Emergence and establishment of canopy-tree seedlings

The fern understory as an ecological filter: emergence and establishment of canopy-tree seedlings

Lisa O. George

INTRODUCTION

Population biologists often view the physical environment as a sieve that filters out individuals from an initial seed population during successive life stages and ultimately determines the success of a seed in producing more seeds (Harper 1977). Through its influence on populations of individual species, the environment can also act as an ecological sieve that structures communities (Van der Valk 1981). An ecological filter “sifts” through the initial seed populations of a whole community and determines emergent community characteristics such as species composition, productivity, and spatial distribution patterns. We propose that the herb and shrub stratum of the forest understory is an important ecological filter in many temperate and tropical communities that influences the composition and structure of the overstory canopy. Seedlings of canopy trees must establish in and grow through the understory herb and shrub stratum in order to reach the overstory canopy. Through differential influence on tree-seedling emergence, growth, and survival, the understory filter may determine which individuals and species will survive in and ultimately penetrate the understory stratum.

Interest in the structure and organization of forests at the community and landscape levels has concentrated on the role of the physical environment (e.g., topography and soils) as determinants of tree recruitment and growth (Collins 1987, 1990, Clark et al. 1995). However, in closed-canopy forests, the physical environment as perceived by tree seeds and seedlings can be highly modified by the biotic environment, particularly the established members of the forest understory. The forest understory stratum is a spatially variable mosaic composed of plants of different species, densities, and degrees of mixture. Because of their clonal nature and ability to resprout after disturbance, many understory herbs and shrubs commonly form dense, monospecific patches that are relatively temporally stable. Clonal plants can be extremely long-lived, with [TABULAR DATA FOR TABLE 1 OMITTED] life spans commonly exceeding those of individual trees (Watt 1947, Oinonen 1967a, b). For example, clones of the fern Osmunda claytoniana growing in our study site date from the period of land abandonment in New England and are presently 150-200 yr old (Knight 1964). Populations of clonal plants may be resilient to disturbance and leave strong imprints on patterns of resource heterogeneity in the forest (Halpern 1989, Hughes and Fahey 1991).

Herbs and shrubs of the understory intercept much of the remaining light in a closed-canopy forest, altering light quality and reducing light quantity beneath their canopies (Messier et al. 1989). Germination and establishment of tree species have been shown to be sensitive to temperature and moisture (Fenner 1985, Burton and Bazzaz 1991) and to quantity and quality of litter (Collins 1990, Facelli and Pickett 1991). To the extent that the understory modifies these microenvironmental variables, temperature, soil moisture, and litter distribution may also play an important role in interactions between the understory and tree seedlings. Understory interference with tree-seedling emergence and establishment may occur not only directly through resource competition or modification, but may occur indirectly through the understory’s influence on the behavior of seed predators. For example, higher rodent densities and acorn removal rates below understory bamboo cover result in low densities of oak tree seedlings in dense bamboo patches in Japanese temperate forests (Wada 1993).

Interactions between ground vegetation and invading tree species have been well studied in old-field environments (Warner and Harbeck 1982, Burton and Bazzaz 1991, 1995, De Steven 1991), and both positive and negative interactions have been demonstrated to occur. Not only has the presence of ground vegetation been shown to modify the microenvironment of potential establishment sites, but different old-field vegetation types uniquely modify microenvironmental conditions below their respective canopies (Burton and Bazzaz 1995). Herbaceous understory plants have additionally been shown to interfere with regeneration of commercially important species after timber harvest primarily through reduction of light levels and production of potential allelopathic substances (Drew 1988, Horsley 1988, 1993, McWilliams et al. 1995). The importance of interactions among herbs, shrubs, and tree seedlings in open environments suggests that herbs and shrubs of the forest understory may similarly be an important determinant of tree regeneration in closed forest environments.

Because the herb and shrub stratum of the forest understory represents one of the first potential barriers to tree regeneration, it can be viewed as a filter that all tree seedlings must pass through (e.g., Gilliam et al. 1994). The understory filter may be nonspecific and regulate only the density of the seedling bank, or it may differentially influence emergence and survival of tree species, thereby controlling species composition and spatial structure of the seedling bank. The objective of this study was to explore the nature of the fern understory as an ecological filter in New England deciduous forests. We monitored emergence of planted seed, and emergence and establishment of natural tree-seedling recruitment in plots where the fern understory was experimentally manipulated in order to: (1) determine how the presence of a well-developed fern understory influences emergence and establishment of canopy-tree species, (2) assess the importance of the identity of understory species in dictating the nature of interactions between the understory and tree seedlings, (3) relate measurements of microenvironmental conditions associated with or imposed by the understory to emergence and establishment of tree seedlings, and (4) explore the importance of using multiple experimental field sites to the interpretation of the nature of the understory filter. We hypothesize that the fern understory selectively filters tree regeneration by differentially influencing emergence and establishment of tree-seedling species, and that the selectivity of this filter will determine the abundance, relative species composition, and spatial distribution of the seedling bank.

METHODS

Study area

Research was conducted in central Massachusetts at the Harvard Forest (42 [degrees] 30[minutes] N, 72 [degrees] 15[minutes] W), which is located in the Transition Hardwoods-White Pine-Hemlock forest type (Westveld et al. 1956). Acer rubrum L. (red maple) and Quercus rubra L. (northern red oak) are the most widespread and abundant overstory species in this region. Betula alleghaniensis Britt. (yellow birch), B. lenta L. (black birch), B. papyrifera Marsh. (white birch), Fraxinus americana L. (white ash), Pinus strobus L. (white pine), and Tsuga canadensis (L.) Carr. (eastern hemlock) are locally abundant. Clonal ferns such as Dennstaedtia punctilobula (Michx.) Moore (hayscented fern), Osmunda claytoniana L. (interrupted fern), Osmunda cinnamomea L. (cinnamon fern) and Thelypteris noveboracensis (L.) Nieuwl. (New York fern) commonly form a dense stratum near the forest floor.

