Forty-Two Years of Succession Following Strip Clearcutting in a Northern Hardwoods Forest in Northwestern Massachusetts Taber D

Forest Ecology and Management 182 (2003) 285–301

Forty-two years of succession following strip clearcutting in a northern hardwoods forest in northwestern Massachusetts Taber D. Allison*, Henry W. Art, Frank E. Cunningham1, Rebecca Teed2 Center for Environmental Studies, Williams College, Williamstown, MA 01267, USA Received 20 July 2002; received in revised form 26 August 2002; accepted 19 January 2003

Abstract

We investigated the effects of strip width, slope position, and soil scarification in a split–split plot design on the regeneration of northern hardwoods in northwestern Massachusetts. Whole plots of 20 and 40 m in width were cut in 1954 in a second growth forest dominated by Betula papyrifera. Slope position and soil scarification were the split and split–split plot treatments, respectively. We measured height for all tree species present in randomly located 4 m2 plots beginning in 1955 and at irregular intervals over the following 42-year period. We measured all trees in the cut strips in 1996. Prunus pensylvanica was the dominant species initially, but had nearly disappeared from the cut strips by 1996. Soil scarification significantly increased initial establishment of B. papyrifera, but density and basal area of this species did not differ by soil treatment in 1996. Tree composition in cut strips was weakly correlated with soil moisture, soil scarification, and initial tree density immediately following cutting, but high spatial variation in species composition and low replication made it difficult to detect any significant correlations among the distribution and abundance of different species and selected environmental variables. The canopy of the cut strips is even-aged; establishment of most canopy trees occurred within 5 years following cutting. A comparison of successional trends in adjacent uncut strips with the trends in the cut strips indicates that cutting has altered the sequence of successional changes in forest composition increasing the abundance of some species that were of low importance prior to cutting. In 1996, Acer rubrum and A. saccharum are replacing B. papyrifera in the canopy of the uncut strips. The canopy of the cut strips consists of a diverse and spatially varying mixture of intermediate hardwoods including Quercus rubra, Fraxinus americana, Betula lenta, Acer rubrum, B. papyrifera, and an understory of late successional hardwoods. # 2003 Elsevier Science B.V. All rights reserved.

Keywords: Clearcutting; Forest succession; Betula papyrifera; Tree age; Tree height; Prunus pensylvanica

1. Introduction

The study of forest succession is one of the oldest * Corresponding author. Present address: Massachusetts Audu- and most interesting topics in forest ecology (e.g. bon Society, 208 South Great Road, Lincoln, MA 01773, USA. Cowles, 1899; Thoreau, 1906). Much of what we have E-mail address: [email protected] (T.D. Allison). 1 Deceased. learned about the pattern of forest succession and 2 University of Maryland, Unit 29216, APO, AE 09102, Maryland, the responses of species to disturbance, especially USA. in the northeastern United States, comes from studies

0378-1127/$ – see front matter # 2003 Elsevier Science B.V. All rights reserved. doi:10.1016/S0378-1127(03)00066-5 286 T.D. Allison et al. / Forest Ecology and Management 182 (2003) 285–301 of forest regeneration following clearcutting (e.g. In addition to examining the effects of the pre- Bormann and Likens, 1979; Marquis, 1967; Leak, scribed treatments on paper birch regeneration we 1991; McClure and Lee, 1993). Studies of logged were interested in the effect of these treatments and forests often provide the longest sequence of forest other non-manipulated factors in the physical envir- change on individual sites as many of these studies onment on forest composition 42 years later. Speci- date from the 1930s and 1940s. Clearcutting studies fically, to what extent could the current composition often are established with controlled, replicated (1996) of the cut forest be predicted from the compo- experimental designs that allow the statistical analysis sition during the establishment phase immediately of forest response under different environmental con- following cutting? It was difficult to determine a ditions. This avoids the problems of pseudoreplication priori, all possible initial conditions that might be or the uncontrolled and confounded comparisons of important. Some of the initial conditions in this study stands compiled along a chronosequence. Data usually were determined by the specific treatments employed are collected on permanent plots established immedi- by the USFS, but disentangling the importance of ately after cutting, and, occasionally, forest composi- other non-controlled factors would readily have tion is measured prior to cutting. degenerated into a fishing expedition with no statis- In this paper we report on the initial response of tical power. To avoid this we asked specific questions strips cut in 1954 in the Hopkins Memorial Forest including whether variation in the abundance of paper (HMF) in northwestern Massachusetts and the subse- birch in 1996 could be predicted from the abundance quent changes in species’ composition and dominance of the initial establishment of paper birch? We also over a 42-year period. Cunningham initiated the clear- asked whether soil moisture or initial density of pin cutting experiment under the auspices of the United cherry (Prunus pensylvanica L.) following cutting States Forest Service (USFS), which owned and man- affected initial stand composition and forest composi- aged the HMF from 1935 to 1968. The experiment’s tion 42 years later. primary objective was to examine the influence of strip width and seedbed scarification on the natural regeneration and early development of paper birch (Betula 2. Methods papyrifera Marsh.) (Cunningham, unpublished manu- script). Clearcutting in strips leaves uncut intervening 2.1. Description of study area strips as a source of seed and shade. Exposure of mineral soil was thought to enhance establishment of light- The study area is located in the HMF now owned by seeded tree species, such as paper birch, and had sub- Williams College, Williamstown, Massachusetts. The sequently been shown to enhance initial establishment HMF has topography typical of the western New of paper and yellow birch (Hutnik and Cunningham, England Highlands. The study area is situated below 1961; Marquis et al., 1964; Marquis, 1965). Because of the Taconic Crest at an elevation ranging from 380 to the topography of the study area, the effect of slope 450 m above sea level. The aspect of the study site is position on paper birch regeneration was also examined. northeast; the slope varies from 0 to 47% and averages Data on abundance of tree species after cutting were 25%. The bedrock of the site is primarily schistose and collected by the USFS from permanent 4 m2 plots in phyllitic rock characteristic of the Taconic Range, and 1955, 1957, 1958, and 1962 at which time the project the soils are primarily thin, stony acid loams (Art and was temporarily abandoned. Allison and Art relocated Dethier, 1986; Scanu, 1988). the plots in 1974 and made several measurements on Just prior to cutting, the study site was an even-aged species composition in 1974, 1975, 1989 (measured stand dominated by paper birch (50% density and by Teed), and 1996. The length of observation of this basal area of trees over 11.7 cm diameter at breast experiment allowed us to examine not only the initial height (DBH)) and red maple (Acer rubrum L.) (25% effect of treatment on paper birch regeneration, and of the basal area and 28% of the density). Sugar maple other tree species, but also whether or not the effect of (A. saccharum Marsh.) was the next most impor- this treatment persisted over the 42-year period of tant species at 9% of the basal area and 11% of the measurement. density. The remainder of the stand consisted of small T.D. Allison et al. / Forest Ecology and Management 182 (2003) 285–301 287 amounts of black birch (Betula lenta L.), yellow birch the HMF. No logging or agricultural activity took (B. alleghaniensis Britton), red oak (Quercus rubra L.), place in the study area after this time (Rosenburg, white ash (Fraxinus americana L.), American beech personal communication). Tree cores removed by (Fagus grandifolia L.), aspen (Populus spp.), Eastern the USFS in 1936, primarily from paper birch and hophornbeam (Ostrya virginiana K. Koch.), and red maple growing in nearby stands, showed tree ages striped maple (A. pensylvanicum L.). at breast height ranging between 40 and 83 years; Massachusetts tax records indicate that the study mean and median ages were 54 and 50 years, res- area was used for pasture and cultivation until the pectively. Permanent plot studies elsewhere in mid-1860s. The study area was forested in 1901 the HMF have shown that paper birch dominates when Amos Lawrence Hopkins purchased the early to mid-successional stands on lands that had land and incorporated it into the property that became been used for cultivation immediately prior to their

