1203

Phytosociology and gradient analysis of a subalpine treed fen in Rocky Mountain National Park, Colorado

J. Bradley Johnson

Abstract: The vegetation of a subalpine fen in Colorado was studied. Insight was sought into the community structure and factors influencing species distribution of a vegetation type heretofore undescribed in the southern Rocky Mountains. A vegetational gradient was evaluated using detrended correspondence analysis (DCA). Four types of vegetation were subjectively defined; these same types were distinguished by the DCA. DCA further revealed marked differences in the vegetation occurring on peat hummocks versus in hollows. Species composition was related to environment using canonical correspondence analysis (CCA). Water-table depth, hummock height, shading, groundwater temperature, and conductivity were significantly correlated with species distribution, accounting for 51 % of the total species variance. Univariate regression was used to examine how tree density varied with environment. The above factors, except for shading and conductivity. were also significantly correlated with tree density. It is suggested that the peat hummocks that form on this moderate fen provide an environment similar to that of an ombrotrophic bog and that these "miniature bogs" form in areas unable to support expansive bogs. Further, these hummocks provide small-scale environmental heterogeneity that exerts a strong control over species composition that would not be evident in studies based on samples of a large areal extent.

Key words: Colorado, gradient analysis, ordination, heterogeneity, peatlands, phytosociology.

Resume: L'auteur a etudie la vegetation d'une tourbicre haute du Colorado. II a eherche a comprendre la structure de la communaute et les facteurs qui influencent la distribution des espcces d'un type de vegetation non-deerit jusqu'ici, du sud des montagnes Rocheuses. 1\ a evalue un gradient de vegetation a raide de l'analyse de correspondance sans tendances (DCA). II a subjectivement deflni qllatre types de vegetation; ces memes types ont ete deceles par la DCA. La DCA montre egalement des differences importantes qui surviennent dans la vegetation venant sur les hummocks versus dans les depressions. La composition en especes est relice a I 'environnement selon l'analysc par correspondance canonique (CCA). La profondeur de la nappe phreatique, la hauteur des hummocks, l'ombrage, la temperature de la nappe phreatique et 13 conductivite sont significativement correles avec la distribution des especes, expliquant 51 % de la variance totale en especes. La regression univariee a ete utilisee pour examiner comment la densite des arbres varie selon I'environnement. Les facteurs enumeres plus haut, sauf pour l'ombrage et la conductivile, sont egalement correles significativement avec la densite des arbres. L'auteur suggere que les hummocks de tourbe qui se forment sur celte tourbiere moderement elevee creent un milieu similaire a celui d'lIne tourbiere ombrotrophe, et que ces "tourbieres miniatures" se forment dans des regions incapables de supporter des tourbieres expansives. Dc plus, ces hummocks conduisent a une heterogeneite environnementale 11 petite echelle qui excn;e un important contr6le sur la composition en especes, qui ne serait pas evidente dans les etudes basce~ sur de grandcs superficies,

Mots eMs: Colorado, analyse de gradient, ordination. hctcrogcneite, tourbieres, phytosoeiologie. {Traduit ~ar la redaction]

Introduction this occurs in the Rocky Mountains, however, where a climate comparable to more northern latitudes is found (Marr

In Europe, peatland studies go back to the roots of ecology 1961). Rocky Mountain wetlands were only recently examined (e.g., Weber 1911; Cajander 1913), whereas North American systematically and few studies have specifically examined peat­peatlands have only received attention since the late 19405. lands. These studies were primarily concerned with describing Peatlands cover vast areas in the boreal regions of the world general phytosociological relationships and wetland classifi­and are found in more southern reaches where the climate is cation (Cooper 1994): investigations of wetland processes suitable. In interior North America, south of the boreal zone, and the factors influencing species composition remain few peatlands occur infrequently because of the comparatively (e.g., Bierly 1972; Cooper 1990). hot summer and inadequate precipitation. An exception to The general absence of information on Rocky Mountain

wetlands, coupled with current concerns about wetland preser­vation, make studies of these areas particularly appealing.

Received September 12, 1995. Additionally, wetland vegetational types. well known from J.B. Johnson. Department of Biology. Colorado State the boreal regions, have either gone unnoticed, or have only University. Fort Collins, CO 80523, U.S.A. recently been observed in these mountains. One such example

Can. J. Bot. 74: 1203 1218 (1996'1. Printed in Canada i Imprimc au Canada

1204

is treed fens, which were recently found in the subalpine zone but have heretofore not been examined in detail. These fens, while not as common as other wetland types, seem to be widespread in the upper subalpine areas (R.L. Dix, unpublished data). This paper reports the findings of a study of one such fen.

The specific objectives of this study were (i) to describe the phytosociology of a subalpine treed fen and its surround­ing vegetation and (ii) to investigate the factors influencing species distribution, and especially to examine the role that fen microtopography plays in determining species distribu­tion and vegetational development.

Background The physiognomic similarity of treed fens with the common spruce-fir upland forest has lent to their obscurity. Among previous studies of southern Rocky Mountain vegetation, only one study mentioned subalpine treed peatlands (Peet 1981). In this study, the treed peatlands were considered to be outliers on the upland forest continuum and were not considered in detail, although several were examined. The tree species composition of these mires is also puzzling. Outside of Colorado, species such as black spruce (Picea mariana) and larch (Larix laricina) typically dominate the upper canopy of peatlands. These species are specially adapted to wet/and environments but do not occur in the southern Rocky Mountains. In Colorado, Engelmann spruce (Picea engelmannii), subalpine fir (Abies lasiocarpa), and to a much lesser degree, lodgepole pine (Pinus contorta) make up the canopy of treed fens in their stead. These species normally inhabit mesic to xeric upland sites, and their presence in these wetlands is enigmatic. Strangely, Pinus contorta, which frequently occurs in boreal peatlands (Neiland 1971; Sjors 1983; Vitt et al. 1990) and which is quite common in the subalpine forest, is only sparsely distributed on the subalpine peatlands. Also, on Alberta fens, which are often surrounded by Picea engelmannii - Abies lasiocarpa forest, these species do not occur. Instead, when trees are present on the fen they are generally either Picea mariana or Larix laricina (Vitt et al. 1975).

Peatland terminology A difficulty inherent in peat/and studies is the confusing and often inconsistent usage of terms. These inconsistencies came about because many tetms have a long-standing colloquial usage and definitions have been imprecise and regionally variable. For the purposes of this paper, I adapted a termi­nology that is general and widely employed.