Six sites, in which the herb/shrub stratum was dominated by the clonal ferns Dennstaedtia punctilobula and Osmunda claytoniana, were randomly located within Harvard Forest property. These sites are abbreviated and named in the text after the Harvard Forest compartment in which they are located: PH3A and PH3B (Prospect Hill compartment 3), PH4A and PH4B (Prospect Hill compartment 4), PH9 (Prospect Hill compartment 9), and TS1 (Tom Swamp compartment 1). While these sites were chosen for similar understory composition, they varied in overstory composition and age, drainage, soil type, and slope (Table 1).

Experimental manipulations

In the following experiments, three types of experimental treatments were implemented within naturally occurring clones of both Dennstaedtia and Osmunda. In one treatment, ferns were left undisturbed as a control (“fern”), and in another ferns were removed from plots (“fern-free”) by applying glyphosate (Round-up, Monsanto, Saint Louis, Missouri, USA) directly to fern fronds in July 1993 at commercially recommended rates (Grossbard and Atkinson 1985). Treatments were implemented in 1-[m.sup.2] study plots and in a 0.5-m-wide buffer zone around the perimeter of each study plot. In order to conserve an intact litter layer and soil horizons, dead fern fronds were cut at the base and removed from the plots; dead roots and rhizomes were not removed from plots. Plots were spaded at the perimeter to restrict horizontal in-growth and were periodically weeded to remove new establishment. Additionally, a “shade-free” manipulation was implemented by pinning fern fronds down horizontally using wire wickets. This method allowed fronds to lie horizontally 5-10 cm above the ground, creating a shade-free area above the fronds while leaving fern roots and rhizomes intact. Shade-free conditions could not be maintained at ground level over the entire area of 1-[m.sup.2] plots. However, fronds could be directed away from target areas within plots, thus creating relatively “shade-free” conditions at ground level in smaller zones. Five replicate plots of each fern x manipulation treatment were established in each of the six study sites (3 manipulations x 2 fern species x 5 replications x 6 sites = 180 plots).

Microsite descriptions

Microenvironmental conditions associated with experimental manipulations implemented within the two fern species were measured in all or a subset of all study plots. Before implementation of experimental treatments, fern canopy height and percentage cover were also measured in each experimental plot in order to characterize the two fern species. Fern canopy height was sampled in four locations within each plot with a meter stick, and percentage cover was estimated using a partitioned sample quadrat placed over the plot.

After treatment implementation, integrated photon flux density (PFD) was evaluated using photosensitive-paper light sensors (Friend 1961, Wayne and Bazzaz 1993) during the growing season from 7 to 14 July 1994 and from 11 to 17 August 1995 and during canopy leaf-out from 25 May to 1 June 1 1995. Light readings in July and August were similar in magnitude and were combined in analysis to present more integrated values of mid-summer light levels. Two light sensors per measurement period were placed in a random location at two heights within each experimental plot: one at 1 m to measure light above the fern canopy and one at 8 cm to measure light at the approximate height of 1-yr-old tree seedlings beneath the fern canopy. The 8-cm height was above the level of the restrained ferns fronds in the shade-free manipulation. Paper light sensor readings were calibrated with measurements of instantaneous photosynthetically active radiation ([R.sup.2] = 0.982-0.994) obtained concurrently using a LI-COR quantum sensor (LI-190SA, LI-COR, Lincoln, Nebraska, USA).

Soil samples (0-10 cm depth) were collected from each experimental plot on 1 July 1995. Soil water content was measured gravimetrically by weighing fresh samples, oven-drying them at 70 [degrees] C for 48 h, and reweighing dry samples. Soil water content (SWC) is expressed as grams of soil water per unit dry mass of soil. The percentage of exposed soil (no litter cover) in experimental plots was estimated by placing a transect through the plot and scoring 10 random points along the transect as exposed or not exposed. Litter depth was measured with a ruler to the nearest millimeter at nine random points in a subset of experimental plots (n = 108 plots). Three-factor ANOVA was used to test whether these microenvironmental measurements varied among sites, fern species, and experimental manipulations. Soil water and organic-matter content measurements were In-transformed and soil exposure measurements were square-root transformed to reduce heteroscedasticity.

Natural recruitment of tree seedlings

Natural recruitment of seedlings of Acer rubrum, Fraxinus americana, Pinus strobus, and Quercus rubra into fern and fern-free plots was recorded during the 1994 and 1995 growing seasons (2 manipulations x 2 fern species x 5 replications x 6 sites = 120 plots). Recruitment of Betula alleghaniensis and B. lenta was recorded together as Betula because the species are indistinguishable at emergence. Natural recruitment data were gathered on a plot-level basis, therefore natural recruitment in shade-free plots was not recorded because the entire area of the plot at ground level, where seedlings were emerging, was not shade-free. Emergence was scored as the visibility of cotyledons or leaves above the litter layer. Seedlings were tagged during the initial survey during the last week of May and new births were tagged at two more subsequent dates during the growing season.

Three-factor ANOVA was performed in order to test the effects of site, fern species, and understory manipulation on the emergence of each tree-seedling species. Because of the severe nonnormality of the data, significance levels of factor effects and interactions were determined using randomization procedures (Manly 1996). Seedling survivorship of the 1994 cohort was monitored until September 1994 (survivorship through the first growing season) and again in June 1995 (survivorship through the first year). Data from all plots of each fern manipulation (fern vs. fern-free) were pooled across fern species and sites (n = 60 plots; 10 plots per site x 6 sites), and chi-square tests were used to determine whether survivorship depended on fern manipulation. A second chi-square analysis, in which plots were pooled across manipulation and sites, tested whether survivorship depended on fern species.