Fig. 1. Field layout of experimental design showing orientation of cut and uncut strips and 16 split–split plots (experimental units). 288 T.D. Allison et al. / Forest Ecology and Management 182 (2003) 285–301 abandonment from agriculture (Art, 1974; Art and amount of scarification. Plots were then assigned to Dethier, 1986). classes based on percent scarification—the scale ranged from 0 to 100% at 25% intervals. 2.2. Experimental design 2.3. Data collection In 1952, Cunningham and colleagues surveyed a portion of the selected stand, measuring 200 and 240 m. Cunningham and USFS colleagues counted and This area was subdivided as follows: four cut strips were measured the height of all tree stems in the ninety- separated from each other by intervening uncut strips six 4 m2 plots in the fall of 1955, 1957, 1958, and 1962; each measuring 20 m in width (see Fig. 1). Two of the in the summers of 1974, 1989, and 1996 counts were cut strips were 20 m wide, and two of the cut strips were made by Allison, Art, and Teed. In 1974, 1989, and 40 m wide. Each strip was 240 m long. Experimental 1996, each 4 m2 plot was enlarged to 16 m2 to accom- treatments were assigned in a split–split plot design; modate the growth of trees. We measured the diameter strip width was the whole plot treatment, slope position of all stems >1.3 cm DBH in the enlarged plots in those was the split plot treatment, and soil treatment was the years. In 1996 we conducted a complete census of trees split–split plot treatment. Each of the four cut, or experi- in the cut strips. Each split–split plot in each strip was mental, strips was subdivided into four sections or plots, surveyed and marked off with string, and we recorded 50 m long. Slope position was assigned to half of each diameters for all trees by species and plot correspond- strip; upslope plots contained the top two sections in ing to treatments established by the USFS in 1954. In each strip. One of the sections in each of the slope 1974 and 1996 we measured the diameter of all trees position plots was selected randomly to be scarified in greater or equal to 2.5 cm DBH in the uncut strips. We order to examine the effect of seedbed treatment on measured heights of all trees in the ninety-six 4 m2 paper birch seedling establishment (see Fig. 1). At the subsample plots in 1975 and in the 16 m2 plots in 1996. time the experimental plots were surveyed the diameter In 1975, we randomly selected 48 4 m2 plots in the of all trees >11.7 cm diameter at breast height (DBH) clearcut strips in order to determine the age structure in the cut and uncut strips was measured and recorded of the clearcut stand. Whenever possible, the age of an by species. For consistency in data comparisons all individual tree within a plot was determined by count- field measurements have been made using English ing the terminal bud scars for seedlings or by obtaining units and converted to metric units for this paper. a core at 0.3 m above the root collar using an incre- Trees larger than 12.5 cm DBH were cut in the ment corer for canopy trees. To obtain ages of indi- strips between February and September 1954; all viduals from those size classes that could not be aged residual material, including seedlings, saplings (i.e. by either of these techniques, a parallel plot was ‘‘advanced reproduction’’), and shrubs was cut by located 4 m from each of the 48 4 m2 subsample plots. Spring 1955. All logging slash resulting from the The parallel plot was randomly located in one of the cutting operation was completely removed from the four compass directions (NW, SW, SE, and NE). A 20 m strips and either removed or placed in small piles different direction was chosen if the first choice fell in the 40 m strips. An occasional large tree, or ‘‘cull within 2 m of other permanent plots or the strip border. hardwood’’, was girdled, poisoned, and left standing. Individuals in the parallel plot that could not be aged Selected plots or sections were scarified with a bog by coring or by the counting of terminal bud scars harrow pulled over the ground by a crawler tractor. were harvested. A disc was cut from the base of each Permanent 4 m2 subsampling plots were randomly harvested tree and taken to the lab where the discs located following cutting and scarification in each were sanded and the rings counted. Ages of trees were clearcut strip. In the 20 m wide strips, four subsam- recorded by species and height. pling plots were established in each of the sections; eight subsampling plots were established in each of 2.4. Data analysis the 40 m wide sections. In all 96 subsampling plots were established. After cutting and soil treatment Changes in tree density and tree height over time USFS personnel examined each 4 m2 plot for its actual and by treatment for common species were analyzed T.D. Allison et al. / Forest Ecology and Management 182 (2003) 285–301 289

Table 1 from data collected from 16 m2plots, which were Sample MANOVA table used in statistical analysis for individual averaged within each of the 16 split–split plots as years of tree count data based on a split–split plot experimental design described earlier, and in 1996 forest composition was estimated with data collected in the complete Source Analysis of variance census of all trees in the 16 split–split plots. d.f. F-ratio We constructed Pearson correlation matrices for selected species and environmental factors and sig- Width 1 MS/strip (width) Slope 1 MS/slope strip (width) nificance was adjusted using Bonferonni’s test. Age Soil treatment 1 MS/error data from 1975 were analyzed with SYSTATANOVA Width slope 1 MS/slope strip (width) procedure. We evaluated differences in mean height Soil width 1 MS/error among different tree species in different years using Soil slope 1 MS/error t-tests with an adjusted probability value to reflect Strip (width) 2 Slope strip (width) 2 the number of comparisons made. We calculated 2 Error 5 Simpson’s diversity index (1 Spi ) for cut strips and uncut strips for all years of measurement where As indicated different error terms were used to assess significance of different effects as defined by SYSTAT version 7 (Wilkinson, pi is the proportion of abundance as described for each 1996). ‘‘Strip’’ refers to the four strips of different widths and the year above. term strip (width) means ‘‘strips within width’’. The remaining We performed canonical correspondence analysis ‘‘error’’ term is derived by summing the sums of squares for all (CANOCO, ter Braak, 1986) on changes in the com- remaining interaction terms of the full model. Degrees of freedom position of the cut strips using tree species percentage are derived from the 16 permanent split–split plots, or experimental data for the ten most abundant taxa from the 16 split– units. When repeated measures were performed, year was added as 2 a within-subjects factor (Wilkinson, 1996). split plots for all years and for the ninety-six 4 m plots in 1955. Environmental variables used in the analysis were strip width, initial tree density, average percent using the General Linear Model procedure in SYSTAT scarification of each section, and a modified topo- (Wilkinson, 1996; Table 1) using a split–split plot graphic soil moisture index (Bormann et al., 1970). model for individual years. Repeated measures ana- We used detrended correspondence analysis (DCA) to lysis was used when examining effects across years. examine temporal trends in species data in cut and The 16 strip sections, or split–split plots (hereafter uncut strips between 1955 and 1996. ‘‘plots’’), were defined as the experimental units— results from counts of the 96 subsample plots were combined to calculate plot averages for analysis for all 3. Results years except 1996. In 1996, results from the complete survey of sections were used in the analysis. All data 3.1. Changes in tree species composition were log10(X þ 1) transformed prior to analysis when following clearcutting necessary to stabilize variances. All analyses were performed on data collected in English units, but In 1955, 1 year after cutting, plots were dominated we are reporting means converted to metric units. by pin cherry, which accounted for nearly half the Significance was defined at P < 0:050; results of density of trees when averaged over 16 plots; paper statistical analyses were considered marginally sig- birch and red maple were next in abundance (Table 2). nificant if P ¼ 0:51–0.10. Together these three species accounted for approxi- Between 1955 and 1962, relative abundance by mately 90% of the tree density in 1955. Ten other species was calculated as the percent density of all species, including striped and sugar maple, American stems in the 16 split–split plot averages calculated from beech, black, yellow, and gray birch (B. populifolia the ninety-six 4 m2 plots. Between 1974 and 1996, we L.), red oak, white ash, aspen, and black cherry calculated relative abundance for each species from accounted for most of the remaining density. the average of percent density and percent basal area. Distribution of all tree species in 1955 was highly For 1974 and 1989, forest composition estimates came variable throughout the cut strips, but little of this 290

