|
|
| Northeastern Naturalist: gap dynamics of canopy trees of a Tsuga canadensis forest community, The |
ABSTRACT - We examined the correlation between canopy gap formation and the initial growth of forest trees by reconstructing the gap history of a Tsuga canadensis (L.) Carr. forest community in southeastern Ohio. We cored each tree (>10 cm dbh, n = 156) in a 40 x 90 m plot and examined the cores for release events. characterized by dramatic increases in radial growth. We identified 80 former gaps in the 79 yr sample period by clustering release events in time and space. Thirteen of the 80 former gaps coincided with the initial growth of trees. These 13 gaps were usually large gaps containing few established trees or gaps undergoing repeat disturbance. Of the 36 trees >10 cm dbh that began growth during the sample, 21 (58%) began growing inside a gap within 6 yr of gap formation-three times the rate predicted by chance (p = 0.001 ). We also measured the distance in time and space between the first year of growth and the closest canopy gap for each tree. We called the inverse of this measure the gap affinity index (GAI). Although we did not find significant differences in GAI among species, the index confirmed qualitative notions about the shade tolerance of six common tree species: intolerant species exhibited high gap affinities, while tolerant species exhibited low gap affinities. We also found a significant, positive correlation between GAI and the area of the closest gap, as well as a significant, negative correlation between GAI and the mean number of releases per species.
INTRODUCTION
Canopy gaps have become a critical element in our understanding of forest dynamics. Gap dynamics have been applied to models of forest structure, spatial pattern, and composition (Brokaw 1985; Denslow 1980, 1987; Jones 1945; Pickett and White 1985; Runkle 1984, 1990). Gaps may actively influence forest dynamics by altering the dispersal patterns, germination rates, recruitment rates, growth rates, and growth forms of tree species (Canham 1988, balling et al. 1988, Denslow et al. 1990, Gray and Spies 1996, Hibbs 1982, Kobe 1999, Poage and Peart 1993, Rebertus and Burns 1997, Schupp 1988, Sipe and Bazzaz 1995, Thompson and Willson 1978, Watt 1925). The gap cycle of disturbance and succession may form an adaptive gradient for forest species, allowing tree species to partition the forest environment on a temporal basis. The differential use of gaps by forest species during gap succession has been called gap partitioning. Grubb (1977) hypothesized gap partitioning may be most prevalent during the recruitment phase of gap succession. For example, the recruitment of tree species intolerant of canopy shade may be more likely in larger gaps, while tree species tolerant of canopy shade may be confined to smaller gaps or the closed canopy (Denslow 1980, 1998; Kneeshaw and Bergeron 1998; McClure and Lee 1993; Runkle 1982).
Although gap partitioning has stimulated a large number of studies, data to support the hypothesis have been elusive. Based on data from neotropical forests, Hubbell et al. (1999) and Tilman (1999) have recently questioned the role of gap partitioning in maintaining forest diversity. One potential problem with gap research is the time scale of the process: gap dynamics may operate over periods that cannot be characterized from short-term studies. Most gap research has been conducted on extant gaps and the dynamics of the system extrapolated back in time (e. g., Runkle 1981, 1982). Less common are long-term studies on extant gaps (Kupfer and Runkle 1996; Poulson and Platt 1996; Runkle 1990, 1998; Young and Perkocha 1994). A few studies have reconstructed the history of small forest communities to examine the nature of past disturbance and its relationship to tree recruitment (Frelich and Graumlich 1995, Henry and Swan 1974, Oliver and Stephens 1977, Parshall 1996, Payette et al. 1990).
To our knowledge, the correlation between individual gaps and the initial growth of individual trees has not been explicitly examined over an extended period of time. We examine this relationship by reconstructing the gap and growth history of the common tree species in a forest community. We ask the following questions: how often do gaps function as sites for the initial growth of trees? Are some tree species more likely to begin growth in canopy gaps than other species? And, if this is the case, do these species exhibit different life history characteristics?
FIELD-SITE DESCRIPTION
Forest communities dominated by eastern hemlock, Tsuga canadensis (L.) Carr., are ideally suited for this research. They are floristically and structurally simple (Griggs 1914, Oosting and Bourdeau 1954). In addition, release and suppression events are clearly evident in the radial growth of Tsuga canadensis trees (Lorimer and Frelich 1994, Marshall 1927). Tree cores have been successfully used to reconstruct release and suppression events in similar forest communities elsewhere (Canham 1985, 1990; Frelich and Graumlich 1994; Parshall 1994; Payette et al. 1990).