A bog is a peatland that is ombrotrophic regardless of other physical or vegetational characteristics (Du Rietz 1949; Sjors 1950b; Moore and Bellamy 1974). A fen is broadly defined as a peatland that is minerotrophic (Du Rietz 1949; Sjors 1950a; Vitt and Slack 1975; Slack et al. 1980). Because of this minerotrophy, fens are more nutrient rich than bogs, and fen species have comparatively high nutrient requirements. Fens are typically dominated by graminoids, especially sedges, and shrubs are usually common. Trees may also be present. Minerotrophic peatlands with a substantial forest canopy were also termed swamps, especially by Canadian workers (e.g., leglum 1987; Jeglum and He 1995). This term will not be used here, as I am using the broad definition of fen that

Can. J. Bot. Vol. 74, 1996

encompasses this type of peatland. This will also be done in an attempt to eliminate possible confusions of this peatland system with other swamp types, particularly the deep-water cypress swamps of the southern U.S.

Fen microtopography The role that microtopography plays in peatland ecology needs further consideration because of its effect on species composition. Differences in vegetational composition between microtopographical positions have been noted for over 80 years (Cajander 1913). These disparities are generally attributed to the differences in the effective depth to water, differences in the oxidation - reduction potential of soil, and differences in water mineral nutrient content brought on by the high cation exchange properties of moss species, especially Sphagnum species (Gorham 1957; Sj6rs 1963; Moore and Bellamy 1974; Mitsch and Gosselink 1993). The effective lowering of the water table seems especially crucial for the establishment and survival of tree seedlings on these sites. In controlled experi­ments, Boogie and Miller (1976) found that the only Pinus contorta seedlings to survive in situations in which the water table was kept close to the surface were those that became established on Sphagnum hummocks. In a similar experi­ment, seedlings growing on hummocks were the only ones that exhibited growth over a period of 2 years (Boogie 1972). Seedling survival was attributed to the growth of adventitious roots, which penetrated the hummock as the moss grew up the seedling's stem. These roots were exposed to a more favorable (aerobic) environment, rather than the saturated, anoxic hollow environment. Boogie (1972) also found that 90 % of seedling roots were located in the top 4 cm of the substrate, suggesting that even small rises could increase seedling survival. Topographically dependent survival was also observed in field studies of peatlands in which woody species, particularly trees, are often confined to rises (e.g., Sjors 1950a; Vitt et al. 1975; Slack et al. 1980; Zoltai and Johnson 1985).

When a plant grows on a hummock large enough so that its root zone is above the capillary fold of the Sphagnum moss, the plant becomes isolated from the minerotrophic groundwater. The environment that these plants experience is therefore essentially that of an ombrotrophic bog. The hummock height necessary for this isolation to occur varies by species, but Karlin and Bliss (1984) found that most shrubs in their study sites produced roots only to a depth of about 15 cm. Thus, rises need not be excessively large for this effect to occur. Several workers noted the presence and described these mini-ombrotrophic bogs that occur on peat­lands that are otherwise classified as poor or even rich fens (e.g., Bellamy and Rieley 1967; Zoltai and lohnson 1985). In the terminology of Sj6rs (1965), peatlands with such heterogeneity may also be called mixed mires.

Colorado peatlnnds No true bogs have yet been found in the southern Rocky Mountains (Cooper 1986). Peet (1981) describes what he calls "bog forest," but this description seems to be a misnomer, as these sites located in valley bottoms would receive con­siderable runoff from melting snow. Therefore, they should be classified as treed fens. Most peatlands in Colorado were described as poor or moderate fens (Bierly 1972; Cooper

1205 Johnson

1986, 1990). Isolated rich fens were described in South Park, Colorado (D.J. Cooper, unpublished data; J .B. Johnson, unpublished data), but these appear to be confined to that region. The majority of Rocky Mountain wetlands were described as soligenous, that is they receive groundwater that has been flowing down slopes (Gore 1983). The major excep­tions to this are the wetlands residing in intermountain basins, such as those of the South Park and the San Luis valley regions.

Poor fens dominate the Front Range because of the area's geology and hydrology. Siliceous rocks are the predominant soil parent material in this region (Marr 1961; Richmond 1974; Rink and Kiladis 1986) and this led to the formation of mineral poor soils. Further, soils in this area tend to be heavily leached by acidic coniferous litter (Marr 1961). These traits led to ionically poor groundwater and subsequently only weakly minerotrophic fens.

Study area and methods

Site description The wetland chosen for this study is located on the west side of the continental divide in Rocky Mountain National Park, Colorado, at an elevation of approximately 2900 m (9560 ft.). This wetland is not named on maps but will be called Spring Fen for convenience. Spring Fen is located about 0.5 km NE of the north end of Big Meadows (Fig. 1; 40020'N, 105°47'W) and sits at the bottom of a glacial valley that runs NW to SE. The wetland covers approxi­mately 9 ha and grades from a sedge fen to a treed fen and finally to upland spruce-fir forest on the northeast side (Fig. 2). On the southwest side, the fen ends abruptly at a lateral moraine.

Spring fen is difficult to classify according to established criteria because it shares attributes of both poor and moderate fens, but it seems to be closer to a moderate fen (J .B. Johnson, unpublished data). A pronounced hummock - hollow topography is present on the fen. The hummocks, which are dome-shaped and made pre­dominantly of Sphagnum peat, are small (avg. height = 17 cm) on the sedge fen but increase in height and diameter in the treed portions of the fen (avg. height = 46 cm). Much of the fen is fed by springs that emerge along the bottom of the southwest-facing slopes. In other portions of the fen, particularly the southwest end, no springs were found.

The climate of the subalpine zone in Colorado is characterized as cool and moist, with long, cold winters and short, cool summers (Baker 1944; Rink and Kiladis 1986). Peak precipitation occurs in the winter as snow. Beginning in May and continuing to late June, the snow pack melts and inundates the area with several months of accumulated precipitation, but much of this water is rapidly drained off because of the steep mo~ntain topography. Sheet flow ceases by late June in most areas, but surface water may be persistent in valley bottoms throughout the summer. Cooper (1990) calculated that Big Meadows receives about 74 cm of precipitation annually; precipita­tion is expected to be similar at Spring Fen. Spring Fen was chosen as a study site for several reasons: (i) it is an excellent example of a treed fen; (ii) the wetland has long been free from large-scale disturbances (e.g., catastrophic fires, avalanches, alteration by beavers); and (iii) it exhibits a striking vegetation gradient.

