Current address: Department of Biology, Swarthmore College, Swarthmore, PA, 19081, USA.
Fine-scale environmental variation and structure of understorey plant communities in two old-growth pine forests
Article first published online: 1 APR 2003
Journal of Ecology
Volume 91, Issue 2, pages 283–293, April 2003
How to Cite
Frelich, L. E., Machado, J.-L. and Reich, P. B. (2003), Fine-scale environmental variation and structure of understorey plant communities in two old-growth pine forests. Journal of Ecology, 91: 283–293. doi: 10.1046/j.1365-2745.2003.00765.x
- Issue published online: 1 APR 2003
- Article first published online: 1 APR 2003
- Received 13 August 2002 revisionaccepted 8 January 2003
- Aster macrophyllus;
- boreal forest;
- understorey plant communities;
- white pine
- 1Although it is well established that nitrogen and light play major roles in structuring plant communities across the landscape, it is not as clear how they structure communities within forest stands. Virtually nothing is known about within-stand structure of understorey communities of herbs and small shrubs in near-boreal forests.
- 2We tested the hypothesis that fine-scale (5–20 m) variability in N and light structure forest-floor plant communities in two old-growth mixed Pinus resinosa and Pinus strobus forests in north-eastern Minnesota, USA.
- 3In each forest, all trees > 1.4 m tall were mapped on a 0.75–1.0 ha area. A grid of subplots 5–10 m apart was established (total n = 147), and N mineralization (µg g−1 soil day−1), soil depth (cm), light (% canopy openness), and percentage cover of all herbs and small shrubs were measured on each subplot.
- 4Cluster analysis showed that the dominant understorey species fall into three groups. Group 1 is unrelated to N and light, and is negatively associated with a midstorey of the small tree Acer rubrum and the most abundant tall shrub Corylus cornuta. Group 2 reaches maximum abundance in places (mostly gaps) with relatively high light, but is unrelated to within-stand variation in N availability. Group 3 consists of a single species, Aster macrophyllus, and reaches maximum abundance in areas with low N availability and low abundance of Corylus, but higher than average abundance of P. strobus and Betula papyrifera overstorey trees.
- 5N and light have a moderate influence on understorey plant community structure. The plant species do arrange themselves along N and light gradients, but the gradients are likely to be too narrow to allow the degree of differentiation seen at the landscape level. Spatial patterning of the species groups is probably influenced by other factors, including disturbance history, chance and neighbourhood effects such as clonal reproduction.
It is clear that environmental gradients of nitrogen and light determine forest species distribution and thus are important in structuring communities across the landscape. In cold-temperate, near-boreal forests of North America, Pinus and Picea, with long-leaf lifespans, typically dominate forests on nutrient poor sites, whereas species with short leaf lifespans (Populus, Betula, and Acer) occupy progressively more nutrient rich sites (Curtis 1959; Grigal & Ohmann 1975; Pastor et al. 1984; Reich et al. 1997), and similar patterns are observed in all forested biomes (Reich et al. 1992). Across the temperate–boreal transition zone of eastern North America, there is also a gradient in species that cast progressively greater shade, which is associated with a successional gradient from Betula papyrifera, to Quercus rubra, Pinus strobus, Acer rubrum and, finally, to Abies balsamea, Picea mariana, Acer saccharum and Tsuga canadensis (Canham et al. 1994; Walters & Reich 1996). Under the canopy of many of these species light levels do not favour their own reproduction, but allow succession to any species later in the list. Hence the landscape is a mosaic, strongly related to light availability, of stands in different stages of succession.
At the landscape level, understorey plant communities in forests also respond to N and light (Carleton & Maycock 1980, 1981; Kurmis et al. 1986; Whitney & Foster 1988; Lichter 1998). Disturbance history is superimposed upon these environmental gradients, and succession in understorey plant communities often parallels that in the overstorey, although the short dispersal distances of herb propagules may delay recolonization of a disturbed site and thus limit the correlation (Whitney & Foster 1988; Matlack 1994; Lichter 1998). In boreal forests, there is significant, but not total correlation between canopy type and understorey composition (Reich et al. 2001), because of the presence of species with the same environmental preference (Carleton & Maycock 1980).
The well documented importance of N and light gradients at the landscape scale may or may not translate to other scales. The degree to which variation in N and light structures the distribution of understorey communities within a single forest stand is unclear. Within stands, variation is at a small spatial scale (on the order of tens of metres), and is also more limited in the total range of measured values than those measured across the landscape (Reich et al. 1997; Finzi et al. 1998; Ferrari 1999). Cold-temperate mixed deciduous–evergreen forests, such as those at our study site, clearly exhibit small-scale variation in N and light availability, which are sometimes higher under deciduous trees (here broad-leaf species) in mixed stands than under evergreens (which are coniferous). This statement cannot, however, be generalized for all forest communities (Mladenoff 1987; Pastor et al. 1987; Boettcher & Kalisz 1990; Canham et al. 1994; Binkley 1995; Ferarri 1999).
The relationship between nutrients or light or overstorey structure and herb growth have been examined individually in boreal forests (Landhauser et al. 1997; Arii & Turkington 2002; Legare et al. 2002; Turkington et al. 2002), and the influence of fire history and forest type on herb communities at landscape scales has also been examined (Ahlgren 1960; Carleton & Maycock 1981). At this point, however, we are unaware of any studies that take the next step and link forest floor community composition with fine-scale variation in light and N in boreal forests.
