Distribution and Dynamics of Two Ferns: Dennstaedtia punctilobula (Dennstaedtiaceae) and Thelypteris noveboracensis (Thelypteridaceae) in a Northeast Mixed Hardwoods-Hemlock Forest

by James D. Hill, John A. Silander, Jr.
Distribution and Dynamics of Two Ferns: Dennstaedtia punctilobula (Dennstaedtiaceae) and Thelypteris noveboracensis (Thelypteridaceae) in a Northeast Mixed Hardwoods-Hemlock Forest
James D. Hill, John A. Silander, Jr.
American Journal of Botany
Start Page: 
End Page: 
Select license: 
Select License

American Journal of Botany 88(51: 894-902. 2001



Department of Ecology and Evolutionary Biology, U-32, University of Connecticut, Storrs, Connecticut 06269-3042 USA

Denir.rtaedtiiz prtt~ctilobula and The1ypteri.c t~o~~ebo,ncer~si.r

are two native specie5 that often arrest forest succession and reduce ~inderstorj diversity. As part of a project to examine the feedback between forest ~inderstory and canopy dynamics, we studied the patterns of distribution and dynamics of these two fern species in an oak-transition hardwoods-hemlock forest. Dennrtcredrin was least abundant under shade-tolerant tree species and most abundant in small (1-2 trees) canopy gaps, but did not show any distinct patterns across the sampled moisture regime. The light response was verified using light manipulation experiments and examination of plant size-abundance patterns across light environments. Thel,.pteris tended to be most prevalent under maple canopies and appeared to be more sensitive to soil moisture regime being restricted to more rnesic sites than Derztisrcledtia. Seasonal and year-to-year changes in abundance of established clones of both fern species were small, suggesting that once established, both species can maintain a strong hold on a site. Further work on the niche requirements of the two species is warranted, but any event that maintains or promotes canopy openness (tree death by disease or windthrow, forest harvesting, or the elimination of a shrub layer by browsing) will promote persistence of Dentismedtia.

Key words: Connecticut; forest regeneration; forest understory dynamics; Great Mountain Forest: hayscented fern; New York

fern: successional dynamics.

The distribution and abundance of herbs, shrubs., and tree seedlings in the forest understory are influenced by a wide variety of environmental factors. These include light avail- ability (Lipscomb and Nilsen, 1990a), water availability (Beals and Cope, 1964; Lipscomb and Nilsen, 1990b), the availability of nitrogen and other soil nutrients (Rice and Pancholy, 1973; Finzi, 1996), and microtopographical variation (Bratton, 1976a, b; Beatty, 1984). Some of the variation in the forest understory environment can be linked to species-specific in- fluences of canopy trees in the quality and quantity of light transmittance (Canham, 1988; Canham et al.. 1994), differ- ences in stemflow quality and quantity (Gersper and Holo- waychuk, 1971; Crozier and Boerner, 1984), and leaf litter chemistry (Lodhi, 1977). Although less studied, biotic inter- actions such as competition anlong understory species within the herb layer may be important in determining the distribution and abundance patterns of understory plants (Maguire and Forman, 1983; Collins and Good, 1987). However, few studies have explored these biotic interactions and even fewer have

Manuscript received 14 March 2000; revision accepted 18 July 2000.

The material is this paper is from a dissertation presented in partial fulfill- ment of the requirement for the degree doctor of philosophy at the University of Connecticut.

The authors thank C. D. Canhain, R. L. Chaydon. S. W. Pacala, C. Craddock. and two anonymous reviewers for advice and comments on an earlier draft of the manuscript; the Childs family for their generous hospitality, and for the use of field sites at the Great Mountain Forest; and the Bridgeport Hydraulic Company for access to additional field sites. This research was funded in part by grants from the National Science Foundation (BSR8918616), Department of Energy (DE-FG02-90ER60933). the National Aero- nautics and Space Administration. and the Department of Ecology and Evo- lutionary Biology of the University of Connecticut.

'Author for correspondence, current address: Yale University. Department of Ecology and Evolutionary Biology, PO. Box 208106, New Haven, Con- uecticut 06520-8106 USA (e-mail: jim.hill@yale.edu).

attempted to examine the feedback between forest understory and canopy dynamics (Hill, 1996).

Although there are many studies of the distribution and dy- namics of temperate forest herbs and shrubs (e.g., Brewer, 1980; Moore and VanKat, 1986; Plocher and Carvell, 1987; Collins and Pickett, 1988), few have examined the role of these understory species in the successional dynamics of the forest (Hill, 1996). Herbs and shrubs can play significant roles in determining the composition and abundance of tree seed- lings in forest stands (Horsley. 1977a, b; Maguire and Forman, 1983; Phillips and Murdy, 1985; Kolb, Bowersox, and Mc- Cormick, 1990). These understory species can also have an important impact on forest dynamics. For example, the com- bined, negative effects of understory species (Denr~staedtia punctilobula, Thelypteris noveboracensi.~, Brachyelytrum er- ectum, and Lycopodiunz obscurunz) often results in the failure of the forests to regenerate (arrested succession) in the Alle- gheny Plateau of Pennsylvania (Horsley, 1977a, b, 1985) and New York (Drew, 1988. 1990).

Hayscented fern (Dentzstaedtin yulzctilobula) and New York fern (The1jpteri.r tzoveboracelzsis) are two prominent species that can interfere wit11 forest regeneration (Horsley, 1977a, b; Drew, 1988, 1990; Hill, 1996): yet our understanding of the ecology of these species is limited. The abundance of tree seedlings under these fern canopies can be dramatically (60- 85%) reduced (Horsley, 1977a; Hill, 1996). Though the det- rimental effects of hayscented fern on tree seedling growth and survivorship have been shown (Horsley and Marquis, 1983; Drew, 1988; Horsley, 1993; Hill, 1996), less work has been conducted on New York fern (Horsley, 1977bj. Since these understory species have the potential to influence forest re- generation patterns, understanding their distribution and dy- namics is critical to understanding the successional dynamics of thc forest.