We located the study in the Hocking Hills region of Ohio, approximately 40 km southeast of Columbus. Physiographically, the region belongs to the unglaciated Appalachian plateau (Braun 1961). Ravine communities in this area are strongly dominated by Tsuga canadensis, reaching relative densities of 60-65%. Common associates include tulip-poplar (Liriodendron tulipifera L.), American beech (Fagus grandifolia Ehrh.), red maple (Acer rubrum L.), sweet birch (Betula lenta L.) and red oak (Quercus rubra L.; nomenclature follows Weishaupt 1971).
METHODS
Overview
We examined the relationship between former canopy gaps and the initial growth of canopy trees in four steps. First, we cored all trees in a Tsuga forest community and examined the cores for release events. Second, we used cluster analysis to group the releases in time and space, and equated the clusters with former canopy gaps. Third, we compared the former gaps to the initial growth of trees entering the community during the past 79 yr, and measured the distance between the site of initial growth and the closest former gap. Finally, we regressed the mean distances for each species against gap size, species growth rate, and number of releases, and used these correlations to examine the role of gaps in the life histories of the tree species.
Coring the trees
All Tsuga communities in the region were surveyed for large areas of mature, homogeneous vegetation containing at least one canopy gap. Of the three communities deemed suitable for this study, a community in Blackjack Hollow, Hocking Hills State Forest, was randomly selected for the reconstruction. This community showed no obvious signs of fire or human disturbance, although land use in the area suggested a history of logging. In the summer of 1991, a 40 x 90 m sampling grid was placed into the center of the stand and all trees >10 cm dbh (n = 156) were cored at ground level on the north, downhill side of the tree. For each core, we counted the annual rings to determine the initial year of growth. We estimated tree height visually using triangulation, and noted the presence or absence of the tree in the canopy.
Reconstructing the gap history
Release events evident in the cores were enumerated using methods similar to Payette et al. (1990) and Parshall (1994). We defined a release as an abrupt, fourfold or greater increase in annual growth rates sustained for at least three years. The fourfold increase was calculated by comparing growth rings with the running mean of the three previous growth rings. Releases less than fourfold were attributed to weather effects and removed from the analysis. Some cores exhibited release growth rates in the earliest growth ring, a condition we called initial release. Although still subject to the fourfold threshold, initial releases were identified by comparing the initial growth rate to the mean annual growth rate of the entire core.
To identify former canopy gaps, release events were clustered by year and location using centroid linkage (Davis 1986). Mathematically, clusters are groups of adjacent trees undergoing relatively synchronous release. Clusters were delineated using two criteria. First, we did not allow two release events within a single core to join if they were more than 6 yr distant. Second, we did not allow a release event to join a cluster if the link bisected a shorter link of non-releasing trees.
The cluster analysis could not be completed, however, until all three axes of the community array were expressed in a common unit. We converted the temporal axis into meters using the mean vertical growth rate (0.4 m/yr) of the tree species that most rapidly fill canopy gaps (primarily Liriodendron but also Betula, Quercus and Acer; see Table 4).
Continued from page 1.
We believe the clusters represent former gaps for the following reasons. First, many clusters were centered around the most pronounced release events and edged by less pronounced release events. Trees growing in the center of a gap should exhibit a more pronounced response to a gap than a tree growing on the edge of the gap. Second, the Blackjack community contained one extant gap, aged independently in the field at 6 yr before present (hereafter abbreviated BP). The only tree > 10 cm dbh in the Blackjack community exhibiting a release during the past 10 yr grew on the northeast edge of this gap; the release occurred 5 yr BP. Third, the reconstruction suggested a widespread disturbance in 1947. Following the reconstruction, we found logging stumps buried in the litter in locations predicted by reconstructed gaps.
The size of the gap was estimated by joining the non-releasing tree trunks surrounding the cluster, an area Runkle (1982) has called the expanded gap. This substantially overestimates the gap area compared to estimates based on crown margins (Runkle 1990), but our interest lay in comparing the functional effects of gaps of different sizes, not the absolute size of the gaps. In addition, expanded gaps could be unambiguously replicated and measured. For gaps completely within the sample grid, we calculated area by overlaying a counting grid. For gaps on the edge of the sample grid, we estimated area by doubling the gap area within the grid. Although this may introduce some errors into our estimates of gap size, arbitrary transects should bisect former gaps in a random manner, producing a distribution in which the mean gap area within the sample grid should approximate the mean gap area lying beyond the sample grid.
