The Forest Communities of Mount Rainier National Park
NPS Logo


Many of the floristic and environmental relationships between the plant communities have already been outlined in the introduction to the classification and discussions of the individual types. This section addresses more systematically both environmental and floristic patterns. The topographic-elevational relationships, as they vary around Mount Rainier, will be considered first. Then the details and adequacy of the classifications will be examined from the perspectives of similarity, discriminant, and principal component analyses.

Topographic-Elevational Patterns

The forest patterns at Mount Rainier are believed to be largely governed by moisture and temperature gradients (see Fig. 11). Moisture variations appear important at low elevations, and complex temperature and snowpack gradients are associated with higher elevations. Because of the substantial climatic and topographic variability in different sectors around the Mount Rainier cone, we constructed a generalized topographic-elevational pattern for each of the major drainage systems (Figs. 34 to 37).

Each of these diagrams (Figs. 34 to 37) presents generalized, simplified forest patterns reduced from far more complex mosaics actually occurring in the landscape. At this scale of generalization, we have overlooked microrelief features of dissected, geomorphically active, local landforms which produce intricate variations of vegetation distribution, including gradations from one forest type to another and inter-fingering of distinct types on uneven slopes and drainages. Slope effects can be envisioned as diagonal boundaries between habitat types as suggested in the caption of Figure 11. We have also generalized through the complex age patterns of forest distributions. The contribution of the complex soils and their nutrient characteristics to the forest composition and structure is unknown. In addition, we can only speculate about the influence of other phenomena, such as dewpoint condensation and cloud cover, on the distribution of tree species from one drainage to the next. We can suggest from our principal component analysis that the topographic-environmental patterns of Figure 34 account for only about 30 to 40 percent of the variation of species distributions on the slopes of Mount Rainier; the remainder must be attributable to microsite, disturbance, and historical factors.

The Ohanapecosh River drainage (Fig. 34) is situated within a comparatively dry and warm sector of the Park (see Fig. 2). At lower elevations here, forests of TSHE/ACTR and TSHE/GASH are common on slopes and benches; they do not occur in appreciable amounts in the other drainages. The complex patterns at higher elevations reflect the forest-topographic relations of the Cowlitz Divide and upper Panther Creek and Laughingwater drainages.

Figure 34. Generalized distribution of forest habitat types in the Ohanapecosh drainage. The horizontal axis depicts a generalized topographic moisture gradient from wet river valleys (left) to dry ridgetops (right). The shape of the overall forested area is determined by topographic features within the Ohanapecosh watershed. Numbers between adjoining habitat types are mean similarities (as percents) suggesting the degree of floristic retationship.

At the other climatic extreme are the watersheds in the wetter, northwestern sector of the Park (Fig. 35). The forest patterns of the Carbon, Mowich, and Puyallup River drainages differ from the Ohanapecosh in major ways. At low elevations on lower slopes, draws, and benches, the TSHE/POMU Association is common; the Abies amabilis phase of this type is found at higher elevations or along north-facing lower slopes adjacent to the normal phase. Open or upper slopes between 1050 and 1200 m (3,500 to 4,000 ft) elevation are often forested with examples of the ABAM/GASH type; comparable landforms in the Ohanapecosh drainage are within the ABAM/BENE or ABAM/TIUN types. In the Tsuga mertensiana Zone, ABAM/RHAL, ABAM/MEFE, and ABAM/XETE types seem to occur more widely; these types are restricted in the Ohanapecosh where ABAM/TIUN and ABAM/RULA types sometimes occupy comparable elevation-topographic positions. The Rubus pedatus phase of the ABAM/VAAL Association is suggested as a high elevation, wetter, cooler environmental variation of this type, and is essentially absent in other sectors of the Park.

Figure 35. Generalized distribution of forest habitat types on northwestern drainages (Carbon, Mowich, Puyallup) of Mount Rainier. The horizontal axis depicts a generalized topographic moisture gradient from wet river valleys (left) to dry ridgetops (right). Numbers between adjoining types suggest the degree of floristic similarity based upon mean percent similarity.

The Nisqually forest patterns are intermediate between the Ohanapecosh and Carbon sectors (Fig. 36). At 900 to 1050 m (3,000 to 3,500 ft), ABAM/BENE forests are common on slopes and ridges; these adjoin ABAM/XETE forests of upper slopes or ridges at higher elevation. Stands of ABAM/VAAL and ABAM/TIUN occupy the modal microenvironments of mesic soils and moderate thermal regimes. Stands of ABAM/OPHO and ABAM/VAAL occupy valley floors, toe slopes, or lower elevation benches. Considerable topographic overlap occurs in the distribution of ABAM/RULA and ABAM/RHAL Associations at higher elevations. Clearly, soil drainage, snowpack, and other environmental features of microsites are also important in the distribution of forest habitats (Long 1976).

