Open Access
Translator Disclaimer
1 January 2015 Plant Species Coalition Groups of Zion National Park: An Individualistic, Floristic Alternative to Vegetation Classification
Jeffrey E. Ott, Stewart C. Sanderson, E. Durant McArthur
Author Affiliations +
Abstract

Vegetation surveys at Zion National Park (Zion), Utah, have contributed to our understanding of plant community patterns and their relationship to environmental factors. Previous authors used vegetation plot data to characterize vegetation types at Zion following conventional procedures that emphasize spatial discreteness and dominant species. We developed and applied an alternative approach for community characterization that emphasizes nondiscrete presence-absence patterns and is compatible with the individualistic concept. We reanalyzed existing plot data from Zion using coalition clustering, an algorithm that identifies groups of positively-associated species referred to as coalition groups. Each species and plot in the data set was linked to each coalition group via an “affinity” value obtained through weighted averaging. Affinity values were used to characterize environmental affinities of coalition groups through regression tree modeling and predictive mapping. We also identified species that frequently co-occurred with coalition groups (affiliate species) and those that frequently co-occurred with high cover (dominant-affiliates), viewing these as alternatives to conventional prevalent and dominant species. Following this approach, we identified 10 coalition groups at Zion that overlapped compositionally and spatially to differing degrees. Mesic environments on a gradient from low-elevation riparian zones through mid-elevation narrow canyons to high-elevation plateaus were represented by 3 overlapping groups. Two groups occupying slickrock and sand environments were detected on the Navajo Sandstone, as well as 2 on mesa tops above it. At lower elevations, 3 intergrading xeric coalition groups were distinguished. When previously classified associations of the National Vegetation Classification were clustered based on shared affinities to coalition groups, the arrangement differed from existing classification schemes but was environmentally interpretable. Although these patterns are contingent on conditions at the time of data collection, they provide a baseline that could be used for evaluating and predicting plant community change in the park. With proper attention to sampling and analysis issues, our community characterization approach could be applied in other settings as an alternative or supplement to conventional vegetation classification.

Zion National Park (Zion), Utah, is best known for its striking geological features but also boasts a high diversity of plant taxa and plant communities. Over 800 native vascular plant species (Fertig and Alexander 2009, Fertig et al. 2012) and 95 vegetation associations (Cogan et al. 2004) are distributed across pronounced edaphic and climatic gradients within the park (Harper 1993, Biek et al. 2003, Cogan et al. 2004, O'Meara 2006, Fowler et al. 2007, Fertig and Alexander 2009). The concentrated and contrasting environmental and botanical diversity of Zion contribute to the park's scientific and conservation value. Given the U.S. National Park Service mandate to protect natural resources for the enjoyment of present and future generations, information on the distribution, composition and environmental attributes of plant communities at Zion is important for science-based resource management (NPS 2006).

Numerous workers have contributed to current knowledge of vascular plant communities at Zion. Plant collection efforts from 1894 to the present have documented the flora of the park and provided insights concerning species distributions and habitat relations (reviewed in Fertig and Alexander 2009, Fertig et al. 2012). Woodbury's (1933) monograph of biotic communities was followed by more detailed descriptions of specific plant communities such as hanging gardens (Malanson 1980, Malanson and Kay 1980, Welsh 1989, Fowler et al. 2007) and isolated mesa tops (Madany and West 1983, 1984). Two park-wide, plot-based vegetation surveys have been carried out at Zion, one in 1987–1989 (Harper 1993, Harper et al. 2001) and the other in 1999–2003 (Cogan et al. 2004). Harper (1993) and Harper et al. (2001) identified 10 major vegetation types at Zion using an informal, expert-knowledge approach, while Cogan et al. (2004) characterized vegetation more formally and in greater detail according to the U.S. National Vegetation Classification (NVC) (Grossman et al. 1998, FGDC 2008, Jennings et al. 2009). These studies have jointly yielded an enviable amount of information but they have not exhausted all that can be known about plant communities at Zion. In this paper we demonstrate that even in the absence of new data, alternative approaches for data analysis and synthesis can provide valuable new information and insights.

Harper et al. (2001) and Cogan et al. (2004) followed conventional vegetation classification practices of delineating spatially discrete vegetation types differentiated primarily by dominant species or vegetation structure. While these practices are widely followed and useful for many purposes (Mueller-Dombois and Ellenberg 1974, Grossman et al. 1998), they result in community characterizations that emphasize certain ecological patterns at the possible expense of others. Continuous gradations of community composition due to individualistic species responses (Gleason 1926, 1939) can be obscured by discrete community models, and distributions of subordinate species might not be well represented by those of dominants. These issues are relevant at Zion because species responses to the park's complex overlay of environmental gradients are likely to be individualistic, and because the number of structurally or numerically dominant species is small relative to the number of species in the flora as a whole (Cogan et al. 2004, Fertig and Alexander 2009).

In this paper we present an alternative plant community characterization for Zion that mirrors the earlier vegetation classification of Harper (1993) but is neither spatially discrete nor built around dominant species. Instead of following Harper's (1993) approach using dominant species as classification criteria, then characterizing the floristic composition of communities thus classified, we essentially reversed the process by first identifying species groups defined by presence-absence patterns, then identifying dominant species and otherwise characterizing the communities and environments where these species groups occur. This was accomplished by reanalyzing the Harper (1993) data set using coalition clustering and other algorithms from the RCLUS computer program (Sanderson et al. 2006) in combination with regression tree modeling (Breiman 1993) and predictive mapping (Ferrier and Guisan 2006). Because RCLUS is relatively new and its algorithms are here applied in a novel way, one of our objectives is to outline the conceptual and methodological basis of our approach. We draw conceptual parallels that align our approach with Curtis's (1959) modal and prevalent species concepts, applied here in a spatially nondiscrete way. We demonstrate the ecological insights gained from our approach by comparing our results directly with those of Harper et al. (2001) and Cogan et al. (2004). We also use our results to build an alternative higher-level classification of the NVC associations described by Cogan et al. (2004) that reflects their position in relation to environmental gradients at Zion.

Methods

Study Area

Zion National Park (Zion), Utah, USA, covers approximately 590 km2 at the western edge of the Colorado Plateau physiographic/floristic area, bordering the Great Basin to the northwest and the Mojave Desert to the southwest (Biek et al. 2003, McLaughlin 1986). Climate at Zion is mostly semiarid and temperate but varies greatly across an elevation gradient of 1128–2660 m (Cogan et al. 2004). Precipitation is concentrated in winter/spring frontal storms (bringing snow to higher elevations) and late-summer monsoons (Woodbury 1933, Mortensen 1977). At Zion Canyon (1234 m), weather records from 1904 to 2012 show a mean annual precipitation of 382 mm and mean temperatures of 5 °C in January and 29 °C in July (WRCC 2014).

Horizontally aligned exposures of sedimentary rocks are prominent at Zion. The oldest strata occur at low elevations on the southwestern side of the park and on slightly higher elevations on the northwestern side where the Taylor Creek thrust fault disrupts their stratigraphic continuity (Biek et al. 2003, O'Meara 2006). The lowermost geologic formations (Kaibab, Moenkopi, Chinle, Moenhave, and Kayenta) contain alternating bands of mudstone, siltstone, sandstone, conglomerate, limestone, and gypsum that erode into badlands, slopes, and cliffs (Biek et al. 2003). These strata are dwarfed by the larger cliffs and ‘slickrock’ exposures of the overlying Navajo Sandstone (the park's primary scenic attraction), which is further overlain by sandstone, limestone, shale, and gypsum deposits of the Temple Cap, Carmel, and Cedar Mountain formations (Biek et al. 2003, O'Meara 2006). The sedimentary sequence is in turn partially covered by alluvial, lacustrine, landslide, talus, residual, eolian, and volcanic (basalt flows and cinder cones) deposits (Biek et al. 2003, O'Meara 2006).

Soils at Zion are generally poorly developed and have been classified as rock outcrops, entisols, aridisols, mollisols, and alfisols (Mortensen 1977). Some soils have distinctive physical and chemical properties, as exemplified by the bentonite shrink-swell clays of the Petrified Forest member of the Chinle Formation (Biek et al. 2003). Cryptobiotic crusts are present on many soil surfaces and contribute to soil stability and fertility (cf. Belnap et al. 2000).

Although most of Zion currently has an undeveloped wilderness character, vegetation has been affected by historic floodplain agriculture, livestock grazing, and fire suppression, resulting in increases in invasive annuals and woody plants at the expense of native grasses and forbs (Woodbury 1933, Madany and West 1983, Cogan et al. 2004). Although livestock numbers have dropped since the early twentieth century, mule deer (Odocoileus hemionus) have increased in areas of high human visitation, resulting in declines in cottonwood (Populus fremontii) and other riparian species (Ripple and Beschta 2006). The troublesome invasive species cheatgrass (Bromus tectorum), red brome (Bromus rubens), ripgut brome (Bromus diandrus), and tamarisk (Tamarix spp.) were already well-established in Zion at the time of the Harper (1993) and Cogan et al. (2004) vegetation surveys.

Vegetation Data

Our primary data source is the vegetation survey carried out in 1987–1989 by Dr. K.T. Harper and associates (Harper 1993, Harper et al. 1992, 2001, 2003), which we refer to as the Harper Survey. This was the first park-wide, plot-based vegetation survey conducted for Zion. The Harper survey followed a systematic sampling strategy in which U.S. Geological Survey (USGS) cadastral section corners and corresponding park boundary markers served as sampling targets (Fig. 1A). In addition to 270 plots located at or near these markers, 18 plots were placed using subjective criteria so as to capture spatially restricted plant communities (riparian areas, hanging gardens, and old fields). Data collected within 100-m2 circular plots (or rectangular, in 6 cases) included vascular plant species composition following Welsh et al. (1987), ocular estimates of species cover classes according to the Braun-Blanquet scale (Braun-Blanquet 1932), and percent cover of bare soil, rock, litter, and cryptogams (Harper 1993, Harper et al. 2001).

Fig. 1.

Plot locations in Zion National Park, Utah, as sampled by 2 vegetation surveys referenced in this paper.

f01_26.jpg

Harper (1993) used the Harper survey data to classify the vegetation of Zion into ten vegetation types that were concisely summarized in Harper et al. (2001). Plots were assigned to vegetation types “commonly used by land managers in the area” based on dominant species and environmental context (Harper et al. 2001). Prevalent and modal species were characterized for each vegetation type following Curtis (1959). Prevalent species are defined as the most frequently encountered species (based on plot occurrences) within a community or vegetation type, where the number of prevalent species is equal to the species density (average number of species per plot) for the community (Curtis 1959, Harper et al. 2001). Modal species are defined as prevalent species that reach their greatest frequency in a given community relative to other communities under study. In Harper et al. (2001), modal species were identified for each vegetation type relative to the others described at Zion.