Phytosociological sampling Species names are according to Weber (1987), but because readers may not be familiar with his recent nomenclatorial revisions, spes;ies names not consistent with Kartesz (1994) are noted parenthetically. The vegetation changes markedly across the fen, from a sedge fen, through an ecotone, to mature treed fen, and finally to typical upland spruce-fir forest. This vegetational gradient runs perpen­dicular to the long axis of the fen (Fig. 2) and is the basis for the

Fig. 1. Map of Colorado showing the location of Rocky Mountain National Park and Spring Fen within the park.

105°47' W

Rocky Mountain

1-70

~

COLO,io

National Pa

Spring F n

. ntal Divid

---'------~-----~--~---

sampling design of this study. To facilitate an even sampling, Spring Fen was divided into three sections, with surficial features partitioning each section. Sections were subdivided into four stands, with one stand being placed in each type of vegetation. A water track type of vegetation (sensu Sjors 1950a), dominated by spike rush (Eleocharis paucijlora), also occurs on the fen. This vegeta­tional zone divides the fen nearly in half and is oriented orthogonally to the other zones. Because the vegetation in this area is quite different than elsewhere on the fen, an additional stand was established within this vegetation. Three stands of the surrounding upland forest adjoining the sections were also sampled, so that differences between the treed fen and the physiognomically similar upland forest could be investigated. A total of 13 stands were examined and the vegeta­tion within each stand was visually homogeneous.

A single sampling protocol was not suitable for the entire fen because of the marked differences in physiognomy across the vegeta­tional types. Therefore, vegetation sampling was carried out using two separate but comparable regimes. Such a blending of methodo­logical approaches was shown to be a valuable way of appropriately sampling vegetation with disparate attributes (e.g., Curtis 1959). Sample points were objectively chosen by tossing a metre stick over the shoulder, walking 30 paces in the indicated direction, and posi­tioning the sample point marker at the location of the last step. In total, 208 points were sampled.

Phytosociological attributes were evaluated using a combination of areal and distance measures. The line-strip method (Lindsey 1955;

1206 Can. J. Bot. Vol. 74. 1996

Fig. 2. Map of Spring Fen, The numbered points correspond to the location of groundwater wells, The cross-hatching identifies vegetation types.

UPLAND FOREST \

•11

N

TREED FEN UPLAND FOREST OPEN-TREED FEN

SEDGE FEN

a SPIKE RUSH FEN

Greig-Smith 1983) was used to measure shrub cover. and either areal quadrats or the point-centered quarter method (Cottam and Curtis 1956) was utilized to measure tree density. The method used depended on the estimated tree density but remained consistent within a given vegetation zone. Tree seedling density and herba­ceous cover were always measured using quadrats,

On the sedge fen and open-treed fen (Fig. 2), a IO-m line was used to measure shrub cover, and tree density was measured using 4 x 10 m quadrats. The species and diameter at breast height (dbh) were recorded for all trees within these quadrats, If the encountered individual was a seedling (dbh < 2.5 cm), only its species was recorded. To estimate herbaceous species cover, two 0,33 x 0.33 m quadrats were used at each sample point. One quadrat was placed on top of the hummock nearest the sampling point, while the other was placed in the nearest hollow. This was done to determine the extent to which microtopography affects vegetational composition. This quadrat size was chosen as it fit within the confines of all but the smallest hummocks or hollows. The height of the hummock on which the quadrat was placed was also measured, Within quadrats, the cover of each species was visually estimated using Braun-Blanquet (1932) cover classes (+, <1 %; 1, 1-5%; 2, 6-25%; 3, 26-50%; 4,51-75%; 5, 76-100%). When shrubs or their canopies occurred. in quadrads, their percent cover was also estimated,

Sampling in the treed fen and upland forest followed this general scheme but varied in two ways. First, in the measurement of shrubs, 15-m rather than IO-m transects had to be used to obtain an adequate sample because of the lower shrub density. Second, in the measurement of tree densities, the point-centered quarter method was employed. This sampling technique was used here because it has been shown to be a quick and accurate way of obtaining density values, provided that individuals are relatively plentiful and randomly distributed (Cottam and Curtis 1956). The assumption of random­ness was verified during a preliminary study. The sampling of seedlings was carried out separately from that of the tree-sized individuals, and 2 x 5 m rectangular area quadrats were used,

Environmental sampling Environmental data were obtained from transects of groundwater wells, located in the three regions ofthe wetland (Fig. 2). Twenty­

•

100 m

one groundwater wells were placed on the wetland, Wells were made out of 3.75 cm diameter polyvinylchloride (PVC) piping. The pipes were capped at one end, perforated using a hand drill, and sunk to a depth of about 90 cm .. Six times throughout the summer of 1992, electrical conductivity, pH, and temperature were measured in situ, using Hach field meters (meter model numbers 44 600 and 1, respectively). Depth to the water table was measured seven times by inserting a steel tape measure into the well.

To characterize groundwater minerotrophy, one water sample was taken from each well. Before samples were extracted, wells were pumped dry and allowed to refill. Samples were filtered and preserved at the Colorado State University Soil Testing Lab, using 0.45-JLm nitro-cellulose filters and nitric acid. Samples were then analyzed for Ca2+, Mg2+, Na+, K+, and NH4 + concentrations using ion spectrophotometry (ICP). These ions were selected for analysis based on the findings of previous studies (e.g., Gorham 1967; Heinselman 1963; Chee and Vitt 1989; Waughman 1980) in which they were shown to be correlated with species distributions. Last, tree density was used as an approximator of shading. While an absolute relationship between these parameters is not given, density should provide an adequate relative measure of shading in each of the quadrats.

To relate these physical parameters to species composition, 5 x 5 m quadrats were centered around each well. Within these quadrats, the percent coverage of each herb and shrub species was visually estimated and recorded using the Braun-Blanquet scale, and all trees, saplings, and seedlings were counted. The dbh of trees and saplings was also recorded.

Data analysis Linear and exponential regression and the coefficient of determina­tion (R2) were used to investigate the relationship between tree species density and environmental factors. As herbaceous and shrub species data were too numerous to examine in such a manner, multi­variate direct and indirect gradient analysis (ordination) were employed (Whittaker 1967; Gauch 1982; Ter Braak 1986; Ter Braak and Prentice 1988). Detrended correspondence analysis (DCA) was used to ordinate species and samples, while canonical correspondence analysis (CCA) was used to relate species composition to environ­

1207 Johnson

mental factors. CCA and DCA are analogous techniques of eigen­analysis, but in CCA stand and species placements are constrained to be linear combinations of the values of environmental variables.