Figure 36. Generalized distribution of the Nisqually drainage of Mount Rainier National Park. The horizontal axis depicts a generalized topographic moisture gradient from wet river valleys (left) to dry ridgetops (right). The shape of the overall forested area is determined by topographic features within the Nisqually watershed. Numbers between adjoining habitat types are mean similarities (as percents) suggesting the degree of floristic relationship.

The White River drainage is the most continental of the forest climates in Mount Rainier (Fig. 37). Forests with high proportions of Picea engelmannii may be found at the upper elevations of glacial valleys, and Pinus contorta occurs in periglacial environments of moraines and rockfields around 1350 m (4,500 ft) elevation. Forests of Abies lasiocarpa are conspicuous at high elevations on warm, dry margins of a subalpine forest climate. At midelevations, 900 to 1200 m (3,000 to 4,000 ft), the sequence from wet to dry environments along a topographic moisture gradient is ABAM/OPHO, ABAM/VAAL, and ABAM/BENE. This is generally comparable to forest distribution in the Ohanapecosh and Nisqually River drainages. The wet, warm TSHE/OPHO forest is restricted to lower valleys of the White River in vicinity of the Park boundary.

Figure 37. Generalized distribution of forest habitat types in the drainages of the White River. The horizontal axis depicts a generalized topographic moisture gradient from wet river valleys (left) to dry ridgetops (right). The shape of the overall forested area is determined by topographic features within the White watershed. Numbers between adjoining habitat types are mean similarities (as percents) suggesting the degree of floristic relationship.

Classification Insights from Similarity, Discriminant, and Principal Component Analyses

Classifying the forests at Mount Rainier is not an easy task. Within a local area the types are sometimes sharply defined with abrupt ecotones. As abstractions, the types we have recognized represent distinctive compositional (including relative species importance) nodes. Both environmental and floristic gradients are typically continuous, however, and the involved forest flora is composed mainly of species with broad ecological amplitudes rather than species of high fidelity to limited environmental conditions. The complexity of the mountain environment at Mount Rainier, with the presence of many highly limited and individualistic site conditions within a limited area, further complicates the job of classification. Under these circumstances, gradual transitions from one type to another are encountered in field situations, and plots intermediate between types do occur. We say this not by way of an apology for the classification but, rather, so that the user is aware of its limitations.

We have used various statistical and analytical procedures in constructing this classification system and can provide some insight into its general validity. These include: (1) similarity analyses, which can show the degree to which types are related; (2) discriminant analyses, which can be used to reexamine classification of old plots as well as assign new plots to types; and (3) principal component analyses, which can be used to examine how individual species and groups of species respond to complex gradients and, in turn, see how these relate to the typal groupings and hypothesized environmental gradients.

Similarity Analysis

We used similarity analysis to develop this classification (explained in step 4 of the "Data Analysis" section in Chapter 4); abstracted results of one analysis at the level of forest type comparisons (contrasted with analysis of individual plots) are shown in Table 11. In constructing this table we recognized that similarity values are conditioned by whatever algorithm is used to compare forest types (for example, Kelsey et al. 1977). These values also depend on the choice of plant species whose average cover or density was used to compute the percentage of similarity. Our similarity values ranged from 12 to 65 percent. Forest types are extremely similar, we feel, for similarities greater than 55 percent; they are highly similar between 47 and 55 percent, and similar in the range of 40-56 percent. These similarity classes are based on forest type comparisons involving the first 400 plots.

The results of Table 11 generally support the discussion of related forest types included in earlier presentations. This table highlights the floristically similar forests whose significance, we believe, is that they represent intergrading environments along complex moisture, temperature, nutrient, or successional gradients (Dyrness et al. 1974, Zobel et al. 1976). The following comparisons are noteworthy:

At lower elevations, the Tsuga heterophylla/Gaultheria shallon Association has the greatest overall similarity to other forest types, especially Tsuga heterophylla/Achlys triphylla, Pseudotsuga menziesii/Viola sempervirens, Abies amabilis/Gaultheria shallon, Abies amabilis/Berberis nervosa, and Abies amabilis/Vaccinium alaskaense Associations. As shown in the topographic-elevational patterns (Figs. 34 to 37), several of the associations highly similar to Tsuga heterophylla/Gaultheria shallon are ecotonal or found on topographically identical positions of the landscape in different sectors of the Park.