Hayscented fern (Dentzstaedtia punctilobula) and New York


fern (Tlzeljpteris noveboracensi.r) are abundant throughout the understory of secondary growth, oak-transition hardwoods-hemlock stands at Great Mountain Forest (GMF), Connecticut. The two species commonly co-occur and thus appear to have similar distributional patterns in the forest understory. How- ever, they differ in their rhizome morphology (R. McCalley and J. D. Hill, unpublished data, Yale University) with hay- scented fern growing -9 crntyr and New York growing -1- 2 cmlyr. Also, hayscented fern has a continuous population of fronds and New York fern has more discrete patches of fronds. In addition, there may be subtle differences in the moisture requirements of the two species (J. D. Hill, personal obser- vation). Both are native species but, based on current distri- bution patterns, hayscented fern appears to be rnore invasive, spreading aggressively in comparison to New York fern.

The objectives of this study were to quantify the patterns of distribution, abundance, and within- and between-season dvnamics of havscented and New York fern in relation to can- opy tree species and canopy tree gaps (1-2 tree gaps) in the forest and to assess whether these patterns are related to light and moisture availability.


Study area-This research was conducted in and around the Great Moun- tain Forest, located in the towns of Norfolk and Canaan, Connecticut (42"00' N, 73"15' W). Great Mountain Forest is a 2500-ha, privately owned tract of oak-transition hardwoods-hemlock forest located on the south end of the Berkshire Plateau. The forest spans an elevational range of 250-550 m and consists mostly of secondary growth stands logged 80-150 yr ago. Soils in the sites are fine sandy loams, primarily Dystrochrepts developed on shallow, glacial till derived from the mica schist and gneiss bedrock of Canaan Moun- tain. Forests in the region are dominated by a mix of species characterizing the conifer-northern hardwood forests of central and northern New Ellgland and the oak forests of southern New England.

Stzrdy species-Hayscented fern (Derzn.utriedrin pllilcrilob~lla (Michx.) Moore) is a hornosporous, leptosporangiate fern native to the eastern United States and Canada (Cody, Hall, and Crompton. 1977). In New England, hay- scented fern is common in deciduous forests. along roadsides, and at the edges of old fields (Hammen, 1993). Hayscented fern colonizes bare rnineral soil via spores, but once established can spread aggressively through underground stems (rhizomes) (Conard, 1908; Horsley, 1984: Groninger and McCormick. 1991). The basic architecture of the rhizome is linear with dichotomous branching at 8-24 cm intervals with fronds arising on alternate sides of the rhizome (Conard. 1908). The rhizome of hayscented fern has the capacity to expand by either centrifugal (due to repeated branching) or linear growth (unbranched rhizome) of the perennial rhizome systern (Harnmen, 1993; Hill, 1996). The perennial rhizomatous habit allows the rapid formation of an early growing season canopy. often putting the species in direct competition with other herbs and tree seedlings (Horsley and Marquis, 1983). Despite being a native species, hayscented fern is often considered a weed (Cody, Hall. and Crompton, 1977; Hammen, 1993), as it out-competes rnore desirable species.

New York fern (TIze1yl1teri.u 17ovel~oracet~sis

(L.) Nieuwl.) has long-creeping rhizomes that are slender and cord-like. Unlike hayscented fern. fronds of New York fern occur closely packed in tufts of three to four fronds or as solitary fronds. Although the rhizomes are linear in form like those of hay- scented fern. rhizome growth rates are much slower in New York fern on the order of 1-2 cm annually (R. McCalley and J. D. Hill, unpublished data, Yale University). New York fern is prevalent in sornewhat shady, moist to dry, rich woods and on swamp edges (Lellinger, 1985). Both hayscented and New York fern occur in moderately acidic soils and appear to be similar in their niche requiremeilts (J. D. Hill. personal observation). Additionally, New York fern appears to be a far less weedy and aggressive species. posing less of a problern with forest regeneration failure

Qltarztifying fern distribution arzd abundarzce-Because hayscented and New York fern are clonal organisms, the strict definition of a genetic indi- vidual (genet) was difficult. To circumvent this problem, we used above- ground ramet density (fronds) to quantify fern abundance. Ramet density was assessed in 105. 1-in2 plots in July and August of 1991. Plots were placed out randomly after stratifying by canopy tree species types (see below). Plots were recensused in 1992. 1993, and 1994 to assess the dynamics of these fern populations (see Fen7 djrlarnrcs below)

To address whether species-specific canopy tree effects were important, census plots (N= 15 per canopy type) were located beneath specific canopy treer (Tsugn canadensis, F0gu.r grnndifolin, Q~rercus r~ibra, Pr~ctz~cs

serotirza, Acer rubrunz, or A. sacckarutn) or in small (1-2 canopy tree gaps) canopy gaps. Plots were chosen in a stratified random fashion. and although they span a range of understory light conditions (0.25-26.0% full sun), soil moisture conditions were in the mesic (6.5-30.0°h moisture) range. The species of fern, the number of mature and immature ramets. and the average ramet height by plot were recorded on each census date.

The patterns of distribution and abundance of hayscented and New York fern with respect to canopy type, light availability, and soil moisture avail- ability were determined using the 1991 data set. Assumptions of normality and homogeneity of variance were evaluated throughout using graphical di- agnostics. If necessary, log or power transformations were utilized, or non- parametric approaches were employed. Distributions in terrns of presence- absence were evaluated using Wilcoxon two-sample tests (SAS. 1987). The analysis of abundance patterns with respect to resources was performed at the canopy type level using one-way ANOVA and Kruskal-Wallis tests (with Bonferroni t tests for comparisons of means) and on a continuous basis with regression.

Quantifj~irzg resource availability-Light availability was assessed using hemispherical-photographs (see Canhain 1988 for detailed methods) taken 1.5 m above the ground in each plot in July of 1991. An index of whole-growing- season light availability (GLI, following Canharn, 1988) was computed for each photograph. The GLI index integrates the seasonal and diurnal move- ments of the sun, the mix of diffuse and beam radiation, and the spatial distribution of canopy openness illto a single index in units of percentage of full sun. This index is correlated with total photosynthetically active radiation (PAR) under closed and open canopies and in gaps (Canham. 1988: Canham et al., 1994). Soil moisture availability (percentage of volurne) was quantified using time dornain reflectornetry (Trase Ilistrurnentation, Soilmoisture Equip- ment Corp., Goleta, California. USA; see Gray and Spies [I9951 for detailed methods) in each of the 1-m? plots. Soil moisture measurements were taken in both 1991 and 1992 at the mid-growing season census, as the average of three measurements (spatial replicates) in each of the I-m' plots.