Species data from the 416 quadrats were used in the DCA ordination. These data were first grouped according to the stand from which they came, and secondarily according to their micro­topographical position (i.e., whether the quadrat was located on a hummock or in a hollow between hummocks). By grouping samples in such a way, it was hoped that disparities in species composition caused by microtopographical position would become evident. Tree data could not be included in these analyses because tree density rather than cover was measured. This is a consequence of the dis­parate life-forms of the species sampled.

Vegetation data for the CCA were obtained from 21 quadrats centered on the wells. A stepwise selection of environmental varia­bles using a Monte Carlo permutation test was initially performed. Because of this selection and the subsequent removal of unneeded factors, no further detrending was necessary (Ter Braak 1990). DCA and CCA scores were also regressed to assess the stand place­ment due to the constraint of making scores linear combinations of environmental values (Prentice and Cramer 1990; Allen and Peet 1990) and also to judge the importance of unmeasured factors. The computer program CANOCO (Ter Braak 1990) was used to carry out the analyses.

Results

Phytosociology of spring fen and ordination of vegetation Four major types of wetland vegetation were subjectively distinguished on Spring Fen: sedge fen, spike-rush fen, open­treed fen, and treed fen. These same divisions were later obtained in quantitative analyses (see below). Figure 2 is a map of the distribution of vegetational types on the wetland. The qualitative characters of the vegetation types defined in this study are briefly presented below; following this, quan­titative aspects are presented.

The upland forest surrounding the fen is dominated by Picea engelmannii and Abies lasiocarpa. Pinus contorta is also present as a minor species. The understory is sparse and patchy, with isolated herbs or prostrate ericaceous shrubs. The physiognomy of the treed fen is similar to that of the upland forest. Picea engelmannii and Abies lasiocarpa make up the canopy, but in many areas the understory tree Alnus inc{lna is also present. The understory in the treed fen is dense and dominated by herbs. In this and following vegeta­tional types, a difference in species composition between hummocks and hollows Was observed, with hydrophytes, such as sedges, inhabiting the hollows and mesophytes growing on hummocks. The open-treed fen vegetation typically consists of three strata: an open canopy of Picea engelmannii, a middle layer of shrubs, and a lush carpet of herbaceous vegetation (mainly sedges). This zone has many species in common with the treed fen and sedge fen and is an ecotone between these two vegetation types.

The sedge fen is typical of wet meadows in the Rocky Mountains. It has a scattering of low willows (Salix planifolia) and bog birch (Betula glandulosa) that form a diffuse canopy layer and a dense carpet of herbaceous vegetation, commonly dominated by Carex aquatiUs. The spike-rush fen forms a track through the center of the wetland and is the major drainage path for the springs that feed the eastern portions of the fen. This area, which has appreciable overland water flow, has few trees and is dominated by Eleocharis quinque flora

Table 1. Density (stems/ha) of tree species in each vegetation type.

Upland Treed Open-treed Sedge Spike-rush Species forest fen fen fen fen

Abies lasiocarpa 5470 2130 90 Picea engelmannii 3320 2340 1640 80 20 Pinus contarta 190 40 30 40

in its low wet areas. Only on dry hummocks and other topographic rises do Pentaphylloides floribunda (Potentilla fruiticosa) and Deschampsia caespitosa dominate instead.

All tree species decrease in density from the upland forest to the sedge and spike-rush fen (Table 1). This quantitative decrease is also visually apparent, except between the upland forest and the treed fen. Pinus contorta is sparsely distributed in most places but becomes more important in the open areas of the fen (although it still has a low density). Shrub species composition also changes, but species are subject to replacement across vegetational types, rather than being Ubiquitous and exhibiting changes in abundance (Table 2). In the upland forest, the shade-tolerant, mesophilic shrubs Juniperus communis and Ribes spp. dominate, and Alnus incana occurs occasionally near rivulets. In wetland areas, the hydrophilic shrubs Salix planifolia and Betula glandulosa dominate (Table 2). A similar pattern is exhibited by herba­ceous species, which show continual species replacement throughout the upland and wetland sites (Table 3).

Figure 3 shows the placement of stands in the DCA ordi­nation. Even though tree data were not included in these analyses, a separation of treed versus untreed stands was obtained. The units of DCA diagrams are standard deviations (SD) and measure the amount of species turnover along gradients (Gauch 1982). Along axis 1, stands from like vege­tation zones tend to cluster together, while each vegetational type is separated. The gradient length for axis 1 is 4.8 SD, which indicates a complete species turnover (i.e., spike-rush fen and upland forest have no species in common) (Gauch 1982). Stands within vegetation types are separated on axis 2, according to the microtopographical position from which samples were taken. Samples taken from hummock tops are consistently placed lower on axis 2 compared with those taken from hollows. This separation may be as much as 1.25 SDs (in the treed fen) and shows that quadrats located spatially close to one another can have more than a 50% difference in species composition due to microtopographical position alone.

Direct gradient analysis The average values of environmental factors in each vegeta­tion type are presented in Table 4. Note that because tree density was used as an index of shading, its values were not included in this table; instead, they can be found in Table I. Among these, peat depth, shading, hummock height, conduc­tivity, water-table depth, and water temperature were found, using CCA, to correlate significantly with species composi­tion (Fig. 4). Together, these six factors account for 51 % of the total variance found within the species data. The first axis was found to be most strongly, positively correlated with shading and hummock height and negatively correlated with a combination of water temperature and water depth.

1208 Can. J. Bot. Vol. 74, 1996

Table 2. Average percent cover of shrub species in each vegetation type.

Upland Treed Open-treed Sedge Spike-rush Species forest fen fen fen fen

Juniperus communis 1.03 Ribes spp. 1.87 0.13 Alnus incana 0.5 15.31 0.6 Betula glandulosa 7.15 1.9 0.06 Salix planifolia 0.15 12.92 13.75 0.35 Pentaphylloides fioribunda 1.71

Table 3. Estimated cover values of herbaceous species in each vegetation type, averaged from 416 quadrats.