At intermediate elevations, the Berberis nervosa phase of Abies amabilis/Vaccinium alaskaense Association bears the greatest overall similarity to most of the other community types or associations and is extremely similar to Abies amabilis/Berberis nervosa Association. This underscores our belief that Abies amabilis/Vaccinium alaskaense is the vegetation of modal environments.

At high elevations, several associations are very difficult to separate floristically. For those species chosen to make similarity comparisons, the Abies amabilis/Menziesia ferruginea Association is extremely similar to the Rubus pedatus phase of Abies amabilis/Vaccinium alaskaense and to the Abies amabilis/Rhododendron albiflorum Associations. These associations, as well as the Erythronium phase of Abies amabilis/Rubus lasiococcus and Chamaecyparis nootkatensis/Vaccinium ovalifolium Associations, doubtless represent subtly intergrading environments of high snow pack and short growing seasons.

Seral community types may show little similarity to the more mature associations. The Alnus rubra/Rubus spectabilis and Abies lasiocarpa/Valeriana sitchensis community types are especially distinctive. These young forests can be nearly monocultures of an early seral tree dominant as well as having pronounced dominance by understory plants (such as Pteridium aquilinium) that decline as the sere develops. The three Pseudotsuga menziesii community types are floristically similar to each other, but only Pseudotsuga menziesii/Xerophyllum tenax has similarity with some of the other associations.

Table 11. Classes of percent similarity between all forest types in Mount Rainier National Park.1

Forest type2
Forest type2




1ab34abababcd8 910111213abcab161718ab

2a TSHE/POMU, TSHE phase2*
2b TSHE/POMU, ABAM phase25*
4 ALRU/RUSP2113*
5a ABAM/OPHO, valley phase23353*
5b ABAM/OPHO, slope phase232545*
6a ABAM/TIUN, climax phase3213345*
6b ABAM/TIUN, seral phase41122345*
7a ABAM/VAAL, VAAL phase2111
7b ABAM/VAAL, BENE phase3331
7c ABAM/VAAL, RUPE phase
7d ABAM/VAAL, CHNO phase1


10 PSME/XETE211111
11 PSME/VISE5232232332511333*


13 ABAM/BENE533223233351142352*
14a ABAM/XETE, TSHE phase111

14b ABAM/XETE, TSME phase

14c ABAM/XETE, seral phase1

15a ABAM/RULA, RULA phase1111122332122113212243*
15b ABAM/RULA, ERMO phase



18 CHNO/VAOV1111134533143
19a ABAM/MEFE, climax phase

19b ABAM/MEFE, seral phase


1Percent similarity (Gauch 1982) computed using relative average cover of all herb and shrub species within each forest type. Seven classes are defined as follows: blank <10%, 10% <- 1 <20%, 20% <- 2 < 30%, 30% <-3 <40%, 40% <-4 < 50%, 50% <-5 < 100%, * 100%.

2Forest type numbers and acronyms correspond with those in Table 1.

3ABAM = Abies amabilis, ABLA2 = Abies lasiocarpa, ACTR = Achlys triphylla, ALRU = Alnus rubra, BENE = Berberis nervosa, CEVE = Ceanothus velutinus, CHNO = Chamaecyparis nootkatensis, ERMO = Erythronium montanum, GASH = Gaultheria shallon, MEFE = Menziesia ferruginea, OPHO = Oplopanax horridum, POMU = Polystichum munitum, PSME = Pseudotsuga menziesii, RULA = Rubus lasiococcus, RUPE = Rubus pedatus, RUSP = Rubus spectabilis, TIUN = Tiarella unifoliata, TSHE = Tsuga heterophylla, TSME = Tsuga mertensiana, VAAL = Vaccinium alaskaense, VAOV = Vaccinium ovalifolium, VASI = Valeriana sitchensis, VISE = Viola sempervirens, XETE = Xerophyllum tenax.

Discriminant Analysis

The discriminant classification, based on 39 discriminant variables, is summarized in Table 12. The 19 groups in the left-most column are the defined habitat and community types. The ecological and floristic characteristics used in defining these 19 forest types were not necessarily the same variables used in computing the discriminant functions. Variables employed in discriminant analysis were chosen from important species considering forest vegetation over the entire forest region, but those used to resolve the 19 forest groups were selected from important species within each of four broad environmental groups (moist, modal, dry, and cold) at Mount Rainier. Therefore, the groups defined by discriminant analysis are less sharply resolved in cases where an important classificatory variable within a subgroup is not included in the analysis. On the other hand, zonal differences between vegetation groups may be more clearly separated in the discriminant classification.