Table 2 Change in absolute and relative abundance of common tree species in uncut strips between 1952 (prior to cutting) and 1996

Species Year 285 (2003) 182 Management and Ecology Forest / al. et Allison T.D.

1952a 1955b 1974c 1989c 1996d

Density Basal Percent Density Percent Density Basal Percent Density Basal Percent Density Basal Percent (stems/ha) area (stems/ha) (stems/ha) area (stems/ha) area (stems/ha) area (m2/ha) (m2/ha) (m2/ha) (m2/ha)

Acer rubrum 169.0 6.5 24.0 16331.1 8.7 1727.8 2.1 15.4 511.6 3.9 20.5 327.4 3.4 20.1 Acer saccharum 81.3 2.6 10.7 2740.4 1.5 863.9 0.4 5.7 217.2 1.4 7.4 111.5 0.6 5.2 Betula 44.5 0.4 4.0 2490.8 1.3 415.1 0.7 4.5 236.5 2.3 9.9 257.9 2.7 15.7 lenta/alleghaniensis Betula papyrifera 361.3 17.1 57.0 46139.2 24.7 955.6 1.3 9.0 135.1 1.0 5.9 108.7 1.7 8.3 Fagus grandifolia 0 0 0 212.5 0.1 212.4 0.1 1.2 101.4 0.2 3.4 130.7 0.3 4.8 Fraxinus americana 0 0 0 1275 0.6 497.1 1.2 6.6 241.3 3.7 16.6 177.0 4.1 17.8 Populus spp. 0 0 0 790.7 0.4 164.1 1.1 4.6 72.4 1.6 6.8 27.5 1.2 4.6 Prunus pensylvanica 0 0 0 105580.1 56.4 1800.2 3.5 20.2 82.0 2.9 3.7 14.5 0.2 1.0 Quercus rubra 8.0 0.9 2.2 501.6 0.3 115.8 0.2 1.1 48.3 0.6 1.9 44.8 0.9 4.0 Othere 12.2 0.7 2.1 9286.2 4.9 3566.6 4.3 31.6 637.1 3.1 23.3 483.7 1.2 17.7

Total 676.3 28.2 100.0 187085.7 98.9 10357.2 15.0 99.9 2278.0 20.7 99.4 1701.9 16.3 99.2

1955 abundance is based on a census of all stems as described in the text and is presented as density. 1974, 1989, and 1996 abundance is presented as density and basal area based on a census of all stems >1.3 cm. All data in the table are based on the 16 split–split plots averaged across treatments. The data source for absolute and relative abundance for each year –

varies and is indicated in the footnotes. In all years other than 1955, relative abundance is the average of relative density and relative basal area. 301 a Density, basal area, and relative abundance based on complete census before cutting of all stems >11.7 cm DBH in experimental strips (see Fig. 1). b Density and relative abundance based on a census of all stems in ninety-six 4 m2 subplots randomly located in 16 split–split plots. c Density, basal area, and relative abundance based on a census of all stems >1.3 cm DBH in 16 m2 sub plots randomly located in 16 split–split plots. d Density, basal area, and relative abundance based on a census of all stems >1.3 cm in 16 split–split plots. e Includes Acer pensylvanicum and Ostrya virginiana. T.D. Allison et al. / Forest Ecology and Management 182 (2003) 285–301 291

Table 3 Mean height in meters for selected species in cut strips by year

Species Year

1955 1958 1962 1975 1996

Acer pensylvanicum/other 0.30 (0.27) c 0.85 (0.7) c 1.5 (1.4) ab 6.1 (3.2) ab 5.6 (4.2) a species Acer rubrum 0.15 (0.03) a 0.55 (0.61) a 0.97 (1.0) a 6.9 (5.1) bc 11.2 (6.2) c Acer saccharum 0.24 (0.18) c 0.97 (0.58) c 1.4 (1.2) ab 4.6 (2.2) ab 7.1 (3.2) ac Betula lenta 0.27 (0.24) c 0.79 (0.61) bc 1.2 (1.1) ab 10.2 (4.5) cd 9.9 (7.2) bc Betula papyrifera 0.18 (0.09) b 0.73 (0.45) b 1.5 (0.97) ab 22.3 (2.7) abc 14.5 (4.4) cd Fagus grandifolia 0.27 (0.21) bc 0.76 (0.45) ac 1.3 (0.27) ab 4.2 (2.4) ab 5.7 (3.3) abc Fraxinus americana 0.30 (0.18) c 1.1 (0.52) c 1.8 (1.0) b 9.4 (4.7) cd 18.0 (6.5) d Populus spp. 0.27 (0.21) c 1.1 (0.52) c 2.2 (0.91) c 13.8 (2.6) d 22.1 (6.7) d Prunus pensylvanica 0.58 (0.3) d 1.5 (0.55) d 2.5 (0.97) c 7.8 (2.5) c 13.8 (0.64) abcd Quercus rubra 0.24 (0.12) bc 0.52 (0.39) ab 0.94 (0.88) a 5.2 (2.7) ab 11.8 (7.2) abcd