Fern dynamics-In addition to the 1991 data set, census data were also collected in June, July, and August of 1992, and July of 1993 and 1994. As with the 1991 census data. species of fern, the number of mature and im- [nature ramets, and the average rarnet height by plot were recorded on each census date.

Abundance patterns within the growing season (seasonal) were extracted frorn the 1991 and 1992 data sets. Paired cornparisons t tests were used to assess differences in the mean abundance (number of fronds per square meter) across census dates. In addition. correlation and regression analysis were used to evaluate changes in frond density with respect to initial density. light avail- ability, soil nloisture availability, and the density of the other fern speciei.

Changes in abundance between growing seasons were cornpared using the entire data set (1991-1994). Patterns of fern dynamics were evaluated using mean midseason (July) abundance for all years in paired comparison t tests. In addition, correlation and regression analyses were used to evaluate the importance of initial frond density, light availability, and soil rnoisture avail- ability to changes in frond density over three growing seasons. All data anal- yses were performed using the SAS statistical progranl (SAS, 1987) unless otherwise indicated.

To firrther assess the importance of light in determining the distribution and abundance patterns of hayscented fern, two smal!-scale experiments were per- formed. The shading experiment explored the dynamics of fern following reductions in light. The invasion experiment involved measuring the dynarnics of fern following increases in light. In the shade experinlent, aboveground raniet density and light incident on the fern canopy were measured in 40 1m2 plots. A shading treatment was then applied to 20 of these plots using a shade box 1.5 X 1.5 m in si?.e, 1.5-2.0 m tall, and consi~ting of 92% shade cloth. In other words. 8% of the light incident on the shade box transmitted through to the fern. Frond density was measured in these plots 1 and 2. yr after initiation of the shading treatment. Light beneath the shade boxes was calculated as a percentage of light inctdenr on the shade box. Pretreatment light levels ranged from 1.4 to 10.1% of full sun: after shading. light levels above the fern ranged from 0.1 to 0.8% oS full sun. Shading treatment did not affect soil moisture as no difference betvieen shaded and control plots (t test, P > 0.05) was observed. Posttreatment (future) fern abundance was expressed as a function oS pretreatment (past) fern abundance and light avail- ability using iliaxiilluni likelihood regression analjsis.

The invasion experiment involved increasing the light over the fern census plots. Light was increased by either cutting experimental gaps in the forest canopy or by locating areas in the forest that had been harvested for timber within the last 1-2 mo. Initial fern abundance was quantified in 30 1-m=plots inuilediately following canopy rernoval. Light was quantified following the remo\.al of the forest canopy using hemispherical photographs and ranged fiom 2. to 37% of full sun. Fern abundance was then quantified I, 2, and 3 yr following the initiation of the canopy removal treatment. Posttreatment (future) fern abundance was modeled as a function of pretreatment (past) fern ahi~ndance and light availability using ~naximum likelihood regression anal- ysis.


Patterns of fern distribution and abzttzdance-Hayscented fern (Drrz17staecltia p~nlctilobula) was differentially distributed with respect to canopy tree species (Fig. 1A). On the other hand, New York (Thelypter-is novehomcensis) fern was more evenly distributed (Fig. 1B). In both 1991 (one-way ANOVA, F,,,, = 17.95, P < 0.0001, r2 = 0.68) and 1992 (one-way ANOVA; F,,s, = 22.56; P < 0.0001; r-' = 0.73). hayscented fern was least abundant beneath eastern hemlock (Tsuga can- udensis) and American beech (Fagus grandtfolia) canopies and most abundant in open areas within the forest (Fig. 1A). These relationships were quite significant as 68 and 73% of the variation in fern abundance (expressed throughout as frond density) was explained by canopy tree species.

New York fern abundance tended to be highest, although not significantly, under red maple (Acer- r-uhrzsrn) and sugar inaple (A. sacclziinnm) canopies and lower under other species or in the open (Kruskal-Wallis test: 1991 and 1992, P > 0.05, Fig. 1B). Clearly. the distributional patterns of hayscented and New York fern were different: In both 1991 (p = -0.2555, P < 0.01) and 1992 (p = -0.2296, P < 0.05), the abundance of hayscented and New York fern were negatively correlated.

Differential distribution of hayscented fern with respect to canopy trees might result from species-specific differences in environmental conditions beneath trees (Canham et al., 1994). As a preliminary test, we looked at the patterns of light and soil moisture availability beneath the different canopy types. There were clearly differences in light availability beneath the canopies of the various tree species (Fig. 2A). Percentage of full sun available was lowest under eastern hemlock and high- est where there was no canopy tree (NONE [I-2 tree gaps]) directly above the sampling point (one-way ANOVA: F,,,, = 20.06; P < 0.0001; r2 = 0.70; Fig. 2A). Furthermore, a subset of areas in the forest without hayscented fern had lower mean light levels (N = 13, mean = 3.13% full sun, SE = 0.94) at

Canopy Species

Fig. 1. Distribution of (A) hayscented fern (Denizstncdtin punctilnbuln) and (B) New York fern (Thelypteris noveboracensis) in relation to canopy tree species in an oak-transition hardwoods-hemlock forest, Bars represent mean fern itbtlndance in 1991 (solid bars) and 1992 (stippled bars), erros bars are 1 SE of the mean. Canopy tree species are as follows: TSCA, Tslcgn cnnndensis; FAGR, Fagus grnnd(fn1in; QURU, Qt1crc.u~ rubrn; PRSE, Pr1tnu.s rerotinn; ACRU. Acer r~rbrtmz; ACSA, Acer sacchnrunz: NONE, gap in Corest canopy. Shared letters indicate mealis that are not signifiiantly different (0.05 level) by Bonferroni t tests performed Ibr each year separately.

1.5 m above the ground than a subset of areas with hayscented fern (N = 92, mean = 6.13% full sun, SE = 0.57) (Wilcoxon two-sample test; x2 = 5.798: df = 1; P < 0.01). The abun- dance of hayscented fern, although quite variable, increased with increasing light availability up to a point, reaching an asymptote at -15% full sun (r2 = 0.341, P < 0.001, Fig. 2B). 111 contrast, light availability had little impact on the distri- bution or abundance (data not shown) of New York fern. In fact, mean light levels for areas without (A7 = 71, mean = 5.47% full sun, SE = 0.61) and with New York fern (N = 34, mean = 6.38% full sun, SE = 0.57) did not differ (Wil- coxon two-sample test; P > 0.05).