An interesting sidelight of the discriminant analysis illustrates one problem of defining vegetation groups. The set of species chosen as variables had low coverage or density values in several of the Abies amabilis/Berberis nervosa plots used in the analysis. Consequently, plots from distinctly dissimilar vegetation types were often incorrectly classified as an Abies amabilis/Berberis nervosa Association if they had low total shrub and herb coverage.

The circular analytic pathway is another consideration for interpreting the discriminant matrix (Table 12). In effect, the computer is told what the initial groups are (left column of Table 12); it then computes optimally discriminating functions, and in a second pass over the plot data reassigns each plot to a group. No matter how poorly or skillfully the initial group classification was made, this method assures that a certain proportion of the forest plots are "correctly" assigned to the original groups. It is conceivable that by careful choice of discriminating variables, all or most of the plots could be correctly classified no matter how valid the a priori classification of groups.

Each of the 19 numbered columns in Table 12 represents a forest group used to calculate discriminant functions by the procedure described by Nie et al. (1975). Plots "correctly" classified are tallied in the main diagonal; those assigned to other than their initial group are tallied in the appropriate off-diagonal column. The proportion of correctly reassigned plots is shown in the right-hand column of Table 12.

One of the benefits of using discriminant analysis is its computational ability to screen the classified plots for possible misclassification. Misclassified plots will appear off the diagonal in such an analysis. However, plots off the diagonal in our analysis are not necessarily misclassified because we used less information to discriminate the groupings than was used to resolve the initial forest groups. To decide whether or not any off-diagonal plot has been misclassified requires examination of the entire plot data (including environmental and successional information) and exercise of ecological judgment. Of all our plots, 25 percent were off diagonal. About 10 percent of those were either misclassified in the initial forest groupings or were intergrades between two groups, thus having about equal probability of assignment to either group.

Table 12 suggests floristic affinities between the forest groups based on the 39 variables used in the discrimination. Groups 5a and 5b (Abies amabilis/Oplopanax horridum, valley and slope phases, respectively) have floristic similarity, for example, because off-diagonal plots occur within both groups. We feel, however, that this expression of floristic relationship is less persuasive than actual similarity measures between the different forest types shown, for example, in Table 11.

For a discussion of the logic, assumptions, and algorithms used in discriminant analysis, see either Nie et al. (1975) or Cooley and Lohnes (1971).

Principal Component Analysis

Principal component analysis (PCA) was applied to plots within each of the four major environmental groupings (step 3 of the "Data Analysis" section, Chapter 4). As a technique of indirect ordination, interpretation of PCA focuses on the meaning of whatever environmental gradients are represented by the component axes. Since most species have curvilinear distributions along complex environmental gradients, only those species whose distributions can be approximated linearly will show high "factor loadings." Tables 13 to 16 show these loadings within each of the four environmental groupings.

Table 12. Summary of discriminant analysis on forest types in Mount Rainier National Park1

Forest type2Forest type2
of plots
1235a5b6789101112 13141516171819



















































1Plots classified "correctly" according to the analysis (total 75 percent) are on the main diagonal; "incorrectly" classified plots are off the diagonal.

2Names and numbers of the types correspond with those in Table 1.

3TSHE/ACTR = Tsuga heteraphylla/Achlys triphylla, TSHE/POMU = Tsuga heterophylla/Polystchum munitum, TSHE/OPHO = Tsuga heterophylla/Oplopanax horridum, ABAM/OPHO = Abies amabilis/Oplopanax horridum, ABAM/TIUN = Abies amabilis/Tiarella unifoliata, ABAM/VAAL = Abies amabilis/ Vaccinium alaskaense, TSHE/GASH = Tsuga heterophylla/Gaultheria shallon, PSME/CEVE = Pseudotsuga menziesii/Ceanothus velutinus. PSME/XETE = Pseudotsuga menziesii/Xerophyllum tenax, PSME/VISE = Pseudotsuga menziesii/Viola sempervirens, ABAM/GASH = Abies amabilis/ Gaultheria shallon, ABAM/BENE = Abies amabilis/Berberis nervosa, ABAM/XETE = Abies amabilis/Xerophyllum tenax, ABAM/RULA = Abies amabilis/Rubus lasiococcus, ABLA2/VASI = Abies lasiocarpa/Valeriana sitchensis, ABAM/RHAL = Abies amabilis/Rhododendron albiflorum, CHNO/VAOV = Chamaecyparis nootkatensis/Vaccinium ovalifolium, ABAM/MEFE = Abies amabilis/Menziesia ferruginea.