In order to explore possible microenvironmental factors responsible for the influence of the fern understory on seedling emergence, Spearman correlation coefficients were calculated between density of emergent seedlings and plot microenvironmental variables. The first correlation analysis was performed using data from both fern and fern-free plots. A second analysis tested correlations between microenvironmental variables and density of emergent seedlings in fern and fern-free plots separately. In species exhibiting an effect of fern manipulation, we wished to more narrowly examine which microenvironmental variables exhibited correlations strong enough to be detected over the natural range of microenvironmental variability within fern manipulations.

Emergence experiments

In order to further explore factors responsible for differential response of tree species to understory manipulation, several field experiments were conducted to test hypotheses generated from known natural-history characteristics of the species and observed patterns of recruitment in the field. “Fern,” “shade-free,” and “fern-free” manipulations were implemented in plots measuring 20 x 30 cm (3 manipulations x 2 fern species x 3 replications x 6 sites = 108 plots) that were located adjacent to a subset of the plots used for measuring microenvironmental conditions and natural recruitment patterns. Planting depth was determined by observing naturally overwintering seed. All seed was planted under the most recent season’s litter layer at the interface between the coarse litter and decomposing humus. Greenhouse germination trials were run concurrently to estimate potential emergence of planted seed.

Quercus rubra. – We placed red oak acorns into fern plots, shade-free plots, and fern-free plot, after fern canopy development to test the hypotheses that differential Quercus emergence was related to light levels and/or seed predation. In April 1994 naturally stratified Quercus acorns were collected from locations adjacent to experimental sites. Two acorns were placed into each plot, and emergence, decay, and predation of acorns was monitored through the growing season and analyzed using categorical modeling procedures (CAT-MOD, SAS Institute 1987).

Acer rubrum. – Because Acer rubrum is the only dominant tree species in the study sites that disperses in spring and can germinate immediately during the current growing season, we wished to evaluate and contrast the effects of the fern understory on germination and emergence of the spring (overwintering) and summer (non-overwintering) seedling cohorts. The majority of the spring cohort of Acer emerges in April and May before fern leaf-out, and the summer cohort emerges after fern leaf-out. The contrast in timing of emergence with respect to fern leaf-out provided a way to test the importance of light reduction by ferns to Acer emergence. In June 1994, 50 recently fallen Acer samaras were planted into each emergence plot. Summer emergence was monitored during 1994 and early spring emergence of the same seed cohort was monitored the following year.

Betula alleghaniensis. – This experiment was designed to test the importance of both light and litter cover in Betula emergence. Because litter cover was expected to exert a strong influence on Betula emergence, two emergence subplots were created within each fern, shade-free, and fern-free experimental plot, one in which litter was undisturbed (“litter”) and one in which coarse litter cover had been removed (“litter removal”). In April 1994 the Betula seeds that had been collected from the Harvard Forest and placed in cold storage were planted in Betula emergence plots (40 seeds/plot). [TABULAR DATA FOR TABLE 2 OMITTED] Seedling emergence was monitored through the 1994 growing season. The influence of understory manipulation on both Acer and Betula emergence was analyzed using nonparametric Kruskal-Wallis [[Chi].sup.2] tests.

RESULTS

Microsite descriptions

Fern cover and fern height, measured before implementation of experimental manipulations, varied by fern species and by site. Mean fern cover was higher in Osmunda plots (66%) than in Dennstaedtia plots (60%) (F = 13.18, df = 1, 168, P [less than] 0.001), and Osmunda was significantly taller (64 cm) than Dennstaedtia (39 cm) (F = 442.13, df = 1,168, P [less than] 0.0001). Fern height and fern cover were lower at site PH9 than all other sites, whereas fern height was greater at site PH4A than all other sites (Table 2).

Midsummer (July-August) light levels below the fern understory (8-cm height) differed among experimental manipulations (Table 3, [ILLUSTRATION FOR FIGURE 1A OMITTED]). Light levels at 8 cm in plots where fern fronds had been pinned back to create a shade-free area were comparable to light levels at 8 cm in plots where ferns had been removed. Midsummer light [TABULAR DATA FOR TABLE 3 OMITTED] levels beneath ferns were [approximately] 32% of light levels in shade-free or fern-free plots. During the growing season when the overstory was fully leafed out, light in the forest at 1 m was 3.4% that of full sun, whereas light below ferns was only 1.1% full sun (1.03 mol[multiplied by][m.sup.-2][multiplied by][d.sup.-1] vs. 0.33 mol[multiplied by][m.sup.-2][multiplied by][d.sup.-1], respectively).

Midsummer light levels measured above the fern canopy at 1 m did not vary with manipulation or fern species (Table 3), which indicates that Osmunda and Dennstaedtia are found in similar light environments. Midsummer light levels below the fern understory did not differ between fern species, indicating that Osmunda and Dennstaedtia reduce light to equivalent levels beneath their canopies. However, because Osmunda initiates and completes frond development earlier than Dennstaedtia, light levels in late May, a period when tree seedlings have leafed out but the overstory has not yet completely leafed out, were lower under Osmunda (2.22 mol[center dot][m.sup.-2][multiplied by][d.sup.-1]) than under Dennstaedtia (3.08 mol[multiplied by][m.sup.-2][multiplied by][d.sup.-1]) (significant fern X manipulation effect, Table 3). Light levels in late May at 1-m height were on average greater (3.93 mol[multiplied by][m.sup.-2][multiplied by][d.sup.1]) than during the growing season (1.03 mol[center dot][m.sup.-2][multiplied by][d.sup.-1]). Although the difference between ferns disappears after the canopies of both ferns are fully developed, the lag in development may allow tree seedlings below Dennstaedtia to accumulate more carbon during a critical period of relatively high light.

Midsummer light levels above and below the fern canopy varied by site (Table 3) and were particularly low at PH9 (Table 2). At site PH9, light levels in shade-free and fern-free manipulations were lower than all other sites, but below ferns the light levels were comparable with all other sites (significant site x manipulation effect, Table 3). Site PH9 had less light reaching the level of the fern canopy. However, the fern canopy was less well developed at this site (i.e., lower in height and percentage cover, Table 2) and therefore removed proportionally less light than the fern stratum of other sites. May light levels also varied by site at 1-m height and at 8-cm height (Table 3) and were relatively high at PH4A and relatively low at PH9 (Table 2).

