), Clements (1916), Watt (1934), and Tansley (1939) described both climatic climaxes and edaphic climaxes, areas of vegetation occurring on different soils within the same climate (poly-climax types). In the tropics, structural profiles of the vegetation were described in detail and physiognomic units characterized the layers (Richards 1952, Beard 1955, Cain et al. 1956).

Mapping standards improved as cartographic techniques summarizing vegetation structure through symbols were developed by Dansereau (1951) and Kuchler (1949, 1967). Kuchler's (1964) work led to a physiognomic vegetation map of the United States that has received widespread use and management application (Klopatek et al. 1979, Crumpacker et al. 1988).

3.2.1.2 Floristic Methods

Whereas most physiognomic methods emphasize attribute patterns of dominant species groups in the vegetation, floristic methods characterize the species themselves. The basic floristic unit is the "association," defined by Flahault and Schroter (1910) as "a plant community of definite floristic composition, presenting a uniform physiognomy, and growing in uniform habitat conditions." This definition implies that associations that share a certain physiognomy would be grouped together into the same formations.

In defining associations, some floristic methods focus on species that occur constantly throughout a set of stands, while others emphasize indicator or diagnostic species, species that are dominant or restricted to these stands. Floristic methods require intensive field sampling, detailed knowledge of the flora, and careful tabular analysis of stand data to determine the constant or diagnostic species groups. Floristic methods reflect local and regional patterns of vegetation and are more detailed than physiognomic methods. They also provide detailed descriptions of biotic communities regardless of their successional stage or origin. As such, they are typically organized by an agglomerative (or "bottom-up") approach, with lower units being combined into higher ones. Floristic composition is often correlated with soil or landform patterns. Thus, floristic units have been used frequently as indicators of ecosystem processes and are a useful component of ecosystem classifications (Mueller-Dombois and Ellenberg 1974, Rowe 1984, and Strong et al. 1990).

Early twentieth century ecologists who favored a strict floristic system included members of what has been termed the Zurich-Montpellier Tradition in central Europe (see Shimwell 1971). The most well known among them is Braun-Blanquet (1928, 1932, 1951), whose work established a formal approach to the floristic classification of vegetation. The Braun-Blanquet system has been explained in detail by Poore (1955), Becking (1957), Whittaker (1962), Mueller-Dombois and Ellenberg (1974), and Westhoff and van der Maarel (1973). Initially, floristic data (composition and cover of species) are collected from stands using plot methods. The plot, a relèvé, is placed in an area of the stand that is considered to be representative of the vegetation of the entire stand. The plot data are then compiled into tables (species by plots), and the species are sorted to identify those that co-occur in certain patterns. Based on this analysis of the plot data, stands can be grouped into associations. The associations can then be compared to one another to determine which groups of species best exemplify the association, either by being dominant or restricted to the association.

Species that are common to several associations can be used to assemble the associations into broader groups. For example, the Braun-Blanquet approach groups plant associations with common diagnostic species into units called "alliances." In this way associations can be arranged into a hierarchy based on floristic composition. Mueller-Dombois and Ellenberg (1974) note that the association concept has become more narrowly defined as more information is gathered in a region. They consider the alliance level, where species with more widespread distribution are used to identify groupings, a more easily defined unit at the regional level and useful for orientation with respect to floristic composition.

Ecologists in northern Europe initially emphasized floristic differences between the vegetation layers rather than the overall floristic list, but they subsequently adopted an approach similar to that of Braun-Blanquet (Whittaker 1962). In England, less effort was expended in formalizing the use of floristics, and more on basic description for the purposes of vegetation dynamics (Tansley 1939, Watt 1947). Recent efforts by Rodwell (1991) emphasized species constancy to define associations, and represents a substantial contribution to a fully developed floristic classification of British vegetation. Until recently, floristic classifications in the United States have only focussed on very local areas.

3.2.1.3 Potential versus Existing Vegetation

When identifying objectives for a classification, it is important to decide whether the classification is intended to portray existing vegetation or potential natural vegetation (PNV). Classifications emphasizing existing vegetation determine their vegetation units based on the current characteristics of the vegetation regardless of the stage of development. Stands are classified according to their characteristics at the time the sample is collected. The selection of the stands for sampling, however, may be weighted to those considered most natural.