As a classification procedure, however, we used PCA to identify sets of species responding in approximately the same quasi-linear manner to the environmental gradients reflected in the component axes. These sets are suggested as ecological groupings under the assumption that each component axis does, in fact, reflect some kind of complex environmental gradient along which various species may be distributed according to their particular tolerances and competitive abilities. If, for example, a species is indifferent to the environmental factors reflected in a particular PCA component, or if it has a marked curvilinear response. then its factor loading will be small. Should a species respond positively and more or less linearly to environmental factors reflected by the component axis, it will have a relatively high factor loading. Conversely, a species having the opposite and more or less linear response will have a high negative factor loading. Species with nearly similar ecological tolerances and similar distributions along environmental gradients should have about the same factor loadings.

In discussing the results of PCA (Tables 13 to 16), we identify species which may be ecologically similar and see how these might coincide with species assemblages characteristic of the forest types. Whenever possible, principal components are also tentatively interpreted as environmental gradients.

Cold or High-Elevation Forests—The four components of Table 13 collectively account for 34 percent of the variation in the R matrix. The first component accounts for 12 percent of this variation. This component reflects some environmental complex along which shrubs Vaccinium ovalifolium and Menziesia ferruginea respond positively, whereas mesic herbs, often of subalpine parkland affinity, respond negatively, as does Abies lasiocarpa. The second component accounts for 9 percent of the variation in the R matrix and seems to reveal a "wetness" gradient along which herbs such as Tiarella unifoliata, Trautvettaria, Gymnocarpium, et al. have positive response. Responding in an opposite way to this component is Vaccinium membranaceum. The third component, with 7 percent of the variation in the R matrix, might be interpreted as an axis of temperature or snowpack duration. Rhododendron albiflorum, Vaccinium membranaceum, and Erythronium montanum respond positively, whereas Xerophyllum tenax and Chimaphila umbellata exhibit opposite responses. The fourth component, with 6 percent of the R matrix variation, is some kind of gradient affecting shrub cover, for Vaccinium alaskaense and V. ovalifolium and Menziesia ferruginea all have positive factor loadings.

The species in Table 13 can be grouped by their similar patterns in each of the four PCA components:

1. Abies lasiocarpa, Rubus lasiococcus, Valeriana sitchensis, Luzula glabrata, Arnica latifolia, Clintonia uniflora, and Vaccinium deliciosum*;

2. Menziesia ferruginea and Vaccinium ovalifolium;

3. Mature Chamaecyparis nootkatensis, Tiarella unifoliata, Trautvettaria grandis, Viola glabella, Gymnocarpium dryopteris, Osmorhiza spp., Streptopus roseus, and Rubus pedatus;

4. Mature Tsuga heterophylla, Xerophyllum tenax, and Chimaphila umbellata;

5. Young Tsuga mertensiana, Rhododendron albiflorum, and Erythronium montanum;

6. Vaccinium alaskaense; and

7. Pyrola secunda, Rubus lasiococcus*, and Clintonia uniflora*.

Table 13. Factor loadings of tree and understory variables on the first four components from principal component analysis, 125 cold or high-elevation plots, Mount Rainier National Park1


Variable212 34 4R2

Abies amabilis, mature

Abies amabilis, young0.42
Abies lasiocarpa, mature-.78

Abies lasiocarpa, young-.61

Chamaecyparis nootkatensis, mature

Tsuga heterophylla, mature

Tsuga mertensiana, young

Arnica latifolia-.46

Chimaphila umbellata

Clintonia uniflora-.38

Erythronium montanum

Gymnocarpium dryopteris

Luzula glabrata-.48

Osmorhiza spp.

Pyrola secunda

Rubus lasiococcus-.71

Rubus pedatus

Streptopus roseus

Tiarella unifoliata

Trautvetteria grandis

Valeriana sitchensis-.68

Viola glabella

Xerophyllum tenax

Menziesia ferruginea.43

Rhododendron albiflorum

Vaccinium alaskaense

Vaccinium membranaceum
Vaccinium ovalifolium.44

1Principal component analysis was performed on raw data from 125 plots.

2Noncorrelated variables: Chamaecyparis nootkatensis (young), Tsuga heterophylla (young). T. mertensiana (mature). Alnus sitchensis, Vaccinium deliciosum, and Viola sempervirens.

3Only the highest loadings have been given for each component.

4Values are the squares of the multiple correlation between each variable and others in the set.