The Hanging Gardens vegetation type described by Harper (1993) and Harper et al. (2001) incorporated data from published studies of Malanson and Kay (1980) and Welsh (1989); however, these data were excluded from the current analysis, which uses only plot data from the actual Harper survey.

We also draw from plot data originating with the USGS—NPS Vegetation Characterization Program at Zion (Cogan et al. 2004), hereafter USGS Survey, which featured 346 circular, square, or rectangular plots (Fig. 1B) ranging from 100 m2 to 400 m2 where plant species were identified following Kartesz (1999). A gradsect strategy (Gillison and Brewer 1985) was used to divide Zion into biophysical units (BPUs) that were mapped and targeted for field sampling. Accessible polygons of each BPU were targeted by field survey teams who placed plots on sites deemed representative of major vegetation types and relatively homogeneous in terms of vegetation structure and composition (Cogan et al. 2004). Plot data were collected in 1999–2000.

Cogan et al. (2004) used the USGS survey data to classify and map the vegetation of Zion following the U.S. National Vegetation Classification System (NVC) (FGDC 2008, Jennings et al. 2009). Each plot was assigned to an association nested within higher levels of the NVC as defined at the time of publication (Grossman et al. 1998, Cogan et al. 2004). Multivariate analysis of species cover class data by strata guided the classification process (Cogan et al. 2004).

We merged the Harper data set, the list of prevalent and modal species by vegetation type from Harper et al. (2001), the USGS data sets containing species composition and NVC classification by plot, and our RCLUS results (described below) into a single database. Taxonomic names from these sources were standardized to Welsh et al. (2003) nomenclature so that they could be matched from one data set to another.

Coalition Clustering

We used coalition clustering, a nonhierarchical agglomerative clustering algorithm included in the program RCLUS (Sanderson et al. 2006), to identify groups of species with similar patterns of presence-absence in the Harper data set. RCLUS was developed as a tool for R-mode (species-based) clustering of community data sets. The name coalition clustering underscores the individualistic variability of species within the clusters (coalitions) that the algorithm builds. Coalition clustering generates clusters of positively associated species here referred to as coalition groups. The term core species, used by Sanderson et al. (2006) to refer to species belonging to a group, is here replaced by the term coalition species.

Different coalition clustering solutions can be obtained by varying 3 user-defined parameters of the algorithm: association coefficient, threshold association value, and minimum number of occurrences for inclusion in groups. Clustering solutions may also vary from one run of the algorithm to another because of the random nature of the agglomerative process (Sanderson et al. 2006). The clustering solution presented here was selected from among a small number of variants obtained using the phi coefficient at a threshold value of 0.15, allowing species occurring in 3 or more plots to be included in groups.

Association coefficients measure the degree to which 2 species show the same pattern of presence and absence in a data set, calculated from the cells of a two-way presence-absence contingency table:

e01_26.gif

The phi coefficient incorporates all 4 cells of the table according to the formula

e02_26.gif

Phi ranges from -1 (negative association) to +1 (positive association) and is independent of the total number of occurrences in the table (Jackson et al. 1989, Tichý and Chytrý 2006).

Coalition clustering agglomerates species whose mean pairwise association is higher than the user-defined threshold. The algorithm initiates the first group with a randomly selected species and then randomly cycles through the remaining species, adding them to existing groups or initiating new groups as appropriate. Species can belong to multiple groups with the exception that species already present in an existing group cannot initiate a new group. The algorithm undergoes several cycles of adding species, removing species and merging clusters allowing the composition of species groups to stabilize. The threshold association value affects group number and size; high thresholds tend to produce many small groups in contrast to few large groups at low thresholds (Sanderson et al. 2006).

Once the Coalition Clustering algorithm is complete, RCLUS saves the mean association—here referred to as affinity—of each species to each group (Sanderson et al. 2006). Affinity values provide a quantitative measure of the relationship between individual species and coalition groups (calculated for all species, not just those belonging to groups). RCLUS also averages affinity values for each group across all species in each plot, yielding values that can be correlated with environmental variables at plot locations (Sanderson et al. 2006). Our use of the term affinity thus has a specific quantitative meaning that encompasses 3 contexts: species affinities to species groups (mean pairwise association with members of the group), spatial affinities of species groups (plot-averaged species affinities), and environmental affinities of species groups (environmental correlates of spatial affinities).

Affiliation and Dominance Calculations

In addition to generating coalition groups, RCLUS contains a function that calculates percent co-occurrence of each species to each coalition group. Percent co-occurrence is calculated by summing joint presences across all pairings of a focal species with the species comprising a coalition group and then standardizing by the maximum possible number of joint presences. Using the notation of presence-absence frequencies, this can be represented as follows:

e03_26.gif

Thus percent co-occurrence of the focal species to the coalition group is the sum of joint presences between the focal species and each of the coalition species in the group (sp1 to spN), standardized by the total number of occurrences of all the coalition species. We have adopted the term affiliation to use interchangeably with percent co-occurrence.

Using percent co-occurrence values from RCLUS, we identified a set of species with high affiliation (affiliate species) for each coalition group. Given that Curtis (1959) used species density, the mean number of species per plot in a community, to determine how many species to include in prevalent species lists, we followed an analogous procedure for determining the number of affiliate species for each coalition group. The species density of a coalition group was calculated as the mean number of species occurring in plots where coalition species occurred, weighting plots by how many coalition species they contained.

Extending the affiliate species concept further to include abundance as well as occurrence, we calculated the abundance of each affiliate species as its mean cover across plots containing coalition species, weighted once again by the number of coalition species per plot. Arithmetic midpoints of cover classes were used in these calculations. In this way we were able to identify the dominant species occurring alongside members of each coalition group. Affiliate species with mean cover >2% were considered dominant-affiliate for our purposes.

Environmental Modeling and Mapping

Plot-averaged affinity values of coalition species groups, obtained as described above for each plot of the Harper survey, were integrated with environmental variables for correlative modeling. We selected 5 variables—elevation, slope, topographic exposure, topographic position, and geologic substrate—that are proxies for the major physical and chemical gradients affecting plant distributions at Zion. GIS layers for these variables were available or easily created and integrated in ArcGIS (ESRI 2003, 2004) (Appendixes 1, 2). Our use of GIS-based variables permitted predictive mapping of species group affinities beyond plot locations (Ferrier and Guisan 2006).

Plot coordinates from the Harper Survey, originally marked on paper maps, were digitized manually with the help of a high-resolution aerial orthophoto of Zion National Park (Cogan et al. 2004), digital USGS topographic maps, and cadastral survey marker coordinates acquired through the USDI Bureau of Land Management (BLM 2006). Hawth's Tools (Beyer 2004) was used to extract environmental data values for each plot based on its coordinates.

Elevation and other topographic variables were obtained from a 10-m digital elevation model (DEM) acquired from the U.S. National Park Service (NPS, personal communication). Slope and exposure were calculated using the slope and hillshade functions, respectively, in ArcGIS Spatial Analyst (ESRI 2004) (Appendix 1). Hillshade was implemented at 215° azimuth and 45° altitude such that exposure values were highest on steep southwest-facing slopes where solar heat loading is expected to be highest (cf. McCune and Keon 2002). Topographic position was defined using the Topographic Position Index (TPI), calculated as the difference in elevation between a pixel and the mean of surrounding pixels within a defined radius (Weiss 2001). We created 2 TPI layers at radii of 50 m (Appendix 1) and 100 m.

A 1:24,000-scale digital geologic map built from multiple maps of the Utah Geologic Survey (O'Meara 2006) was converted to raster format at the same resolution as the DEM and simplified from about 100 map units to 16 geologic substrate classes (Table 1, Appendix 2). We merged units with similar lithology, spatial continuity, and/or topographic characteristics, guided by exploratory analyses of relationships between putative substrate groupings and coalition group affinities. Our groupings include each of the major geologic formations of the park (Moenkopi [+Kaibab], Chinle, Moenhave, Kayenta, Navajo, Temple Cap, and Carmel [+Cedar Mountain]) and Quaternary deposits grouped according to their dominant depositional material (alluvium, colluvium, eolian, lacustrine, mass movement, residuum, and volcanics). Mass movement deposits were further split into 3 classes that occupy different topographic settings in the park: (1) talus, (2) slides/slumps/flows and (3) mass movement/colluvium/alluvial pediment mantle deposits (Table 1, Appendix 2).

Table 1.

Geologic substrate classes used for environmental modeling of plant coalition group affinities at Zion National Park, Utah. For each substrate class, corresponding map units from a digital geologic map of the Utah Geological Survey (O'Meara 2006) are shown in the rightmost column.

t01_26.gif

Environmental affinities of coalition groups (relationships between plot-averaged species group affinities and environmental variables) were modeled using regression tree analysis (Breiman 1993, Crawley 2007) in SPLUS (Insightful Corp. 2005) via the tree function with consistent splitting rules (mincut = 5, minsize = 10, mindev = 0.1). Multiple runs of random 10-fold cross-validation were applied to each model to determine the best number of terminal nodes minimizing total deviance (Crawley 2007), and models were pruned accordingly.

Relating Groups to Previously-described Vegetation Types

We quantified the degree to which vegetation types described by Harper et al. (2001) and Cogan et al. (2004) were compositionally similar to our coalition groups. This was accomplished by averaging, for each coalition group, species affinity values of matching species attributed to each vegetation type. For the Harper et al. (2001) data set, we calculated mean affinity values across modal species of vegetation types. For the USGS (Cogan et al. 2004) data set we averaged affinity values across all matching species recorded in plots, then further averaged across plots within NVC associations. NVC associations sharing similar averaged affinities across all (10) coalition groups were then identified through cluster analysis, using Pearson's correlation coefficient and average linkage hierarchical clustering in R (Crawley 2007, R Development Core Team 2010).

