Ecology, 94(5), 2013, pp. 1165–1175 Ó 2013 by the Ecological Society of America

Thirty-two years of change in an old-growth – forest

1 JAMES R. RUNKLE Department of Biological Sciences, Wright State University, Dayton, Ohio 45435 USA

Abstract. Old-growth forests dominated by understory-tolerant tree species are among forest types most likely to be in equilibrium. However, documentation of the degree to which they are in equilibrium over decades-long time periods is lacking. Changes in climate, pathogens, and land use all are likely to impact stand characteristics and species composition, even in these forests. Here, 32 years of vegetation changes in an old-growth beech (Fagus grandifolia)–sugar maple (Acer saccharum) forest in Hueston Woods, southwest Ohio, USA, are summarized. These changes involve canopy composition and structure, turnover in snags, and development of vegetation in treefall gaps. Stand basal area and canopy density have changed little in 32 years. However, beech has decreased in canopy importance (49% to 32%) while sugar maple has increased (32% to 47%). Annual mortality was about 1.3% throughout the study period. Mortality rates increased with stem size, but the fraction of larger stems increased due to ingrowth from smaller size classes. Beech was represented by more very large stems than small canopy stems: over time, death of those larger stems with inadequate replacement has caused the decrease in beech importance. Sugar maple was represented by more small canopy stems whose growth has increased its importance. The changes in beech and sugar maple relative importance are hypothesized to be due to forest fragmentation mostly from the early 1800s with some possible additional effects associated with the formation of the state park. Snag densities (12–16 snags/ha) and formation rates (1–3 snagsha1yr1) remained consistent. The treefall gaps previously studied are closing, with a few, large stems remaining. Death of gap border trees occurs consistently enough to favor species able to combine growth in gaps and survival in the understory. Key words: Acer; disturbance; Fagus; fragmentation; Hueston Woods, Ohio; long-term study; old growth forest; snag; tree demography; treefall gaps.

INTRODUCTION Oliver and Larson (1996) state that true old-growth Models of forest dynamics, often framed in terms of forests have probably never been common because ecological succession, predict a series of changes in the major disturbances occur quite frequently in most areas forest as it ages, including increases in biomass and compared to the life spans of trees. Pickett et al. (2009) shifts in species composition from fast-growing light- state that stabilization of vegetation composition or demanding species to those better able to survive in the structure is recognized to be a rare outcome of understory until their opportunity to reach the canopy vegetation dynamics at many spatial and temporal occurs (e.g., Whittaker 1975, Perry et al. 2008). If scales and in many habitats. However, in parts of the disturbances are infrequent enough, this process results Lake States and in parts of the central eastern in a steady state or equilibrium community in which the forest, such forests may have been frequent (Runkle dominant canopy species also reproduce well. At that 1996, Frelich 2002, Perry et al. 2008). stage, the community will change relatively little in It is not clear which aspect of forests is the best species composition or ecosystem properties until some indicator of old-growth conditions. Bormann and new disturbance occurs. This simple model, based on Likens (1979) used stand biomass as an indicator: in ideas from Clements (1936) and their subsequent old-growth forests, biomass fluctuations should be development (Pickett et al. 2009), provides a useful minor compared to the mean value. They proposed a framework. However, it leaves several questions unan- shifting-mosaic steady state model in which the ecosys- swered. tem as a whole is in equilibrium although at any one Are there locations and forests in which disturbances location it changes with the establishment, growth, and are rare enough that a steady state community exists? death of individual canopy trees. Oliver and Larson (1996) consider true old growth to occur when all trees that invaded immediately after the last major distur- Manuscript received 8 December 2011; revised 25 October bance have died. They do not assume that species 2012; accepted 13 November 2012; final version received 6 December 2012. Corresponding Editor: J. J. Battles. composition will remain constant at that time but are 1 E-mail: [email protected] unclear under what conditions it might change. Frelich 1165 1166 JAMES R. RUNKLE Ecology, Vol. 94, No. 5