Means are based on all trees recorded in ninety-six 4 m2 plots in 1955 through 1975 and in ninety-six 4 m2 plots in 1996. In 1975 and 1996, mean height is based on all trees >1 m. Numbers in parentheses are standard deviations. Groups of means within years having the same letter are not significantly different at P < 0:001 by t-test. The grouping of similar means is complicated by uneven variances and widely varying numbers of individuals for each species. variation was significantly related to treatments applied and 1.68 m2/ha versus 0.87 m2/ha (P < 0:04)). by the USFS. Pin cherry density in 1955, for example, By 1996 the effect of scarification on paper birch ranged from 2000 to almost 500,000 stems/ha, and abundance had disappeared; paper birch density was averaged 147,617 stems/ha on the 40 m strips and 95 and 125 stems/ha on unscarified versus scarified 66,094 stems/ha on the 20 m strips. These differences, plots,andpaperbirchbasalareawas1.67and1.64 m2/ha however, were not significant (P > 0:30). Pin cherry on unscarified versus scarified plots. Average DBH density was concentrated in plots 1, 2, and 5 in the of paper birch was significantly greater on unscarified northeast corner of the study area (see Fig. 1, mean ¼ plots than on scarified plots (13.9 cm versus 12.0 cm, 387; 621 stems/ha). Removing these three plots from respectively; P < 0:001). the calculations of averages across all plots would When paper birch abundance (relative or absolute reduce mean pin cherry density in 1955 to approxi- density) was compared to mean percent scarification mately 33,000 stems/ha from the calculated mean of the 16 plots by linear regression, there was a signi- of nearly 106,000 stems/ha (Table 3). Total tree ficant positive relationship in 1955 (R2 ¼ 0:362). density in 1955 in the 16 plots ranged from 49,000 Forty-two years later there was no relationship to 580,000 stems/ha; there were no significant between paper birch relative abundance, density, or differences by treatment. Variation in pin cherry den- basal area, and scarification level (e.g. for basal area in sity accounted for 86% of the variation in total tree 1996, R2 ¼ 0:022). In general, there was no significant density. correlation between paper birch abundance by plot in Paper birch density in 1955 was significantly in- 1955 and its abundance in 1996. There also was no creased by scarification (P ¼ 0:04). Mean paper birch significant relationship between abundance of black density in 1955 in scarified plots was 56,609 stems/ha and yellow birch and soil scarification in 1955 or 1996. versus 17,901 stems/ha in unscarified plots. Paper Combined basal area of all birches in 1996 was not birch density was also higher on upper slope plots significantly related to mean percent scarification in 1955—an effect that persisted into 1996. This (R2 ¼ 0:003). difference was marginally significant (P < 0:08). In Total tree density and densities for several species, 1974, paper birch density and basal area in scarified including pin cherry, red maple, and paper birch, plots were still significantly greater than in unscarified decreased significantly over time (P < 0:001) as plots (1162 stems/ha versus 352 stems/ha (P ¼ 0:04) would be expected with stand thinning following 292 T.D. Allison et al. / Forest Ecology and Management 182 (2003) 285–301 disturbance. Relative abundance of pin cherry, for versus unscarified plots; this difference was margin- example, declined over 98% in 42 years and absolute ally significant (P < 0:10). density declined 99.9% from its initial abundance. In Sugar maple, American beech, and red oak also 1996, paper birch increased slightly in relative abun- have increased substantially in relative abundance dance, reversing a decreasing trend that continued over 42 years. In 1955, relative abundance of these through 1989. When averaged over years between three species combined was 1.9% and by 1996 this 1955 and 1996 scarification had a marginally signifi- abundance had increased to 14% (Table 2). All three cant effect on paper birch density (P ¼ 0:10). There species, especially beech, have patchy distributions was a significant soil treatment by year interaction throughout the cut strips. reflecting the greater decrease in paper birch abun- We examined species dominance in 1955 and 1996 dance in scarified plots versus unscarified plots over on a plot-by-plot basis. Dominants were defined as the 42-year period. There was a significant soil treat- species accounting for a minimum of 20% density in ment by year interaction for total density and for 1955 and 20% basal area in 1996. We believed the red maple indicating that the decrease in density for latter measure was a reasonable predictor of domi- red maple density was greater in unscarified plots nance into the 21st century. Using these criteria, plots (P < 0:01). Red maple density in 1996 was higher can have one or more dominants. In 1955, pin cherry in wide strips than in narrow strips, and, across years, was dominant or codominant in 13 plots, paper birch red maple density was higher on unscarified plots. in 8 plots, and red maple in 3 plots. These three Total density of all trees was higher on lower slope, species, in 1955, accounted for >80% of the density unscarified plots in 1996. These differences in red in 13 plots. No other species was dominant in any plot maple density and total density were only marginally in 1955. In 1996, white ash was dominant or codo- significant (P < 0:10). minant in nine plots, red maple in seven plots, black White ash, black birch, and aspen were at low birch in four plots, paper birch and aspen in two plots, relative abundance initially following cutting, but and sugar maple and striped maple in one plot. An became major components of the forest by 1996 average of 3.1 species accounted for 74.1% of the continuing a trend first observed in 1974. In 1955 basal area per plot in the cut strips; the composition of these three species accounted for 2.3% of relative tree this group of dominants varied from plot to plot. abundance when averaged over 16 plots. In 1974, DCA (ter Braak and Smilauer, 1998) of data from this abundance was 15.7%, and in 1996 abundance cut plots in 1955, 1974, and 1996 clearly shows a increased to 38.1% primarily due to increases in tree strong temporal trend in plot distribution (Fig. 2). Plot size (e.g. Table 2). Density in these species declined locations along axis 1 are weighted heavily by the from 1076 stems/ha in 1974 to 462 stems/ha in 1996, abundance of pin cherry and paper birch on one end of while mean DBH increased from 3.4 cm in 1974 to the axis and ash, black birch, and beech on the other 8.8 cm in 1996. end. Plots further differentiate along axis 2 corre- The density of striped maple was not recorded sponding to differences in abundance of white ash separately from ‘‘other species’’ in early measure- and black birch. Axis 1 appears to correspond to ments in the cut strips; in 1974, the first year this temporal trends in the data, while axis 2 indicates species was distinguished separately, the species spatial variation possibly related to variation in soil accounted for 95% of the density of the category of moisture. Immediately following cutting, plots are species originally classified as ‘‘other’’ by the USFS more tightly clustered toward the left side of axis 1 (Table 2). Species in the ‘‘other’’ category also reflecting the dominance of pin cherry and paper included eastern hophornbeam. Between 1974 and birch. The successional trend in the cut strips is 1996, striped maple density and basal area decreased clearly visible as pin cherry’s importance declines approximately 82 and 28%, respectively; most of this and the importance of other hardwoods, such as ash decline occurred between 1974 and 1989. Striped and black birch increases. It is also apparent that maple in 1996 accounted for roughly 4% of the basal spatial variation within the cut strips has increased area and 23% of the density in the cut strips. Basal area with time reflecting increased spatial variation in of striped maple in 1996 was greater in scarified plots species’ dominance. T.D. Allison et al. / Forest Ecology and Management 182 (2003) 285–301 293

Fig. 2. Biplot of species and plots on the ﬁrst two axes of detrended correspondence analysis illustrating temporal trends in forest composition in cut strips. Different symbols correspond to a summary of percent abundance from 16 split–split plots in 1955, 1974, and 1996.