A number of studies have looked at within-stand interactions between nutrients, light or overstorey trees and forest-floor herbs in more southerly forests dominated by genera such as Quercus, Acer, Fagus, Tsuga, Tilia and Lirodendron (Bratton 1976; Hicks 1980; Maguire & Forman 1983; Beatty 1984; Crozier & Boerner 1984; Marino et al. 1997; McKenzie et al. 2000; Lezberg et al. 2001; Rankin & Tramer 2002). These authors concluded that microhabitats are important and that there are some significant relationships between canopy type immediately overhead and herb composition, due to alteration of the nutrient, light and physical environment of the forest floor by the overstorey.
No clear patterns or principles emerge to allow us to predict the influence of N and light gradients on within-stand forest-floor community structure. Growth rates may be more affected than presence or absence of species at a given location on the forest floor, as many species can persist in what is, for them, a poor environment (Lezberg et al. 2001). Many herbs reproduce vegetatively, and/or have very short seed dispersal distances (Cain et al. 1998), or could expand in clonal fashion independently of small-scale variability of N and light within a stand.
One can therefore reasonably entertain alternative hypotheses that: (i) N and light, together with their complex relationship to canopy tree composition and soil, will explain much of the observed understorey community structure, or (ii) that colonies of forest-floor plants may form patches that are independent of small-scale environmental differences. We used spatial analyses to assess the relative importance of canopy, N and light for the structure of forest floor herb and small shrub communities in two near-boreal, old-growth Pinus strobus and Pinus resinosa stands. All trees were mapped and forest floor communities and environmental variables were measured on a systematic grid. Tree seedling dynamics and moss communities are considered as factors that could potentially affect the herb and small shrub community but their responses will be considered elsewhere. Our study sites are bedrock-controlled with shallow soils, whose moisture availability is linked with soil depth. Soil depth was therefore also characterized to investigate its interaction with N and light.
We considered the following questions: (i) do the understorey species arrange themselves according to local canopy composition, (ii) do they arrange themselves along N and light gradients, and (iii) are the understorey species individualistic, or are they arranged in definite communities?
field and laboratory methods
The study sites are in the Great Lakes Pine-Hardwood Forest/boreal forests transition zone of north-eastern Minnesota (92°00′ W, and 48°00′ N). They are covered by a mixture of Pinus strobus, Pinus resinosa, Betula papyrifera and Acer rubrum, representing the pine-hardwood forest, with Abies balsamea and Picea mariana representing the boreal element. Soils in the region are shallow and rocky, consisting of highly variable sandy to loamy glacial deposits from 0 to 0.5 m thick over granitic bedrock. The region has a continental boreal climate (Ahlgren 1969). Winters are long and cold (mean January temperature of −17 °C and snow covers the ground from mid-November to mid-April) and summers are short and cool (growing season length is < 100 days with a mean July temperature of 17 °C).
All trees > 1.4 m tall were mapped, measured for d.b.h., and species identified on two macroplots. Hegman Lake (0.76 ha, n = 1248 trees) was mapped in 1994 and Kawishiwi Pines (0.98 ha, n = 1243 trees) in 1996. The Hegman Lake stand is dominated by large Pinus resinosa and P. strobus that originated after a severe fire in 1822, while the Kawishiwi stand is dominated by P. strobus that recruited after a severe fire in 1854 (Heinselman 1973). Neither stand has experienced commercial logging, and both are old enough to have completed the stem-exclusion stage of development (Frelich 2002), and are now in the early and middle phases, respectively, of transition to multi-aged stands.
Within each macroplot, a grid of subplots, centred at 5-m intervals, was established during 1995–97. A subset of these subplots was chosen for detailed environmental measurements, including N mineralization rate, soil depth and light, which were carried out during the summers of 1995–97 (Machado 1999). This subset was randomly chosen at Hegman (94 subplots, with many subplots at both 5 and 10 m intervals), but every other subplot was chosen at Kawishiwi (53 subplots, creating a 10 m grid with only a few exceptions). Percentage cover was determined at mid-summer 1997 by an experienced field crew that had been doing similar work for at least 3 summers (Reich et al. 2001). There are no spring ephemerals in this area, and all work was done within 2 weeks to ensure comparable phenological stage among subplots. The subplots had nested radii of 1.1 m, within which cover of all shrub species was determined, and 0.75 m, within which cover of vascular herbaceous plants, mosses and tree seedlings was determined using Gleason & Cronquist (1991) as the source for taxonomic nomenclature of vascular plants. The vast majority of plants were identified to species although Viola spp., Fragaria spp. and some Lycopodium spp. had to be grouped on some occasions when early blooming season or lack of resolution of taxonomic status prevented identification. Ordinal cover classes were defined as follows: class 1, absent; class 2, < 1%; class 3, 1–5%; class 4, 6–25%; class 5, 26–50%; class 6, 51–75%; and class 7, 76–100% cover.