Unlike light availability, there were no significant differ- ences in soil moisture availability among the various canopy tree species (one-way ANOVA; P > 0.05, Fig. 3). There were no differences in the soil moisture of areas in the forest with and without hayscented fern in either 1991 or 1992 (Wilcoxon two-sample test; P > 0.05). Furthermore, hayscented fern abundance did not relate systeinatically to soil moisture avail- ability (data not shown). In contrast, mean soil inoisture was higher where New York fern was present (1991 : N = 34, mean


Canopy Species

Light (% Full Sun)

Fig. 2. (A) Mean light available beneath various canopy tree species in an oak-transition hardwoods-hemlock forest (error bars are 1 SE). See Fig. 1 for identification of canopy tree species abbreviations. Shared letters indicate means that are not significantly different (0.05 level) by Bonferroni t tests.

(B) Distribution of hayscented fern (Dennstnedtia puizctilohula) in relation to light availability. Points represent frond densities in 1-m2 plots. Light was quantified using hemispherical photographs.

= 20.01, SE = 0.85; 1992: N = 34, mean = 32.03, SE = 1.17) than where New York fern was absent (1991: N = 71, mean = 18.09, SE = 0.57; 1992: N = 71, mean = 28.81, 0.62) (Wilcoxon two-sample test; 1991: x2 = 3.89; df = 1; P < 0.05; 1992: x2 = 6.21; df = 1; P < 0.01). Abundance patterns of New York fern, however, did not relate systemat- ically to soil moisture availability (data not shown).

Relationship between height, abundance, and resources- Both height (Fig. 4A) and density (Fig. 2B) of hayscented fern fronds increased as light availability increased, but again were not related to soil moisture. Furthermore, frond height and frond density were positively correlated, except in areas of extremely high frond density (in excess of 150 fronds/m2) (Fig. 4B). In contrast, New York fern frond height did not relate to light, frond density, or soil moisture availability (data not shown).

Within-season patterns of fern dynamics-A within-season

Canopy Species

Fig. 3. Mean soil moist~ire available beneath various canopy tree species in an oak-transition hardwoods-hemlock forest (error bars are 1 SE). See Fig. I for identification of canopy tree species abbreviations. Shared letters indicate means that are not significantly different (0.05 level) by Bonfe~~oni

t tests.



0 5 10 15 20 25

Light (% Full Sun)

Frond Density (no.lm2)

Fig. 4. Relationship between (A) average frond height (by plot) and light availability and (B) average frond height (by plot) and frond density for hay- scented fern (Denrtsfaedtin purictilobirla) in 1991.



Frond Density (June 1992)

Fig. 5. (A) Mean abundance of haysce~ited fern (Deniutaedtiu prolctiloh- ula) in relation to time d~rring the growing season. Bars represent mean fern abundance in 1991 (solid bars) and 1992 (stippled barsj (error bars are 1 SE). Shared letters indicate means tha~ are not significantly different (0.05 level) by paired comparison t tests performed for each year -;eparately. (B) End of the growing season (,August) fern abundance in relation to beginning of grow- ing season (June) for hayscented fern. The thin line represents no change in abundance, cmd the thick line represents the regression line.

census of fern ab~mdance was conducted rn both 1991 and 1992 to determine if ramet density varied across the duration of the growing season and to pinpoint the timing of future annual censuses. Hayscented lern abundance peaked in Ju14 and then declined during the latter past of the growing season (Fig. 5A). This seasonal change showed little relationship to canopy tree types (Table 1). In 1991, mean frond densities of hayscented fern decreased from July to August (paired com- parisons f test, t = -7.08. P < 0.0001). Tn 1992, mean frond densities did not differ significantly between June and July (t = 1.02, P > 0.05) but did differ between July and August (t

= -8.03, P < 0.0001).

Overall, hayscented fern frond density was -19% lo\ver at the end of the growing season than at the beginning (1992 data set; Fig. 5B). Thls pattern is primarily duc tit the th~nning of mature fronds, as no new fronds are initiated after the fern canopy closes (Hill, 1996), and presumably relates to decreas- es in light over the course of the growing season. Hayscented

TABLE1. Seasonal change (June-August 1992) of fen1 frond density inunlber per square meter) for hayscented (Deizilstcredtia punctiloh- zrla) and New York fern (Tlzelpteris noveboracensis) by canopy tree species. Nurnbers indicate mean increase or decrease in frond density. Numbers in parentheses are SE. Shared letters indicate means that are not significantly diffcrcnt (0.05 level) by Bonferroni t tests.

Scaronnl change in frolid density

Ca101,y tyvz Hdyscenled fern New York fern

Acrr rubr~cnz -1 1 .4Yh (4.28) 10.8@ (5.40) i2cer succharrrm -7.00" (3.42) -5.6Zh (5.94) Fu,grts grundifi~lia 0.38" (0.68) 0.00" (0.00) No canopy -1-2.12b (4.82) -9.88h (4.19) Pnnn~tssei-otiwrr -1O.5Wh (3.42) -0.7Sa'"(O.75) Querrts rubtr~ -8.3Ph (3.08) -0.7Yb (0.75) Tszrga cana~fensi.r -0.20" (0.13) O.6CP1'(0.34)

fern \host-term dynanlics (i.e., seasonal) appem to be related to initial frond density (p = -0.6227, P < 0.0001, N = 94) and light availability (p = -0.4218, P < 0.0001, N = 94), but not soil moisture (p = 0.05760, P > 0.05, N = 94). Over- all, the reduction in frond density over the course of the grow- ing season was sonlewhat higher on denw plot\ (regre\sion; F = 164.215; P < 0.0001; r.' = 0.6384; Y = -0.163465 X [Initial frond density] ). The seasonal population dynamics of hayscentcd fern were not correlated with the abundance pat- terns of New York fern (p = 0.0746, P > 0.05, N = 94).

New York fern dynamics had little relationship to canopy tree tlpe (Table 1). In 1991. mean frond denrities oSNew York fern decreased from July to August (paired comparisons t test, t = -2.1 4, P < 0.05). In 1992, mean frond densities did not differ bignificantly between June and July (r = 1.85, P > 0.05) but did differ between July and August (t= -2.26, P <0.05). Unlike hayscented fern, however, density at thc end of thc growing season (August) did not differ from the beg~lliling of the season (June) (paired comparison t test, P > 0.05, Fig. 6).