Averaged across all sites, there was a significant understory manipulation effect on soil exposure (Table 3, [ILLUSTRATION FOR FIGURE 1B OMITTED]): soil exposure was greater in fern-free plots than in either shade-free or fern plots. However, the effect of understory manipulation depended on the site (significant site X manipulation effect, Table 3). Soil exposure was greater in the fern-free manipulation only at sites PH4B, PH3B, TS1, and PH4A. In fern-free plots at these sites, there was no vegetation to hold litter in place against wind, which resulted in large areas of soil exposure in fern-free plots. In fern and shade-free plots, intact fern fronds held litter in place at their bases, resulting in low amounts of soil exposure. The most sloped sites (PH4B and PH3B) had particularly high levels of soil exposure in fern-free plots. At sites PH9 and PH3A soil exposure was equivalent among understory manipulations. These sites, in addition to being relatively less exposed (PH9: base of a sheltered cove; PH3A: no slope), also differed from other sites in the structure and dominant species in the litter layer, which may help explain their resistance to soil exposure. Litter at site PH3A was composed of a deep layer of pine needles, and litter at site PH9 was composed of a high proportion of matted birch leaves, in contrast to the dominantly oak and maple litter of other sites.

Litter depth varied with understory manipulation and by site (Table 3). Litter depth was highest in fern plots, followed by shade-free plots, and was lowest in fern-free plots [ILLUSTRATION FOR FIGURE 1C OMITTED]. Average litter depth was particularly high at site PH3A, where the litter layer was composed of a high proportion of pine needles, and at site TS1 where the litter layer was composed of a high proportion of relatively slowly decomposing oak leaves (Table 2).

Soil water content was not influenced by experimental manipulation [ILLUSTRATION FOR FIGURE 1D OMITTED] but did differ among sites (Table 3). Similarly, soil organic-matter content varied only by site (Table 3). Site PH4B had higher soil water content and soil organic material than all other sites, presumably because of the occurrence of a hard pan not present at other sites (Table 2). Site TS1 had lower soil organic material than all other sites.

Microenvironmental variables generally did not differ by fern species (Table 3). However, at site PH9, which is located near a stream at the base of a cove, soil water content was higher in Osmunda plots than in Dennstaedtia plots (significant site x fern effect, Table 3). Soil water content was higher in plots where Osmunda was removed as well as in plots where Osmunda [TABULAR DATA FOR TABLE 4 OMITTED] was left intact, which indicates that Osmunda occupies moister microsites than Dennstaedtia at this site. This was the only other case where fern species was significantly related to any microenvironmental variable in addition to May light levels.

Natural recruitment of tree seedlings

The density of emergent seedlings of each species varied by site (Table 4) and presumably was at least partially related to differences in site overstory composition and thus to differential seed input. Density of emergent seedlings of each tree species also varied between 1994 and 1995. No emergence of Quercus or Pinus was observed in 1995 due to lack of a seed crop the previous year, and no summer recruitment of Acer was observed in 1994 or 1995. Total Betula recruitment in 1995 was approximately double the value observed in 1994.

Tree species responded differently to understory manipulation, thus providing evidence for selectivity of the fern filter. There was no effect of the presence of a fern understory on the emergence of Acer or Fraxinus seedlings in 1994 or 1995 [ILLUSTRATION FOR FIGURE 2 OMITTED]. Average emergence across all sites of Quercus, Pinus, and Betula in 1994 and 1995 was lower in fern plots compared to fern-free plots [ILLUSTRATION FOR FIGURE 2 OMITTED].

The effect of understory manipulation on Pinus and Betula emergence depended on the site (Table 4). Pinus emergence was only higher in fern-free plots at sites PH3A, PH4A, and PH9; emergence was negligible in both manipulations at sites TS1, PH4B, and PH3B [ILLUSTRATION FOR FIGURE 3A OMITTED]. This pattern appears related to expected seed input based on basal area of Pinus in the overstory canopy. Sites TS1 and PH3B have no overstory pine in the stand, and site PH4B has only a few individuals, so little seed input was expected in these sites. In contrast, PH3A, which has a dominantly pine overstory, had the highest emergence in fern and fern-free plots compared to all other sites. The contrast between the manipulations increased with overall expected seed input to the site.

Betula emergence was only higher in fern-free plots at sites PH4B and PH3B in 1994 and additionally at TS1 in 1995 [ILLUSTRATION FOR FIGURE 3B OMITTED]. Betula emergence patterns differ from Pinus in that they do not bear a clear relationship with stand basal area of Betula [ILLUSTRATION FOR FIGURE 3B OMITTED]. Under the fern canopy Betula emergence was nearly completely precluded in all sites regardless of expected seed input. Betula emergence in fern-free plots closely follows the pattern of soil exposure: highest at PH4B followed by PH3B followed by TS1 and all other sites (Table 2). Emergent seedlings within plots were visibly concentrated in areas of soil exposure. Sites with little or no soil exposure in fern-free plots showed no significant differences in Betula emergence among manipulations.

Survivorship of tree-seedling natural recruitment through the 1994 growing season in Quercus, the largest-seeded species, was comparable between fern and fern-free plots (Table 5). All of the smaller-seeded species [TABULAR DATA FOR TABLE 5 OMITTED] exhibited lower survivorship below ferns. The magnitude of the difference in survivorship between understory manipulations increased with decreasing seed size [ILLUSTRATION FOR FIGURE 4A OMITTED]. By the beginning of the second growing season, survivorship of all seedling species was lower below ferns (Table 5). Survivorship in fern-free plots reflected trends in seed size (highest survivorship in large-seeded Quercus, intermediate survivorship in species of intermediate seed size, and lowest survivorship in small-seeded Betula), [TABULAR DATA FOR TABLE 6 OMITTED] but survivorship beneath ferns no longer reflected seed size [ILLUSTRATION FOR FIGURE 4B OMITTED].