To evaluate the distribution of light environments on the two macroplots we measured percentage canopy openness using the LAI-2000 Plant Canopy Analyser (Li-Cor, Lincoln, Nebraska, USA) at 1.0 m above ground level on each subplot. Such measurements provide a good surrogate of mean daily percentage photosynthetic photon flux density in these forests (Machado & Reich 1999). Two canopy openness measurements were taken at the centre of each subplot when the sky was uniformly overcast or within 1 hour after dawn or 1 hour before dusk and compared with open sky values simultaneously measured with a second LAI 2000 in a large clearing less than 1 km away. The values reported here represent percentage open sky in an imaginary inverted cone with the vertex at the sensor and sides that extend upward through the forest canopy at an angle of 60% from vertical. Details of methodology are given by Machado & Reich (1999).
N mineralization was determined with a modified semi-open core in-situ incubation method (Grigal & Homann 1994) between July 1–8 and August 5–11, 1995. We sampled the 0–15 cm horizon since we found the majority of roots at this depth. At the beginning of the incubation period, PVC tubes were used to remove six 2.5 cm diameter × 15 cm deep cores at each subplot for determination of initial values. Simultaneously, six PVC tubes were pounded into the ground to a depth of no more than 15 cm, within 10 cm of the initial cores, and left to incubate for 34 or 35 days. The PVC tubes were covered with rubber corks to prevent water flow through the soil column and had air holes at the top to facilitate ventilation. At the end of the incubation period the cores were refrigerated and transported to the laboratory. The six cores from each subplot for each time point were pooled and cleaned of roots and woody debris. We oven-dried the soils at 105 °C to determine water content, after removal of a sample for determination of 2 m KCl extractable NH4-N and NO3-N. NH4-N and NO3-N in the extract solutions were determined by the University of Minnesota Department of Soil Science Research Analytical Laboratory by conversion to nitrate with salicylic acid and subsequent copper cadmium reduction, followed by colorimetric analysis. N mineralization potential of each sample was calculated as µg N g−1 soil day−1 using the difference in total N concentration between the initial and incubated core samples.
Soil depth was measured with a 50-cm probe at 5 points arranged at the corners and in the middle of a 50 × 50 cm square near the centre of each subplot. The sites are on bedrock of the Canadian Shield, and measurements, which reflect depth to bedrock or large boulder, were averaged to give an index of soil depth at each subplot. Those few subplots with deeper soil are entered into the data set as > 50 cm.
Initial examination of the data showed that 58 species of herbs and small shrubs occurred on the two macroplots. Twenty-nine of these species occurred on a relatively large number of subplots, while the remainder were very rare (most occurring in low abundance only on 1 or 2 subplots), thus precluding any analysis of their response to environmental variables. These rare species collectively accounted for only 5.4% of cover, using the midpoints of the cover classes recorded in the field to convert values to percentages.
An overall description of the subplot data was achieved with non-metric multidimensional scaling (nMDS) ordination with PC-ORD software (McCune & Mefford 1999), using the relative abundance of each of the 29 important species on all 147 subplots. The ‘slow and thorough’ procedure, which carries out 40 runs with the real data, each of which steps though dimensions from 6 down to 1, and 50 runs with randomized data, was employed to determine the best solution. The best final configuration is that which has the highest number of dimensions but still has a stress statistic lower than in 95% of the randomized runs, and is also the lowest among the 40 runs with real data for the chosen number of dimensions (McCune & Mefford 1999). The resulting scatter plots were used to see if any structure exists among the 147 subplots, and whether the subplots cluster tightly for each macroplot, or form a continuum.
A hierarchical, agglomerative clustering method, using Ward's method for group linkage on the ordinal cover classes was employed to assign the 29 important species into groups (JMP software, version 3.2.2, SAS Institute, Cary, NC, USA). The percentage cover for all species within each group was then pooled for each subplot for further analysis. NMDS ordinations (using relative abundance of each species group in an analogous procedure to that for describing the two macroplots) were also done using the species groups as variables, to check the validity of the cluster analysis.
Ordinal logistic regression (Minitab Statistical Software, Release 12, 1998, State College, Pennsylvania, USA) was used to evaluate whether quantitative environmental factors play a significant role in determining the distribution of a given species group among ordered categorical levels of cover. This method assumes that samples are independent (i.e. that the plants on one subplot do not influence the plants on another subplot – see below for spatial autocorrelation tests). Percentages for species groups were calculated by converting the cover class of each species to a percentage and then summing for all species in the group. Five levels of abundance − level 1 (0–1% cover), level 2 (1.1–5% cover), level 3 (5.1–20% cover), level 4 (20.1–50% cover, and level 5 (> 50% cover) − were used in ordinal logistic regressions which were done separately for each species group, using N mineralization, light, soil depth, percentage cover of tree seedlings, percentage cover of mosses, basal area of each of six important tree species within 5 m of each plot centre (Picea mariana, Pinus resinosa, P. strobus, Abies balsamea, Betula papyrifera and Acer rubrum) and percentage cover of the predominant tall shrub species (Corylus cornuta) as variables.
A second method of examining the relationship between species groups and their environment, which we call maximum abundance analysis, was to compare the environment of subplots on which each species group was maximally abundant to all remaining subplots. We developed a two-step procedure to examine the correlation of each species group with each variable. First we examined a histogram of species group abundance among all subplots, and looked for a break in the distribution that separated subplots where the group was a minor component from those relatively few subplots where it achieved high cover. We then separated those subplots with high cover and used one-way anova to test whether their environmental variables were significantly different from the other subplots. One can then, for example, characterize a given species group as reaching maximum abundance on sites with high, average or low N. The nonparametric Kruskal–Wallis anova from Minitab Release 12 was used because all distributions were highly non-normal.