Results

Coalition Groups and Their Environmental Affinities

Eleven coalition groups were identified through coalition clustering at a threshold value of phi = 0.15. One group containing 6 species (Artemisia dracunculus, Elymus spicatus, Sisymbrium altissimum, Galium aparine, Physalis hederifolia, and Tragopogon dubius) was excluded from further analysis because of its limited distribution in disturbed low-elevation environments. The remaining 10 groups contained 9–36 species each and collectively captured 169 of the 511 species recorded by the Harper vegetation survey (Table 2, Appendix 3). These coalition groups can be readily described in terms of their environmental affinities at Zion, and we have given each group a name that describes the environmental setting for which it demonstrates greatest affinity (Table 2, Figs. 25). Coalition clustering's allowance for overlapping group membership resulted in 34 cases of species belonging to 2 groups and one case of a species (Amelanchier utahensis) belonging to 3 groups. Environmental affinities of coalition groups also overlap to varying degrees, as depicted in predictive maps accompanying regression tree models (Figs. 25). In the group descriptions that follow, we highlight these cases of compositional and environmental overlap and interpret their ecological significance.

Three coalition groups were linked to mesic conditions at Zion: Streambank, High Plateau, and Crevice Canyon. The Streambank group (Table 2A) contained several species, including Baccharis emoryi, Salix exigua, Tamarix chinensis, and Populus fremontii, with affinities for low-elevation riparian zones at Zion. Regression tree analysis confirmed the riparian setting: affinities were highest on alluvium or lacustrine substrates below 1356 m with low topographic positions (TPI at 50-m radius < -0.51) (Fig. 2A). The Streambank group also had moderately high affinities for low topographic positions on other substrates (likely intermittent streams) and at high elevations (>2321 m), presumably because mesic conditions at Zion's highest elevations are favorable for some riparian species. The High Plateau group, which had highest affinity at elevations above 2243 m (Fig. 2B), shared 3 coalition species (Bromus carinatus, Elymus trachycaulus, and Poa pratensis) with Streambank (Table 2A–B). Coalition species unique to the High Plateau group include Stellaria jamesiana, Lupinus sericeus, Prunus virginiana, Stipa lettermanii, Rosa woodsii, and Juniperus scopulorum. (Table 2B).

The Crevice Canyon group (Table 2C, Fig. 2C) occupied an environmental setting intermediate between Streambank and High Plateau: mid-elevation Navajo Sandstone canyons where topographic shading and intermittent water flow produce cool mesic conditions. Sites with these conditions have extremely low TPI values (i.e., TPI at 50-m radius < -35; Fig. 2C). Outside of these canyons, the Crevice Canyon group had higher affinity for higher elevations (Fig. 2C), exhibiting an affinity pattern very similar to that of the High Plateau group (Fig. 2B). The canopy dominant species Abies concolor and the herbs Thalictrum fendleri and Taraxacum officinale were coalition species of both Crevice Canyon and High Plateau (Table 2B–C). In addition, the Crevice Canyon group shared 4 species (Acer negundo, Elymus canadensis, Equisetum hymenale, and Agrostis exarata) with the Streambank group (Table 2A) and 5 (Heuchera rubescens, Zauschneria latifolia, Holodiscus dumosus, Selaginella underwoodii, Brickellia grandiflora, and Erigeron sionis) with the Slickrock group (Table 2D).

As their names suggest, the Slickrock and Upland Sands groups were differentiated by substrate preferences despite their common affinity for Navajo Sandstone (Fig. 3). While having 4 coalition species in common (the tree dominant Pinus ponderosa plus Arenaria fendleri, Chrysopsis villosa, and Linanthastrum nuttallii) (Table 2D–E), the Slickrock group was characterized by species typical of exposed sandstone outcrops (e.g., Petrophytum caespitosum, Cercocarpus intricatus, Castilleja scabrida, Muhlenbergia thurberi, and Ivesia sabulosa) (Table 2D), whereas Upland Sands species included the shrub dominant Arctostaphylos patula plus several sand-loving species (e.g., Cryptantha cinerea, Abronia fragrans, Artemisia campestris, Penstemon laevis, and Tradescantia occidentalis) (Table 2E). The high affinity of Upland Sands for Navajo Sandstone can be attributed to the group's presence on pockets of sand smaller than the mapping scale of the geologic layer (O'Meara 2006). Plot data support this observation: in plots mapped to Navajo Sandstone, Slickrock affinity was positively correlated with percent rock cover (Spearman's rho = 0.50), while Upland Sands affinity was negatively correlated (Spearman's rho = -0.15). The Upland Sands group also had relatively high affinity for eolian sand deposits large enough to have been mapped by O'Meara (2006), as well as the sandstone-dominated Temple Cap formation. Both Upland Sands and Slickrock had higher affinities for higher elevations (>1753–1782 m) within the Navajo Sandstone zone (Fig. 3). Beyond the Navajo Sandstone, the Slickrock group had its highest affinity value in sites with low TPI (TPI at 100-m radius > -21.24) that corresponded to talus deposits immediately below Navajo Sandstone cliffs (Fig. 3A).

Table 2.

Coalition and affiliate species of Zion National Park based on coalition clustering of 1987–1989 vegetation survey data. Occ. = number of plot occurrences out of 288 total. Affin. = affinity of a coalition species to its coalition group, calculated as the mean value of the phi coefficient of association between the species and others in the group. Affil. = affiliation of an affiliate species with coalition species, calculated as percent co-occurrence. Cvr. = mean cover of affiliate species weighted by co-occurrences with coalition species. Species in boldface are both coalition species and affiliate species for a given group. Underlined species are dominant affiliates (weighted mean cover ≥ 2%).

t02a_26.gif

Continued.

t02b_26.gif

Continued.

t02c_26.gif

Continued.

t02d_26.gif

Continued.

t02e_26.gif

The group with highest affinity for the steep sloping terrain below Navajo Sandstone cliffs was given the name “Rocky Slopes.” Fraxinus anomala, Erigeron utahensis, Poa fendleriana, and Quercus turbinella were prominent members of this group (Table 2F). The first split of the Rocky Slopes regression tree model identified substrate as an important affinity predictor, with lower affinity for Carmel, alluvium, colluvium, eolian, lacustrine and residuum than for other substrates (Fig. 4A). Rocky Slopes affinity values were low in the deep canyon zone (TPI at 100-m radius < -35) (Fig. 4A) in contrast to the Crevice Canyon group (Fig. 2C), but higher in other low TPI areas adjacent to cliffs. The highest Rocky Slopes affinities were associated with the Kayenta formation which is particularly extensive in the Taylor Creek thrust fault area of northwestern Zion (Fig. 4A). The broad zone of relatively high affinity above, below, and around the Kayenta formation can be attributed to the broad distribution of several members of the Rocky Slopes group, including Poa fendleriana, Amelanchier utahensis, and Arabis perennans. These species occupy a wide range of environments at Zion but form a coalition with more narrowly distributed species such as Shepherdia rotundifolia (Table 2F).

The Rocky Slopes group shared 3 coalition species (Opuntia macrorhiza, Senecio multilobatus, and Erysimum asperum) with the Upland Sands group (Table 2E) and 3 others (Pinus monophylla, Juniperus osteosperma, and Gilia inconspicua) with the lower-elevation Arid Lowlands group (Table 2G). Pinus monophylla and J. osteosperma are well-known dominants of Great Basin woodlands (West 1999) and epitomize the Great Basin affinities of many of the Arid Lowlands coalition species. Other coalition species of the Arid Lowlands group, including Coleogyne ramossima, Baileya multiradiata, and Eriastrum eremicum, have geographic affinities for the Mojave Desert and lower-elevation Colorado Plateau rather than the Great Basin (Welsh et al. 2003). Included in this mixture is the common Mojave Desert invasive species Bromus rubens as well as its Great Basin counterpart Bromus tectorum. Gutierrezia spp. (G. sarothrae and G. microcephala), indicators of grazing disturbance history (Ladyman 2004), also had high affinity for the Arid Lowlands group (Table 2G).

The arid character of sites occupied by the Arid Lowlands group was verified by its regression tree model which indicated higher affinity for sites below 1596 m elevation (Fig. 4B). Within this low-elevation zone the Arid Lowlands group had lower affinity for deep canyons (TPI at 100-m radius < -22), substrates associated with riparian areas (alluvium and lacustrine) and substrates concentrated at higher elevations (Kayenta, talus, Navajo and eolian), except where these substrates were highly exposed (exposure > 215). At somewhat higher elevations (1596–1818 m) the Arid Lowlands group likewise had higher affinity for more exposed sites (exposure > 130) (Fig. 4B) where the moisture/temperature regime was presumably similar to lower elevations.

Fig. 2.

Regression tree models and maps for Streambank, High Plateau and Crevice Canyon coalition groups at Zion National Park, Utah. Mean affinity values at terminal nodes are in units of the phi coefficient. Predicted distribution of affinity values are mapped using shading shown at tree tips. TPI (50/100 m) = topographic position index (50- or 100-m radius). See Table 1 for key to substrate codes.

f02_26.jpg

Fig. 3.

Regression tree models and maps for Slickrock and Upland Sands coalition groups at Zion National Park, Utah. Mean affinity values at terminal nodes are in units of the phi coefficient. Predicted distribution of affinity values are mapped using shading shown at tree tips. TPI (100-m) = topographic position index at 100-m radius. See Table 1 for key to substrate codes.

f03_26.jpg

Lowland Flats, another low-elevation species group, had high affinities for sites below 1388 m, especially alluvium and Moenkopi substrates with TPI (50-m radius) > -0.5 (Fig. 4C). In Zion Canyon, these conditions correspond with floodplains and benches above the zone where the Streambank group is concentrated (Fig. 2A). These portions of Zion Canyon have been subject to intensive agriculture in the past and bear the marks of this land-use history. A combination of native species (e.g., Lycium pallidum, Atriplex canescens, Sphaeralcea grossulariifolia, Elymus smithii) and exotics associated with disturbance (Erodium circutarium, Tragopogon dubius, Lactuca serriola) typified the Lowland Flats group (Table 2H). Various combinations of these species also occurred throughout Zion at the lower elevations. Above 1388 m elevation, higher Lowland Flat affinities were associated with flatter areas (slope < 6%) on mostly unconsolidated substrates (Fig. 4C).

Fig. 4.

Regression tree models and maps for Rocky Slopes, Arid Lowlands and Lowland Flats coalition groups at Zion National Park, Utah. Mean affinity values at terminal nodes are in units of the phi coefficient. Predicted distribution of affinity values are mapped using shading shown at tree tips. TPI (50/100 m) = topographic position index (50- or 100-m radius). See Table 1 for key to substrate codes.

f04_26.jpg

Fig. 5.