Species with asterisks (*) were added to the group only when the masking effects of very strong correlations associated with dominance of Abies lasiocarpa were removed by deleting a lasiocarpa-dominated plots from PCA. The seven groupings above were analogous to the assemblages of species that were identified with forest types recognized by the following similarity analysis:

Principal component and
factor loadings (+ or -)
Forest Type
Abies lasiocarpa
Abies lasiocarpa
Abies lasiocarpa/Valeriana sitchensis1-
Abies amabilis/Rubus lasiococcus, Rubus lasiococcus phase
Abies amabilis/Rubus lasiococcus, Erythronium montanum phase3+3+
Abies amabilis/Rhododendron albiflorum3+3+
Abies amabilis/Menziesia ferruginea4+2-
Chamaecyparis nootkatensis/Vaccinium alaskaense3-3-

Thus, many species of the Abies lasiocarpa/Valeriana sitchensis Community Type showed negative responses to the first PCA component. Certain species of the Rubus lasiococcus phase of the Abies amabilis/Rubus lasiococcus Association had positive factor loadings to the second PCA component of the reduced analysis without Abies lasiocarpa stands. The third PCA component (in both analyses) appears to be coincident to both the Abies amabilis/Rhododendron albiflorum Association and the Erythronium montanum phase of the Abies amabilis/Rubus lasiococcus Association.

In general, the above tabulation suggests that the complex environmental space of the R matrix, defined in part by each of the major PCA components, roughly resembles the environmental stratification of the Tsuga mertensiana Zone defined by the associations.

Forests of Valleys, Wet Slopes, and Benches—The factor analysis of plots of streamsides, wet slopes, and benches is presented in Table 14. Most plots were characterized by an abundance of Oplopanax horridum and were very rich in herbaceous assemblages. Trees restricted to these environments were Abies grandis and Picea sitchensis at low elevations.

The first axis generally appears to be an elevational gradient. Negative factor loadings are seen for such low elevation species as Thuja plicata, Polystichum munitum, and Berberis nervosa. Higher elevation plants have positive factor loadings, and include young Abies amabilis, Vaccinium ovalifolium, Vaccinium alaskaense, and such herbs as Clintonia uniflora, Rubus lasiococcus, and Streptopus roseus.

Eight percent of the R matrix variation is accounted by the second axis, which clearly relates to streamside Alnus rubra environments (the Alnus rubra/Rubus spectabilis Community Type). A very large share of variance of Alnus rubra is centered on this axis. Understory species of high factor loadings are Rubus spectabilis, Achlys triphylla, and Pteridium aquilinum.

The third axis is difficult to interpret, but seems to highlight low-elevation seral plots associated with Pseudotsuga menziesii. Species relating to this unknown environmental gradient include Abies grandis, Cornus canadensis, Viola sempervirens, and Vaccinium parvifolium.

The following ecological groupings in wetter valley and lower-slope environments are revealed in Table 14:

1. Young Abies amabilis, Clintonia uniflora, Streptopus roseus, Rubus lasiococcus, Smilacina stellata, Vaccinium alaskaense, Vaccinium ovalifolium, and Vaccinium membranaceum;

2. Tsuga heterophylla, mature Thuja plicata, Polystichum munitum, and Berberis nervosa;

3. Alnus rubra, Rubus spectabilis, Pteridium aquilinum, Montia sibirica, Achlys triphylla, and Circaea alpina; and

4. Pseudotsuga menziesii, Abies grandis, Viola sempervirens, Cornus canadensis, and Vaccinium parvifolium.

Important noncorrelated species of these wet environments include such widespread dominants as Gymnocarpium Oplopanax, and Tiarella spp. Locally dominant, noncorrelated species are Blechnum spicant, Corydalis scouleri, and Oxalis oregana. The more widespread and constant species reveal no apparent (linear) trend with elevation, and are not specific to seral environments. The locally dominant species occur too sporadically to indicate trends or environmental preferences.

Mesic Forests of the Abies amabilis Zone—Results of PCA from plots occurring mostly on Abies amabilis/Vaccinium alaskaense and Abies amabilis/Tiarella unifoliata habitats are given in Table 15. The three components collectively account for 31 percent of the R matrix variation. The first component differentiates mesic herbs (positive factor loadings) from the shrub Vaccinium alaskaense. In other words, this component represents an environmental axis that essentially separates the Abies amabilis/Tiarella unifoliata and Abies amabilis/Vaccinium alaskaense Associations, and might be thought of as an "herb cover gradient." The following herbs had high factor loadings: Tiarella unifoliata, Streptopus roseus, Achlys triphylla, Valeriana sitchensis, Gymnocarpium dryopteris, Smilacinia stellata, and Viola glabella.