Upland Treed Open-treed Sedge Spike-rush Species Abbreviation forest fen fen fen fen

Arnica cordifolia ARN COR I + Vaccinium scoparium VAC SCO 2 2 + + Ribes lacustre RIB LAC 1 Erigeron perigrinus ERI PER + Osmorhiza depauperata OSM DEP + Senecio triangularis SEN TRI + Trollius albijlorous (Trollius laxus) TRO ALB + Moneses unijlora MON UNI + + Lycopodium annotinum LYC ANN + Listera cordata LIS COR + Limnorchis spp. (Habenaria spp.) LIM SPP + Geranium richardsonii GER RIC + Epilobium lactijlorum EPI LAC + + + Pseudocymopterus montanus PSE MON 1 1 Streptopus fassettii (Streptopus amplexifolius) STR FAS I + Rhizomnium pseudopunctatum RHI PSE 2 Vaccinium myrtillus VAC MYR 2 Pyrola spp. PYR SPP + + + Crunocallis chamissoi (Montia chamissoi) CRU CHA + Micranthes odontoloma (Saxifraga odontolomo) MIC ODO 1 Mitella pentandra MIT PEN Equisetum pratense EQU PRA I Carex disperma CAR DIS + 3 Luzula parvijlora LUZ PAR + + Fragaria vesca FRA VES + + + Cardamine cordifolia CAR COR 1 Viola adunca VIO ADU + + + Carex canescens CAR CAN I 1 I Carex aquatilis CAR AQU 3 4 3 2 Galium trifidum GAL TRI + + + Epilobium hornemannii EPI HOR + + + Clementsia rhodantha (Sedum rhodanthum) CLE RHO + 1 1 + Calamagrostis canodensis CAL CAN + 2 2 + + Pedicularis groenlandica PED GRO 1 1 Psychrophila leptosepala (Caltha leptosepala) PSY LEP 2 1 Salix planifolia SAL PLA + 2 1 Carex utriculata (Carex rostrata) CAR UTR + + 1 2 Viola renifolia VIO REN 2 Deschampsia cespitosa DES CES 1 Eleocharis quinque flora ELO QUI + 2 3

Note: Raw cover values were grouped into six cover classes: +, <1 %; 1, 1-5%; 2, 6-25%; 3, 26-50%; 4,51-75%; 5,76-100%.

The second axis represents the interaction of peat depth and unduly influenced the placement of species and stands and conductivity. Upland forest stands were not included in these obscured the gradients occurring on the wetland. analyses because the lack of peat substrate and peat hummocks Eigenvalues of the axes in the constrained analysis (CCA)

Johnson 1209

Fig. 3. Detrended correspondence analysis ordination of stands, after samples were grouped according to their topographical position, i.e., whether a sample was located on a hummock or in a hollow. Stands from the same vegetation type and topographical position were grouped subjectively in the diagram. The topographical position from which samples were taken is indicated by the labeled arrows, and the vegetation type is identified at the top of the diagram. The gradient length of axis I is 4.8 SD, and that of axis 2 is 1.85 SD. For axis I, eigenvalue 0.62; for axis 2, eigenvalue = O.IS.

-.~.--.................. - .__..__.._----_.....•.........~.---- ..~. -~ ~---------,

SEDGE OPEN-TREED UPLANDTREED FEN FEN FEN FOREST

3

2 HOLLOW VEGETATION

,/, • ,~o

1 J @) .......­ HUMMOCK VEGETATION b7 ¥. -ra, oo +---~--. -----..--, - ---""""""T-+"'~--"-1

o 2 3 4 5

Axis 1

Table 4. Environmental parameters (±SD) for each vegetation type for the summer of 1992.

Parameter Upland forest Treed fen Open-treed fen Sedge fen Spike-rush fen

Calcium (mg/L) 2.05±0.05 2.13±1.0S 2.60±0.SS 1.90±O.3S 2.97±0.74 Conductivity 38.3±O.1O S1.9±2.90 S1.7±2.00 42.9±1.50 51.8±3.80 Mqgnesium (mg/L) O.SS±O.OS 0.97±O.4S 1.13 ±0.47 O.70±0.IS 1.20±0.16 Potassium (mg/L) 0.S2±O.27 0.72±O.49 0.53±O.40 O.35±0.27 0.40±O.21 pH 6.40±O.48 6.37±O.47 6.41 ±O.S4 6.34±0.70 6.70±O.S3 Peat depth (cm) O.O±O.OO 123.2±44.60 IS3.2±36.S0 16S.0±0.OO 16S.0±0.OO Shading (see text) 0.90±0.06 0.48±0.24 O.l8±O.OS 0.01 ±0.01 0.01 ±O.OO Sodium (mg/L) 2.05±0.3S 2.3HO.19 2.27±0.29 I.7HO.3S 1.93±0.31 Water-table depth (cm) -17.30±IS.80 -3.80±4.49 +3AO±2.30 +2.S0±2.25 +9.80±2.9S Water temperature (0C) 7.20±1.20 8.20± 1.89 11.60±3.90 13.00±S.OO 14.0±6.20 Hummock height (cm) O.OO±O.OO 4S.80± 16.6 30AO±7.70 17AO±4.70 16.30±4.S0

are only slightly lower than the eigenvalues in the uncon­strained (DCA) analysis (Table 5), indicating that the environ­mental variables included in the analysis captured much of the variance inherent in the species data (Ter Braak 1986; Ter Braak and Prentice 1988). The species-environment correlation, which provides a measure of how well chahges in environmental parameters mirror changes in species compo­sition, is high in this analysis and so re-enforces this assertion.

Species and stands have a narrow distribution on axis 2 (Figs. 4 and 5), but at about 1.25 on axis 1, there is a disper­sion of species scores on axis 2. This dispersion takes place

in the region corresponding to treed fen vegetation and signi­fies that in the treed fen, the secondary gradient of conduc­tivity and peat depth becomes more important than it is elsewhere on the fen.

Tree density was regressed against each environmental factor, and the regression lines were then tested for zero slope (i.e., a significant correlation). Four environmental components correlate significantly with tree density: depth to the water table, hummock height, peat depth, and water temperature. Table 6 presents the diagnostics from these regressions. Tree density is negatively correlated with all of

1210 Can. J. Bot. Vol. 74, 1996

Fig. 4. Canonical correspondence analysis of herb and shrub species and significant environmental factors. The vectors indicate the direction of maximum change of environmental variables, and the relative vector lengths convey the importance of the factor. The species abbreviations are given in Table 3. For axis 1, eigenvalue = 0.56; for axis 2, eigenvalue = 0.17.

2.5,-------------------------,---------------------------,

OXYFEN.

JUNCOM1.5 •

PEAT DEPTH Eau PRA. .

ELE aUI CAR DIS• CAR.CANWATER TEMPERATURE N EPI HOR ••

• MICODOIJ) 0.5 • SHADING ~ RIB

WATER TABLE~A';,TRI,

-0.5 CONDUCTIVITY

VAC$CO. -1.5 .+-----+----~--.-+---+-__+-___+--.. ,_+-_+~-+__-_+_---+--_+_--_+_--_1

-3.5 -2.5 -1.5 -0.5 0.5 1.5 2.5 3.5 Axis 1

Table 5. Diagnostics for the canonical correspondence analysis (CCA) and detrended correspondence analysis (DCA) of well-transect vegetation data.