Classifications emphasizing potential natural vegetation use vegetation characteristics that represent the most mature and stable endpoints of vegetation development. In the words of Tuxen (1956, in Mueller-Dombois and Ellenberg 1974), potential vegetation becomes "the vegetation structure that would become established if all successional sequences were completed without interference by man under the present climatic and edaphic conditions (including those created by man)." Thus the vegetation units are hypothetical units that are thought to indicate a site's potential for developing certain kinds of vegetation. These units are based on known current relationships between vegetation and site characteristics, such as soils or landform. They can be used to great advantage by land managers faced with a landscape where much of the vegetation has been removed. However, PNV units are limited by the current knowledge of vegetation-site relationships, and the ability of vegetation per se to infer site characteristics. They also emphasize hypothesized climax vegetation, a concept fraught with theoretical difficulties.

The best known portrayal of potential natural vegetation is that of Kuchler (1964), who mapped the potential natural vegetation of the United States at a scale of 1:3,168,000 and (in 1985) at a scale of 1:7,500,000. This map is limited in its focus to only mature types. Thus, for example, extensive natural stands of trembling aspen are not portrayed on the map because these are not considered climax types.

3.2.2 Site Classifications

3.2.2.1 Site Classifications Emphasizing Vegetation

Site classifications are intended to reflect the potential of a particular site to support various types of vegetation. A number of different site classification systems have used vegetation only to determine the site potential, usually with reference to successional trends or productivity. In this sense, these systems focus on potential natural vegetation.

Site classifications emphasizing vegetation have been developed in concert with the development of physiognomic and floristic classifications. Cajander (1909, in Shimwell 1971) noted how the same understory composition could occur under different canopy dominants in a system of "forest site types." He inferred that ground vegetation is more representative of site factors than are canopy dominants and worked with others to describe ecological series of communities along environmental gradients.

A widespread approach to site classification using vegetation is that of the habitat type classification system (Daubenmire 1952, Pfister and Arno 1980, Kotar et al. 1988). This system focuses on natural climax or near climax vegetation with an emphasis on all understory species as a faithful reflection of site characteristics. Relationships between vegetation and the soils or landform factors are established during and after the classification process, but these factors are not used to define the vegetation units (Komarkova 1983). The units described are natural ones, but the emphasis is on determining vegetation units that represent "ecologically equivalent landscapes" (Kotar et al. 1988). Insofar as they describe the floristic composition of part of the natural vegetation, namely climax stands, the units of the habitat type are fairly equivalent to the plant association concept (Komarkova 1983). The intent is to use these descriptions to visit sites that are not at climax and, by examining their understory composition, to infer their ecological potential.

Somewhat different from the habitat type approach is that of ecological species groups, which are species that show similar "ecological behavior." Generally these species belong to the same layer of vegetation (e.g., the herb layer, nonvascular layer, or shrub layer). The method presumes that communities are combinations of plant species whose composition is dependent on the local environment (Mueller-Dombois and Ellenberg 1974). The community unit identified can, at times, be very similar to the plant association level, whereby the ecological species groups are the diagnostic species for the association. However, it is also possible that the same association could contain several ecological species groups (Mueller-Dombois and Ellenberg 1974). The ecological species group information can either be used by itself to indicate site characteristics, in which case the system partially resembles the habitat type system, or it can be integrated with other measured site factors as part of an ecosystem classification (Pregitzer and Barnes 1982, Cleland et al. 1994).

3.2.2.2 Site Classifications Emphasizing Multiple Factors

Site classification systems that use multiple factors have as their focus the subdivision of land into major and minor land types or landscape ecosystems. They have been developed primarily for land managers who need to integrate resource management, biological conservation, and restoration planning. They are also used for comparisons of productivity, species distributions, and interactions. These systems are most appropriate for classifying ecosystems, defined by the dynamic interactions of the biotic and physical components. Ecosystems are treated as "layered, volumetric segments of the biosphere" (Barnes et al. 1982, Rowe 1984). As with vegetation classifications, emphasis is placed on units that are more or less homogeneous both as to form and structure, but in this case with respect to all factors of the land and the vegetation supported thereon (Rowe 1961).

An ecosystem approach to classification, namely that the plant community is considered together with its environment, was implicit in Clements work (1916), but was defined explicitly by Tansley (1935) and similarly by Sukachev (1945) as "biogeocoenosis." Central to the application of the approach is that all parts of the system are included. In some systems, each part vegetation, soils, climate and landform - is first studied independently and then combined (Jones et al. 1983, Sims et al. 1989, Driscoll et al. 1984). For others, it is considered essential that the parts be combined at the outset, since it is their joint interactions on the landscape that define the units. It is difficult to bring together all of the multiple factors jointly beyond the local level and understand their interactions. Thus, the units are considered hypotheses in need of further testing (Albert et al. 1986). Mapping is a key step in the process (Rowe 1984, Zonneveld 1989). Bailey's ecoregional map of the United States (1976, 1994) is more like the independent approach, as he relies heavily on separate climatic, physiographic, and vegetation maps and then reconciles their boundaries. The work of Albert et al. (1986) and Cleland et al. (1994) represent more of the combined approach.