(2002) uses the definition of Clements (1936) that the The opportunity for measuring these features of old- climax forest will be able to reproduce its own growth mesic forests in some forest types is very limited. composition in the absence of severe disturbance. Perry Few old-growth remnants exist in some areas of fertile et al. (2008) also characterize the steady state forest as agricultural soils (Davis 1996). Available sites are one in which species composition becomes relatively further restricted by several recently introduced and stable if species reproduce under their own canopies or spreading pathogens and pests affecting such important in small gaps. trees as American beech (Fagus grandifolia), ash It is uncertain how much change occurs in the final (Fraxinus spp.), eastern hemlock (Tsuga canadensis), stage. At that stage, most tree death occurs as single and, to a lesser extent, sugar maple (Acer saccharum) treefalls (referred to here as gaps) or, at most, small (Griffin et al. 2003, Small et al. 2005, Watmough 2010, clusters of treefalls. The frequency of such deaths Garnas et al. 2011, Kashian and Witter 2011). Such undoubtedly varies from year to year. It is not known increased recent mortality obscures the disturbance how variable over short (a few years) to medium (a few regimes under which current tree species have evolved. decades) time spans are mortality rates, stand charac- Hueston Woods State Park in southwestern Ohio, teristics such as total basal area and total density, and USA, is an ideal place to examine the nature of an old- species composition. growth mesic forest. It is dominated by two long-lived In many parts of eastern North America, the mesic species, beech and sugar maple. There is no indication of a major overstory disturbance for several disturbance regime is dominated by individual treefalls tree generations. It is one of the very few old-growth and replacement in the resultant canopy gaps (Runkle forest stands in this forest type and has not yet been 1990, 1998, Lorimer et al. 2001, Busing 2005, D’Amato impacted by invasive tree pathogens. My 32-yr study and Orwig 2008, Gravel et al. 2010). Several generaliza- there allows me to distinguish between trends and tions about this disturbance regime are possible. fluctuations and allows me to quantify variations Average annual mortality is about 1% for trees around average conditions. I will examine (1) the corresponding to average canopy residence times of dynamics of canopy trees, including their rates of about a century for many tree species (Runkle 1996, growth, mortality, and replacement overall, by species Tyrrell et al. 1998). Less clear is the degree to which and by size class; (2) the dynamics of snags (dead stems mortality varies over short to moderate time periods. It .1.5 m high); and (3) changes in the vegetation of also is not clear what deviations around that 1% figure treefall gaps originally sampled in 1977. entail and why mortality in some forest types is consistently greater than 1% while in other forest types METHODS it is consistently less. Study area Related to the stand structure of living trees is the The study site is a 67-ha old-growth stand in Hueston stand structure of dead trees and other coarse woody Woods State Park, Ohio, USA (398340 N, 848450 W). debris (see Plate 1). Those components of the forest are That stand has remained relatively undisturbed since its important for many species, including birds and other purchase in 1797. Its overstory is dominated by beech vertebrates, fungi and other decomposers, and plants and sugar maple with white ash (Fraxinus americana), (Harmon et al. 1986, Hura and Crow 2004, Garber et al. tulip tree (Liriodendron tulipifera), and black cherry 2005, Siccama et al. 2007). Some information on these (Prunus serotina) as codominants in some areas. Soils resources is accumulating (Tyrrell et al. 1998) but some are Russell silt loams (a Typic Hapludalf) and are deep forest types, including the beech–maple forest type, have and well drained, having formed in silt-mantled glacial been the focus of relatively few studies. Standing dead till overlying limestone (Lerch et al. 1969). Summers are trees (snags) are of particular interest with regard to hot (mean July temperature 23.78C), winters are cold forest change because they are ephemeral. Some trees die (mean January temperature 2.08C), and precipitation and immediately become fallen logs. Others die standing occurs year-round (mean annual precipitation 102.5 cm or broken off and gradually deteriorate. The time during [averages 1950–1988 for Hamilton, Ohio; National which they remain standing is an important component Climatic Data Center]). Braun (1950) included it in her of the forest habitat for many species. Long-term studies characterization of the beech–maple forest region and are necessary to determine the rates at which snags form, recognized three topographic areas: upland, slope, and the rates at which they deteriorate and the variation in lowland; elevation ranges from 268–296 m above sea the availability of this resource through time. level. More recently it has been the subject of several A third component of forest change is the filling of studies on forest dynamics (Vankat et al. 1975, Runkle treefall gaps. It is unclear how rapidly gaps close and 1981, 1982, 1990, 2000, Moore and Vankat 1986, Peters whether they do so primarily from the height growth of and Poulson 1994). Currently it is not impacted by saplings or the height and lateral growth of surrounding beech bark disease and it has few invasive shrubs. Its subcanopy or canopy stems. It also is not clear the western half suffered from minor anthropogenic distur- extent to which new disturbances near old ones affect bances roughly 75–100 years ago, though was never the process. completely cleared. A narrow (,15 m wide) road May 2013 TEMPORAL CHANGES IN AN OLD-GROWTH FOREST 1167 covered by a closed canopy and predating the establish- Runkle (1981, 1982). Woody stems 1 m tall were ment of the state park goes through the woods. A larger measured if they were in the polygon marked by the road and parking lots were constructed in the 1950s and bases of canopy trees bordering the canopy opening. 1960s with the establishment of the state park. Both The species of the border trees were listed. In 2001, the types of roads have some local impact on species dbh (diameter at breast height) and approximate composition (Kupfer and Runkle 1996). Hiking trails location of these border trees were measured to allow through the old-growth area may have a small influence. more precise resampling of the gap saplings and estimates of border tree growth and mortality. Field methods I sampled canopy trees (stems 25 cm dbh ¼ diameter Analytical methods at breast height, i.e., 137 cm) using the point-centered Canopy density and basal area were calculated from quarter method in 1977, 1991, 1997, 2001, 2005, and the equations developed by Cottam and Curtis (1956). 2009. The point-centered quarter method was used to These calculations are not as accurate as plot samples characterize the canopy composition of Hueston Woods (Mark and Esler 1970) but allow comparisons over the and other sites in my original sample (Runkle 1981, course of the present study. Pearson product moment 1982), in which I sampled 102 points at randomly chosen correlations were used to check for significant changes in intervals (17–33 m apart) along 21 transects totaling stand and species importance values over time. 3390 m. Because I recorded the stems in a standard Annual tree diameter growth measurements were order and recorded for each its dbh and distance from calculated as differences in dbh between samples divided each point and because configurations of trees of given by the number of years in the sample. An analysis of sizes, species composition, and distances from a variance was used to relate annual tree diameter growth common point are unique within the small areas to specific tree (trees might have as many as five associated with points known distances apart along measures of annual growth, depending on the number transects of known starting positions and compass of times they were sampled), to species, to initial dbh, directions, I have been able to resample these same and to the sampling interval. Values from species trees. Compared to forest resampling using fixed plots sampled less than 10 times were eliminated from the and tagged trees, this technique involves more precise data set. A Tukey’s test identified significant differences measurements of the location of individual trees along between pairs of species and sampling intervals. A transects and within points but has the advantage of separate analysis of variance for just beech and sugar focusing only on canopy trees and allowing a larger area maple related annual tree diameter growth to tree, to be characterized by its canopy composition, impor- species, initial dbh, sampling interval, and the interac- tant for my original study, which focused on gaps. For tion of dbh and species. Regressions of annual tree each resample previous trees were measured for dbh and diameter growth vs. dbh were run to indicate the were recorded as (1) living and still the stem 25 cm dbh direction of that relationship if significant. Significance closest to the original point, (2) dead, or (3) living but levels for the parameters were based on the sum of replaced by another stem closer to the point that had squares of each parameter added last to the model. grown to be 25 cm dbh since the previous sample. Annual mortality rates were calculated as exponential Where the original stem had died or been replaced, the decay rates: new stem 25 cm dbh now closest to the point was sampled for dbh and distance to the point. Changes over ð1=yÞ Average annual mortality rate ¼ 1 ðS=N0Þ time are due to these changes in trees at individual points. I recorded the approximate height of dead trees where S is the number of survivors, N0 is the original or indicated that they had uprooted. All trees, including number of stems, and y is the number of years between dead trees and trees that had been replaced by nearer samples (Runkle 2000). These samples are based on 102 stems for each subsequent sample, were resampled. I points (408 point-centered quarter trees). The signifi- used changes in height of dead trees to measure the cance of variation in mortality rates among different size dynamics of snags, defined here as stems 25 cm dbh classes was calculated using chi-square tests, comparing and 1.5 m in height. the numbers of trees that died and the number surviving I sampled 36 treefall gaps in Hueston Woods in 1977, for each size class. Beech and sugar maple mortality 1981, 1985, 1989, 1993, 1997, 2001, 2005, and 2009. rates also were compared using chi-square tests for each Gaps were initially identified as openings in the canopy size class (25–40, 40–60, 60–80, .80 cm dbh) and time caused by the death of large branches or one or a few interval. canopy trees that occurred over the line transects Snags were assumed to follow an exponential decay mentioned in the last paragraph. A short sampling rate, with a constant chance of deteriorating to a height interval allowed these gaps to be positively identified for of ,1.5 m, at which point it was treated as a stump or each sample using their distance along transects, their uproot (Runkle 2005). Thus, species and sizes of border trees, and characteristics of rt their causal tree. Gap size and age were estimated as in Nt ¼ N0e 1168 JAMES R. RUNKLE Ecology, Vol. 94, No. 5