Patterns of fern dynamics between growing seasons-Be- cause mean fern abundance was hlghect at mid-growing sea- son for both species, July census data were used to m. 'ik e com- parisons of hayscented and New York fern abundance across years. The mean abundance of hayscented fern at mid-growing season was higher in 1991 (Fig. 7A, N = 105, mean = 72 fronds/m2, SE = 5.0) than in 1992 (iV = 105. mean = 63 frondstm" SE = 4.5). Mean abundance decreased from 1991 to 1992 (paired colnparison t test, t = -5.35, P < 0.00011, increased from 1992 to 1993 (paired comparison t test, t =

4.32. P < 0.0001), and then decreased from 1993 to 1994 (paired cornparison t test, t = -3.12, P < 0.005). Compari- sons of mean abulldance of hayxented fern in 1991 and 1993 and In 1992 and 1994 revealed no significant differences (paired compasison t tests. P > 0.05). Over the course of this study (1991-1994), there was on average roughly a 9% seduction in frond density for hayscented fern (Fig. 7B).

Mean New York fern abundance at mid-growing season tended to be highest in 1993 (Fig. 7A, N = 105, mean = 25 fronds/m2, SE = 7.0) and lowest in 1994 (hT = 102, mean = 12 fronds/m2, SE = 3.0), although no statistically significant patterns were revealed. The low density in 1994 probably re- flects the loss of two plots with high New York fern densities to tree fall debris. The year-to-year changes in New York fern abundailce did not appear to be cyclic like those of hayscented



Hayscented Fern New York Fern a

June July August Month

Frond Density (June 1992)

Fig. 6. ('4) Mean abundance of New York fern (TI~el~preris

rio~,eboi-(Icensis) in relation to time during the growing season. Bars represent inean fern abundance in 1991 (solid bars) and 1992 (stippled bars) (error bars are 1 SE of the mean). Shared letters indicate means that are not significantly different (0.05 level) by paired comparison t tests performed for each year separately. (B) End of the growing season (August) fern abundance in relation to beginning of growing season (June) for New York fern. The thin line represents no change in abundance. and the thick line represents the regression line.

fern (Fig. 7A), and mean abundance did not differ year to year (paired comparison t test, P > 0.05). For both fern species, change in frond density over the course of three growing seasons was not related to canopy tree type (data not shown).

Year-to-year dynamics of hayscented fern appear to be correlated with initial frond density (Table 2). Overall, the reduction in frond density over the course of the three growing seasons was somewhat higher on denser plots (regression; F = 18.518;P < 0.0001; rb0.1722; Y = -0.091848 X [lnitial frond density]). There was little relationship between

change in frond density and initial light conditions (p = -0.01894, P > 0.05) or change in frond density and initial soil moisture conditions (p = -0.14695, P > 0.05) (Table 2).

The year-to-year population dynamics of hayscented fern were not correlated with the abundance of New York fern (Table 2).

The year-to-year dynamics of New York fcrn were related


Frond Density (July 1991)

Fig. 7. (A) Mean midseaso11abundance for hayscented fern (Der~nsmetltici pz~nctilohl~l<l)(solid bars) and h'ew York fern iT11eI~pter.i~i7ol~eborzrceizsi.\) (stippled bars) over se\eral growing seasons. Bars represent niean fern densities (error bars are 1 SE of the mean). Shared letters indicate means that are not significantly different (0.05 level) by paired co~npnrisonr tests performed for species of fern separately. (B) Fern abundance in July 1994 in relation to July 1991 for hayscented fern. The thin line represents no change in abundance, and the thick line represents regressioil line.

TABLE2. Correlations (r-), between the between-season change (July 1991-July 1994) of fern frond density for hayscented fern (Dennstnedtic~pu~~ctilohula)and New York fern (Tlrel\ptei-is r~ovehorczcen.tis in relation to initial Sern density, light availability, and soil moisture availability. Numbers indicate the colrelation coefficient and the probability of the correlation coefficient (NS = P > 0.05). The number of plots (N) included in the correlation analysis was


Change in frond den\it) (19'11-199-0

Hayccrnted fern    Ncu. khik fcrn
Factor r ii    r I>
Hayscented Sern frond    
density (June 1991) 0.25464 <0.02    0.15773 NS
New York fern frond    
density (June 1991) 0.16100 NS    0.46920 <0.0001
Light (% full sun) 0.01893 NS    -0.00882 NS
Soil tnoisture (96) 0.14695 NS    0.02593 NS
Fern Density [fl

Fig. 8. (A) Estimation of tile decrease in hayscented fern iDritiislriedtiri


abundaiice f~llowing decreased light availability. Hayscented fern was quantified on 20 shade plots prior to shading and 1---2qr followitig shading. (B) Estiination of the increase in hayscented fern ahundarice follow- ing iricreases in lieht availability. Hayscentetl fern was quantified on a series of' invasion plots (40 plots) over 1 y~ fi: = (178.8 X I..i~lit)/~3.558-&-Liglit) and represeiits a light-speciiic equilibrium fern density (derived from the re- lationsl~ip presented in Fig. 2R).

to initial frotid density, but not to light or soil moisture avail- ability (Table 2). Overall, the reduction in frond density over the course of the grouing season was sornewhat higher oil high-density plots than on low-density plots (regress~on; F = 24.396; P < 0.0001: r*' = 0.2151; Y = -0.258455 X [Initial frond density]). (data not shomn). The year-to-year population dynamics of New Yorh fern were not correlated with hay- sccnted fern abundance (Table 2).

Patterns of fern dyr~amics followirzg light manipulation- In the shadtrig experiment, abovegfound ramet densit! mea-cured 1 and 2 yr posttrcatnient wab -73% of that prior treat- lnent (Fig. 8'4) indicating x very slow decline in abundance. On the other hand, when light availability was tncreased, as in the invaiion experimcnt, a fairly rapid Increase In abun- dance war obserjed (Fig. 8B), with frond denbity doublmg nithin 3-5 yr. For plots with high initial frond dens~ty, this rnp~d increase mas absent, suggesting a saturation effect.