Regression tree models and maps for Upper and Lower Mesa Top coalition groups at Zion National Park, Utah. Mean affinity values at terminal nodes are in units of the phi coefficient. Predicted distribution of affinity values are mapped using shading shown at tree tips. See Table 1 for key to substrate codes.

f05_26.jpg

Coalition Clustering yielded 2 groups associated with the Carmel formation that caps the Navajo Sandstone on the eastern and northern sides of Zion. We interpreted one group (Upper Mesa Top) as indicating slightly more mesic settings than the other group (Lower Mesa Top). Highest affinity values for both groups were found in a zone defined by the Carmel formation above 4794 m, but outside of this zone the Upper Mesa Top group's affinities are shifted toward higher elevations than Lower Mesa Top's (Fig. 5). The 2 Mesa Top groups share a set of prominent woody species (Pinus edulis, Quercus gambelii, Amelanchier utahensis, and Peraphyllum ramosissimum) as well as 2 herbaceous species (Carex rossii and Swertia radiata), but the affinity order of these species is almost exactly reversed for the 2 groups, i.e., affinity of P. ramosissimum > S. radiata > A. utahensis > C. rossii > P. edulis for Upper Mesa Top (Table 2I), whereas the opposite sequence is true for Lower Mesa Top (Table 2J). Two members of these groups overlap with other groups: Symphoricarpos oreophilus links Upper Mesa Top to the High Plateau group (Table 2B) and Amelanchier utahensis links both Mesa Top groups to the Rocky Slopes group (Table 2F).

Affiliate Species of Coalition Groups

Many but not all coalition species were also affiliate species of their coalition groups (Table 2, bold), meaning that they co-occurred with high frequency. The percentage of coalition species that were also affiliate species ranged from 36% for Crevice Canyon group (Table 2C) to 83% for Streambank (Table 2A). Each group also had 10–16 affiliate species that were not coalition species (Table 2), meaning that they frequently co-occurred but did not share overall distribution/ association patterns with members of the group. The total number of affiliate species, which reflects the mean species density in plots where coalition species were concentrated, ranged from 21 for the Lower Mesa Top group (Table 2J) to 27 for Crevice Canyon (Table 2C).

As a general pattern, the most widely distributed species had the highest affiliation values and were affiliated with the most coalition groups. Poa fendleriana, the most frequently encountered species at 190 occurrences, was an affiliate of all 10 described groups (Table 2). Other widely distributed species with multiple affiliations included Quercus gambelii, Amelanchier utahensis, Bromus tectorum, Arabis perennans, Senecio multilobatus, and Opuntia macrorhiza (Table 2). In each of these cases, the species was affiliated with several additional groups beyond the one(s) where it belonged as a coalition species. Some species, including Artemisia ludoviciana, Artemisia tridentata, Elymus elymoides, Eriogonum racemosum, Machaeranthera canescens, Phacelia heterophylla, Sporobolus cryptandrus, Stipa comata, Stipa hymenoides, and Yucca angustissima, had the distinction of being affiliate species without belonging to any of the coalition groups.

Weighted mean cover ranged from a fraction of a percent for most affiliate species (Table 2) to 23% for Quercus gambelii in the High Plateau group (Table 2B), followed by Bromus tectorum in the Lowland Flats group at 18% (Table 2H). A total of 25 affiliate species met our criteria for dominance (weighted mean cover > 2%) in relation to one or more groups (Table 2, underlined entries). Ten of these were dominant in only one group, while 5 species were dominant in 5 or more groups (Quercus gambelii, Poa fendleriana, Amelanchier utahensis, Arctostaphylos patula, and Juniperus osteosperma) (Table 2).

Relations to Previously Described Vegetation Types

To a certain extent, our 10 coalition groups resembled the 10 vegetation types previously described by Harper et al. (2001). Some coalition groups were compositionally very similar to modal species of Harper et al.'s (2001) vegetation types (Fig. 6, Appendix 4). For example, 7 of the 9 coalition species of the Lowland Flats group were modal species of the Abandoned Fields type (Appendix 4). The Arid Lowlands, Slickrock, High Plateau, and Streambank groups corresponded relatively closely with the Blackbrush, Rock Crevice, Douglas Fir-White Fir, and Riparian vegetation types, respectively (Fig. 6). However, the correspondence was not exact, and some groups had high correspondence with multiple vegetation types, namely, Upland Sands (with Ponderosa Pine, Mountain Brush, and Rock Crevice), Crevice Canyon (with Riparian, Douglas/White Fir, and Hanging Gardens), and the Mesa Top groups (with Douglas/ White Fir, Ponderosa Pine and Juniper-Pinyon) (Fig. 6).

When clustered according to coalition group affinities, NVC associations were arranged into groups that reflected environmental affinities but did not have uniform physiognomy (Fig. 7, Appendix 5). The 11 clusters highlighted in Fig. 7 were named for their predominant environmental (i.e., coalition group) affinities and the mixture of physiognomic classes they contained. These clusters were nested in 2 major groups (Xeric and Mesic Zion Vegetation) and 4 major subgroups (Fig. 7). Some associations emerged as outliers whose environmental affinities seemed poorly matched with their neighboring clusters (Needle-and-Thread Great Basin Herbaceous, Coyote Willow/Barren Shrubland, and Green Rabbitbrush/Kentucky Bluegrass Semi-natural Shrubland) and these associations were excluded from the 10-cluster classification (Fig. 7, Appendix 5). These outliers may be associations sampled by the USGS survey that were missed or poorly captured by the Harper survey (see discussion below).

Fig. 6.

Correspondence between vegetation units of Harper et al. (2001) (rows) and coalition species groups of the current study (columns) at Zion National Park, Utah. Values shown are mean affinities (in units of the phi coefficient) of modal species of each vegetation unit in relation to each coalition group (see also Appendix 4).

f06_26.jpg

Discussion

Our analysis of plant community variation and community-environment relationships at Zion revealed several patterns that were overlooked or obscured in previous community characterizations. Important patterns revealed by our analysis include (A) floristic similarities between communities of low-elevation riparian zones (Streambank) and high elevations (High Plateau), linked by mid-elevation communities of narrow sandstone canyons (Crevice Canyon), (B) contrasting communities of sandstone bedrock (Slickrock) and sand deposits (Upland Sands) on the Navajo Sandstone Formation, (C) subtle community differentiation related to elevation within the Carmel Formation (Upper and Lower Mesa Top), and (D) continuity of communities containing Mojave Desert and Great Basin species within a xeric low-elevation zone (Arid Lowlands).

Our unique results can be attributed to the way in which we defined and characterized communities as coalition groups rather than conventional vegetation types. Although we used the (nearly) identical dataset that Harper (1993) and Harper et al. (2001) used to characterize vegetation types at Zion, our characterization was based on coalition clustering of species association patterns as opposed to site classification based on dominant species and vegetation structure. Our 10 coalition groups are comparable in scale to the 10 vegetation types described by Harper et al. (2001) but lack discrete spatial boundaries and reflect multispecies distribution patterns independent of species commonness or abundance—although we were also able to identify the most common and abundant species associated with the coalition groups (affiliate and dominant-affiliate species) in secondary analysis steps. Our coalition groups likewise differ from the NVC associations described by Cogan et al. (2004), although the latter represent a finer scale of characterization that we were able to cross-walk to our broader coalition group scheme.

Species Groups and the Individualistic Concept

Coalition groups can be considered species groups according to most definitions of the term as used by plant community ecologists. They are (phyto-) sociological species groups comprised of species with similar association patterns (Doing 1969, Holzner et al. 1978) and also qualify as statistical species groups derived through multivariate analysis (Bruelheide and Chytrý 2000, McCune and Grace 2002). Although our coalition groups are not ecological species groups as defined by European schools of ecosystem classification (Mueller-Dombois and Ellenberg 1974, Kashian et al. 2003) they are clearly “ecological” in the sense of being correlated with environmental variation, such that species within groups can be interpreted as having similar environmental affinities. Given the strength and interpretive value of these environmental correlations (Fig. 25), we opted to use environmentally descriptive names instead of following the widespread convention of naming species groups based plant taxa (e.g. Holzner et al. 1978, Kashian et al. 2003).

Fig. 7.

Cluster dendrogram of National Vegetation Classification (NVC) associations of Zion National Park, Utah (Cogan et al. 2004), based on compositional affinities to 10 coalition groups presented in this paper. Names are assigned to clusters at 3 levels. A = Xeric Zion Vegetation, B = Mesic Zion Vegetation. A1 = Mesa Top/Upland Sands/Slickrock/ Rocky Slopes, A2 = Lowlands/Rocky Slopes/Upland Sands, B1 = High Plateau/Crevice Canyon, B2 = Streambank/ High Plateau.

f07_26.jpg

We propose that the term individualistic species group also applies to coalition groups because their nondiscrete spatial patterns, overlapping composition and fuzzy attributes reflect notions of species individualism and community continuity espoused by Gleason (1926, 1939). Gleason is well known for having pointed out difficulties of discrete community classification stemming from individualistic species distributions, having noted that “every species of plant is a law unto itself, the distribution of which in space depends upon its individual peculiarities” (Gleason 1926, p. 26). But he also recognized that some species are sufficiently aligned in their distributions to warrant group recognition:

The [individualistic] concept is by no means opposed to the recognition of the synusia, or union, defining it as a group of plants whose physiological demands are so similar that they are regularly selected by the same environment and consequently regularly live together (Gleason 1939, p. 108).

Hence an individualistic approach to community characterization permits the dual objective of identifying spatially-associated species possessing similar ecological requirements, while also recognizing and quantifying species differences.

Unlike methods such as TWINSPAN or COCKTAIL which characterize species groups in relation to discrete site groups (Bruelheide and Chytrý 2000), the coalition clustering approach extracts groups lacking discrete boundaries (beyond the overall boundaries of the chosen study area). Any clustering method carried out in R-mode (clustering species but not sites) (McCune and Grace 2002) would likewise yield spatially nondiscrete groups. Coalition groups have an added distinction of being compositionally nondiscrete (i.e., overlapping in their species membership). The only sense in which coalition groups are “discrete” is that any one group is composed of a distinct set of member species, but even this form of discreteness is diluted by our use of affinity values to characterize degrees of membership. We note conceptual and mathematical similarities between our affinity calculations and fuzzy set theory: species affinity is essentially identical to fuzzy membership, group-averaged affinity to relative cardinality, and plot-averaged affinity to fuzzy species richness as described by Olivero et al. (2011).