Moran's I (Sokal & Oden 1978) was calculated at 5-m intervals to determine if the plant species groups on adjacent subplots were independent of one another. The computer program for Moran's I was written by the senior author, following the formulas and significance tests from Sokal & Oden (1978). These spatial analyses were done separately for each macroplot, which have independent grid-coordinate systems. Analyses of spatial clustering of the environmental variables N, light and soil, had already been done for a previous study (Machado 1999), whose data are used for comparison to the species group data.
general pattern of forest floor herbs and small shrubs
Average aggregated cover for the 29 species of herbs and small shrubs studied was 22% for the two macroplots pooled. Cover was significantly different between the two macroplots (median 9.5% at Hegman, and 25.5% at Kawishiwi, Kruskal–Wallis H = 31.5, P < 0.001). The lower median cover at Hegman was due to some subplots having essentially no vascular plant community on the forest floor, either because they were located on two bedrock knobs, most of which were covered by mosses (n = 15), or because of very dense coniferous cover or dense clones of the tall shrub Corylus in addition to a coniferous tree canopy (n = 8 with less than 2.0% cover for the 29 species added together). An nMDS ordination showed that the bedrock plots were somewhat separate from the main cluster of points, within which representatives from the two macroplots were mixed together (Fig. 1).
The species richness of all forest floor vascular plants (species per 1.77 m2 subplot), ranged from 1 to 19, and the distribution among subplots was normal with a mean of 9.9 and standard deviation of 3.2. Species richness for the 29 species studied in detail ranged from 1 to 13, and the distribution was also normal with a mean of 6.3 and standard deviation of 2.54. The only significant correlations with species richness (Pearson correlation, n = 147 subplots, for the 29 common species) were for N mineralization (r = −0.265, P = 0.001), light (r = −0.221, P = 0.007), and abundance of Corylus (r = −0.186, P = 0.023).
The 29 species studied in detail included 12 summer-green herbs, six small shrubs, five ferns and allies including clubmosses, two evergreen herbs, two miniature vines, one vine, and one grass. Summer-green herbs had the highest percentage cover of any growth form on 120 of the 147 subplots and were present on all but one subplot. One-way anova revealed that N mineralization and light levels were higher on 10 subplots where small shrubs dominated than on subplots where other growth forms dominated (1.97 µg N g−1 soil day−1 vs. 0.76 for other growth forms, F = 16.8, d.f. 1,143, P < 0.001 and 21% canopy openness vs. 7.6% for other growth forms, F = 51.2, d.f. 1,145, P < 0.001). No other significant effects were detected, probably because the preponderance of summer-green herbs weakened any relationships between growth form and environmental variables.
Cluster analysis revealed that the 29 common species could be split into three (possibly five) groups (Fig. 2). To determine which grouping scheme worked best, nMDS ordinations were run using three and five species groups as variables. Relative abundance, based on total cover of all species within a group on each subplot, was used in the ordinations. For both ordinations PC-ORD recommended retaining two axes, and the resulting final plot of the axes revealed a triangular structure. For both ordinations the vertices of the triangle corresponded with subplots dominated by the three-species grouping, which was therefore used for subsequent analyses (Fig. 3).
Group 1 is a taxonomically and morphologically diverse group of six perennials including two summer-green herbs (Aralia nudicaulis and Maianthemum canadense), a grass (Oryzopsis asperifolia), a miniature vine in the rose family (Rubus pubescens), a miniature vine in the honeysuckle family (Linnaea borealis), and a clubmoss (Lycopodium clavatum). Group 2 is the largest, with 22 species, and comprises several summer-green herbs (Clintonia borealis, Trientalis borealis, Fragaria spp., Polygala paucifolia, Galium triflorum, Pyrola rotundifolia, Viola spp., Cornus canadensis and Streptopus roseus), two evergreen herbs (Chimaphilla umbellata and Gaultheria procumbens), two small ericaceous shrubs (Vaccinium myrtilloides and V. angustifolium), two small honeysuckle-family shrubs (Lonicera canadensis and Diervilla lonicera), two small rosaceous shrubs (Rosa acicularis and Rubus strigosus), a fern (Pteridium aquilinum), three club mosses (Lycopodium obscurum, L. annotinum and Lycopodium sp.), and a vine in the honeysuckle family (Lonicera hirsuta). The third group consists of a single summer-green herb, Aster macrophyllus, commonly called large-leaved aster, whose dense clonal habit often excludes other species.