Gradient Species -environmentEigenvalue Total inertia length correlation

Axis DCA CCA DCA CCA DCA CCA rs

1 0.620 0.560 4.425 0.958 1.917 0.979 0.914 2 0.175 0.169 1.761 0.792 0.246

Note: Gradient lengths are in units of SDs and are only provided in DCA. The species-environment correlation measures how well changes in environmental parameters reflect changes in species composition. Total inertia is the sum of all the eigenvalues in an ordination and therefore does not correspond to a single axis. r, is Spearman's rank correlation of the DCA and CCA stand scores.

I

Table 6. Diagnostics for the regression of tree densities (all Discussion species) against significant environmental factors. The open-canopied portions of Spring Fen have a vegetation

typical of other subalpine fens, but the forested segments Environmental factor Regression formula N have a canopy composition more reminiscent of upland

Water temperature y = -0.094x + 1.30 0.61 (0.003) 12 spruce-fir forest. This unusual situation has not been previ­Peat depth Y = -O.OORx + 1.42 0.77 (0.001) 10 ously described, and questions concerning possible causal Depth to water table y = -0.027x + 0.27 0.58 (0.004) 12 factors of the vegetational composition and its patterns of Hummock height Y = 0.002 x 1.144X 0.68 (0.001) 10 zonation warrant discussion.

Several gradients influence the distribution of species on Spring Fen; these were summarized using both direct and

the environmental variables except hummock height. Addi­ indirect gradient analysis. Two primary components of the tionally, the relationship between tree density and each axis 1 gradient are water temperature and shading (Figs. 4 environmental variable is linear, again with the exception of and 5). These factors are highly negatively correlated, non­hummock height, for which it is exponential. orthogonal (Palmer 1994), and are functionally interdependent.

1211 Johnson

Fig. 5. Canonical correspondence analysis diagram showing placement of stands with regard to significant environmental factors. The symbols denote the vegetation type in which the quadrat was located: ., spike-rush fen; +, sedge fen; ., open-treed fen; ... , treed fen. Open symbols indicate the average score from stands within a given vegetation type. For axis 1, eigenvalue 0.56; for axis 2, eigenvalue = 0.17.

3,----------------,----------------,

2

&+ water eat Depth Shading

C'\I

'xtfJ

& «

D.0 + water Table

Depth Hummock & & Height-1

&

• -2~----,-----r---~----_,----_.----~

-3 -2 -1 o 2 3 Axis 1

This suggests that at least part of this gradient is a reflection of their interaction. That is, the water-temperature vector may simply be the negative extension of the shading vector (or vice versa) rather than an additional biologically important factor. Shade is known to be an important regulator of species distri­butions in wetlands, as well as in other ecosystems (e.g., Harper 1977; Vitt et al. 1990), and the strong change in the level of shading more than likely affects species composition.

The effects of temperature on the growth and survival of acclimated plants are less well known. Running and Reid (1980) found there to be an abrupt change in the water relations of Picea engelmannii at 7 - 8°C, and this threshold is found within the temperature gradient on Spring Fen. However, Running and Reid's (1980) study focused on water uptake, and (lack of) water is not a limiting factor on the fen, although processes such as nutrient uptake could be similarly affected. It is, therefore, unclear what role water tempera­ture plays in the determination of the species composition of Spring Fen.

The other factors correlated with axis 1 are depth to the water table and peat depth (Figs. 4 and 5). Numerous studies found that depth to water is frequently the most impor­tant environmental factor in wetlands (e.g., Gorham 1957; Heinselman 1963; Slack et al. 1980; Mitsch and Gosselink 1993; Jean and Bouchard 1993), and this gradient is probably biologically important on Spring Fen. Peat depth, however, was variously shown to be highly, moderately, or uncorre­lated with site "quality" or vegetation (Maclean and Bedell 1955; Jean and Bouchard 1993; leglum and He 1995; Heinselman 1963; Averell and McGrew 1929). In the case of Spring Fen, peat depths are generally deeper than the root zone of plants, so rooting through the peat and into richer

mip.eral soil can be ruled out as a major factor causing the differential growth and survival of individuals. However, peat depth may be biologically important in another way. Various workers showed that water becomes less ion rich as it passes through peat (e.g., Gorham 1957; Moore and Bellamy 1974) and there may be evidence of this on Spring Fen. As water flows from the treed and open-treed fen areas on the peatland (where the springs emerge) ~t onto the sedge fen, a decrease in ion concentration is commonly found (Table 4). This decrease is not found in spike-rush fen, perhaps because the high flow volume prevents much of the filtering by peat. This decrease may also be caused in part by the uptake of these ions by the vegetation.

The way in which peat hummocks affect the species composition of Spring Fen and the way in which vegetation affects the genesis and growth of these hummocks is complex, and general ideas about this were reviewed in the discussion. On Spring Fen, microtopography clearly exerts a strong influence on species composition (Fig. 3), i.e., hummocks are favorable microsites for tree seedlings and other meso­phytic plants that otherwise do not occur on the fen. Although there were recent reports on the distortive qualities of DCA (Palmer 1994), in this case the technique seems to have performed well, and distinctions, visually evident in the field, are also shown in the DCA. The strength of vegetation­topography association is further demonstrated by a regres­sion of stand scores versus average hummock height within vegetation zones that yielded an R2 = 0.85 (not shown). The relationship of vegetation with microtopography was observed in many other peatland studies (e.g., Sjors 1959, 1%la, 1961b; Heinselman 1969, 1970; Vitt et al. 1975; Glaser et al. 1981; Eurola 1984). Hummocks seem to be especially important to tree seedlings. On Spring Fen, virtually no tree seedlings grew directly out of the peat substrate; rather, they were always found on hummock tops (J .B. lohnson, personal observation). Sjors (1 950a) made a similar observation for Canadian peatlands,

The relationship between average hummock height in the different parts of the fen and tree density is an exponential function and positive, suggesting that there is a threshold of hummock height below which hummocks do not improve tree survivorship. I observed that as Sphagnum hummocks grow up the stems of Picea engelmannii, the trees produce adventi­tious roots that penetrate the hummock mass. This is similar to Boogie's (1972) observations of Pinus contorta seedlings. Such adventitious growth places roots in an aerobic environ­ment where metabolism can occur throughout the growing season. The growth of adventitious roots, although not syste­matically sampled for, was not observed on Abies lasiocarpa and could partially explain why it did not grow on the wetter portions of the fen. When Abies lasiocarpa was present on the fen, it grew only on the tallest, driest hummocks.