I11 gcneral, hayscented fern (Dennst~edticr pwctilobuln) was more responsive to changes in conditions than New York fern (7he!\.preris novehnracensis). Hayscented fcrn was differen- tially distributed with respect to forest canopy tree species, whereas New York fern was not. Hayscented fern was most abundant in areas of high light (>jt?c full sun), although it could also persist at low frond densities in areas with low light availability (<2% fill1 sun), whereas New York fern was spo- radic in occurrence, indicating abundance did not relate to light, but was generally in greater abundance under interme- diate soil moisture conditions.

Both ferns decreased in abundance over the course of a growing season. with New York fern varying little in abun- dance from year to year. On the other hand, there was signif- icant variation in hayscented fern abundance. For hayscented fern, those plots with the highest initial abundance had the greatest seasonal and between-year decrease in froad density. The seasonal and betweeti-year dynamics of hayscented fern did not relate to the dy~lamics of New York fern and vice \;ersa, suggesting independence in the dyliamics of these two species over the limited nutnber of plots (iV = 35) where thc two species co-occui~ed.

Distribution and abundartce-Hayscented fern was least abundant under canopies doriiinated by easterii hemlock (Ts~c- ga canndrnsis) and most abundant within canopy gaps. Many shade-tolerant species, such as Eastern hemlock and Atnesicail beech (Fagus grancl(folblia), allow only very low quantities of light (<29 full sun) to be transmitted through their canopies (Canham et al., 1994; this study), suggesti~lg the importaiice of light availability in determining distributional patterns. Hammen (1993j also found a positive correlation betcveen light availability and frond density in hayscented fern popu- lations in Rhode Island.

In contrast, in our study abundarice patterns of Kew York fesn were not related to light availability but rather tended to increase with soil n~oisture availability. Nitrogen availability, which was not assessed in this study, may also be an important factor in determining New York fern distribution. Red maple (Acer r~,tbnmz) and sugar maple (A. sacclzn~.un?) trees, under which New York fern tended to be most prevalent, have been shown to be associated with higher soil nitroretl miner. 'i I'iza


tion rates than for other tree species at Great Mountain Forest (Finzi. 1996). At present neither our data nor published data are sufficient to adequately assess the distribution and abun- dance patterns of New York fern.

Distribution patterns might also be caused by differences in other abiotic or biotic conditions beneath different canopy tsee species as different canopy tree species have been show11 to for1n different understory ~nicrohabitats (Crozier and Boerner, 1984). Several studies have noted lower species richness in the urrderstoiy of hemlock compared to non-hemlock areas (Hicks. 1980; Reatty, 1984). This lower species diver\ity has becn attributed to loucr soil pH, poorer llght quality and qLrd11- tity. lower soil moistL~re availability. and thicker organic layers found beneath hen~lock (Hicks, 1980; Beatty, 1984).


the general pattern of increased abundance at in- creaseci light levels, the observed density of hayscented fcrn at any given light level \\as quite variable. This variability might be related to several factors. First, there is potentla1 for phj siologlcal integration anioilg ramets (as suggested by I-lam- men, 1993). Clonal integration among shaded and un\haded ranlets might result 1x1 sharing of assimilate\, allowing fiond density to be higher than can be rnaintained by the local light environment. Rhizomes of hayscented fern can often persist


for long periods of time, in excess of 10 yr (Cody, Hall, and Crompton, 1977: Hammen; 1993; J. D. Hill, personal obser- vation), and suggest the potential for long-term physiological integration among fronds in a heterogeneous environment. Al- though there are no published accounts of physiological inte- gration in hayscented fern, data from our shading experiment provide circumstantial evidence. In a shading experiment, fern census plots (1 m2) were subjected to extreme shade (92% shade, such that all plots received <1% of incident radiation), but ramet connections were left intact (no severing). The rate of decline in frond density we observed was far slower than expected based on known fern abundance-light availability re- lationships (see Fig. 2B). Clonal integration among shaded and unshaded rarnets is one explanation for the slow rate of decline in these experimental shade plots.

Secondly, the ability of ferns to use sunflecks (Hollinger, 1987; Gildner and Larson, 1992; Brach, McNaughton, and Raynal. 1993) may be important. Even if ramets are not phys- iologically linked, there is strong evidence that ferns can use sunflecks to maintain positive net photosynthetic rates. Hollin- ger (1987) showed that 68% of daily photosynthesis in brack- en fern (Pteridium oql~ilinum) occurred during sunflecks (>I00 pmol.n~-~.s-~).

Ferns may in fact respond to sunflecks extremely rapidly with little or no induction period as in Po- lypodiuin virginianum (Gildner and Larson, 1992). It has been demonstrated experimentally that shade-grown hayscented fern can have significantly higher net photosynthetic rates (per dry mass) than sun-grown conspecifics (Brach et al., 1993). One explanation of this pattern is the ability of shade leaves to use periodic sunflecks (Chazdon and Pearcy, 1991; Gildner and Larson, 1992).

Lastly, the ability to respond to sunflecks effectively, with or without physiological integration among ramets, may allow hayscented fern to persist in the forest understory under low light conditions. This persistence coupled with the ability of many temperate forest herbs to reproduce vegetatively rather than sexually (Sobey and Barkhouse, 1977; Bierzychudek, 1982) may allow colonization of favorable and retreat from

inf favorable environments. In hayscented fern, vegetative spread via the rhizome system (Cody, Hall, and Crompton, 1977; Hammen, 1993; Hill, 1996) might result in rapid colo- nization of canopy openings as they become available. For example, in our study, we found that local frond density in- creased dramatically from one season to the next when light was increased (invasion experiment) suggesting the potential for rapid expansion of existing clones. Likewise, in large can- opy openings at Hubbard Brook Experimental Forest, Hughes and Fahey (1 99 1) found that hayscented fern abundance was -5 times higher just 3 yr following canopy opening.

Seasonal and between-year dynamics-Both hayscented and New York fern decreased in abundance over the course of a growing season, with New York fern abundance varying little from year to ycar, and hayscented fern abundance show- ing a cyclic pattern. The lack of dramatic changes in frond density over the short-term might indicate that abundance is at or near light-specific equilibrium densities or that the dy- namics are slow. However, the fact that plots with highest abundance had the greatest seasonal and between-year decrease in frond density challenges this conclusion. In fact, it has been shown that in areas of low fern density (such as recently cut forests) hayscented fern density and leaf size in- crease rapidly (this study; Collins and Pickett, 1988; Hughes and Fahey, 1991). Similarly, position within clone influences the dynamics as areas at the edge of clones might experience rapid increases in abundance (Hammen, 1993; Hill, 1996).