There was no significant spatial autocorrelation (using Moran's I on percentage cover for each species group) for any of the three species groups, on either macroplot, for distance classes from 5 m to 45 m. Thus, it was valid to assume independence of plant communities on subplots for the logistic regression and Kruskal–Wallis tests. Subplots would be perfectly correlated with themselves, and the lack of a significant correlation at 5 m means that the correlation drops from 1.0 to 0.0 somewhere between 0 and 5 m. Our grid of sample points was not therefore fine enough to detect any spatial structure for each species group. Although there appeared to be no significant patch structure, subplots with maximum abundance of particular species groups occurred in spatially separated areas of both macroplots (Fig. 4). The nMDS ordination (Fig. 3) also showed separation of subplots with high abundance of the different species groups with a limited overlap, indicating that the groups reached maximum abundance on different subplots, while coexisting on others.
forest floor community vs. tree seedlings
Ordinal logistic regression using tree seedling cover as a predictor of the occurrence of the three plant groups indicated that tree seedlings were independent of the herb and shrub community and did not influence its structure (P = 0.948, 0.524 and 0.599 for plant groups 1, 2, and 3, respectively).
forest floor community vs. moss cover
Moss cover did not predict the presence of plant groups 1 and 2 (P = 0.594 and 0.661, respectively), but was negatively associated with cover of group 3 (P = 0.011). These results were expected because the moss community is heavily concentrated in mats on rock outcrops that cannot support significant levels of vascular plant cover, whereas Aster macrophyllus is found primarily on deep soil areas (see below). Overall, there is no indication that the moss community has a systematic influence on the structure of this vascular plant community, with the only correlation probably explained by soil depth.
forest canopy vs. the forest floor plant community, light and n
Overall, there was little association between the forest-floor plant community and the local occurrence of seven major tree species or the tall shrub Corylus (i.e. within 5 m of the subplot centres). Abundance of species group 1 was however negatively associated with Acer and Corylus (P = 0.037 and 0.035, respectively, for the two macroplots pooled). Species group 2 showed no significant response to any forest canopy attribute. Group 3 was positively associated with mature Pinus strobus and Betula (P = 0.006 and 0.044, respectively) and, like group 1, negatively with Acer and Corylus (P = 0.018 and 0.012, respectively).
Where there was a significant relationship with the understorey communities, tree and shrub species data were converted to quartiles which were used as ordered categories in an ordinal logistic regression against light and N. There were no significant relationships between N and presence of any tree or tall shrub species within a 5 m of the centres of the subplots (P-values from 0.372 to 0.879), but Corylus and Pinus did have significant negative relationships with light (P < 0.001 and 0.002, respectively).
Species groups 1 and 2 attained maximum abundance on subplots whose median rates of N mineralization were not significantly different from the rest of the subplots, for macroplots analysed separately or pooled (Table 1). Aster macrophyllus (group 3) reached maximum abundance on subplots where median N mineralization rates were significantly lower than on the rest of the subplots on the Hegman macroplot and for both sites pooled, but there was no significant effect on the Kawishiwi macroplot.
|Macroplot/analysis||Plant species group|
|Median N high abundance subplots||0.930||1.185||0.455|
|Median N other subplots||0.767||0.951||1.06|
|Median N high abundance subplots||0.505||0.497||0.424|
|Median N other subplots||0.497||0.479||0.503|
|Hegman + Kawishiwi:|
|Median N high abundance subplots||0.453||0.623||0.424|
|Median N other subplots||0.574||0.537||0.591|
|Ordinal logistic regression P-value||0.127||0.480||0.018|
The ordinal logistic regression analyses for the pooled macroplots, were consistent with these findings. P-values for the N parameter for groups 1 and 2 were well above 0.05, indicating no sensitivity to N availability, whereas for group 3, the P-value was 0.018, and the odds ratio was also well above 1.0, indicating a preference for low-N environments (Table 1).
The subplots with high N availability (as well as high canopy openness) were those with shallow soils on the Hegman macroplot on the domes of bedrock, which were covered with mosses, but were largely unoccupied by the herbs and small shrubs studied.
Canopy openness just above the herb/seedling layer (at 1.0 m above ground) was significantly higher, and was also more variable on the Hegman macroplot. The brightest areas had canopy openness up to 35%. Analyses of subplots with maximum abundance indicates that, for both macroplots, groups 1 and 3 showed no response to light, whereas group 2 reached maximum abundance in areas with significantly higher canopy openness. The ordinal logistic regression confirmed this trend, with a significant P-value only for group 2 (Table 2).
|Macroplot/analysis||Plant species group|
|Median openness high abundance subplots||9.11||18.0||9.38|
|Median openness other subplots||9.72||9.21||9.72|
|Median openness high abundance subplots||7.46||9.08||8.05|
|Median openness other subplots||6.34||6.04||6.04|
|Hegman + Kawishiwi:|
|Median openness high abundance subplots||7.25||11.4||8.03|
|Median openness other subplots||6.66||6.13||6.44|
|Ordinal logistic regression P-value||0.176||0.008||0.094|
Soil depths were similar on the Hegman and Kawishiwi macroplots (Table 3). Analyses of subplots with maximum abundance showed that, for both macroplots when pooled, none of three species groups had a significant preference (Table 3). However, species group 3, Aster macrophyllus, showed a significant preference for relatively deep soil on the Hegman macroplot, which probably reflects the avoidance of bedrock knobs on that plot. Ordinal logistic regression results were generally consistent with this trend, with no significant P-values for the soil parameter for any species group.
|Macroplot/analysis||Plant species group|
|Median depth high abundance subplots||26.0||21.3||32.6|
|Median depth other subplots||23.6||24.0||22.2|
|Median depth high abundance subplots||19.2||19.4||23.4|
|Median depth other subplots||23.4||23.5||21.4|
|Hegman + Kawishiwi:|
|Median depth high abundance subplots||17.4||26.3||24.8|
|Median depth other subplots||23.7||23.4||23.5|
|Ordinal logistic regression P-value||0.862||0.188||0.913|
Discussion and conclusions
Individual species abundances corresponded poorly to environmental gradients. As is typical of biological abundance data, relationships had a triangular shape, with the upper edge of the triangle indicating the potential abundance for each species. Most subplots have abundances falling far below the potential maximum because of interactions with other environmental variables (known and unknown), and because of sharing abundance with other species. In a system with 29 important species few, if any, species can approach the maximum potential abundance allowed by the environment at a given spot on the forest floor.