The biogeoclimatic zone system of Krajina (1965, in Mueller-Dombois and Ellenberg 1974) is another system in which vegetation is emphasized in defining landscape or ecosystem units. These zones are defined as geographic areas that are predominantly controlled by the same macroclimate and contain similar soils and (climatic climax) vegetation. The definition of the zones at lower scales utilizes vegetation units that are defined by the plant association concept (Pojar et al. 1987). At higher levels, climatic zones and topographic position are used to help group vegetation units into the biogeoclimatic zones.

3.2.3 Land Cover Classifications

Land cover classifications are primarily intended for land management or resource planning. They emphasize conspicuous features of the land surface, and can be combined with land-use maps to convey an overall perspective on what is visually present on the land. As such, they often rely on characters that can be seen by remote sensing images (Witmer 1978).

To a certain extent, land cover classifications can draw from units defined by physiognomic classifications (Anderson et al. 1972). For example, forest cover types are a "descriptive classification on forest land based on present occupancy of an area by tree species. They are named by characteristic dominants that recur over tens of thousands of hectares," (Eyre 1980). Since physiognomic units also emphasize the dominant features of the vegetation (see above), there is some overlap in perspective.

3.2.4 Combined Classification Approaches

There are many commonalities among these classification systems. For example, site classifications include considerable vegetation information that is collected in the same way that would be used for vegetation classification (Mueller-Dombois and Ellenberg 1974, Pregitzer and Barnes 1982). Similarly, habitat type classifications define plant associations in a similar manner to that of the floristic system of Braun-Blanquet. Furthermore, site classifications that bring together independent vegetation, soil, and landform classifications rely on the independent classification of these variables as their starting point (Jones et al. 1983, Sims et al. 1989).

3.2.4.1 Physiognomic-Floristic Approaches

The principle underlying physiognomic classification is that each specific life form has a strategy (Stearns 1976) which has been selected under similar ecological pressures, and that the composition of life forms in a vegetation type is governed by these strategies (Monsi 1960, Raunkier 1904, Walter 1973, Whittaker 1975). Since physiognomic attributes are borne by individual species, recognition of a physiognomic assemblage depends on the co-occurrence of species in a given area. The co-occurrence of species leads to specific physiognomic vegetation types that can be delineated as discrete units in the landscape. As such, the physiognomic types can be related to floristic classifications that include the total composition.

The advantages of the separate components of the physiognomic and floristic approaches to classifying vegetation have been presented above. An important reason for combining these approaches is that vegetation is most thoroughly described by both structure and floristic composition. Physiognomic systems are easily recognized in the field, can be applied with little knowledge of the flora, permit generalizations of vegetation patterns over large areas, and can be linked to remote sensing signals to facilitate vegetation mapping. These attributes allow the identification of patterns where little is known about an area, or more detailed survey is impractical. Floristic information, however, is almost always used for detailed site analyses, whether for studying environmental gradients, ecological site factors, or describing and forming classification units. Patterns of succession, disturbance, history (including paleo-ecology), and natural assemblages are better assessed through floristic composition than physiognomy.

A fully developed classification is most readily developed by combining physiognomy and floristics. This type of system allows the geographic orientation of physiognomic characters to be tied to the more local site specific information of the floristic characters. In combination, these systems can satisfy a broader range of objectives for use of the classification system. In particular, the combined physiognomic floristic approach has the desirable attribute of producing mappable units with significant ecological meaning.

The rationale for such a coupling of systems has been developed over the years (e.g., Rubel 1930, Ellenberg 1963, Webb et al. 1970, Wergner and Spangers 1982, Westhoff 1967, Westhoff and Held 1969, Borhidi 1991). These studies have found a very good fit between floristic and physiognomic classifications of the same areas because both types of attributes are borne by individual species. Whittaker (1962, p. 137), despite his hesitation on the usefulness of vegetation classifications, provided guidelines on the development of a physiognomic floristic system, when such systems were warranted. He fully expected that plant associations, ecological species groups, and habitat types could be used to develop flexible, but consistent community units. In the United States, Driscoll et al. (1984) recommended the development of a joint system using the physiognomic units of UNESCO (1973) and the floristic units of habitat types, of which an example has recently been provided by Dick Peddie (1993) in New Mexico. Strong et al. (1990) in Canada also proposed a combined physiognomic floristic approach. The list of plant communities which was used to map the vegetation of Australia's National Parks and Reserves was developed by Specht et al. (1974) using a joint physiognomic floristic approach.