The date of gap formation was estimated by the mean release date of the trees in the cluster. Although the gap history of the stand extended 165 yr, we confined our analysis of gap recruitment to the gaps occurring within the most recent 79 yr BP (1912 through 1991). This was the longest period that did not exhibit a significant correlation (p = 0.05, linear regression model) between yr BP and total gap area. Positive, significant correlations (p
The relationship between gaps and the initial growth of trees.
Reconstructed gaps exhibited a variable number of existing trees. To determine the effect of existing trees on the initial growth of new trees, we calculated the density of existing trees by dividing the number of trees by the area of the gap. We then compared the mean density of gaps allowing the initial growth of a tree against the mean density of gaps that did not allow the initial growth of a tree using a t-test.
Because gaps exist in both time and space, they can be visualized as cylindrical volumes within the three-dimensional community array. To test if trees were more likely to begin growth within the canopy gaps than the closed canopy, we calculated the total volume of the array undergoing gap disturbance during the sample period, estimating the life of a canopy gap at 6 yr. If trees entered the array at random, the trees should enter the gap volumes in the same proportion that the gaps occupy in relationship to the closed canopy portion of the array. We tested these proportions using a Chi square test.
Regressing gap distances against community and life history parameters.
Trees that successfully begin growth only in canopy gaps during the year of formation clearly depend upon gaps for recruitment and establishment. As the distance in time and space between initial growth and gap formation increases, gaps would appear to be less important in the recruitment and establishment of tree species. As a result, we used the mean distance between the site of initial growth and the closest gap to examine the role of gaps in the recruitment and establishment of tree species. For each tree in the sample period, we calculated the Euclidean distance between the date and location of initial growth to the edge of the closest gap. We converted the temporal axis to meters using the 0.4 cm/yr estimate of growth rate employed in the reconstruction. We considered all trees inside the spatial boundaries of the gap to be simply inside the gap, and set the horizontal distance for these trees to zero.
We called the distance between the site of initial growth and the closest gap the mean gap distance (hereafter abbreviated MGD). We transformed MGD into a gap affinity index using the formula
GAI^sub i^ = 1 / (MGD^sub i^ + 1) where GAI = the gap affinity index of species i and MGD = the mean gap distance in meters for species i. We found GAI more useful than MGD for three reasons. First, GAI correlates positively with the gap use of a tree species: increases in GAI correspond to an increase in gap recruitment. Second, GAI varies between one and zero, where an index value of one means a tree species begins growth only inside canopy gaps during the year of gap formation. Third, GAI can be interpreted as a percentage value roughly corresponding to the propensity of a species to begin growth in a gap.
We calculated GAI for the six most common tree species in the Blackjack community (n > 4 in the n = 156 database). To be consistent with the gap data, we confined the GAI sample to trees beginning growth during the 79 yr sample period (n = 36 trees in nine species, but n = 32 trees in the 6 most common species). We used ANOVA to test differences among the GAI of the individual species.
To analyze the relationship between GAI and natural history attributes such as growth rate and number of release events, we regressed the attributes of the six species against GAI. We used a step-wise regression to examine the correlation between GAI and the mean area of the gap closest to initial growth, the mean number of releases, and the mean growth rate. For each tree in the GAI sample, we calculated gap area directly from the reconstructed gaps. For releases, we used only release events evident in the cores that also corresponded to a reconstructed gap. To determine growth rate, we divided the height of each tree by the age of the tree. We recognize growth rate is not a linear function of tree age, but sought to minimize the effects of age-specific growth rates by using only the trees in the GAI sample. These trees exhibited a narrow range of ages [41-79 yr], and did not include trees of extreme age.
Sipe and Bazzaz (1995) hypothesized large gaps may be hostile environments for shade tolerant species. If this hypothesis is correct, shade tolerant species should be less common in large gaps than predicted by chance. Tree species unaffected by gap environments, however, should exhibit a random distribution. To examine this hypothesis, we arbitrarily divided gaps from the 79 yr sample into two categories of equal numbers: small gaps (145 m^sup 2^, n = 40). For each gap, we counted the number of trees beginning growth inside the gap boundary and within 6 yr following gap formation (n = 23 trees). We used the ratio of gap area (3:1) to calculate the predicted ratio of trees for Chi square analysis because the large gap category contained three times the total area of the small gap category. To increase the sample sizes, we combined the two most intolerant species, Betula and Liriodendron (n = 9), the two mid-tolerant species, Quercus and Acer (n = 6), and the two most tolerant species, Tsuga and Fagus (n = 8; Wenger 1984, Burns and Honkala 1990).