Although coalition clustering does not require or produce spatially discrete community units, it is influenced by discontinuities in species distribution patterns. Multiple species can have coinciding distributions marked by quasi-discrete distribution boundaries due to abrupt environmental transitions, even if these species would otherwise show unaligned individualistic environmental responses (Austin and Smith 1989). Relatively abrupt transitions in hydrology, substrate and elevation are present at Zion and are reflected in our coalition groups. The Streambank and Crevice Canyon groups, for example, were concentrated in mesic environments immediately adjacent (at the scale of 10 × 10-m pixels) to xeric environments dominated by the Lowland Flats, Arid Lowlands or Slickrock groups (Fig. 24). The Navajo Sandstone cliff zone forms a natural break in the Zion landscape that was detectable as a dividing line in regression tree models, e.g., the first split of the Arid Lowlands model at 1596 m elevation lies at approximately the base of the Navajo Sandstone cliffs in the main section of the park (Fig. 4B), while the first split of the Mesa Top groups (1791 m) lies near the top of this same set of cliffs (Fig. 5).

In other instances, our coalition clustering results indicate gradual rather than abrupt changes in community composition and environmental conditions. The 2 Mesa Top groups overlapped extensively in composition (Table 2I–J), spatial distribution and environmental affinities (Fig. 5), leading us to conclude that they represent subtle distinctions along a continuum. The Arid Lowlands group also appears to represent a continuum that was not subdivided at our selected threshold level. We note, however, that at a threshold of phi = 0.2 (instead of 0.15), 2 Arid Lowlands subgroups emerged, one with Great Basin/higher-elevation affinities and the other with Mojave Desert/lower-elevation affinities (Fig. 8). Like the 2 Mesa Top groups, these 2 Arid Lowlands subgroups had many species in common, and species with higher affinity for one group tended to have lower affinity for the other (with the exception of Bromus rubens, which was the highest-affinity species for both groups) (Fig. 8).

Fig. 8.

Two Arid Lowlands coalition groups at Zion National Park, Utah, obtained through coalition clustering in RCLUS at a threshold affinity value of phi = 0.2. Species are in descending order of group affinity and lines connect species shared by both groups. The left group has Great Basin (higher elevation) affinities and the right group has Mojave Desert (lower elevation) affinities.

f08_26.jpg

Coalition Clustering Methodology

We emphasize that our coalition groups portray one of many possible coalition clustering solutions, as illustrated by the alternative Arid Lowlands groupings discussed above. Through experimentation with different association coefficients and threshold levels, we arrived at a solution at the desired scale that agreed with our field experience at Zion. Thus our approach differs from clustering approaches that seek to identify an optimal set of clusters with minimal within-group and maximal between-group variation (Kaufman and Rousseeuw 1990, Aho et al. 2008). Cluster evaluation techniques (Aho et al. 2008) could conceivably be applied to find an optimal threshold level resulting in a set of coalition groups with minimal species overlap. However, given our assertion that overlapping composition contributes to the individualistic character of coalition groups, we argue that flexibility in selecting a threshold level enhances the usefulness of coalition clustering as an exploratory tool.

By applying coalition clustering to our data using the phi coefficient of association we were able to generate clusters containing mixtures of common and rarer species. In many cases, species belonging to the same coalition group did not actually co-occur in any plots, but the negative pairwise associations of these species were overridden by their shared positive associations with other species belonging to the group. Species occurring in multiple plots provided a glue that held species occurring in fewer plots together in the same group. Our use of the phi coefficient contributed to this result because phi incorporates joint absence information and is less sensitive to differences in species frequency than presence-only coefficients such as the Jaccard index (Jackson et al. 1989, Sanderson et al. 2006). Species clustering based on presence-only coefficients tends to group species according to their frequency unless additional standardization steps are followed (Jackson et al. 1989, McCune and Grace 2002). The phi coefficient has the draw-back of being undefined when 2 species have perfectly overlapping distributions or when one species' distribution is nested within the other, but we did not encounter either of these situations in our data set. Phi is always negative when 2 species never occur together, but can also be negative when 2 species occur together with few joint absences (e.g., a pair of ubiquitous species). In our analysis, co-occurring ubiquitous species would justifiably have been excluded from coalition groups, but this situation was minimally relevant because the most ubiquitous species (Poa fendleriana) occurred in 66% of plots but no other species exceeded 50% occurrence. Our results would therefore likely have been similar had we used a coefficient that increases monotonically with increasing joint presence, such as that of Baroni-Urbani and Buser (1976).

In contrast to Harper et al. (2001), whose community characterization began with dominant species followed by prevalent and modal species sensu Curtis (1959), our characterization began with a conceptual equivalent of modal species (coalition species) followed by equivalents of prevalent and dominant species (affiliate and dominant-affiliate species, respectively). Modal species are conceptually similar to what other authors have termed faithful, diagnostic, differential or character species, all of which denote concentration of species occurrences in one or more communities relative to others (Braun-Blanquet 1932, Aho et al. 2008, Jennings et al. 2009). Coalition species likewise express this concept of community fidelity, but in relation to a species group rather than a set of spatially defined communities. Unlike strict definitions of faithful species, coalition species can be “faithful” to more than one group. Prevalent species as defined by Curtis (1959) are essentially synonymous with what others have termed constant species (Braun-Blanquet 1932, Jennings et al. 2009), with the added criterion of using species density (average number of species per plot) to determine the number of prevalent species for a community (Curtis 1959, Harper et al. 2001). Our procedures for identifying affiliate and dominant-affiliate species followed the logic of the prevalent species concept in a way that, to our knowledge, is novel and could be applied to any set of species groups (coalition groups or otherwise). Although our measures of affiliation (percent co-occurrence) and weighted mean cover are not directly comparable with constancy and mean cover calculated for discrete communities, their values can be readily used for comparisons within and across species groups.

Perspectives on Vegetation Classification

By characterizing dominant species in relation to species groups, we gained a different perspective of dominant species than previous authors who focused on them a priori for vegetation classification. Our analysis identified as dominant-affiliates many of the same species that defined vegetation types of Harper et al. (2001), but some of these species were spread among multiple coalition groups. Pinus ponderosa, which defined the Ponderosa Pine vegetation type of Harper et al. (2001), emerged as a dominant-affiliate species of 4 coalition groups (Upper and Lower Mesa Top, Upland Sands and Slickrock); likewise Abies concolor of the White Fir-Douglas Fir vegetation type was a dominant-affiliate of both High Plateau and Crevice Canyon; and dominants of Juniper-Pinyon (Pinus edulis, Pinus monophylla, Juniperus osteosperma) and Mountain Brush (Quercus gambelii, Amelanchier utahensis) vegetation types were dominant-affiliates of 2 or more coalition groups each (Table 2). These mismatches highlight the discordance between abundance patterns of single species and presence-absence patterns of broader species sets. Because dominant species and growth forms such as ponderosa pine, white fir, juniper-pinyon and mountain brush are generalists that occupy multiple environmental settings at Zion, their use as classification criteria by Harper et al. (2001) led to environmentally ambiguous vegetation types. Some of these ambiguities were clarified in a subsequent paper (Harper et al. 2003) that split the Juniper-Pinyon vegetation type into 3 subtypes: a higher elevation subtype containing Pinus edulis (like the Mesa Top groups), a mid-elevation subtype with Pinus monophylla (like Rocky Slopes) and a low-elevation sub-type with Juniperus osteosperma but no Pinus (like Arid Lowlands). Cogan et al. (2004) differentiated pinyon-juniper vegetation in a similar way at the NVC alliance level.

Our hierarchical rearrangement and reclassification of NVC associations on the basis of coalition group affinities (Fig. 7) differs substantially from the way these associations have been treated under the NVC system. According to the NVC classification hierarchy followed by Cogan et al. (2004), associations were nested within 6 major higher levels (alliance, formation, formation subgroup, formation group, formation subclass, formation class), each defined by specific characteristics including dominant species, substrate, hydrologic regime, anthropogenic disturbance, leaf phenology, leaf structure, plant habit, and life form (Grossman et al. 1998). The NVC hierarchy was recently revised through the addition of new levels between alliance and formation (group, macrogroup and division) and changes to classification criteria designed to make groupings more ecologically meaningful (FGDC 2008, Faber-Langendoen et al. 2009, NatureServe 2013, USNVC 2013). Upper levels of the new NVC hierarchy reflect general growth form differences indicative of broad-scale environments or “global macroclimatic factors”, while successively lower levels reflect finer physiognomic and floristic distinctions indicative of finer-scale climate, geology, substrate, hydrology, moisture/nutrient factors, and disturbance regimes (Faber-Langendoen et al. 2009). The NVC hierarchy might thus be expected to portray environmental affinities at Zion in a manner similar to our coalition groups; however, our coalition group-based hierarchy was poorly aligned with both the older and newer NVC classification (Appendixes 6, 7). Some NVC groupings were concentrated within certain clusters of our hierarchy—e.g., Singleleaf Pinyon — (Utah Juniper) Woodland Alliance in the Rocky Slopes Woodland/Shrub cluster—but others were spread across multiple clusters, as illustrated by the mixture of associations with multiple physiognomies (forest, woodland, shrubland, etc.) within groupings of our 3-tiered classification (Fig. 7). These differences can be attributed to the emphasis placed on physiognomy and dominant species at all levels of the NVC hierarchy, in contrast to our strictly floristic, presence/absence-based approach to classification.

Comparisons can also be made between our coalition group-based hierarchy of NVC associations (Fig. 7) and the Terrestrial Ecological Systems Classification (TESC) that has been applied at Zion (Comer et al. 2003, Comer and Schulz 2007, NatureServe 2013). The TESC was developed to complement the NVC by grouping associations or alliances that “tend to naturally co-occur with similar environmental settings, ecological dynamics, and/or environmental gradients” (Comer and Schulz 2007, p. 326), forming “ecological systems” that can more easily be mapped and related to environmental variation at landscape to regional scales (Comer et al. 2003, Comer and Schulz 2007). An ecological system can include multiple physiognomic types that occur in a similar environment, such as successional stages ranging from herbaceous vegetation to shrubland to woodland within a riparian flood-plain (Comer and Schulz 2007). Given these criteria, the TESC approach could potentially yield ecological groupings similar to those we identified at Zion. Upon cross-walking our coalition-group based hierarchy with ecological systems (Appendix 8), we did indeed find general similarities, such as alignment of the North American Warm Desert Lower Montane Riparian Woodland and Shrubland system with the Streambank group. However, the concordance between the 2 classifications is far from exact, and we note that the ecological systems contain less physiognomic variability than clusters in our hierarchy (Fig. 7, Appendix 8). Like the NVC hierarchy, the TESC as currently applied to vegetation types at Zion appears to emphasize physiognomic/structural uniformity at the expense of environmental similarity. These observations are not intended as criticisms of the NVC and TESC approaches but rather highlight their strengths and weaknesses in relation to our approach. No single community characterization approach is likely to serve all purposes; hence our characterization is finely tuned to represent conditions at Zion, but unlike the NVC and TESC it cannot be straightforwardly transferred to other landscapes with different combinations of species and environments.