3.2. Variation in tree height and age oak). Paper birch’s representation in the taller height classes also appeared to be negatively affected by the To visually examine distribution of trees by height surge in importance of these species in 1975, but paper we combined absolute measurements of tree height birch increased in abundance in the taller height into four height classes (HCs) for 1955, 1962, 1975, classes in 1996. The late successional species, sugar and 1996. The range of height classes varied by maple and American beech, began to form a distinct measurement reflecting the increase in mean tree understory in 1975. Striped maple was well repre- height, and the cutoff between height classes is some- sented in all height classes in 1962 and 1975, but by what arbitrary. Consequently, differences in abun- 1996 its relative abundance had declined, and it had dance of species in adjacent height classes are become an understory tree. imprecise. In 1975 and 1996, trees <1 m in height There were no treatment effects on height either were excluded from the analysis. Ranges of each across years or within years for paper birch, pin cherry, height class are listed in the caption for Fig. 3. red maple, or sugar maple, four species that were In 1955, pin cherry constituted approximately 95% found in all 16 plots between 1955 and 1975. Trees of height classes 3 and 4; the vast majority of paper of these four species did get significantly taller over birch and red maple were in height class 1. By 1975, time, and mean height of the species differed signi- pin cherry had lost its dominance in the taller height ficantly when all species are considered, although classes and was overtopped by intermediate hard- the rankings changed over time (see Table 3). Pin woods (red maple, white ash, black birch, and red cherry was initially the tallest species, but it was 294 T.D. Allison et al. / Forest Ecology and Management 182 (2003) 285–301

Fig. 3. Distribution of relative tree density by height class for 1955, 1962, 1975, and 1996. Data are based on measurement of tree height in 96 4m2 plots in 1955, 1962, and 1975, and 96 16.2 m2 plots in 1996. Trees were then assigned to four height classes. For 1955, height class (HC) 1 < 0:3m,HC2¼ 0:3–0.6 m, HC 3 ¼ 0:6–0.9 m, and HC 4 > 0:9 m. For 1962, HC 1 < 1:2m,HC2¼ 1:2–2.4 m, HC 3 ¼ 2:4–3.7 m, and HC 4 > 3:7 m. For 1975, HC 1 ¼ 1:0–3.0 m, HC 2 ¼ 3:0–6.1 m, HC 3 ¼ 6:1–10.7, and HC 4 > 10:7 m. For 1996, HC 1 ¼ 1:0–6.1 m, HC 2 ¼ 6:1–12.2 m, HC 3 ¼ 12:2–18.3 m, and HC 4 > 18:3 m. Early hardwoods includes B. populifolia and Populus spp. Intermediate hardwoods includes B. lenta, B. alleghaniensis, F. americana, P. serotina, and Q. rubra. Late hardwoods includes A. saccharum and F. grandifolia. Weed hardwoods includes A. pensylvanicum and O. virginiana. surpassed in mean height in 1975 by aspen, black statistical analyses of age–height relationships, trees birch, and white ash. taller than 1 m were combined into the same four Mean height varied significantly from plot to plot height classes that we created for analysis of height by (P < 0:04), and many species were present from the species. Seventy-nine percent of the trees taller than understory to the canopy. Height of black birch in 3 m in 1975 established within 5 years after cutting. It 1996, for example, ranged from 2.4 to 27.7 m. Shade is also apparent that presence in the tallest height class intolerant species (e.g. aspen and paper birch) had a in 1975 was not dependent on germinating (or sprout- smaller coefficient of variation in height (28% of the ing) immediately following cutting. Age of seedlings mean) than shade tolerant species (e.g. sugar maple, (<1 m in height) ranged from 1 to 21 years old; the striped maple, American beech; CV ¼ 58:5%) or median age was 2 years for those individuals whose species of intermediate tolerance (red maple, black age could be determined with confidence. Seventy- birch, red oak, and white ash; CV ¼ 53:6%). nine percent of the year old seedlings were striped and Mean ages of trees in 1975 differed significantly by red maples. Striped maple, which sprouts prolifically tree height (Table 4a); taller trees are older. For following cutting, on average was significantly older T.D. Allison et al. / Forest Ecology and Management 182 (2003) 285–301 295

Table 4 Mean tree age at stump height by (a) height class and (b) species in 1975, 21 years after cutting

N Mean age (years) Range Median

(a) Tree height class (m) 0.3–3.0 47 16.6 (3.3) a 8–21 18 3.0–6.1 66 16.8 (4.0) a 4–23 18 6.1–10.7 85 18.6 (2.7) b 10–27 19 >10.7 45 20.4 (1.6) c 16–27 21 (b) Species Acer pensylvanicum 70 19.3 (3.0) a 8–27 20 Acer rubrum 49 16.9 (4.4) b 4–21 18 Acer saccharum 26 17.9 (3.0) ab 9–21 18.5 Betula lenta 10 18.9 (3.0) ab 11–21 19.5 Betula papyrifera 24 16.6 (3.4) b 10–21 17 Fagus grandifolia 7 17.1 (2.6) ab 14–21 18 Fraxinus americana 14 18.4 (2.4) ab 12–21 18 Populus spp. 6 20.7 (0.8) ab 19–21 21 Prunus pensylvanica 28 18.7 (2.3) ab 15–21 19

One standard deviation is listed in parentheses after mean age. Different letters in mean age column correspond to means that differ significantly at P < 0:05. than red maple or paper birch. Several striped maples Paper birch density and basal area in 1996 were were older than 21 years indicating their presence in positively correlated with 1955 pin cherry density the understory prior to cutting. No other differences in (R2 ¼ 0:48 and 0.62, respectively), although this cor- tree age were statistically significant. relation appears to be due primarily to two plots that had very high pin cherry density in 1955. There was no 3.3. Relationship between tree species composition significant correlation observed between pin cherry and environmental variables and paper birch density in 1955 (R2 ¼ 0:003). Beech abundance was negatively correlated with plot moist- Pairwise comparisons between species abundance ure index and sugar maple abundance. in 1996, and moisture indices and initial pin cherry Canonical correspondence analysis of 1955 data density yielded few significant correlations when using 16 plots resulted in a marginally significant adjusted for the number of comparisons. The follow- first canonical axis (P < 0:10), which correlated ing results are based on non-adjusted significance most strongly with initial tree density. When per- levels of P < 0:05 to highlight some possible relation- formed on the 96 4 m2 subsample plots, the first ships. canonical axis was significant (P < 0:005). In this White ash density and basal area were positively analysis, canonical axis 1 correlated most strongly correlated with plot moisture index (R2 ¼ 0:24 and with soil moisture index and distance of the sub- 0.27, respectively). This relationship between ash sample plots from the strip edge. Axis 2 correlated abundance and the plot moisture index was not appar- most strongly with initial tree density. Initial tree ent in 1955 (R2 ¼ 0:07). In 1996, white ash was most density and the soil moisture index were negatively abundant in three plots 9, 10, and 13 (Fig. 1) located at correlated. the base of slopes where moisture seeps out onto the surface. White ash 1996 basal area and density were 3.4. Comparison of cut and uncut forest negatively correlated with 1955 pin cherry density successional trends according to a decay function with R2 ¼ 0:53 and 0.40, respectively. White ash abundance was nega- In the uncut strips, density and basal area of tively correlated with paper birch abundance in 1996. paper birch stems >11.7 cm declined substantially 296 T.D. Allison et al. / Forest Ecology and Management 182 (2003) 285–301