Hummocks may also affect a plant'S immediate environ­ment by inhibiting direct contact between roots and minerally rich groundwater. And even though surface water is readily drawn up into hummocks by capillary action, the high cation exchange ability of the Sphagnum makes a large proportion of the cations present unavailable to the plants (Gorham 1957). Furthermore, the cation exchange also lowers hummock pH, so they are acidic relative to the surrounding hollows. Through

1212

these interactions the hummock environment becomes similar to that of an ombrotrophic bog, and while no expansive bogs have yet been located in the Rocky Mountains, it is asserted that hummocks can act as miniature bog analogs on these fens. Similar situations were observed by several other workers (Bellamy and Rieley 1967; Karlin and Bliss 1984; Zoltai and Johnson 1985). Because of the great environmental differ­ence between hummocks and hollows, environmental, and therefore floristic, heterogeneity is greatly increased. The development of hummock - hollow topography is also thought to playa key role in the development of the forest vegetation on Spring Fen.

While Sphagnum hummocks are beneficial to trees and other mesophytes, the shade provided by these plants is in turn favorable for Sphagnum growth because of a reduction of evapotranspiration (Moore and Bellamy 1974; Cooper 1990). In areas with high precipitation, Sphagnum may grow well in full sunlight, but in the mesic southern Rocky Moun­tains, the shade provided by woody vegetation seems quite important for moss growth.

Because of their exacting environmental requirements, fens containing these miniature bogs are less common than typical sedge fens, but their narrow environmental range may pro­vide a guide to the location of other such peatlands. These fens are located at relatively high altitudes, above approxi­mately 2900 m. For hummock building to occur, a compen­satory mechanism for the dry continental climate needs to be provided; in this case, a number of perennial springs are present. Lastly, this type of fen needs to be situated in rela­tively narrow valleys so it can receive large amounts of surface runoff. A number of fens similar to Spring Fen are located in areas with these conditions, although none as large. So while large-scale factors such as regional climate and geomorphic position provide the overriding regulators of species composition, small-scale heterogeneity and autogenetic processes also have a marked effect on community composi­tion and allow species that normally inhabit mesic areas to grow on this peatland.

Acknowledgment.

I am indebted to R.L. Dix and T. Cottrell for their assistance during every stage of this project. D. Steingraeber, D. Cooper, H. Sjors, R.K. Peet, and two anonymous reviewers provided many valuable commen,ts. Audrey Oberlin provided inspira­tion and much support, and T. Gerhardt assisted in field work. This project was partially supported by M. and S. Derynck.

Aeterence.

Allen, R.B., and Peet, R.K. 1990. Gradient analysis of the forests of the Sangre de Cristo Range, Colorado. Can. J. Bot. 68: 193-201.

Averell, J.L., and McGrew, P.C. 1929. The reaction of swamp forests to drainage in N Minnesota. Minnesota Department of Drainage and Waters, St. Paul, Minn.

Baker, F.S. 1944. Mountain climates of the western United States. Ecol. Monogr. 14: 223-254.

Bellamy, D.J., and Rieley, 1. 1967. Some ecological statistics of a "miniature bog." Oikos, 18: 33 -40.

Bierly, K.F. 1972. Meadow and fen vegetation in Big Meadows, Rocky Mountain National Park. M.Sc. thesis, Department of Botany, Colorado State University, Fort Collins, Colo.

Can. J. Bot. Vol. 74, 1996

Boogie, R. 1972. The effect of water-table height on root develop­ment of Pinus contorta on deep peat in Scotland. Oikos, 23: 304-312.

Boogie, R., and Miller, H.G. 1976. Growth of Pinus contorta at different water-table levels in deep blanket peat. Forestry, 49: 123-131.

Braun-Blanquet, J. 1932. Plant sociology; the study of plant com­munities. McGraw-Hill Publishing Co., New York.

Cajander, A.K. 1913. Studien tiber die Moore Finnlands. Acta For. Fenn.2.

Chee, W., and Vitt, D.H. 1989. The vegetation, surface chemistry and peat chemistry of moderate-rich fens in Central Alberta, Canada. Wetlands, 9: 227-261.

Cooper, D.J. 1986. Community structure and classification of Rocky Mountain wetland ecosystems. In An ecological charac­terization of Rocky Mountain montane and subalpine wetlands. Edited by C.A. Segelquist. U.S. Fish Wildl. Servo BioI. Rep. 86(11): 66-147.

Cooper, D.J. 1990. Ectdogy of wetlands in Big Meadows, Rocky Mountain National Park, Colorado. U.S. Fish Wildl. Servo BioI. Rep. 90(15).

Cooper, D.J. 1994. Patterns of vegetation and water chemistry in peatlands of the west-central Wind River Range, Wyoming, U.S.A. Can. J. Bot. 72: 1586-1597.

Cottam, G., and Curtis, J.T. 1956. The use of distance measures in phytosociological sampling. Ecology, 37: 451-460.

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

Du Rietz, G.E. 1949. Huvudenheter och huvudgriinser i svensk myrvegetation. Sven. Bot. Tidskr. 43: 274-309.

Eurola, C. 1984. Finnish mires. In European mires. Edited by Peter Moore. Academic Press, New York.

Gauch, H.G., Jr. 1982. Multivariate analysis in community ecology. Cambridge University Press, Cambridge, U.K.

Glaser, P.H., Wheeler, G., Gorham, E., and Wright, H., Jr. 1981. The patterned mires of the Red Lake peatland, northern Minnesota: vegetation, water chemistry and landforms. J. Ecol. 69: 575 - 599.

Gore, A.J.P. 1983. Introduction. In Mires: swamp, bog, fen and moor. Edited by A.J.P. Gore. Ecosyst. World, 4a: 1-34.

Gorham, E. 1957. The development of peatlands. Q. Rev. BioI. 32: 145 166.

Gorham, E. 1967. Some chemical aspects of wetland ecology. Annu. Muskeg Reg. Conf. Proc. 12: 20-38.

Greig-Smith, P. 1983. Quantitative plant ecology. University of California Press, Berkeley, Calif.

Harper, J.L. 1977. The population biology of plants. Academic Press, New York.

Heinselman, M.L. 1963. Forest processes and peatland types in the glacial lake Agassiz region, Minnesota. Ecol. Monogr. 33: 327-371.