However, we did find evidence that the forest floor is occupied by three species groups, which could potentially form a mosaic of communities. These groups were verified by: (i) separation in the cluster analysis (Fig. 2); (ii) separation of subplots with the highest level of dominance by each of the three species groups in an nMDS ordination (Fig. 3); and (iii) the fact that niches occupied by the three species groups are consistent, and consistently distinct, across two different macroplots (Tables 1, 2 and 3, Figs 3 and 4).
the role of canopy, nitrogen and light in community structure
The main objective of this paper was to test whether N and light are major forces in structuring the understorey community within a stand. A sub-hypothesis is that the canopy type above a given spot on the forest floor may influence the light and N environment, in turn influencing the understorey communities. The results show that there was a detectable, statistically significant yet modest effect of within-stand variability of N and light on community structure.
Clearly, variation in N is not providing the degree of control over community structure within these two stands that it exerts at the landscape scale. We found a negative relationship between N mineralization and species richness, which would seem to agree with the results of Turkington et al. (2002) who found reduced understorey herb diversity after 10 years of fertilization in a boreal forest in north-western Canada. However, reduced diversity at high N in our study is not due to competitive exclusion, but rather to the interaction of soil depth with N, whereby high N occurs on mats of moss on rock outcrops that support few vascular plants.
The locations with maximum abundance of each group had a characteristic environment. For group 1, the combination of no significant difference in N and light levels between those subplots where it reaches maximum abundance and other subplots, and the lack of significance in ordinal logistic regression, indicate that it was insensitive to within-stand variation of N and light levels (Tables 1 and 2). However, there was a negative effect of the presence of Acer or Corylus in the midstorey (Fig. 5).
Species group 2 was insensitive to within-stand variation in N levels, and was found at moderate light levels, but only reached high abundance at what are high light levels for this forest (Table 2, Fig. 5). The group did not exhibit the negative association with Acer and Corylus shared by groups 1 and 3, nor the preference for neighbourhoods with Pinus strobus and Betula shown by group 3. High light environments often occur in this forest where gaps in the upper canopy are filled with Acer saplings or clusters of Corylus, which have only a limited effect on light penetration. Areas with a layer of Acer or Corylus in addition to the upper tree canopy were nearly devoid of forest floor plants. Group 2 contains all six species of small shrubs, the only growth form shown to be positively related to light, which is consistent with the overall response of the group to light, and with its role as a gap community.
Species group 3, Aster macrophyllus, occupied areas with less than average and average N availability on the Hegman and Kawishiwi macroplots, respectively (Table 1, Fig. 5). The average N mineralization rate was higher on the Hegman macroplot than on Kawishiwi, and inspection of the data in Table 1 shows that the median N mineralization rate for those subplots where Aster macrophyllus attains maximum abundance at Hegman was similar to the mean for all subplots at Kawishiwi (0.455 and 0.503 µg g−1 soil day−1, respectively). It therefore seems likely that Aster macrophyllus was selecting environments with low N availability, regardless of whether that level is average or below average within a given forest. This species also had a significant positive association with canopy Pinus strobus and Betula, and a negative relationship with the tall shrub Corylus and the midstorey tree Acer. Ordinal logistic regression of the effects of these four canopy and midstorey species showed no significant relationship with N, so that N availability and neighbourhood abundance of these species were independent. Therefore, the ideal niche for Aster macrophyllus is where crowns of canopy Pinus strobus and Betula occur over soils with low N availability, and there is no Corylus and/or Acer midstorey. Legare et al. (2002) found, however, that forest canopy type did not influence Aster macrophyllus abundance, although differences in scale and forest type make it difficult to compare the two sets of results. Aster macrophyllus abundance did not show sensitivity to light levels in this study, in agreement with Schulz & Adams (1995), who pointed out the broad range of light tolerance for this species.
Canopy type within a 5-m radius around the centre of subplots explained surprisingly little of the variation in forest-floor community structure. Pinus strobus is apparently independent of N supply in this forest and its positive effect on species group 3, which favours a low-N environment, must have an alternative explanation. One possibility is that a high density of Pinus strobus may provide just enough shade to exclude the mid-tolerant Corylus and Acer, which in turn disfavour species group 3, as well as group 1. Because Acer abundance was shown to have no relationship with light and N, its mechanism for excluding groups 1 and 3 is unknown, but may be related to smothering of herbs by litter fall. Acer species in general are known to have this effect on forest herbs and seedlings (Beatty & Sholes 1988).