RESULTS
Gap dynamics at Blackjack Hollow
We determined the age and release histories of 153 of the 156 trees >10 cm in the Blackjack Hollow community. Three trees were hollow and discarded from the analysis. Thirty-one percent of the trees were 100 years or older, and none of the trees was less than 41 years old (Fig. 1). The ages of the trees indicated the community had been dominated by hemlocks throughout the 79 yr sample. Thirty-six trees began growth during the 79 yr sample.
We clustered 251 releases into 114 former canopy gaps, 80 of which occurred during the sample period. The mean area of all gaps = 109.8 m^sup 2^ (n = 80; SD = 83.m^sup 2^, range = 8 to 420 m^sup 2^, expanded gap area). Most of the gaps were quite small; 22 of the 80 gaps were less than 50 m^sup 2^, and 43 of the gaps were less than 100 m^sup 2^ (Fig. 2). We interpret these small gaps as release events associated with subgaps: disturbances caused by something smaller than the fall of a mature canopy tree. The total area of the community undergoing gap disturbance varied through time, reaching a peak in the late 1940's (Fig. 3).
Several of the gaps overlapped earlier gaps, a process called repeat disturbance (Runkle and Yetter 1987). We considered repeat disturbance to have occurred when expanded gap areas spatially overlapped within 6 yr. Seventeen of the 80 gaps underwent repeat disturbance, a rate consistent with similar forests in the region (Runkle 1998).
Continued from page 2.
Repeat disturbance alters the dynamics of a canopy gap, primarily by extending the life of the initial gap. A gap lineage can be defined as two gaps of different ages that overlap in time and space. Subsequent disturbance may extend the lineage through three or more gaps. Lineages may also branch. A gap overlain by two subsequent gaps that do not overlap produces two separate lineages. Gaps undergoing repeat disturbance developed 17 gap lineages. The mean lineage extended 12.0 yr (range = 9 to 15 yr) from formation of the initial gap to the sixth year of the terminal gap.
Thirteen of the 80 gaps were the site of the initial growth of a tree. Many of the gaps, however, contained trees at the time of formation. Gaps associated with the initial growth of a tree exhibited significantly more gap area per tree (181 m2 per tree) than gaps that did not allow the initial growth of a tree (92 m2 per tree, p = 0.0001, t-test). Following the successful growth of new trees, the tree density of gaps associated with the initial growth of a tree was not significantly different from the mean density of gaps not associated with the initial growth of a tree (p = 0.75). These results suggest gaps must contain circa 80-100 m^sup 2^ of "unoccupied" space to allow the successful growth of a new tree.
Repeat disturbance may also increase the likelihood of tree recruitment and growth. Of the 34 gaps involved in gap lineages, 10 gaps were associated with the initial growth of a tree, a rate significantly higher than gaps that did not undergo repeat disturbance (6 gaps associated with recruitment in 46 non-lineage gaps; p = 0.0001; Chi square test).
The initial growth of trees
Eleven of the 36 trees that began growth during the period of study began growth the same year as formation of the nearest gap. Twenty-six trees (72%) began growth between two years before and two years after gap formation. In this study, the 13 canopy gaps investigated in Blackjack Hollow appeared to remain open to the initial growth of a tree for 5 yr (Table 1). Oliver (1981) estimated gaps in New England forests remained open to recruitment for 6 yr.
Twenty-three trees (64%) began growth at least 2 m inside the expanded gap boundary (Table 2). Three additional trees were found straddling the boundary of the expanded gap (0 rn distance).
We have presented the spatial and temporal distances independently. To begin growth within a canopy gap, however, trees must begin growth within both the gap boundaries and the time period the gap remains open to successful invasion by trees. Twenty-one of the 36 trees (58%) beginning growth during the study period began growth within the expanded gap area and within 6 yr following gap formation. Of the remaining 15 trees, five began growth within the gap area before gap formation. Ten trees began growth outside the gap area; five before gap formation, five following gap formation.
The area and life span of a gap define a three-dimensional volume within the sample array. The percentage of trees beginning growth in gaps due to chance should be proportional to the percentage of the community array undergoing gap disturbance. The gap disturbance volume is equal to the mean annual area undergoing gap disturbance (3.1%) multiplied by the life of a gap (6 yr). Gap disturbance, therefore, occupies roughly 18.6% of the volume of the sample array. The observed rate of tree recruitment and establishment in gaps (58%) was significantly higher than the rate expected by chance (p = 0.001, Chi square test).