In our study, the seasonal and between-year dynamics of hayscented fern and New York fern were unrelated, and there was no evidence of negative cosselations between abundance and dynamics of the two species, suggesting no current com- petition between the two species. However, initial abundance of New York and hayscented fern are negatively correlated, and the two species co-occurred on a small number of plots (N = 34). These observed differences in distribution might be explained by past competition resulting in niche differentia- tion, species-specific differences in response to disturbance history, or the limited sample size of the current study. In any case, further detailed studies on the niche requirements of the two species and their response to disturbance will be required to resolve this issue.

Long-term dynamics-Contributing to the success of hay- scented fern at the landscape level is the influence of browsing by white-tailed deer (Odocoile~u virginionus) (Horsley and Marquis, 1983). Deer do not browse the unpalatable hayscent- ed fern (Tilghman, 19891, but browse potential competitors. In the Allegheny Plateau, Hough (1965) and Rooney and Dress (1997) documented a shift in forest understory domi- nance by hobble bush (ViDzvnlrn~ ~lln(foliz~m) to ferns and her- baceous plants. This shift was attributed to increased browsing on the more palatable competitors of the fern, namely shrubs and tree seedlings (see Horsley and Marquis, 1983). Thus, over time, increased deer densities have effectively released the fern from competition. This release coupled with increased light in the understory due to forest harvesting and disturbance has resulted in the landscape level increase of hayscented fern. On the othcr hand, New York fern appears to be less respon- sive to conditions brought on by disturbance, and this leads to slower dynamics, because of slower rhizome growth rates (R. McCalley and J. D. Hill, unpublished data, Yale University) and thus less rapid spread through the forest understory.

In conclusion, the long-term dynamics of hayscented fern will be dependent on several factors. In the initial stages of succession in a secondary forest, hayscented fern might be quite prevalent (Whitney and Foster, 1988), if there is a local source of spores or low-level populations from which vege- tative spread can occur. Howevel; as succession proceeds: if canopy closure and the douninallce of the forest by more shade-tolerant canopy species occurs (Eastern hemlock and American beech), the abundance of hayscented fern in the un- derstory would diminish (Hill, 1996). This pattern of gradual elimination of hayscented fern from the forest is predicted from the fact that shade-tolerant tree canopies cast deep shade (4%

full sun) with little contribution from sunflecks (<10% of PZ4R)(see Canham et al., 1994). However, any event that maintained or promoted openness in the forest canopy such as canopy tree death by direase or windthrow, forest harvesting, or the elimination of a shrub layer by browsing will promote conditions to allow the persistence of hayscented fern.


BEALS.E. W.. AND .I.B. COPE. 1964. Vegetation and soils in an eastern Indiana woods. Ecology 45: 777-792.

BEATTY,S. LV. 1984. Influence of microtopography and canopy species on spatial patterns of forest understory plants. Ecoiogj 65: 1406-1419.


BIE~YCHUDEK,1982. Life histories and demography of shade-tolerant


temperate forest herbs: a review. h'ew P11jtologi.sr 90: 757-776. BRACH,A. R., S. 1. MCNAIIGHTON,

AND D. J. RAYNAL. 1993. Photosynthetic adaptability of two fern species of a northern hardwood forest. Aiiiericni~ Fern Joui.rin1 83: 47-53.


S. F? 1976a. Resource division in an understory herb community:

OF BOTANY [Vol. 88

. 1977b. Allelopathic inhibition of black cherry. 11. Inhibition by

woodland grass, ferns, and club moss. Cnrzr~diciil Jouri~nl of Forest Re-

.rearch 7: 5 15-519.

. 1984. Hayscented fern rhizome development in uncut and thinned

Allegheny hardwood stauds. Airlericciiz Jo~iri~(il

of Botany 71 : 80-81 (Ab- stract).

responses to temporal and niicrotopographic gradieilts. Americnri hrrzrtc-


1985. Reforestation of orchard stands and savannahs of Pennsyl-

rnlist 110: 679-693.

. 1976b. The response of u~iderstory herbs to soil depth gradients in high and low diversity communities. Biilleriiz of the Torrey Botanical Club 103: 165-172.

BREW~R, 1980. A half-century of changes in the herb layer of a climax

R. deciduous forest in Michigan. Journnl c/JlEco/ogj 68: 823-832. CANHAM.C. D. 1988. An index for understory light levels in and around canopy gaps. Ecology 69: 1634-1638.

, A. C. FINZI, S. W. PACALA, AND D. H. BURBANK.1994. Causes and consequences of resource heterogeneity in forests: interspecific var- iatio~i in light transmission by canopy trees. Caiindinn Joiiri~ril of Forest Resenrclz 24: 337-349.

CHAZJIOK,R. L., AND R. W. PEARCY. 1991. The importance of sunflecks for forest understory plants. BioScierice 41: 760-766.

CODY,U: J.. I. V. HALL,AND C. W. CROMPTON. 1977. The biology of Canadian weeds. 26. Dennsmeclricr punctilob~~la (Michx.) Moore. Caizutlinrl Jouriinl of Plr~rlr Scieizct? 57: 1159-1 168.

COLLINS.B. S., AND S. A. PICKETT. 1988. Demographic responses of herb layer species to experimental canopy gaps in a northern hardwoods for- est. Jo~iri~al

of Ecology 76: 437-450. COLLI~S,

S. L., AND R. I. GOOD. 1987. Canopy-ground layer relationsliips of oak-pine forests in the New Jersey pine barrens. An~ericnri iMidlarld Nnt~trrilisr1 17: 280-288.

CONARII,H. S. 1908. The structure and life-history of the hay-scented fern.

Cariiegie Ii~stiriire of bt'a.r/n.rlzi~~gton

Publicntioiz 94: 1-56.

CROZIER, C. R., AKD R. E.J. BOLRKER. 1984. Correlatiolls of uriderstory herb distribution patterns with rnicrohabitats under different tree species ill a mixed 111esophytic forest. Oecologia 62: 337-333. DREW, A. l? 1988. Interference of black cherry by ground flora of the Al- legheuy uplands. Cai7iiiliatl Joilriznl of Fore.rr Resenrcii 18: 652-656. 1990. Fern and aster effects on blaclc cherry shelterwood regener- ation. Carzndinn Joiinziil of Forest Resetrrclz 20: 1513-1 5 14. FINZI. il. C. 1996. Causes and consequences of soil resource heterogeneity in a transition oak-northern hardwood forest. Ph.D. dissertation, Uni- versity of Connecticut, Storrs, Connecticut, USA. GERSPER,P L., AND N. HOLOWAYCHUK.