Sampling Strategy and Modeling Issues

Our ability to extrapolate coalition group affinities beyond park boundaries or interpolate them beyond plot locations within the park was limited by sampling and modeling constraints (cf. Elith and Leathwick 2009). Environmental conditions such as higher or lower elevations in the surrounding region beyond the park were not sampled, and important community-environment variation within the park may have also been missed. Furthermore, the environmental variables we used posed limitations on predictive modeling; for example, the lacustrine substrate class noticeably overpredicted Streambank (Fig. 2A) and underpredicted Rocky Slopes (Fig. 4A) affinities in the Coalpits Wash area (south-western side of park), where—unlike in Zion Canyon—lacustrine deposits are not associated with riparian areas.

Ideally, a different set of more natural landscape boundaries could have been used to delimit our study area; but given that national park boundaries were used, the mostly systematic sampling strategy followed by the Harper survey was advantageous for capturing community and environmental variation in a relatively unbiased way. Dismissing the unlikely possibility that the approximately 1.6-km distance between Harper survey plots coincides with periodic environmental variation at Zion, the more-or-less evenly spaced plot arrangement (Fig. 1A) can be viewed as an area-proportional sample of Zion's environments.

Environmental conditions that were more frequent in the landscape were captured with greater frequency by the Harper survey, in contrast to the USGS survey, which sought to capture the range of environmental variation without regard for frequency through a gradsect sampling strategy (Cogan et al. 2004). The effects of these different sampling strategies can be illustrated by histograms of plot-environment combinations compared proportionally with background environmental variation at Zion (Figs. 9, 10) (Calenge 2006). The Harper data set had an overrepresentation of low elevations, low slopes, moderate exposures (Fig. 9A), and alluvial substrates (Fig. 10A) that can be attributed at least in part to the nonsystematic riparian and old field plots included in the survey. On the other hand, excessively steep or bare plot locations were avoided or adjusted by Harper survey workers, leading to an under-representation of high slopes and Navajo Sandstone substrates (Fig. 9A, 10A). However, these same patterns of environmental over- and underrepresentation were present to an even greater degree in the USGS data set (Fig. 9B, 10B). Consequently the Harper data set has a larger, comparatively more representative sample of common environments (such as the Navajo Sandstone), while the USGS data set has better representation of rarer, more unique environments and communities. The Harper survey captured “transitional” areas with heterogeneous vegetation that would have been avoided by the USGS survey, while USGS likely captured spatially restricted vegetation types that were missed by Harper. These differences between survey data sets, as well as differences due to year of sampling, plot size, and data collection techniques, must be kept in mind when comparing our results with results derived from the USGS survey (i.e., Fig. 7, Appendixes 58).

Fig. 9.

Histograms showing background frequencies of topographic variables at Zion National Park, Utah (solid border bars), compared proportionally with plot representation of the same variables (shaded bars) by 2 different vegetation surveys (left and right). Based on 50-m-resolution generalizations of 10-m-resolution grids covering the park. Derived using function “histniche” of the adehabitat package (Calenge 2006) in R 2.8.0 (R Development Core Team 2008).

f09_26.jpg

Fig. 10.

Histograms showing background frequencies of geologic substrate variables at Zion National Park, Utah (solid border bars), compared proportionally with plot representation of the same variables (shaded bars) by 2 different vegetation surveys (left and right). See Table 1 for key to substrate codes. Based on 50-m-resolution generalizations of 10-m-resolution grids covering the park. Derived using function “histniche” of the adehabitat package (Calenge 2006) in R 2.8.0 (R Development Core Team 2008).

f10_26.jpg

Temporal Scale, Fire, and Weather

Our community characterization is best viewed as a snapshot that captures community patterns in relation to environment conditions at a specific area within a specific timeframe. Conditions at Zion in the late 1980s when the Harper survey was conducted included the effects of previous decades of agricultural disturbance, livestock grazing, and fire suppression (Woodbury 1933, Madany and West 1983) but not the effects of more recent wildfires within the park, such as the 4256-ha Kolob Fire of 2006 (Brisbin et al. 2013). Most Harper Survey plots were classified by field workers as late-seral or climax with respect to time since fire, although 2 plots on the Horse Pasture Plateau burned in 1988 and were sampled later the same year. These recently-burned plots, despite their reduced cover relative to surrounding woodlands, readily blended into the Mesa Top coalition groups in our analysis, suggesting that the fire had a minimal effect on species composition as measured by presence-absence, at least initially. This supports the view that species presence-absence patterns are less sensitive to short-term environmental fluctuations and disturbances than abundance patterns, making presence-absence a more stable measure indicative of longer-term ecological processes than abundance (Allen and Starr 1982, Wilson 2012). However, the stability of presence-absence patterns can be expected to vary due to variation in plant longevity and sensitivity to disturbance. Other studies in plant communities similar to those at Zion have reported shifts in species presence-absence due to fire, often characterized by the loss of fire-intolerant shrubs and trees combined with an influx of annual species (Koniak 1985, Paysen et al. 2000, Ott et al. 2001, Huisinga et al. 2005, Engel and Abella 2011). Some species that might persist initially after fire could later die out, if for example invasive Bromus species become dominant and cause recurring fires (Bagchi et al. 2013). The recent fires at Zion could have therefore altered some of the community patterns we have presented here, though this does not diminish the value of our characterization as a baseline reference.

Long-term fire history has been implicated as a factor determining current distributions of pinyon-juniper woodland versus mountain brush in areas that could support either vegetation type, with mountain brush occupying sites where the fire-return interval has been shorter (Floyd et al. 2000). Although our coalition clustering analysis did not reveal discrete differences between these 2 vegetation types, the 2 Mesa Top coalition groups were partially aligned with them. The fire-sensitive species Pinus edulis and Cercocarpus montanus had higher affinity for the Lower Mesa Top group than fire-tolerant mountain brush species Quercus gambelii, Amelanchier utahensis, Balsamorhiza sagittata, and Symphoricarpos oreophilus (Floyd et al. 2000, USDA-FS 2013), but in the Upper Mesa Top group the fire-tolerant species had higher affinity (Table 2I–J). These 2 Mesa Top groups might thus reflect differences in long-term fire history that have left a subtle imprint on species presence-absence patterns.

The patterns we detected at Zion also reflect weather conditions and fluctuations during the years leading up to the time of data collection in 1987–1989. Weather data from the park show periods of above-average precipitation during the years 1978–1983, near-average from 1984–1988, and below-average in 1989 (Witwicki 2013). Annual species sensitive to yearly precipitation fluctuations may have been sparser in 1989; however, most sampling of low elevations where most of these annual species occur took place in 1987–1988.

Since the late 1980s, Zion has experienced large fluctuations in annual precipitation and a gradual warming trend (Witwicki 2013) that is projected to continue (Polley et al. 2013). In light of evidence from elsewhere in south-western North America (Kelly and Goulden 2008, Fellows and Goulden 2012, Brusca et al. 2013), upward elevational shifts in plant species distributions could have occurred at Zion during this period. Relative to other areas with less abrupt elevation gradients, plant species at Zion might be expected to easily disperse to higher elevations, although species with affinities for certain horizontally-layered geologic substrates might find their preferred substrate unavailable higher up. The degree to which coalition groups will remain associated through continued climate change is unknown, but from a probabilistic perspective, we suggest 2 general rules: (1) species will likely become increasingly disassociated with the passage of time as their individualistic responses to climate change emerge, yet (2) species currently more strongly associated will likely remain associated for longer than more weakly associated species. We suggest that species association patterns such as those we have quantified for Zion have predictive value in both space and time.

Acknowledgments

We thank the many individuals mentioned by Harper et al. (1992) and Cogan et al. (2004) who were involved in different stages of their respective vegetation surveys. The original Harper survey was partially funded by the U.S. National Park Service, and the work presented here was partially funded by the U.S. Forest Service, Rocky Mountain Research Station. Additional support came from the University of North Carolina at Chapel Hill through use of their facilities during the lead author's tenure as a graduate student and visiting scholar. Robert Peet, Peter While, Todd Vision, Aaron Moody, and Paul Manos provided input on early versions of the manuscript.

Literature Cited

1.

K. Aho , D.W. Roberts , and T. Weaver . 2008. Using geometric and non-geometric internal evaluators to compare eight vegetation classification methods. Journal of Vegetation Science 19:549–562. Google Scholar

2.

T.F.H. Allen , and T.B. Starr . 1982. Hierarchy: perspectives for ecological complexity. University of Chicago Press, Chicago, IL. Google Scholar

3.

M.P. Austin , and T.M. Smith . 1989. A new model for the continuum concept. Vegetatio 83:35–47. Google Scholar

4.

S. Bagchi , D.D. Briske , B.T. Bestelmeyer , and X.B. Wu . 2013. Assessing resilience and state-transition models with historical records of cheatgrass (Bromus tectorum) invasion in North American sagebrush-steppe. Journal of Applied Ecology 50:1131–1141. Google Scholar

5.

C. Baroni-Urbani , and M.W. Buser . 1976. Similarity of binary data. Systematic Zoology 25:251–259. Google Scholar

6.

J. Belnap , R. Reynolds , M. Reheis , and S.L. Phillips . 2000. What makes the desert bloom? Contribution of dust and crusts to soil fertility on the Colorado Plateau. Pages 147–153 in E.D. McArthur and D.J. Fairbanks , compilers, Shrubland ecosystem genetics and biodiversity: proceedings. Proc. RMRS-P-21, USDA Forest Service, Rocky Mountain Research Station, Ogden, UT. Google Scholar

7.

H.L. Beyer 2004. Hawth's Analysis Tools for ArcGIS.  http://www.spatialecology.com/htools  Google Scholar

8.

R.F. Biek , G.C. Willis , M.D. Hylland , H.H. Doelling , D.A. Sprinkel , T.C. Chidsey Jr ., and P.B. Anderson . 2003. Geology of Zion National Park, Utah. Pages 107–137 in Anonymous, editor. Geology of Utah's parks and monuments. Utah Geological Association and Bryce Canyon Natural History Association. Google Scholar

9.

J. Braun-Blanquet 1932. Plant sociology: the study of plant communities [English translation]. McGraw-Hill, New York, NY. Google Scholar

10.

L. Breiman 1993. Classification and regression trees. Chapman and Hall, New York, NY. Google Scholar

11.