Survivorship differences in Fraxinus seedlings were exhibited between Dennstaedtia and Osmunda plots (Table 5). Not only was higher survivorship observed in plots where Osmunda was left intact compared to where Dennstaedtia was left intact, but survivorship was higher in Osmunda plots where ferns had been removed in comparison to Dennstaedtia-removal plots. Half of the total Fraxinus seedlings observed were found at site PH9, the site where Osmunda plots exhibited higher soil moisture values than did Dennstaedtia plots. When seedlings from PH9 were analyzed separately, higher seedling survivorship was found in Osmunda plots at PH9 (September 1994: [[Chi].sup.2] = 45.03, df = 1, P [less than] 0.001; June 1995: [[Chi].sup.2] = 67.40, df = 1, P [less than] 0.001, n = 371 seedlings), but no survivorship differences were found between Dennstaedtia and Osmunda plots at all other sites (September 1994: [[Chi].sup.2] = 1.89, P [greater than] 0.05, df = 1; June 1995: [[Chi].sup.2] = 0.90, P [greater than] 0.05, df = 1, n = 374 seedlings). By the beginning of the second growing season a difference in Acer survivorship was observed between Dennstaedtia and Osmunda plots (Table 5). Because survivorship was lower than expected in Osmunda fern-removal plots compared to Dennstaedtia fern-removal plots, we attribute this survivorship difference to undetected environmental factors or perhaps residual factors (e.g., allelopathic substances) rather than to a direct detrimental result of the presence of Osmunda.

With the exception of Quercus emergence, the majority of tree-seedling emergence occurred before complete leaf-out of the fern canopy, which suggests that factors in addition to light reduction by the fern canopy were important in producing emergence patterns observed for these species. Ninety percent of Fraxinus and 87% of Acer had germinated and emerged above the litter in April and May before complete leaf-out of the fern canopy. By the time of full fern-canopy development, 83% of Pinus seedling emergence and 75% of Betula emergence was complete. Although all Quercus seeds had germinated and their radicles were actively growing, only 35% of Quercus seedlings’ shoot systems had emerged above the litter by the time of complete fern-canopy development.

Correlations between seedling emergence and microenvironmental variables for species responding to understory manipulation indicate that tree species may be responding to different environmental factors (Table 6). Quercus and Pinus emergence were correlated with light levels during and subsequent to fern canopy development. Betula emergence was also positively correlated with light levels, soil exposure, and soil moisture, and was negatively correlated with litter depth. When environmental correlations were examined in fern-free plots alone, Quercus and Pinus emergence retained a correlation with light levels and Betula exhibited a strong positive correlation with soil exposure and soil moisture. No correlations were found between seedling emergence and microenvironmental variables in fern plots alone.

Experimental studies of seedling emergence

Quercus emergence depended on manipulation of the fern understory ([[Chi].sup.2] = 9.60, df = 2, P [less than] 0.01, [ILLUSTRATION FOR FIGURE 5A OMITTED]). As in the natural recruitment study, emergence was lower in fern plots than in fern-free plots. However, seedling emergence in shade-free plots was more equivalent to emergence in fern plots, which indicates that reduced light was not the cause of lower emergence under ferns. Seedling predation also varied with fern manipulation ([[Chi].sup.2] = 7.00, df = 2, P [less than] 0.05). Predation was lower in the fern-free manipulation than under fern cover and in the shade-free manipulation. Seedling predation also varied by site ([[Chi].sup.2] = 19.68, df = 5, P [less than] 0.01) from a low of 25% of all acorns at site PH3a to a high of 78% at site PH9. Numbers of decayed acorns were comparable among understory manipulations; decay appeared to be the result of infection of the radicle, leading to failure of the newly germinated seedling to emerge above the litter layer.

Summer emergence of Acer differed among experimental manipulations of the fern understory (Kruskal-Wallis [[Chi].sup.2] = 7.14, df = 2, P [less than] 0.05). No emergence of planted seed was observed in fern plots, while emergence in shade-free and fern-free plots was comparable [ILLUSTRATION FOR FIGURE 5B OMITTED], indicating sensitivity of Acer summer emergence to reduced light levels under ferns. Although emergence of experimental Acer seed in greenhouse trials was 44%, total experimental emergence in the field was very low (0-4%), which suggests that field conditions were not conducive to summer emergence of Acer in 1994. Planted seed was allowed to over-winter and germinate the following spring. Emergence the following spring was equivalent in all three understory manipulations [ILLUSTRATION FOR FIGURE 5B OMITTED], which reflects the natural pattern of spring recruitment of Acer.

Where litter was undisturbed, Betula emergence differed among experimental manipulations (Kruskal-Wallis [[Chi].sup.2] = 6.32, df = 2, P [less than] 0.05, [ILLUSTRATION FOR FIGURE 5C OMITTED]). Emergence was higher in fern-free plots than in fern plots, which reflects natural recruitment patterns. Emergence in shade-free plots was more equivalent to emergence in fern plots, which suggests that soil exposure and litter depth were more important to Betula emergence than was the distribution of light levels among manipulations. Where litter cover was removed, overall emergence of Betula improved. Emergence differed among experimental manipulations (Kruskal-Wallis [[Chi].sup.2] = 21.57, df = 2, P [less than] 0.0001) and was again lowest in fern plots [ILLUSTRATION FOR FIGURE 5C OMITTED]. In contrast to undisturbed litter treatments, emergence was more equivalent in shade-free and fern-free plots. Although litter cover appeared to be the primary impediment to Betula emergence, in areas where litter cover was not a factor, Betula emergence was also reduced by low light levels under the fern understory.