These environmental patterns leave much of the spatial distribution of species groups unexplained. Forest understorey species do respond to N and light gradients in the white and red pine forests of our study sites, and there is a modest but significant degree of separation into communities. However, the species groups coexist on some subplots. One possible explanation is that a larger difference in environment than is found within these two forests is required to cause more than minor differentiation of separate plant communities. Our three species groups could then be called ‘potential plant communities’, which do not differentiate fully because of the great amount of overlap in the local environment. Second, there might be more community separation than we can detect. Some subplots (which at 1.77 m2 are relatively large compared to some published studies of forest-floor herbs) may overlap the boundaries between two communities, thus blurring community distinctness.
Neighbourhood effects, as well as historical factors, probably also contribute to the spatial patterns found in these two forests. Previous analyses (Machado 1999) indicate spatial autocorrelation on the Hegman macroplot of 30–80 m for N, 10–30 m for light, and 20–50 m for soil depth. Similar values for the Kawishiwi were 3–12 m for N, 50–60 m for light and 10–30 m for soil (Machado 1999). The grain of the species groups, however, is smaller than our subplot spacing (< 5 m). This mismatch in the grain of the environmental and community mosaics is an indicator that other types of patch dynamics than merely matching up with the appropriate environment are taking place. Clonal reproduction or very short seed dispersal distances are common among our plant species and may provide an alternative explanation for patch formation. These types of neighbourhood effects are known to cause spatial structure in other herbaceous communities (Belsky 1986). Forest tree communities in the Great Lakes region of the USA have also been shown to be partly structured by neighbourhood effects as well as by environmental differences between stands (Frelich et al. 1993, 1998; Frelich & Reich 1995, 1999).
Historical factors, such as where plants were established after the last major disturbance (1822 and 1854 on our study areas), could also interact with the clonal nature of the plant species. Forest understorey plants can establish after major disturbance and then maintain themselves through periods of poor environment and, as stand development proceeds, they may be able to respond to gap formation and other stand development processes that allow an increase in abundance by vegetative growth (Ahlgren 1960; McKenzie et al. 2000; Lezberg et al. 2001; Rankin & Tramer 2002). We propose to look at the hypothesis that the overstorey and understorey communities in this forest have independent neighbourhood-effect patch dynamic systems that contribute to community structure. It would be logical to pursue detailed mapping of individual plants on a section of forest floor, giving us more spatial resolution than we had in this study, as a next step to test more fully the relative importance of N, light, soil and neighbourhood effects in structuring forest floor plant communities in near-boreal white and red pine forests.
We gratefully acknowledge financial support from the Wilderness Research Foundation, Chicago, Illinois, The Minnesota Department of Natural Resources, The National Council of the Paper Industry on Air and Stream Improvement, and US National Science Foundation (BSR-9524078). We thank Peter Bakken, Daren Carlson, Martin Dovciak, Steve Friedman, Dave Grigal and Dave W. Peterson for field and laboratory help with the study. The US Forest Service and Minnesota Department of Natural Resources, Natural Areas Program, provided logistical support and permission to work on their property.
- 1960) Some effects of fire on reproduction and growth of vegetation in northeastern Minnesota. Ecology, 41, 431–445. (
- 1969) Eighteen years of weather in the Boundary Waters Canoe Area, Quetico Superior-Research Center. Minnesota Agricultural Experiment Station Miscellaneous Report 88, St. Paul, Minnesota, USA. (
- 2002) Do nutrient availability and competition limit plant growth of herbaceous species in the boreal forest understorey? Arctic, Antarctic and Alpine Research, 34, 251–261. & (
- 1984) Influence of microtopogaphy and canopy species on spatial patterns of forest understory plants. Ecology, 65, 1406–1419. (
- 1988) Leaf litter effect on plant species composition of deciduous forest treefall pits. Canadian Journal of Forest Research, 18, 553–559. & (
- 1986) Population and community processes in a mosaic grassland in the Serengeti, Tanzania. Journal of Ecology, 74, 841–856. (
- 1995) The influence of tree species on forest soils: processes and patterns. Proceedings of the Trees and Soil Workshop, Lincoln University 28 February – 2 Marine 1994 (ed. Mead, D.J. & Cornforth, I.S.). Agronomy Society of New Zealand Special Publication no. 10. Lincoln University Press, Canterbury. (
- 1990) Single-tree influence on soil properties in the mountains of eastern Kentucky. Ecology, 71, 1365–1372. & (
- 1976) Resource division in an understory herb community: responses to temporal and microtopographic gradients. American Naturalist, 11, 679–693. (
- 1998) Seed dispersal and the Holocene migration of woodland herbs. Ecological Monographs, 68, 325–347. , & (
- 1994) Causes and consequences of resource heterogeneity in forests: interspecific variation in light transmission by canopy trees. Canadian Journal of Forest Research, 24, 337–349. , , & (