Table 5 Density, basal area, and relative abundance or value of tree species in uncut strips

Species 1952 1974 1996

Density Basal Value Density Basal Value Density Basal Value (stems/ha) area (stems/ha) area (stems/ha) area (m2/ha) (m2/ha) (m2/ha)

Acer pensylvanicum 4.9 0.07 0.5 4.9 0.05 0.4 5.8 0.1 0.7 Acer rubrum 229.8 9.0 31.3 209.2 12.7 31.9 145.0 13.3 33.4 Acer saccharum 73.3 2.1 8.7 184.5 6.4 21.8 145.8 6.6 23.6 Betula alleghaniensis 15.6 0.8 2.5 11.5 0.2 1.3 9.1 0.4 1.4 Betula lenta 41.2 1.7 5.8 65.1 3.3 9.0 74.9 3.1 11.6 Betula papyrifera 310.5 15.6 48.0 182.9 13.4 31.0 93.1 7.8 20.4 Fagus grandifolia 0.8 0.01 0.1 7.4 0.2 0.8 15.6 0.4 2.1 Fraxinus americana 1.7 0.08 0.3 5.8 0.3 0.8 12.4 0.5 1.9 Ostrya virginiana 10.7 0.2 1.1 21.4 0.4 2.1 11.5 0.2 1.4 Populus spp. 0.8 0.09 0.2 0 0 0 1.7 0.03 0.2 Quercus rubra 7.4 0.7 1.6 5.8 0.4 1.0 17.3 1.0 3.1 Total 696.8 30.4 100.1 698.5 37.4 100.1 532.9 33.5 99.8

Results are based on an average of a complete sample of all stems >11.7 cm DBH in three 0.4 ha strips. Value is the average of percent density and percent basal area. in between 1952 and 1996 while red maple and sugar The change in location in ordination space of the uncut maple increased in abundance (Table 5). Total stand strips between 1952 and 1996 illustrates the decline of basal area of the uncut strips increased 23% between paper birch and the increased abundance of red and 1952 and 1974 and then declined 10% between 1974 sugar maple. Conversely, immediately after cutting in and 1996. Total density increased slightly between 1954, the location of the cut strips reﬂects the initial 1952 and 1974 despite a 41% decline in paper birch dominance of pin cherry, but as pin cherry declines the density. Sugar maple density increased 152% during location in ordination space of the cut strips does not that period, nearly compensating for the decline in move back toward the location of the uncut strips. paper birch density. Black birch density increased Instead, the trend in changes in species composition 58%. The increase in basal area in these two species of the cut strips has been altered. These different was also due to increases in tree size. Average tree trends reﬂect the greater abundance of other hardwood diameter for sugar maple increased from 18.8 to species, primarily white ash and black birch, but also 21.0 cm; black birch DBH increased from 22.9 to aspen, red oak, and beech in the cut strips. This 25.4 cm. Paper birch continued its decline in 1996 divergence between successional trends in the cut while red maple, sugar maple, and black birch and uncut strips, apparent in 1974 has continued continued to increase in importance accounting for and expanded by 1996. 68.6% of the relative abundance in the uncut strips up Simpson’s diversity index was comparable in cut from 45.8% in 1952, prior to cutting. Other species, strips and uncut strips prior to cutting. The index including American beech, red oak, and white ash increased in both stands between 1952 and 1996, increased slightly in importance over the 42 years of but the increase was greater in cut strips. The increase observation. in Simpson’s index in the uncut strips is apparently Trends in changes in species composition for cut due primarily to the decline in the dominance of paper strips and uncut strips have not been the same over the birch. The increase in the index in cut strips appears to 42-year period. These trends are depicted visually by a be due largely to the increased importance of species biplot of the results of detrended correspondence that were relatively unimportant prior to cutting, such analysis (Fig. 4). Composition in both cut and uncut as white ash, black birch, aspen, and American beech strips was essentially identical in 1952 prior to cutting. (Fig. 5). T.D. Allison et al. / Forest Ecology and Management 182 (2003) 285–301 297

Fig. 4. Biplot of species and plots on the ﬁrst two axes of detrended correspondence analysis comparing temporal trends in cut strips vs. uncut strips between 1952 and 1996. Methods are as described in the text. Symbols correspond to relative abundance based on averages of 16 plots within cut strips and a complete census of all trees greater than 11.7 cm in each uncut strip. The 1952 measurement for cut strips is based on an average of trees counted in all four strips.

Fig. 5. Simpson’s diversity index vs. time for cut and uncut strips. Simpson’s index is calculated as described in the text. 298 T.D. Allison et al. / Forest Ecology and Management 182 (2003) 285–301

4. Discussion ‘‘advanced growth’’, that remained in the understory after cutting. Clearcutting in this northwestern Massachusetts Correlation analysis suggests other patterns in for- stand has resulted in a more diverse mixture of north- est composition that do not reflect treatments imposed ern hardwoods and apparently has altered the succes- by the USFS in 1954. A causal relationship is sug- sional trend in progress in the uncut forest. This is a gested when the correlations involve environmental common result following clearcutting of northern factors such as soil moisture. For example, 1996 hardwoods in the northeastern US of any successional white ash abundance was positively correlated with age (e.g. Wang and Nyland, 1996; Smith and Ashton, soil moisture, confirming our field observations. 1993). The overall impression of the cut strips in 1996 White ash is known to do well on moist/wet sites is a mosaic of small stands each dominated by differ- (Schlesinger, 1990). Hypothesizing about causal rela- ent species. In the absence of cutting, forest composi- tionship when the correlations are between species tion appears to be continuing a trend toward is more problematic. For example, the negative cor- dominance by red maple and sugar maple. Forty- relation between ash and pin cherry abundance may two years of succession following cutting, however, reflect an underlying environmental variable such as has resulted in stands with a stratified mixture of tree soil moisture to which the two species respond inver- species; the canopy is dominated by varying abun- sely, or it could reflect the effects of interspecific dances of paper birch, red maple, black birch, and competition. white ash, and the understory is dominated by sugar The strongest correlation was found between 1996 maple, striped maple, and red maple. paper birch abundance and 1955 pin cherry density; There is considerable plot-to-plot variation in the paper birch had its greatest abundance on plots that species that are dominant 42 years after cutting, but had the highest density of pin cherry. Two plots, which most of this variation in composition cannot be were dominated by paper birch in 1996, had an ascribed with statistical certainty to any particular average density of more than 450,000 pin cherry experimental treatment, such as soil disturbance, stems/ha in 1955. Plots with low pin cherry density or other initial condition, such as soil moisture or or that were dominated by paper birch in 1955 were initial tree density. The low statistical power of the dominated in 1996 by a mixture of intermediate hard- experiment, related to low replication and high vari- woods. It is possible that high pin cherry densities ance, limits our ability to say definitively that the inhibited competing hardwoods facilitating the various treatments, such as strip width and soil scar- dominance of paper birch when pin cherry died out. ification do or do not affect species distribution and This conclusion is consistent with observations of the abundance. effects of pin cherry density on hardwood regeneration Soil scarification was imposed on plots to enhance from New York and Pennsylvania (Heitzman and paper birch abundance, and initially this treatment did Nyland, 1994; Ristau and Horsley, 1999), but contra- have the intended effect. This effect persisted for 20 dicts the findings of an inhibitory effect of high pin years, but had disappeared by 1996 with the statisti- cherry densities on paper birch reported for New cally significant and greater decline of paper birch on Hampshire (Leak, 1988; Safford and Filip, 1974). scarified plots, a result counter to that suggested for Alternatively this relationship, too, may reflect some yellow birch at Hubbard Brook (Thurston et al., 1993). underlying environmental factor that we did not The larger size of paper birch on unscarified plots measure. suggests that growth was better perhaps because The changes in forest composition observed in this of the higher nutrient availability in the residual study are broadly consistent with those observed in organic layer (e.g. Marquis, 1965). Several other other clearcutting studies in the northeast. Most of trends in composition are apparent as described ear- these studies have been conducted in places other than lier, particularly the negative effect of scarification Massachusetts including New Hampshire (Marquis, on the abundance of red and striped maple. In this 1967; Leak, 1991), Connecticut (Smith and Ashton, case, scarification may have reduced the density of 1993; Liptzin and Ashton, 1999), New York (Wang these species, possibly by destroying seedlings, or and Nyland, 1996), and Pennsylvania (Marquis, 1982; T.D. Allison et al. / Forest Ecology and Management 182 (2003) 285–301 299