DISCUSSION

The presence of a well-developed fern understory influences the forest-floor microenvironment in ways that lead to differential reduction of tree-seedling species emergence and establishment. This differential reduction represents a mechanism by which the fern understory may act as a selective filter that influences future forest composition through reducing density, altering species composition, and determining spatial distribution of the seedling bank. Fern cover reduced emergence of Betula, Pinus, and Quercus as well as summer emergence of Acer, but had no effect on spring emergence of Acer and Fraxinus. The degree to which seedling density was reduced under fern cover also differed among tree species. Seedling densities of Pinus and Quercus were reduced slightly more than 50%, whereas Betula seedling density was reduced by 94% under ferns and no summer emergence of Acer was observed at all under fern cover.

The herb and shrub stratum of the forest understory modifies the abiotic and biotic environment of the forest floor in several ways. Ferns reduce light levels below their canopies to 32% of the already low light levels existing below the overstory canopy (3.4% full sun) and severely reduce red/far red ratios below their canopies (Horsley 1993). The understory of Asian forests (characterized by a bamboo stratum) and Appalachian forests (characterized by high rhododendron cover) may reduce light levels transmitted through the overstory canopy to even lower percentages (10-12% by bamboo, Nakashizuka 1987, Taylor et al. 1995; 1435% by rhododendron, Clinton et al. 1994). The litter layer below ferns is deeper compared to fern-free areas and may act as a more impenetrable mechanical barrier for seeds reaching the soil and for seedlings emerging from below the litter mat, particularly small-seeded species. Light levels under the litter mat decrease exponentially as the amount of litter increases (Facelli and Pickett 1991). The lack of herbaceous or woody structure to trap and to hold litter results in greater areas of soil exposure in fern-free areas in sloped sites. Removal of the insulating litter layer also contributes to elevated soil-surface temperatures (Facelli and Pickett 1991). In contrast to the significant effects of the fern understory on light and litter structure, fern cover did not affect soil water content or soil organic matter in this study or other studies of the fern stratum (Horsley 1993, Carlton and Bazzaz 1998). Dennstaedtia cover does not affect rates of ammonium or nitrate production, soil ammonium and soil nitrate concentrations, or availability of soil nitrogen (Horsley 1993). In terms of the biotic environment, a dense herb/shrub stratum provides habitat and shelter for both seed and seedling predators, which can lead to reduced seedling densities in both forests (Wada 1993) and old fields (Gill and Marks 1991, Reader 1991).

The mechanism by which fern cover reduced emergence was not uniform among tree species. Summer emergence of Acer and emergence of Pinus was primarily affected by reduced light levels under ferns. Emergence of Betula was also sensitive to light levels, but was primarily inhibited by litter cover below the fern understory. Quercus emergence was lower beneath the fern understory due to higher levels of predation. Because of the species-specific nature of fern interference with emergence, the ultimate effect of the fern filter on species composition of the seedling bank cannot be predicted based only on the relative shade tolerances of tree species.

In contrast to the generally strong and clear effects of the presence of understory cover on the microenvironment and seedling emergence patterns, effects of the species identity of understory vegetation were more subtle and complex. Light levels were lower below Osmunda than below Dennstaedtia for approximately a 2-wk period in May because of earlier frond development in Osmunda. However, this phenological difference between fern species did not lead to a detectable effect of fern species on tree-seedling emergence. In Illinois old fields, more distinctive patterns of microenvironmental variation in terms of temperature, soil moisture, and litter depth have been found among patches of different herbaceous and woody species, which resulted in differential tree-seedling emergence among patch types (Burton and Bazzaz 1995). The two fern species did not segregate along measured environmental gradients such as light or soil moisture, except at the cove site (PH9) where Osmunda was found in moister plots than Dennstaedtia. Higher seedling emergence of Fraxinus was correlated with the higher soil moister in Osmunda plots, but this pattern was not the effect of the specific presence of Osmunda. In systems where dominant understory species occupy different microsites, the activity of an understory filter must be interpreted in the context of this environmental heterogeneity.

Although they were bounded by the natural distribution of Dennstaedtia and Osmunda, our six study sites varied in canopy composition and multiple edaphic factors. The extent to which the activity of the fern filter was consistent across sites illustrates the strength and predictable nature of interactions between the understory and tree seedlings. Low light levels at site PH9 and high soil moisture at site PH4B did not qualitatively affect the activity of the fern filter. Other site-specific factors such as slope and overstory composition influenced the strength of the contrast between tree-seedling emergence in fern and fern-free plots. As illustrated by patterns of Pinus emergence, the contrast between seedling density in fern and fern-free areas was strongest at sites with the highest expected seed rain. Emergence of Betula was correlated with soil exposure, and exhibited high values in fern-free plots at sloped sites with a high incidence of soil exposure. The fern filter severely reduced potential Betula emergence at these sites, but had little effect at sites where litter structure already impeded Betula emergence. Generally, the fern filter may cause its most notable effects in terms of proportional reduction of seedling emergence at sites where potential seedling emergence is high, as a result of high seed input or favorable environmental conditions.

The fern understory differentially influenced 1st-yr survival among tree-seedling species, thus representing another mechanism by which the understory may selectively filter tree seedlings and influence forest composition. The fern filter had a different effect on each seedling species at the emergence and at the establishment stages, which reflects different critical abiotic and biotic mediators at each stage. Survival through the first growing season of all species except Quercus was reduced under fern cover. The magnitude of that reduction increased with decreasing seed size and was greatest for small-seeded Betula. A positive relationship between seed size and 1st-yr seedling survival in shade has been documented for North American temperate trees (Grime and Jeffrey 1965). The establishment of smaller-seeded species may be more sensitive to understory cover because they face reduced resource levels under ferns sooner as they exhaust their maternal reserves. Trees that are initiated as sprouts may be even more tolerant of the competitive effects of the fern understory for longer periods than seedlings, particularly if large reserves are available to them. By the beginning of the second growing season, seedling survivorship in fern-free plots still reflected seed size, but seedling survivorship below ferns did not. Similarly, for tree seedlings invading patches of old-field vegetation, large-seeded species were more tolerant of competition during the 1 st yr of establishment than smaller-seeded tree species; however, in subsequent years tolerance of competition was independent of seed size (Bazzaz 1996).