All 36 trees beginning growth during the study period exhibited one or more releases coinciding with a reconstructed canopy gap at some point during their lives (Table 3). Twenty-one of the 36 trees exhibited only a single release, suggesting successful recruitment and growth through a single gap event. Of the 21 trees beginning growth within a gap, thirteen exhibited release growth rates in the earliest growth ring. Five of the six Betula trees and four of the five Liriodendron trees entering the Blackjack community were associated with gap lineages, while only one of the 12 Tsuga trees and none of the Fagus trees were associated with gap lineages.
Gap affinity index
Gap affinity index (GAI) varied from 0.633 for Betula and 0.565 for Liriodendron to 0.301 for Fagus and 0.292 for Tsuga (Table 4). Although ANOVA did not reveal significant differences among species GAI (p = 0.48), a step-wise regression identified mean number of releases and mean area of recruitment gap as significant variables (p
The effects of gap size
Tolerant trees (Tsuga and Fagus) were significantly more likely to begin growth in small gaps than large gaps (p = 0.001, observed ratio 6:2, Chi square test). Intolerant trees (Betula and Liriodendron) exhibited no significant differences between small gaps and large gaps (p = 0.48, observed ratio 6:3, Chi square test). Mid-tolerant species (Quercus and Acer) exhibited no significant differences between large and small gaps (p = 1.00, observed ration 3:3, Chi square test).
DISCUSSION
The Tsuga community at Blackjack Hollow was characterized by high rates of gap formation throughout the 79 yr study period. Gap size was consistent with the gap sizes of similar ecosystems in the region (Table 5). The rate of gap formation, however, was substantially higher in Blackjack Hollow: 3.05% yr^sup -1^ compared to rates of 0.5-0.9% yr^sup -1^ for similar forests elsewhere (Frelich and Lorimer 1991, Parshall 1995, Runkle 1985, 1990). Some of this discrepancy was methodological. First, we reported rates based on expanded gap area, not the area defined by the canopy margins. Runkle (1990) cautioned that measures of expanded gaps may double size estimates in comparison to margin gaps. This would suggest the more comparable rate in Blackjack Hollow was one half the reported rate, or approximately 1.5% yr^sup -1^. Second, the Blackjack Hollow community was probably logged in 1947, producing 12 artificial gaps, and therefore potentially inflating the gap rate by 15%. Third, we sampled the functional response of trees to disturbance. Gaps as small as 15 m^sup 2^ have been shown to increase growth in tolerant hardwood (Canham 1988, Poage and Peart 1993). Our procedure may have allowed us to measure small canopy disturbances not evident using more traditional methods. If this was the case, smaller disturbances may have contributed one third of the effective canopy disturbance area within the Blackjack Hollow community. Van der Meer and Bongers (1996) have noted that 35% of fallen trees and 42% of damaged trees in a tropical rain forest did not produce an evident canopy gap.
Most trees entering Blackjack Hollow during the study period began growth in or near a canopy gap. In general, advance regeneration was not successful. Although all tree species had individuals that began growth prior to gap formation, only two of the 21 trees associated with gaps began growth more than three years before gap formation. Following gap formation, competition must close the gap quickly. Only five of the 21 trees associated with gaps successfully entered a gap more than three years following formation, and only two trees entered a gap more than five years after formation. Although most trees began growth inside canopy gaps, approximately one-third of the trees in the study began growth 0-3 meters outside a gap boundary.
Our data are consistent with forest models based on gap partitioning (Grubb 1977). Gap affinity index (GAI) varied substantially among the tree species and correlated well with qualitative notions of shade tolerance (Wenger 1984): Betula lenta and Liriodendron tulipifera are intolerant species exhibiting high GAls, while Tsuga canadensis and Fagus grandifolia are tolerant species exhibiting low GAls. GAI also correlated well with life history parameters that presumably reflect shade tolerance: high GAI species required larger gaps for recruitment and may have grown faster than low GAI species. High GAI species also exhibited fewer releases, although the converse of this measure-surviving fewer suppressions-may be the more appropriate interpretation. In addition, high GAI species may have been more dependent upon repeat disturbance than low GAI species.
ACKNOWLEDGMENTS
The University of Toledo, Department of Biology, provided financial and logistical support. The ODNR-Division of Forestry granted permission to use the field site. F.C. MacLeod and C. Foster provided field support, while N.G. Lewis provided love and domestic support for the senior author. S.T.A. Pickett, L. Jones, J. Runkle, J. Harrell and the late P. Fraleigh provided advice throughout the project.
LITERATURE CITED
BRAUN, E.L. 1961. The Woody Plants of Ohio. Ohio State University Press, Columbus, OH. 323 pp.
|
|