- 1980) Vegetation of the boreal forests south of James Bay: non-centered component analysis of the vascular flora. Ecology, 61, 1199–1212. & (
- 1981) Understorey-canopy affinities in boreal forest vegetation. Canadian Journal of Botany, 59, 1709–1716. & (
- 1984) Correlations of understory herb distribution patterns with microhabitats under different tree species in a mixed mesophytic forest. Oecologia, 62, 337–343. & (
- 1959) The Vegetation of Wisconsin. The University of Wisconsin press, Madison, Wisconsin, USA. (
- 1999) Fine-scale patterns of leaf litterfall and nitrogen cycling in an old-growth forest. Canadian Journal of Forest Research, 29, 291–302. (
- 1998) Canopy tree soil interactions within temperate forests: species effects on soil carbon and nitrogen. Ecological Applications, 8, 440–446. , & (
- 2002) Forest Dynamics and Disturbance Regimes. Cambridge University Press, Cambridge, England. (
- 1993) Patch formation and maintenance in an old growth hemlock-hardwood forest. Ecology, 74, 513–527. , , & (
- 1995) Neighborhood effects, disturbance, and succession in forests of the Western Great Lakes Region. Ecoscience, 2, 148–158. & (
- 1999) Neighborhood effects, disturbance severity and community stability in forests. Ecosystems, 2, 151–166. & (
- 1998) Neighbourhood effects in forests: implications for within stand patch structure and management. Journal of Ecology, 86, 149–162. , , , & (
- 1991) Manual of Vascular Plants of Northeastern United States and Adjacent Canada, 2nd edn. New York Botanical Garden, New York. & (
- 1994) Nitrogen mineralization, groundwater dynamics, and forest growth on a Minnesota outwash landscape. Biogeochemistry, 27, 171–185. & (
- 1975) Classification, description, and dynamics of upland plant communities within a Minnesota Wilderness Area. Ecological Monographs, 45, 389–407. & (
- 1973) Fire in the virgin forests of the Boundary Waters Canoe Area, Minnesota. Quaternary Research, 3, 329–382. (
- 1980) Intrastand distribution patterns of southern Appalachian cove forest herbaceous species. American Midland Naturalist, 104, 209–223. (
- 1986) Plant communities of Voyageurs National Park, Minnesota, USA. Canadian Journal of Botany, 64, 531–540. , & (
- 1997) Photosynthetic strategies of summergreen and evergreen understory herbs of the boreal mixedwood forest. Oecologia, 112, 173–178. , & (
- 2002) Influence of forest composition on understory cover in boreal mixedwood forests of western Quebec. Silva Fennica, 36, 353–366. , & (
- 2001) Clonal development of Maiathemum dilitatum in forests of differing age and structure. Canadian Journal of Botany, 79, 1028–1038. , & (
- 1998) Primary succession and forest development on coastal Lake Michigan sand dunes. Ecological Monographs, 68, 487–510. (
- 1999) The role of resource availability in the growth, distribution, survival and physiology of cold temperate tree species. PhD Thesis, University of Minnesota, St. Paul. (
- 1999) Evaluation of several measures of canopy openness as predictors of photosynthetic photon flux density in deeply shaded conifer-dominated forest understory. Canadian Journal of Forest Research, 29, 1438–1444. & (
- 1983) Herb cover effects on tree seedling patterns in a mature hemlock-hardwood forest. Ecology, 64, 1367–1380. & (
- 1997) Influence of sunlight and soil nutrients on clonal growth and sexual reproduction of the understory perennial herb Sanguinaria canadensis L. Journal of the Torrey Botanical Society, 124, 219–227. , & (
- 1994) Plant-species migration in a mixed-history forest landscape in eastern North America. Ecology, 75, 1491–1502. (
- 1999) PC-ORD. Multivariate Analysis of Ecological Data, Version 4. MjM software design, Gleneden Beach, Oregon, USA. & (
- 2000) Overstory influences on herb and shrub communities in mature forests of western Washington, USA. Canadian Journal of Forest Research, 30, 1655–1666. , & (
- 1987) Dynamics of nitrogen mineralization and nitrification in hemlock and hardwood treefall gaps. Ecology, 68, 1171–1180. (
- 1984) Aboveground production and N and P cycling along a nitrogen mineralization gradient on Blackhawk Island, Wisconsin. Ecology, 65, 25–28. , , & (
- 1987) Successional changes in nitrogen availability as a potential factor contributing to spruce declines in boreal North America. Canadian Journal of Forest Research, 17, 1394–1400. , , & (
- 2002) Understory succession and the gap regeneration cycle in a Tsuga canadensis forest. Canadian Journal of Forest Research, 32, 16–23. & (
- 2001) Influence of logging, fire, and forest type on biodiversity and productivity in southern boreal forests. Ecology, 82, 2731–2748. , , , , & (
- 1997) Nitrogen mineralization and productivity in 50 hardwood and conifer stands on diverse soils. Ecology, 78, 335–347. , , & (
- 1992) Leaf life-span in relation to leaf, plant, and stand characteristics among diverse ecosystems. Ecological Monographs, 62, 365–392. , & (
- 1995) Effect of canopy gap light environment on evaporative load and stomatal conductance in the temperate forest understory herb Aster macrophyllus (Asteraceae). American Journal of Botany, 82, 630–637. & (
- 1978) Spatial autocorrelation in biology. I. Methodology. Biology Journal of the Linnean Society, 10, 199–228. & (
- 2002) The effects of fertilization and herbivory on the herbaceous vegetation of the boreal forest in northwestern Canada: a 10-year study. Journal of Ecology, 90, 325–337. , , & (
- 1996) Are shade tolerance, survival, and growth linked? Low light and nitrogen effects on hardwood seedlings. Ecology, 77, 841–853. & (
- 1988) Overstorey composition and age as determinants of the understorey flora of woods of central New England. Journal of Ecology, 76, 867–876. & (