Table 14. Factor loadings of tree and understory variables on the first three components from factor analysis, on 78 streamside or lower slope plots, Mount Rainier National Park1

Variable2123 3h2

Abies amabilis, young0.75

Abies grandis

Alnus rubra
Pseudotsuga menziesii, young

Pseudotsuga menziesii, mature

Thuja plicata, mature-.31

Tsuga heterophylla, mature-.32-.57

Achlys triphylla
Berberis nervosa-.26

Circaea alpina
Clintonia uniflora.76

Cornus canadensis

Montia sibirica
Polystichum munitum-.40

Pteridium aquilinum

Rubus lasiococcus.60

Smilacina stellata.45

Streptopus roseus.68

Viola sempervirens


Rubus lasiococcus
Vaccinium alaskaense.59

Vaccinium membranaceum.45

Vaccinium ovalifolium.69

Vaccinium parvifolium


Percent variation accounted for9.68.06.3

1Factor analysis was performed on raw data from 78 plots.

2Noncorrelated variables: Picea spp., Acer circinatum, Blechnum spicant, Corydalis scouleri, Gymnocarpium dryopteris, Oplopanax horridum, Oxalis oregana, Tiarella unifoliata, and Tiarella trifoliata.

3The proportion of variance for each variable accounted for by a rank 4 model.

The second component is roughly an elevational axis through the Abies amabilis/Vaccinium alaskaense environmental complex. Species of warmer sites include Linnaea borealis and Cornus canadensis; those of cooler, higher elevations are Vaccinium ovalifolium, Rubus pedatus, and Menziesia ferruginea. This environmental axis might pass through the Berberis nervosa phase of the Abies amabilis/Vaccinium alaskaense Association at lower elevations and the Rubus pedatus phase at higher elevations.

Table 15. Factor loadings of tree and understory variables on the first three components from principal component analysis, 98 plots on mesic slopes and benches at intermediate elevation, Mount Rainier National Park1


Variable2123 3h2

Abies amabilis, young
Chamaecyparis nootkatensis

Tsuga mertensiana

Achlys triphylla0.70

Cornus canadensis
Gymnocarpium dryopteris.54
Linnaea borealis
Rubus pedatus
Smilacina stellata.56
Streptopus roseus.75
Tiarella unifoliata.82
Valeriana sitchensis.62
Viola glabella.56

Viola sempervirens


Menziesia ferruginea
Vaccinium alaskaense-.55.34
Vaccinium ovalifolium

Percent variation accounted for12.89.97.9

1Principal component analysis was performed on raw data from 98 plots.

2Noncorrelated variables; Abies procera, Pseudotsuga menziesii, mature, Acer circinatum, Berberis nervosa, Clintonia uniflora, Oplopanax horridum, Rubus lasiococcus, Vaccinium membranaceum, and Xerophyllum tenax.

3Proportion of variance for each variable accounted for by the first five principal components.

The third component is difficult to interpret. It separates herbs into two categories. Smilacina stellata and Viola sempervirens (as well as Chamaecyparis nootkatensis and Menziesia ferruginea) have positive factor loadings, whereas Gymnocarpium dryopteris, Valeriana sitchensis, Tiarella unifoliata, and Streptopus roseus respond in the opposite manner. This component is some obscure complex gradient through both Abies amabilis/Tiarella unifoliata and phases of the Abies amabilis/Vaccinium alaskaense environment.

Forests of Warm or Dry Sites—The PCA components of Table 16 segregate five vegetation groups:

1. Mesic herbs with positive factor loadings on the first component and negative loadings on the second: Viola sempervirens, Achlys triphylla, Tiarella unifoliata, and Smilacina stellata;

2. Other mesic herbs of positive factor loadings on the first component: Cornus canadensis, Rubus lasiococcus, Rubus ursinus, Linnaea borealis, and Gaultheria ovatifolia;

3. Species with positive loadings along the second axis: young Pseudotsuga menziesii, Vaccinium membranaceum, Xerophyllum tenax, and Gaultheria ovatifolia;

4. Low shrubs and a woody herb with negative factor loadings on the third axis: Berberis nervosa, Chimaphila umbellata, and Arctostaphylos uva-ursi; and

5. Trees and shrubs with positive factor loadings on the fourth component: Abies amabilis, Tsuga heterophylla, Vaccinium alaskaense, and Vaccinium parvifolium. These species groupings also generally coincide with species optima in various community types:

Community type
Principal component and factor loading (+ or -)
Tsuga heterophylla/Achlys triphylla, Pseudotsuga menziesii/Xerophyllum tenax1 +
Pseudotsuga menziesii/Xerophyllum tenax2 +
Pseudotsuga menziesii/Ceanothus velutinus3 +
Abies amabilis/Berberis nervosa3 -
Abies amabilis/Vaccinium alaskaense, Berberis nervosa phase4 +

The first component reflects an environmental gradient affecting herb richness, perhaps in soil nutrition. Our major herb-dominated, low elevation habitats generally encompass this component. The second PCA axis takes into account much of the variation of Vaccinium membranaceum, Xerophyllum tenax, and Gaultheria ovatifolia, whereas mesic herbs (Achlys triphylla, Smilacina stellata, Tiarella unifoliata, and Viola sempervirens) respond in opposite manner. This axis might be a complex elevational-topographic moisture gradient reflecting, perhaps, intensities of summer soil drought. The third component may reflect an environmental (possibly elevational) gradient of rather dry, slope environments involving both the Pseudotsuga menziesii/Ceanothus velutinus Community Type and Abies amabilis/Berberis nervosa Association. The fourth component, accounting for 7 percent of the R-matrix variation, apparently spans an environmental range approximating the Berberis nervosa phase of the Abies amabilis/Vaccinium alaskaense Association.

Table 16. Factor loadings of tree and understory variables an the first four components from principal component analysis, 94 plots on warm or dry sites at low elevations, Mount Rainier National Park1


Variable21234 4R2

Abies amabilis, young-0.34

Pseudotsuga menziesii, young

Tsuga heterophylla, young


Achlys triphylla.51-.58

Arctostaphylos uva-ursi

Berberis nervosa

Chimaphila umbellata

Cornus canadensis.60

Gaultheria ovatifolia.40.51

Linnaea borealis.59
Rubus lasiococcus.65

Rubus ursinus.52

Smilacina stellata.34-.36

Tiarella unifoliata.41-.46

Viola sempervirens.64.50


Xerophyllum tenax


Vaccinium alaskaense

Vaccinium membranaceum

Vaccinium parvifolium


Percent variation accounted for11.

1Principal component analysis was performed on raw data from 94 plots.

2Noncorrelated Variables: Abies procera, Thuja plicata, Acer circinatum, Clintonia uniflora, Fragaria, spp., Gaultheria shallon, Hieracium albiflorum, Pachistima myrsinites, Pteridium aquilinum, Rosa gymnocarpa, Symphoricarpos mollis, and Trientalis latifolia.

3Only the highest loadings have been given for each component.

4Values are the squares of the multiple correlation between each variable and others in the set.

General Conclusions from PCA—Several general conclusions are possible from the PCA. In view of the species groupings there is strong coincidence with analogous groupings of some of our forest types. This is particularly so in Abies amabilis/Rubus lasiococcus, Chamaecyparis nootkatensis/Vaccinium ovalifolium, Alnus rubra/Rubus spectabilis, Tsuga heterophylla/Polystichum munitum, Abies amabilis/Tiarella unifoliata, and Pseudotsuga menziesii/Xerophyllum tenax communities. Other forest types are not so clearly revealed, but species groupings are at least suggestive of some. We feel that PCA is less effective for habitat and community type resolution than similarity analysis and plot groupings based upon tabular procedures (Shimwell 1972). One shortcoming of PCA is its limitation to species showing linearity along environmental axes. This often eliminated dominant species whose distributional modes coincide with associations and community types. On the other hand, PCA does present insights into individualistic patterns of species distribution. For example, Rhododendron albiflorum was shown to respond differently than Menziesia ferruginea along principal components at high elevations. Herbs of wet environments are separated from herbs of mesic or drier environments. We made no attempt to resolve species distances along ordination axes.

PCA gave insights into possible environmental gradients affecting species distribution. But until environmental measurements permit direct ordination, these gradients remain hypothetical. Generally, environmental factors affecting shrub cover differ or oppose those favoring herb cover. Our analyses suggested a variety of contrasting, complex environmental gradients affecting species distribution, including elevation, soil moisture, snowpack duration, and successional or microclimatic gradients. Our results also show, however, that at most only about 30 percent of the variation in the species correlation matrix can be accounted for by the principal components or gradients. This further illustrates the complexity of environmental factors affecting species distributions within each of the four major forest groupings. Some of the remaining variation in species distribution might also be reduced by use of more powerful interspecies association measures.

<<< Previous <<< Contents >>> Next >>>

Last Updated: 06-Mar-2007