The selectivity of the fern filter with respect to both seedling emergence and establishment shows that the fern understory has the capacity to influence the density and species composition of the seedling bank. These factors in turn influence the spatial structure of the seedling bank and initially define a seedling’s competitive and genetic neighborhoods (Bazzaz 1983, 1996). The selectivity of the fern filter results in Betula seedlings being highly concentrated in fern-free areas relative to areas with fern cover. Quercus and Pinus seedlings are more common in fern-free areas, and Acer and Fraxinus exhibit a nearly random spatial distribution with respect to fern cover. In terms of absolute seedling numbers in the 1994 and 1995 cohorts, recruitment in fern-free areas was dominated by Betula, and recruitment under fern cover was dominated by Acer and Fraxinus. The spatial structure of the forest will significantly impact a variety of forest processes such as competitive and reproductive dynamics (Bazzaz 1983, 1996). The spatial aggregation of species into monospecific competitive neighborhoods may lead to coexistence of species at the community level and impact forest diversity (Hibbs 1982, Pacala 1986, Silvertown and Law 1987, Pacala et al. 1993).

Pre-disturbance interactions between tree seedlings and the understory shape the nature of the seedling bank, which is the starting capital available for future forest regeneration occurring after small- and large-scale canopy disturbance. After disturbance, herbs and shrubs of the understory stratum (if present) may continue to filter the seedling bank, although it may operate differently in the context of new resource flux (e.g., Horsley and Marquis 1983, Drew 1988, 1990). Tree growth after disturbance enters a more rapid phase, and particularly after emergence from dense understory cover, tree-to-tree competition becomes intense (Bormann and Likens 1979, Oliver 1981). However, post-disturbance tree-tree competition may still be influenced by the legacy of the understory. Relative seedling and sapling size as influenced by understory conditions can affect the outcome of tree competition and canopy replacement after release (Canham 1988). The outcome of plant competition depends also on the density and identity of competitors (Bazzaz 1996), features of a seedling’s competitive neighborhood that may also be shaped by the understory filter.

Pre-disturbance interactions among tree seedlings and herbaceous and woody understory plants may have far-reaching consequences for the dynamics and trajectory of future forest regeneration. The herb and shrub stratum of the forest understory is an ecological filter that filters tree recruitment from the earliest stages of seedling emergence and establishment. The selectivity of the understory filter results from individual tree-species response to the biotic and abiotic microenvironment imposed by the understory stratum. In order to more precisely understand the effects of the understory filter on the density, species composition, size structure, and spatial distribution of the seedling bank, understory effects on emergence and establishment must be integrated with effects on seedling growth and survival in subsequent years. The contribution of the understory to future forest composition must then be interpreted in the context of pre-disturbance as well as post-disturbance processes.

ACKNOWLEDGMENTS

We wish to thank Lori Biederman for her invaluable field assistance and Charles Canham for his guidance on nearly every aspect of this study. Art Allen and John Buck kindly assisted in gathering Harvard Forest site information. We thank Diane De Steven, William Bossert, Sebastian Catovsky, Peter Wayne, and two anonymous reviewers for reading drafts of this manuscript. This research was supported by NSF Dissertation Improvement Grant DEB-94-11785, a fellowship from the American Association of University Women (AAUW) to Lisa George, and by NSF Long Term Ecological Research (LTER) Grant DEB-94-11795.

LITERATURE CITED

Bazzaz, F. A. 1983. Characterizations of populations in relation to disturbance in natural and man-modified ecosystems. Pages 259-277 in H. A. Mooney and M. Godron, editors. Disturbance and ecosystems – components of response. Springer-Verlag, Berlin, Germany.

—–. 1996. Plants in changing environments. Linking physiological, population, and community ecology. Cambridge University Press, Cambridge, UK.

Bormann, F. H., and G. E. Likens. 1979. Pattern and process in a forested ecosystem Springer-Verlag, New York, New York, USA.

Burton, P. J., and F. A. Bazzaz. 1991. Tree seedling emergence on interactive temperature and moisture gradients in patches of old-field vegetation. American Journal of Botany 78:131-149.

Burton, P. J., and F. A. Bazzaz. 1995. Ecophysiological responses of tree seedlings invading different patches of old-field vegetation. Journal of Ecology 83:99-112.

Canham, C. D. 1988. Growth and canopy architecture of shade-tolerant trees: response to canopy gaps. Ecology 69: 786-795.

Carlton, G. C., and F. A. Bazzaz. 1998. Resource congruance and forest regeneration following an experimental hurricane blowdown. Ecology 79:1305-1319.

Clark, D. A., D. B. Clark, R. M. Sandoval, and M. V. Castro. 1995. Edaphic and human effects on landscape-scale distributions of tropical rain forest palms. Ecology 76:25812594.

Clinton, B. D., L. R. Boring, and W. T. Swank. 1994. Regeneration patterns in canopy gaps of mixed-oak forests of the Southern Appalachians: influences of topographic position and evergreen understory. American Midland Naturalist 132:308-319.

Collins, S. L. 1987. The seedling regeneration niche: habitat structure of tree seedlings in an oak-pine forest. Oikos 48: 89-98.

—–. 1990. Habitat relationships and survivorship of tree seedlings in hemlock-hardwood forest Canadian Journal of Botany 68:790-797.

De Steven, D. 1991. Experiments on mechanisms of tree establishment in old-field succession: seedling emergence Ecology 72:1066-1075.

Drew, A. P. 1988. Interference of black cherry by ground flora of the Allegheny uplands. Canadian Journal of Forest Research 18:652-656.

—–. 1990. Fern and aster effects on black cherry shelterwood regeneration. Canadian Journal of Forest Research 20:1513-1514.