H. Brisbin , A. Thode , M. Brooks , and K. Weber . 2013. Soil seed bank responses to postfire herbicide and native seeding treatments designed to control Bromus tectorum in a pinyon-juniper woodland at Zion National Park, USA. Invasive Plant Science and Management 6:118–129. Google Scholar

12.

H. Bruelheide , and M. Chytrý . 2000. Towards unification of national vegetation classifications: a comparison of two methods for analysis of large data sets. Journal of Vegetation Science 11:295–306. Google Scholar

13.

R.C. Brusca , J.F. Wiens , W.M. Meyer , J. Eble , K. Franklin , J.T. Overpeck , and W. Moore . 2013. Dramatic response to climate change in the Southwest: Robert Whittaker's 1963 Arizona mountain plant transect revisited. Ecology and Evolution 3:3307–3319. Google Scholar

14.

[BLM] Bureau of Land Management. 2006. Geographic Coordinate Data Base,  http://www.blm.gov/wo/st/en/prog/more/gcdb.html  Google Scholar

15.

C. Calenge 2006. The package “adehabitat” for the R software: a tool for the analysis of space and habitat use by animals. Ecological Modelling 197:516–519. Google Scholar

16.

D. Cogan , M. Reid , K. Schulz , and M. Pucherelli . 2004. Zion National Park, Utah: 1999–2003 vegetation mapping program, final report. Report Technical Memorandum 8260-03-01, Remote Sensing and GIS Group, Technical Service Center, Bureau of Reclamation, Denver, CO.  http://www.usgs.gov/core_science_systems/csas/vip/parks/zion.html  Google Scholar

17.

P. Comer , D. Faber-Langendoen , R. Evans , S. Gawler , C. Josse , G. Kittel , S. Menard , M. Pyne , M. Reid , K. Schulz , et al. 2003. Ecological systems of the United States: a working classification of U.S. terrestrial systems. NatureServe, Arlington, VA.  http://www.natureserve.org/library/usEcologicalsystems.pdf  Google Scholar

18.

P.J. Comer , and K.A. Schulz . 2007. Standardized ecological classification for mesoscale mapping in the southwestern United States. Rangeland Ecology and Management 60:324–335. Google Scholar

19.

M.J. Crawley 2007. The R book. John Wiley & Sons Ltd., West Sussex, England. Google Scholar

20.

J.T. Curtis 1959. The vegetation of Wisconsin: an ordination of plant communities. University of Wisconsin Press, Madison, WI. Google Scholar

21.

H. Doing 1969. Sociological species groups. Acta Botanica Neerlandica 18:398–399. Google Scholar

22.

J. Elith , and J.R. Leathwick . 2009. Species distribution models: ecological explanation and prediction across space and time. Annual Review of Ecology Evolution and Systematics 40:677–697. Google Scholar

23.

E.C. Engel , and S.R. Abella . 2011. Vegetation recovery in a desert landscape after wildfires: influences of community type, time since fire and contingency effects. Journal of Applied Ecology 48:1401–1410. Google Scholar

24.

[ESRI] Environmental Systems Research Institute. 2003, 2004. ArcGIS. Versions 8 and 9.  http://www.esri.com/software/arcgis/index.html  Google Scholar

25.

D. Faber-Langendoen , D.L. Tart , and R.H. Crawford . 2009. Contours of the revised U.S. National Vegetation Classification Standard. Bulletin of the Ecological Society of America 90:87–93. Google Scholar

26.

[FGDC] Federal Geographic Data Committee. 2008. National Vegetation Classification Standard, version 2.  http://www.fgdc.gov/standards/projects/FGDC-standards-projects/vegetation/  Google Scholar

27.

A.W. Fellows , and M.L. Goulden . 2012. Rapid vegetation redistribution in southern California during the early 2000s drought. Journal of Geophysical Research-Biogeosciences 117:G03025. Google Scholar

28.

S. Ferrier , and A. Guisan . 2006. Spatial modelling of biodiversity at the community level. Journal of Applied Ecology 43:393–404. Google Scholar

29.

W. Fertig , and J. Alexander . 2009. Annotated checklist of vascular flora: Zion National Park. Natural Resource Technical Report NPS/NCPN/NRTR-2009/157, USDI National Park Service, Natural Resource Program Center, Fort Collins, CO. Google Scholar

30.

W. Fertig , S. Topp , M. Moran , T. Hildebrand , J. Ott , and D. Zobell . 2012. Vascular plant species discoveries in the Northern Colorado Plateau Network: update for 2008–2011. Natural Resource Technical Report NPS/NCPN/NRTR—2012/582. U.S. National Park Service, Fort Collins, CO. Google Scholar

31.

M.L. Floyd , W.H. Romme , and D.D. Hanna . 2000. Fire history and vegetation pattern in Mesa Verde National Park, Colorado, USA. Ecological Applications 10:1666–1680. Google Scholar

32.

J.F. Fowler , N.L. Stanton , and R.L. Hartman . 2007. Distribution of hanging garden vegetation associations on the Colorado Plateau, USA. Journal of the Botanical Research Institute of Texas 1:585–607. Google Scholar

33.

A.N. Gillison , and K.R.W. Brewer . 1985. The use of gradient directed transects or gradsects in natural-resource surveys. Journal of Environmental Management 20:103–127. Google Scholar

34.

H.A. Gleason 1926. The individualistic concept of the plant association. Bulletin of the Torrey Botanical Club 53:7–26. Google Scholar

35.

H.A. Gleason . 1939. The individualistic concept of the plant association. American Midland Naturalist 21:92–110. Google Scholar

36.

D.H. Grossman , D. Faber-Langendoen , A.S. Weakley , M. Anderson , P. Bourgeron , R. Crawford , K. Goodin , S. Landaal , K. Metzler , K. Patterson , et al. 1998. International classification of ecological communities: terrestrial vegetation of the United States. Volume 1, The National Vegetation Classification System: development, status, and applications. The Nature Conservancy, Arlington, VA.  http://www.natureserve.org/library/vol1.pdf  Google Scholar

37.

K.T. Harper 1993. Zion National Park vegetation: summary report. Zion National Park, Springdale, UT. Google Scholar

38.

K.T. Harper , S.C. Sanderson , and E.D. McArthur . 1992. Riparian ecology at Zion National Park. Pages 32–42 in W.P. Clary , E.D. McArthur , D. Bedunah , and C.L. Wambolt , compilers. Proceedings—symposium on ecology and management of riparian shrub communities. General Technical Report INT-289, USDA Forest Service, Intermountain Research Station, Ogden, UT. Google Scholar

39.

K.T. Harper , S.C. Sanderson , and E.D. McArthur . 2001. Quantifying plant diversity at Zion National Park. Pages 318–324 in E.D. McArthur and D.J. Fairbanks , compilers, Shrubland ecosystem genetics and biodiversity: proceedings. Proc. RMRS-P-21, USDA Forest Service, Rocky Mountain Research Station, Ogden, UT. Google Scholar

40.

K.T. Harper , S.C. Sanderson , and E.D. McArthur . 2003. Piny on-juniper woodlands in Zion National Park, Utah. Western North American Naturalist 63: 189–202. Google Scholar

41.

W. Holzner , M.J.A. Werger , and G.A. Ellenbroek . 1978. Automatic classification of phytosociological data on the basis of species groups. Vegetatio 38:157–164. Google Scholar

42.

K.D. Huisinga , D.C. Laughlin , P.Z. Fule , J.D. Springer , and C.M. McGlone . 2005. Effects of an intense prescribed fire on understory vegetation in a mixed conifer forest. Journal of the Torrey Botanical Society 132:590–601. Google Scholar

43.

Insightful Corp. 2005. S-PLUS [software]. Google Scholar

44.

D.A. Jackson , K.M. Somers , and H.H. Harvey . 1989. Similarity coefficients—measures of co-occurrence and association or simply measures of occurrence? American Naturalist 133:436–453. Google Scholar

45.

M.D. Jennings , D. Faber-Langendoen , O.L. Loucks , R.K. Peet , and D. Roberts . 2009. Standards for associations and alliances of the U.S. National Vegetation Classification. Ecological Monographs 79:173–199. Google Scholar

46.

J.T. Kartesz 1999. A synonymized checklist and atlas with biological attributes for the vascular flora of the United States, Canada, and Greenland. First edition. In : J.T. Kartesz and C.A. Meacham , editors, Synthesis of the North American flora, Version 1.0. North Carolina Botanical Garden, Chapel Hill, NC. Google Scholar

47.

D.M. Kashian , B.V. Barnes , and W.S. Walker . 2003. Ecological species groups of landform-level ecosystems dominated by jack pine in northern Lower Michigan, USA. Plant Ecology 166:75–91. Google Scholar

48.

L. Kaufman , and P.J. Rousseeuw . 1990. Finding groups in data, an introduction to cluster analysis. John Wiley & Sons, New York, NY. Google Scholar

49.

A.E. Kelly , and M.L. Goulden . 2008. Rapid shifts in plant distribution with recent climate change. Proceedings of the National Academy of Sciences of the United States of America 105:11823–11826. Google Scholar

50.

S. Koniak 1985. Succession in pinyon-juniper woodlands following wildfire in the Great Basin. Great Basin Naturalist 45:556–566. Google Scholar

51.

J.A.R. Ladyman 2004. Gutierrezia sarothrae (Pursh) Britt, and Rusby. Pages 367–369 in J.K. Francis , editor, Wildland shrubs of the United States and its territories: tharnnic descriptions, volume 1. General Technical Report IITF-GTR-26, USDA Forest Service, International Institute of Tropical Forestry, San Juan, PR, and Rocky Mountain Research Station, Fort Collins, CO. Google Scholar

52.

M.H. Madany , and N.E. West . 1983. Livestock grazing fire regime interactions within montane forests of Zion National Park, Utah. Ecology 64:661–667. Google Scholar

53.

M.H. Madany , and N.E. West . 1984. Vegetation of two relict mesas in Zion National Park. Journal of Range Management 37:456–461. Google Scholar

54.

G.P. Malanson 1980. Habitat and plant distributions in hanging gardens of the Narrows, Zion National Park, Utah. Great Basin Naturalist 40:178–182. Google Scholar

55.

G.P. Malanson , and J. Kay . 1980. Flood frequency and the assemblage of dispersal types in hanging gardens of the Narrows, Zion National Park, Utah. Great Basin Naturalist 40:365–371. Google Scholar

56.

B. McCune , and J.B. Grace . 2002. Analysis of ecological communities. MjM Software Design, Gleneden Beach, OR. Google Scholar

57.

B. McCune , and D. Keon . 2002. Equations for potential annual direct incident radiation and heat load. Journal of Vegetation Science 13:603–606. Google Scholar

58.