Heinselman, M.L. 1970. Landscape evolution, peatland types, and the environment in the Lake Agassiz peatlands natural area, Minnesota. Ecol. Monogr. 40: 235 -261.

Jean, M., and Bouchard, A. 1993. Riverine wetland vegetation: importance of small-scale and large-scale environmental varia­tion. J. Veg. Sci. 4: 609-620.

Jeglum, J.K. 1987. The use of TWINSPAN, tabular analysis and Finnish/Swedish concepts in classifying wooded peatlands in Ontario. In Proceedings of Symposium 87, Wetlands/Peatlands, Edmonton, Alta., 23-27 August 1987. CompiledbyC.D.A. Rubcc and R.P. Overend. The Canadian Society for Peat and Peatlands, Dartmouth, N.S. pp. 383-391.

Jeglum, J.K., and He, F. 1995. Pattern and vegetation: environment relationships in a boreal forested wetland in northeastern Ontario. Can. J. Bot. 73: 629-637.

Karlin, E.F., and Bliss, L.C. 1984. Variation in substrate chemistry along microtopographical and water-chemistry gradients in peat­lands. Can. 1. Bot. 62: 142 153.

1213 Johnson

Kartesz, J.T. 1994. A synonymized checklist of the vascular flora of the United States, Canada, and Greenland. Timber Press, Portland, Oreg.

Lindsey, A.A. 1955. Testing the line-strip method against full tallies in diverse forest types. Ecology, 36: 484 -494.

Maclean, D.W., and Bedell, G.H.W. 1955. Northern clay belt growth and yield survey. Can. Dep. North. Affairs Natl. Resour. For. Branch Res. Div. Tech. Note No. 20.

Marr, J.W. 1961. Ecosystems of the front range in Colorado. Univ. Colo. Stud. Ser. Bio!. 8.

Mitsch, W.1., and Gosselink, J.G. 1993. Wetlands. Van Nastrand Rienhold Co., New York.

Moore, P.D., and Bellamy, D.J. 1974. Peatlands. Springer-Verlag, New York.

Neiland, B.J. 1971. The forest-bog complex of south-east Alaska. Vegetatio, 22: 1-64.

Palmer, M.W. 1994. Putting things in even better order: the advantages of canonical correspondence analysis. Ecology, 74: 2215 - 2230.

Peet, R.K. 1981. Forest vegetation of the Colorado Front Range. Vegetatio, 45: 3-75.

Prentice, H.C., and Cramer, W. 1990. The plant community as a niche bioassay: environmental correlates of local variation in GypsophilaJastigiata.1. Eco!. 78: 313-325.

Richmond, G.M. 1974. Raising the roof of the Rockies. Rocky Mountain Nature Association, Inc., Estes Park, Colo.

Rink, L.P., and Kiladis, G.N. 1986. Geology, hydrology, climate, and soils ofthe Rocky Mountains. In An ecological characterization of Rocky Mountain montane and subalpine wetlands. Edited by C.A. Segelquist. U.S. Fish Wild!. Servo Bio!. Rep. 86(11): 42 -65.

Running, S.W., and Reid, C.P. 1980. Soil temperature influences on root resistance of Pinus contorta seedlings. Plant Physio!. 65: 635-640.

Sjors, H. 1 950a. Regional studies in North Swedish mire vegetation. Bot. Not. 1950: 173-222.

Sjors, H. 1950b. On the relation between vegetation and electrolytes in north Swedish mire waters. Oikos, 2: 241-258.

Sjors, H. 1959. Bogs and fens in the Hudson Bay lowlands. Arctic, 12: 2-19.

Sjors, H. 1961a. Forest and peatland at Hawley Lake, northern Ontario. Nat!. Mus. Can. Bul!. 171: 1-31.

Sjors, H. 1961b. Surface patterns in boreal peatlands. Endeavour (Eng!. Ed.), 80: 217-224.

Sjors, H. 1963. Bogs and fens on Attawapiskat River, northern Ontario. Bull. Nat!. Mus. Can. 186: 45-133.

Sjors, H. 1965. Northern mires. In The plant cover of Sweden. Acta Phytogeogr. Suec. 50: 180-188.

Sjors, H. 1983. A comparison between mires of southern Alaska and Fennoscandia. Aquilo Ser Bot. 21: 89-94.

Slack, N.G., Vitt, D.H., and Horton, D.G. 1980. Vegetation gradients of minerotrophically rich fens in western Alberta. Can. J. Bot. 58: 330-350.

Ter Braak, C.1.F. 1986. Canonical correspondence analysis: a new eigenvector technique for multivariate direct gradient analysis. Ecology, 67: 1167 -1179.

TerBraak, C.J.F. 1990. CANOCO: a FORTRAN program for canon­ical community ordination by [partial] [detrended] correspondence analysis, principal components analysis and redundancy analysis (version 3.10). Agricultural Mathematics Group, Wageningen, The Netherlands.

Ter Braak, c.1.F., and Prentice, C. 1988. A theory of gradient analysis. In Advances in ecological research. Edited by M. Begon, A.H. Fitter, E.D. Ford, and A. Macfadyen. Academic Press, New York. pp. 271-317.

Vitt, D.H., and Slack, N.G. 1975. An analysis of the vegetation of Sphagnum-dominated kettle-hole bogs in relation to environmental gradients. Can. J. Bot. 53: 332-359.

Vitt, D.H., Achuff, P., and Andrus, R.E. 1975. The vegetation and chemical properties of patterned fens in the Swan Hills, north central Alberta. Can. J. Bot. 53: 2776-2795.

Vitt, D.H., Horton, D.G., Slack, N.G., and Maimer, N. 1990. Sphagnum-dominated peatlands of the hyperoceanic British Columbia coast: patterns in surface water chemistry and vegeta­tion. Can. J. For. Res. 20: 696-711.

Waughman, G.1. 1980. Chemical aspects of the ecology of south German peatlands. 1. Eco!. 68: 1025 - 1046.

Weber, C.A. 1911. Das Moore. Hannoversche Geschichtsblatter, Hanover, Germany.

Weber, W.A. 1987. Colorado flora, western slope. Colorado Asso­ciated University Press, Boulder, Colo.

Whittaker, R.H. 1967. Gradient analysis of vegetation. Bio!. Rev. Camb. Philos. Soc. 49: 207 - 264.

Zoltai, S.C., and Johnson, J.D. 1985. Development ofa treed bog island in a minerotrophic fen. Can. 1. Bot. 63: 1076-1085.