Facelli, J. M., and S. T. A. Pickett. 1991. Plant litter: its dynamics and effects on plant community structure. Botanical Review 57:1-32.

Fenner, M. 1985. Seed ecology. Chapman & Hall, New York, New York, USA.

Friend, D. T. C. 1961. A simple method of measuring integrated light values in the field Ecology 42:577-580.

Gill, D. S., and P. L. Marks. 1991. Tree and shrub seedling colonization of old fields in central New York. Ecological Monographs 61:183-205.

Gilliam, F. S., N. L. Turrill, S. D. Aulick, D. K. Evans, and M. B. Adams. 1994. Herbaceous layer and soil response to experimental acidification in a central Appalachian hardwood forest. Journal of Environmental Quality 23:835844.

Grime, J. P., and D. W. Jeffrey. 1965. Seedling establishment in vertical gradients of sunlight. Journal of Ecology 53: 621-642.

Grossbard, E., and D. Atkinson. 1985. The herbicide glyphosate. Butterworth, London, UK.

Halpern, C. B. 1989. Early successional patterns of forest species: interactions of life history traits and disturbance Ecology 70:704-720.

Harper, J. L. 1977. Population biology of plants Academic Press, New York, New York, USA.

Hibbs, D. E. 1982. White pine in the transition hardwood forest. Canadian Journal of Botany 60:2046-2053.

Horsley, S. B. 1988. How vegetation can influence regeneration Pages 38-55 in H. C. Smith, A. W. Perkey, and W. E. Kidd, editors. Guidelines for regenerating Appalachian hardwood stands. United States Department of Agriculture, Forest Service, Morgantown, West Virginia, USA.

—–. 1993. Mechanisms of interference between hay-scented fern and black cherry. Canadian Journal of Forest Research 23:2059-2069.

Horsley, S. B., and D. A. Marquis. 1983. Interference by weeds and deer with Allegheny hardwood reproduction. Canadian Journal of Forest Research 13:61-69.

Hughes, J. W., and T. J. Fahey. 1991. Colonization dynamics of herbs and shrubs in a disturbed northern hardwood forest. Journal of Ecology 79:605-616.

Knight, G. J. 1964. Distribution of Osmunda cinnamomea L. and Osmunda claytoniana L. in relation to natural soil drainage. Thesis. Harvard University, Cambridge, Massachusetts, USA.

Manly, B. F. J. 1996. RT – a program for randomization testing, version 2.0. The Centre for Applications of Statistics and Mathematics, University of Otago, Dunedin, New Zealand.

McWilliams, W. H., S. L. Stout, T W. Bowersox, and L. H. McCormick. 1995. Adequacy of advance tree-seedling regeneration in Pennsylvania’s forests. Northern Journal of Applied Forestry 12:187-191.

Messier, C., T. W. Honer, and J. P. Kimmins. 1989. Photosynthetic photon flux density, red:far-red ratio, and minimum requirement for survival of Gaultheria shallon in western red cedar-western hemlock stands in coastal British Columbia. Canadian Journal of Forest Research 19: 1470-1477.

Nakashizuka, T 1987. Regeneration dynamics of beech forests in Japan. Vegetatio 69:169-175.

Oinonen, E. 1967a. Sporal regeneration of bracken (Pteridium aquilinium (L.) Kuhn.) in Finland in light of the dimensions and the age of its clones. Acta Forestalia Fennica 83(1): 1-96.

—–. 1967b. Sporal regeneration of ground pine (Lycopodium complanatum L.) in southern Finland in the light of the size and the age of its clones. Acta Forestalia Fennica 83(3):76-85.

Oliver, C. D. 1981. Forest development in North America following major disturbances. Forest Ecology and Management 3:153-168.

Pacala, S. W. 1986. Neighborhood models of plant population dynamics. II. Multi-species models of annuals. Theoretical Population Biology 29:262-292.

Pacala, S. W., C. D. Canham, and J. A. Silander, Jr. 1993. Forest models defined by field measurements. I. The design of a northeastern forest simulator. Canadian Journal of Forest Research 23:1980-1988.

Reader, R. J. 1991. Control of seedling emergence by ground cover: a potential mechanism involving seed predation. Canadian Journal of Botany 69:2084-2087.

SAS Institute. 1987. SAS/STAT guide for personal computers. Version 6 edition. SAS Institute, Cary, North Carolina, USA.

Silvertown, J., and R. Law. 1987. Do plants need niches? Some recent developments in plant community ecology. Trends in Ecology and Evolution 2:24-26.

Taylor, A. H., Z. Qin, and J. Liu. 1995. Tree regeneration in an Abies faxoniana forest after bamboo dieback, Wang Lang Natural Reserve, China. Canadian Journal of Forest Research 25:2034-2039.

Van der Valk, A. G. 1981. Succession in wetlands: a Gleasonian approach. Ecology 62:688-696.

Wada, N. 1993. Dwarf bamboos affect the regeneration of zoochorous trees by providing habitats to acorn-feeding rodents. Oecologia 94:403-407.

Warner, P. A., and A. M. Harbeck. 1982. The pattern of tree seedling establishment relative to staghorn sumac cover in Michigan old fields. American Midland Naturalist 108: 124-132.

Watt, A. S. 1947. Contributions to the ecology of bracken (Pteridium aquilinium). IV. The structure of the community. New Phytologist 46:97-121.

Wayne, P. M., and F. A. Bazzaz. 1993. Birch seedling responses to daily time courses of light in experimental forest gaps and shadehouses. Ecology 74:1500-1515.

Westveld, M., R. I. Ashman, H. I. Baldwin, R. P. Holdsworth, R. S. Johnson, J. H. Lambert, H. J. Lutz, L. Swain, and M. Standish. 1956. Natural forest vegetation zones of New England. Journal of Forestry 54:332-338.

COPYRIGHT 1999 Ecological Society of America

COPYRIGHT 2000 Gale Group