S.P. McLaughlin 1986. Floristic analysis of the southwestern United States. Great Basin Naturalist 46:46–65. Google Scholar

59.

V.L. Mortensen 1977. Soil survey of Washington County area, Utah. USDA Soil Conservation Service, Washington, DC. Google Scholar

60.

D. Mueller-Dombois , and H. Ellenberg . 1974. Aims and methods of vegetation ecology. Wiley, New York, NY. Google Scholar

61.

[NPS] National Park Service. 2006. Management policies 2006.  http://www.nps.gov/policy/MP2006.pdf  Google Scholar

62.

NatureServe. 2013. Natures erve explorer: an online encyclopedia of life. [Accessed 21 October 2013, 28 October 2013].  http://www.natureserve.org/explorer  Google Scholar

63.

J. Olivero , R. Real , and A.L. Márquez . 2011. Fuzzy ehorotypes as a conceptual tool to improve insight into biogeographic patterns. Systematic Biology 60: 645–660. Google Scholar

64.

S. O'Meara 2006. Geologic map of Zion National Park and vicinity, Utah. Utah Geology Survey. Google Scholar

65.

J.E. Ott , E.D. McArthur , and S.C. Sanderson . 2001. Plant community dynamics of burned and unburned sagebrush and pinyon-juniper vegetation in west-central Utah. Pages 177–191 in E.D. McArthur and D.J. Fairbanks , compilers, Shrubland ecosystem genetics and biodiversity: proceedings. Proc. RMRS-P-21, USDA Forest Service, Rocky Mountain Research Station, Ogden, UT. Google Scholar

66.

T.E. Paysen , R.J. Ansley , J.K. Brown , C.J. Gottfried , S.M. Haase , M.G. Harrington , M.G. Narog , S.S. Sackett , and R.C. Wilson . 2000. Fire in western shrubland, woodland, and grassland ecosystems. Pages 121–159 in J.K. Brown and J.K. Smith , editors, Wildland fire in ecosystems: effects of fire on flora. General Technical Report RMRS-GTR-42-vol. 2, USDA Forest Service, Rocky Mountain Research Station, Ogden, UT. Google Scholar

67.

H.W. Polley , D.D. Briske , J.A. Morgan , K. Wolter , D.W. Bailey , and J.R. Brown . 2013. Climate change and North American rangelands: trends, projections, and implications. Rangeland Ecology and Management 66:493–511. Google Scholar

68.

R Development Core Team. 2008, 2010. R: a language and environment for statistical computing. Versions 2.8.0 and 2.10.1. R Foundation for Statistical Computing, Vienna, Austria.  http://www.r-project.org/  Google Scholar

69.

W.J. Ripple , and R.L. Beschta . 2006. Linking a cougar decline, trophic cascade, and catastrophic regime shift in Zion National Park. Biological Conservation 133:397–408. Google Scholar

70.

S.C. Sanderson , J.E. Ott , E.D. McArthur , and K.T. Harper . 2006. RCLUS, a new program for clustering associated species: a demonstration using a Mojave Desert plant community dataset. Western North American Naturalist 66:285–297. Google Scholar

71.

L. Tichý , and M. Chytrý . 2006. Statistical determination of diagnostic species for site groups of unequal size. Journal of Vegetation Science 17:809–818. Google Scholar

72.

[USDA-FS] United States Department of Agriculture-Forest Service. 2013. Fire effects information. [Accessed 16 November 2013].  http://www.fs.fed.us/database/feis/plants  Google Scholar

73.

[USNVC] United States National Vegetation Classification. 2013. The U.S.National Vegetation Classification: your guide to inventorying natural and cultural plant communities. [Accessed 21 October 2013].  http://usnvc.org  Google Scholar

74.

A.D. Weiss 2001. Topographic position and landforms analysis. The Nature Conservancy and Jenness Enterprises, Seattle, WA.  http://www.jennessent.com/arcview/TPI_Weiss_poster.htm  Google Scholar

75.

S.L. Welsh 1989. Hanging gardens of Zion National Park: final report. Endangered Plant Studies, Inc., Orem, UT. Google Scholar

76.

S.L. Welsh , N.D. Atwood , S. Goodrich , and L.C. Higgins . 1987. A Utah flora. 1st edition. Brigham Young University, Provo, UT. Google Scholar

77.

S.L. Welsh , N.D. Atwood , S. Goodrich , and L.C. Higgins . 2003. A Utah flora. 3rd edition. Brigham Young University, Provo, UT. Google Scholar

78.

N.E. West 1999. Juniper-pinyon savannas and woodlands of western North America. Pages 288–308 in R.C. Anderson , J.S. Fralish , and J.M. Baskin , editors, Savannas, barrens, and rock outcrop plant communities of North America. Cambridge University Press, Cambridge, England. Google Scholar

79.

[WRCC] Western Regional Climate Center. 2014. [Accessed 4 August 2014].  http://www.wrcc.dri.edu/climatedata/climsum  Google Scholar

80.

J.B. Wilson 2012. Species presence/absence sometimes represents a plant community as well as species abundances do, or better. Journal of Vegetation Science 23:1013–1023. Google Scholar

81.

D. Witwicki 2013. Climate monitoring in the Northern Colorado Plateau Network: annual report 2011. Natural Resource Technical Report NPS/NCPN/NRTR—2013/664, U.S. National Park Service, Fort Collins, CO. Google Scholar

82.

A.M. Woodbury 1933. Biotic relationships of Zion Canyon, Utah Math special reference to succession. Ecological Monographs 3:151–245. Google Scholar

Appendices

Appendix1.

Abiotic environmental variable layers derived from a 10-m-resolution digital model of Zion National Park, Utah.

fA01_26.jpg

Appendix2.

Geologic substrate classes derived from a 1:24,000-scale digital geologic map of Zion National Park, Utah, and vicinity (O'Meara 2006).

fA02_26.jpg

Appendix 3.

Vascular plant species recorded by the Harper vegetation survey (1987–1989) at Zion National Park, Utah, and their affinities to coalition species groups. # Occ. = number of plot occurrences out of 288 plots total. Affinities are mean pairwise association values (in units of the phi coefficient) between a species and members of a coalition group. For species that have at least 3 occurrences, values >0.15 indicate membership in a group. Green shading by cell value is scaled separately for each column. Nomenclature follows Welsh et al. (2003).

tA03a_26.gif

Continued.

tA03b_26.gif

Continued.

tA03c_26.gif

Continued.

tA03d_26.gif

Continued.

tA03e_26.gif

Continued.

tA03f_26.gif

Continued.

tA03g_26.gif

Continued.

tA03h_26.gif

Continued.

tA03i_26.gif

Continued.

tA03j_26.gif

Continued.

tA03k_26.gif

Continued.

tA03l_26.gif

Continued.

tA03m_26.gif

Continued.

tA03n_26.gif

Continued.

tA03o_26.gif

Continued.

tA03p_26.gif

Continued.

tA03q_26.gif

Continued.

tA03r_26.gif

Continued.

tA03s_26.gif

Appendix 4.

Modal species of vegetation types described by Harper et al. (2001) for Zion National Park, Utah (rows), and their affinities to coalition groups (columns; see also Appendix 3). Modal species are prevalent species assigned to the vegetation type where their percent occurrence is highest (Curtis 1959). Affinities are mean pairwise association values (in units of the phi coefficient) between a species and members of a coalition group. Species are listed in descending order of percent occurrence within vegetation types. Green shading by cell value is scaled separately for each column. Nomenclature follows Welsh et al. (2003).

tA04a_26.gif

Continued.

tA04b_26.gif

Continued.

tA04c_26.gif

Continued.

tA04d_26.gif

Continued.

tA04e_26.gif

Appendix 5.

Associations of the U.S. National Vegetation Classification System at Zion National Park, Utah, as recognized by Cogan et al. (2004), placed within an alternative classification hierarchy based on affinities to coalition groups (see Fig. 7 in paper). Affinity values (right 10 columns) are in units of the phi coefficient. The values are calculated for each species in relation to each coalition group (see Appendix 3) and then averaged across plots and associations. Green shading by cell value is scaled separately for each column.

tA05a_26.gif

Continued.

tA05b_26.gif

Continued.

tA05c_26.gif

Continued.

tA05d_26.gif

Appendix 6.

Vegetation units of the U.S. National Vegetation Classification at Zion National Park, Utah, as recognized by Cogan et al. (2004), and their placement in alternative classes based on coalition group affinities (see Fig. 7 in paper).

tA06a_26.gif

Appendix 6. Continued.

X. Class fi01_26.gif X.X. Subclass fi01_26.gif X.X.X. Formation Group fi01_26.gif X.X.X.X.x. Formation.

tA06b_26.gif

Continued.

tA06c_26.gif

Continued.

tA06d_26.gif

Continued.

tA06e_26.gif

Continued.

tA06f_26.gif

Appendix 7.

Associations of the U.S. National Vegetation Classification (NYC) at Zion National Park, Utah (Cogan et al. 2004), arranged according to the revised classification heirarchy (FGDC 2008), and their placement in alternative classes based on coalition group affinities (see Fig. 7 in paper). NYC classes were obtained or inferred from NatureServe (2013) or USNVC (2013).

tA07a_26.gif

Appendix 7. Continued.

X - Class fi01_26.gif X.X Subgrass fi01_26.gif X.X.X - Formation fi01_26.gif X.X.X.Xx - Division fi01_26.gif Macrogroup fi01_26.gif Group.

tA07b_26.gif

Continued.

tA07c_26.gif

Continued.

tA07d_26.gif

Continued.

tA07e_26.gif

Continued.

tA07f_26.gif

Continued.

tA07g_26.gif

Appendix 8.

Numbers of associations of the U.S. National Vegetation Classification (NVC) at Zion National Park, Utah (Cogan et al. 2004), classified to land cover classes and ecological systems (obtained or inferred from 2013) and alternative classes based on coalition group affinities (see Fig. 7 in paper).

tA08a_26.gif

Continued.

tA08b_26.gif
© 2015
Jeffrey E. Ott, Stewart C. Sanderson, and E. Durant McArthur "Plant Species Coalition Groups of Zion National Park: An Individualistic, Floristic Alternative to Vegetation Classification," Monographs of the Western North American Naturalist 8(1), 26-97, (1 January 2015). https://doi.org/10.3398/042.008.0102
Received: 25 February 2014; Accepted: 18 August 2014; Published: 1 January 2015
JOURNAL ARTICLE
72 PAGES


Share
SHARE
RIGHTS & PERMISSIONS
Get copyright permission
Back to Top