Marquis and Bjorkbom, 1982). All demonstrate to dormant pin cherry seeds in the soil seed bank all varying degrees the temporal trends in tree species change with age (Marks, 1974; Tierney and Fahey, composition observed here including an initial impor- 1998). As observed here, clearcutting in relatively tance in fast growing, early successional species, vary- young second growth hardwoods has resulted in ing abundances of later successional species depending higher densities of pin cherry and intermediate hard- on the age of the stand at the time of cutting, and the woods, such as white ash and black birch, and lower development of a stratified mixture of species approxi- densities of later successional species, especially mately 20 years after cutting. beech and sugar maple. The primary difference between these and other The observed response of striped maple to cutting is studies is in the relative importance of certain species particularly interesting. Although no measurements during the ‘‘reorganization phase’’ (sensu Bormann of seedling abundance were made immediately prior and Likens, 1979) immediately following cutting. For to cutting, results of a 1936 survey on nearby plots by example, yellow birch plays a more prominent role in USFS indicate an abundance of striped maple in regeneration following clearcutting in New Hamp- the understory on the site at that time. This species shire and New York, and red oak and black cherry increased in relative importance, peaking approxi- (P. serotina L.) play a more prominent role in Con- mately 20 years after cutting, and tree age data indicated necticut and Pennsylvania, respectively. In HMF, that striped maple continued to establish in the unders- black birch, white ash, and red maple are the pre- tory of the cut forest several years after cutting. The dominant intermediate hardwoods; eastern hemlock decline of striped maple abundance between 1974 and (Tsuga canadensis (L.) Carr.) and eastern white pine 1989 was second only to that of pin cherry and compar- (Pinus strobus L.) are relatively unimportant in the able to that of paper birch indicating a life expectancy Hopkins Forest (Art and Dethier, 1986). comparable to pin cherry. This species, however, is still The initial importance and, in many cases, dom- prevalent in the understory and seedling layer of the cut inance, of pin cherry immediately following cutting strips, unlike the other two species. Striped maple has been observed in studies of forest response to seems to possess an ability to respond rapidly to canopy clearcutting throughout the northeast, and this process openings that is unusual in a shade tolerant species has been thoroughly described by Marks (1974). Pin (Wilson and Fischer, 1977; Erickson, 1977). cherry was not recorded in the stand prior to cutting, The abundance of birches and ash indicate that but individuals were present as buried seeds—in very establishment from seed was an important mechanism high densities in some areas. The area of highest pin in the response to cutting. Sprouting also was probably cherry density following cutting was next to the north- a very important means of tree ‘‘establishment’’ fol- east boundary of the HMF (Fig. 1) an area rich in bird lowing cutting in 1954, and relative sprouting ability roosts such as hedgerows and fence lines. The dom- may have influenced the composition of the clearcut inance of pin cherry began declining almost immedi- strips. At least 38% of the trees taller than 9 m in 1975 ately after cutting, and within 20 years pin cherry trees (Fig. 3) were of sprout origin, and most of these trees were being overtopped by other, taller, and longer- were red maple and paper birch. No estimates of sprout lived species (Fig. 3). The length of observation origin were recorded between 1955 and 1962, but it is described in this paper is nearly long enough to likely that aspen, if present at the time of cutting encompass the ‘‘life cycle’’ of this species following established from root suckers rather than from seed a large disturbance. We expect that pin cherry will (Perala, 1990). Seedlings, or advanced reproduction, disappear from the aboveground portion of these were mowed at the time of cutting, which may have stands within the next 5–10 years, although we expect stimulated sprouting, particularly of red maple and the species will persist in the seed bank. striped maple (Jacobs, 1974). The relative increase Stand age at the time of cutting has been shown to of beech and decrease of sugar maple in the cut strips affect the composition of regeneration following cut- may also reflect differences in the sprouting ability of ting. The composition of the forest understory, parti- these two species following cutting. cularly in the abundance of later successional species The analysis of 1975 tree age suggests that there is (e.g. Leak and Wilson, 1958), and the density of a 5-year window of establishment after cutting for 300 T.D. Allison et al. / Forest Ecology and Management 182 (2003) 285–301 individual trees that reach the canopy 10–20 years We thank William B. Leak and one anonymous later. An ‘‘establishment window’’ of this length has reviewer for their careful reading of this paper. This been observed elsewhere (Greene, 2000). Understory research was funded by an NSF-URP grant to Henry trees have established over a longer period of time, but Art, the Hubbard Brook Ecosystem project, the this layer also includes individuals that established Howard Hughes Medical Institute grant to the Williams shortly after cutting. This then is a relatively even- College Biology Department, grants from the Williams aged forest, and the stratification by species observed College Bronfman Science Center, and a grant to the in the clearcut strips is due largely to the differing Williams College Center for Environmental Studies growth rates and tolerances of the different tree spe- from an anonymous foundation. cies, a characteristic of the mixed hardwoods of the northeast (Marquis, 1967; Oliver and Stephens, 1977; Bormann and Likens, 1979). References We close with two final points. First, many of the patterns observed in 1996 in the cut strips could not Art, H.W., 1974. Thirty-five years of change in a western New have been predicted from patterns observed in 1955 or England forest. Bull. Ecol. Soc. Am. 55 (2), 40. Art, H.W., Dethier, D.P., 1986. Influence of Vegetative Succession shortly thereafter. Examples include, the current dom- on Soil Chemistry of the Berkshires, vol. 153. Water Resource inance of ash in wetter plots, the positive correlation Research Centre, UMASS, Amherst, p. 167. of paper birch abundance in 1996 with initial pin Bormann, F.H., Likens, G.E., 1979. Pattern and Process in a cherry abundance in 1955, and the lack of correlation Forested Ecosystem. Springer, New York. between paper birch abundance in 1996 and paper Bormann, F.H., Siccama, T.G., Likens, G.E., Whittaker, R.H., 1970. The Hubbard Brook Ecosystem Study: composition and birch abundance in 1955. None of these patterns were dynamics of the tree stratum. Ecol. Monogr. 40, 73–388. apparent the first year after cutting. Second, and Cowles, H.C., 1899. The ecological relations of the vegetation on related to the first point, the results of this study the sand dunes of Lake Michigan. Bot. Gaz. 27, 95–117, 167– indicate the importance of observing succession fol- 202, 281–308, 361–391. lowing clearcutting and other disturbances over long Erickson, J.T., 1977. The Response of Striped Maple Acer pensylvanicum L. to a Natural Disturbance in the Hopkins periods of time. Most published clearcutting studies Memorial Forest. Honors Thesis, Biology Department, Wil- showing effects of initial conditions, such as soil liams College, p. 102. disturbance or seedling density immediately before Greene, D.F., 2000. Sexual recruitment of trees in strip cuts in or after cutting, contain results from the first one or eastern Canada. Can. J. For. Res. 30, 1256–1263. two decades after cutting (e.g. Thurston et al., 1992). Heitzman, E., Nyland, R.D., 1994. Influences of pin cherry (Prunus pensylvanica L. f.) on growth and development of young even- Long observations are sometimes necessary to fully aged northern hardwoods. For. Ecol. Manage. 67, 39–48. appreciate the importance of initial conditions in Hutnik, R.J., Cunningham, F.E., 1961. Silvical Characteristics of determining forest composition, especially when pat- Paper Birch, Betula papyrifera Marsh, Paper No. 141. North- terns may take decades to appear. If, for example, we eastern Forest Experiment Station, p. 24. had ceased observation of this clearcutting experiment Leak, W.B., 1988. Effects of weed species on northern hardwood regeneration in New Hampshire. Northern J. Appl. For. 5, in 1975, 20 years after cutting, our conclusions regard- 235–237. ing the effects of the initial degree of soil disturbance, Leak, W.B., 1991. Secondary forest succession in New Hampshire, or scarification, on the abundance of paper birch USA. For. Ecol. Manage. 43, 69–86. would have been erroneous. Leak, W.B., Wilson Jr., R.W., 1958. Regeneration After Cutting of Old-Growth Northern Hardwoods in New Hampshire, Paper No. 103. USDA Northeastern Forest Experiment Station, p. 8. Liptzin, D., Ashton, P.M.S., 1999. Early-successional dynamics of Acknowledgements single-aged mixed hardwood stands in a southern New England forest, USA. For. Ecol. Manage. 116, 141–150. We would like to thank Dawn Biehler, Tim Billo, Marks, P.L., 1974. The role of pin cherry (Prunus pensylvanica L.) Candy Cox, Susan Halbach, Jodie Knight, Ben Mon- in the maintenance of stability in northern hardwood ecosys- tems. Ecol. Monogr. 44, 73–88. tgomery, Bill Peterson, Richard J. Peterson, Taylor Marquis, D.A., 1965. Regeneration of Birch and Associated Schildgen, Elinor Shoreman, Amy Smith, Elizabeth Hardwoods After Patch Cutting, Paper NE-32. USDA Forest Titus, and Chuck Wall for assistance in the field. Series Experiment Station, p. 13. T.D. Allison et al. / Forest Ecology and Management 182 (2003) 285–301 301