1971. Some effects of stemflow from forest canopy trees on chemical properties of soils. Ecology 52: 691-702.

GILDKER,B. S., AKD D. IV. LARSON. 1992. Photosyntlietic response to sun- flecks in the desiccation-tolerant fern Polypoiliiiri~ virgininriuin. Oecolo- gin 89: 390-396.

GRAY.A. N., AVD T. A. SPIES. 1095. Water content in forest soils and de- cayed wood using lime domain reflectornetry. Curlndiail Jo~rrrznl @For- est Resenirh 25: 376-385.


J. LV., AND L. H. MCCORMICK. 1991. Effects of sulfometuron on hay-scented fern spore emergence. Cnizndiciiz Jouriinl of Forest Re- senrch 21 : 942-943.


S. C. L. 1993. Density-dependent phenotypic variation in the hay- scented fern. Dein7stnedtin pui~ctilobuln. Bulletiiz qf'rhe Torrey Botni~ical Clirb 120: 392-396.

HICKS, D. J. 1980. Intra~tand distribution patterns of souther11 ilppalachian cove forest herbaceous species. Arnericniz ~Vfidlni~d A'iituraii,rt 104: 209-


HILL, J. D. 1996. Population dynamics of hayscented fern (Denn.rtnedtin ~~il~ictilol~~llri)

and its impact on forest composition, structure and dynam- ics. P11.D. dissertation, University of Connecticut. Storrs. USA. HOLLIKGER,D. Y. 1987. Photosy~itllesis and stornatal conductance patterns of two fern species from different forest understoreys. Jo~irnal of Ecologq.

75: 925-935.

HORSLEY,S. B. 1977a. Allelopathic inhibition of black cherry by fern. grass. goldenrod, and aster. Canadiniz Joi~ri~nl of Foreat Resenrcll 7: 205-216.

vania's Allegheny Plateau. hrorther71 Joiirizcil ofApplieil Forestry 4: 22-

26. . 1993. Mechanisms of interference between hayscented fern and black cherry. Caiziidinn Jour.iinl of Forest Research 23: 2059-2069.

-. AND D. A. MARQIIIS. 1983. Interference by weeds and deer with Allegheny hartla,ood reproduction. Cnrinclinrl Jouri7al of Foresr Resenizh

13: 61-69. HOUGH, A. E 3965. A twenty-year record of understory vegetational change in a virgin Pennsylvailia forest. E(,ologj 46: 370-373.

HUGHES,J. W., AND T. J. FAHEY. 1991. Colonization dynamics of herbs and shrubs iii a disturbed northern hardwood forest. Joto.rznl of Ecology 79: 605-6 16.

KOLB, T. E., T. W. BOWERSOX,AND L. N.MCCOIIMICK. 1990. Influences of light intensity on weed-induced stresses of tree seedlings. Canridicin Joruizt11 qf Forest Research 20: 503-507.

LELLINGER,D. B. 1985. A field manual to the ferus and fern-allies of the United States and Canada. Smithsonian Institution Press, Washington, D.C., USA.

Ll~scoh?r%,bl.V., AND E. T. NILSEN. 1990a. Environmental and physiological factors influencing the uatural distribution of evergreen and deciduous ericaceous shrubs on northeast and southwest slopes of the southern Ap- palachian Mountains. I. Irradiance tolerance. Anierictirz Joil~.rzal cjfBotnny

77: 108-115.

, AKD --------. 1990h. Environmental and physiological factors influ- encing the natural distribution of evergreen arid deciduous ericaceous shrubs on nortlieast and south\vest slopes of the southern Appalachian Mountains. 11. Water relations. Aiizericnn Jo~trizcil qf Botnilj 77: 517-


LODHI, M. A. K. 1977. The influence and comparison of individual fores~ trees 011 soil properties and possible inhibition of nitrification due to intact vegetation. Airiericnn Journcil qf Botciity 64: 260-264.

MAGUIRE.D. A., AND R. T. T. FORRIAN. 1983. Herb cover effects on the tree seedling patterns in a mature l~emlock-hardwood forest. Ecology 64: 1367-1380.

MOORE.M. R.. AND J. L. VANKAT. 1986. Responses of the herb layer to the gap dynamics of a mature beech-maple forest. Ai7rericar1 Mirilarrd iVnt- ~lmlist115: 336-347.

PIIILLIPS,D. L., AND W. H. MURDY. 1985. Effects of rhododendron (Rkodoilrildron innxinlurn L.) on regeneration of southern Appalachian hard- woods. Forest Science 31: 226-233.

PL.OCHER,A. E., AND K. L. CARVELL. 1987. Population dynamics of rosebay rhododeudron thickets in the souther11 Appalachians. Bztlletiri of rlze Tor- rey Botiinicai Club 114: 121-126.

RICE, E. L., AND S. K. PANCHOLY.1973. Inhibition of nitrification by climax vegetation. 11. Additional evideuce and possible role of tannins. Ainericait Joltrnal of Botnily 60: 691-702.

ROONEY,T. P., AKD W. J. DRESS. 1997. Species loss over sixty-six years in the ground layer vegetation of heart's content, an old-growth forest in Pennsylvania, USA. ilinrurnl Arecis Jounzai 17: 297-305.

SAS. 1987. SASISTAT guide for personal computers. version 6 edition. SAS Institute, Cary, North Carolina, USA. SOBEY.D. G., AKD P BARKIIOL~SE.

1977. The structure and rate of growth of the rhizomes of some forest herbs and dwarf shrubs of the New Bruns- wick-Nova Scotia border region. Caizndiciri Field Nnriiralist 91: 377-

383. TILGH~I.~N,

N. G. 1989. Impacts of white-tailed deer on forest regelleration in northwestern Pennsylvania. Jortrrzal of Wildlife h.1ancigenleizt 53: 524-

532. WHITNE~,

G. G., AN13 D. R. FOSTER. 1988. Overstorej~ composition and age as determinants of the understorey flora of woods of central New Eng- land. Jourrinl of'Ecology 76: 867-876.

  • Recommend Us