Marquis, D.A., 1967. Clearcutting Northern Hardwoods: Results Hardwoods, USDA Forest Service Agricultural Handbook # After 30 Years, Paper NE-85. USDA Forest Series Research, 654, vol. 2. US Government Printing Office, Washington, DC, p. 13. pp. 333–338. Marquis, D.A., 1982. Effect of Advance Seedling Size and Vigor Smith, D.M., Ashton, P.M.S., 1993. Early dominance of pioneer on Survival After Clearcutting, Paper NE-498. USDA Forest hardwood after clearcutting and removal of advanced regenera- Series Research. tion. Northern J. Appl. For. 10, 14–19. Marquis, D.A., Bjorkbom, J.C., 1982. Guidelines for Evaluating ter Braak, C.J.F., 1986. Canonical correspondence analysis: a new Regeneration Before and After Clearcutting Allegheny Hard- eigenvector technique for multivariate direct gradient analysis. woods, Note NE-307. USDA Forest Series Research. Ecology 67, 1167–1179. Marquis, D.A., Bjorkbom, J.C., Yelenosky, G., 1964. Effect of ter Braak, C.J.F., Smilauer, P., 1998. CANOCO Reference Manual seedbed condition and light exposure on paper birch regenera- and User’s Guide to Canoco for Windows: Software for tion. J. For. 62, 876–881. Canonical Community Ordination, version 4. Microcomputer McClure, J.W., Lee, T.D., 1993. Small-scale disturbance in a Power, Ithaca, NY, p. 352. northern hardwoods forest: effects on tree species abundance Thoreau, H.D., 1906. In: Francis, H.A., Bradford, T. (Eds.), The and distribution. Can. J. For. Res. 23, 1347–1360. Writings of Henry D. Thoreau, vols. 1–20. Houghton Mifflin, Oliver, C.D., Stephens, E.P., 1977. Reconstruction of a mixed Boston. species forest in central New England. Ecology 58, 562–572. Thurston, S.W., Krasny, M.E., Martin, C.W., Fahey, T.J., 1992. Perala, D., 1990. Populus tremuloides Michx, quaking aspen. In: Effect of site characteristics and 1st- and 2nd-year seedling Burns, R.M., Honkala, B.C. (Eds.), Silvics of North America, densities on forest development in a northern hardwood forest. Hardwoods, USDA Forest Service Agricultural Handbook # Can. J. For. Res. 22, 1860–1868. 654, vol. 2. US Government Printing Office, Washington, DC, Tierney, G.L., Fahey, T.J., 1998. Soil seed bank dynamics of pin pp. 555–569. cherry in a northern hardwood forest, New Hampshire, USA. Ristau, T.E., Horsley, S.B., 1999. Pin cherry effects on Allegheny Can. J. For. Res. 28, 1471–1480. hardwood stand development. Can. J. For. Res. 29, 73–84. Wang, Z., Nyland, R.D., 1996. Changes in the condition and Safford, L.O., Filip, S.M., 1974. Biomass and nutrient content of 4- species composition of developing even-aged northern hard- year-old fertilized and unfertilized northern hardwood stands. wood stands in central New York. Northern J. Appl. For. 13, Can. J. For. Res. 4, 549–554. 189–194. Scanu, R.J., 1988. Soil Survey of Berkshire County, Massachusetts. Wilkinson, L., 1996. SYSTAT 7.0 for Windows. USDA NRCS, p. 216. Wilson, B.F., Fischer, B.C., 1977. Striped maple (Acer pensylva- Schlesinger, R.C., 1990. Fraxinus americana L. white ash. In: nicum): shoot growth and bud formation related to light Burns, R.M., Honkala, B.C. (Eds.), Silvics of North America, intensity. Can. J. For. Res. 7, 1–7.