Gap regeneration values simply summed totals for all gaps at different sampling times. The significance limit for statistical tests was set at P 0.05. Statistical calculations were done using the SAS statistical packet (SAS Institute 2002–2008).

RESULTS Canopy tree dynamics Stand basal area decreased nonsignificantly (P ¼ 0.17) from 36.7 m2/ha in 1977 to 35.1 m2/ha in 2009, an overall decrease of only 0.14% per year while stand density decreased significantly (P 0.001) from 171 to 153 stems/ha, or 0.35% per year. Despite this near constancy in stand values, beech decreased significantly (P 0.001) from 49.3% to 31.7% (1.4% per year) in importance (average of relative density and relative basal area) while sugar maple increased significantly (P 0.001) from 31.6% to 46.9% (1.2% per year; Fig. 1). White ash (P ¼ 0.30) and tulip tree (P ¼ 0.10) showed no significant trends. Individual tree diameter growth averaged 0.39 cm dbh/yr. The model based on individual trees (n ¼ 415 trees), common species (n ¼ 7 species), dbh, and FIG. 1. Relative importance value (average of relative sampling interval (n ¼ 5 intervals) was significant (P density and relative basal area, percentages) of the four most 0.001, R2 ¼ 0.54, df model ¼ 424, df error ¼ 1177, F ¼ important tree species. 3.31). Significant differences were found for each parameter of the model. The main four species differed where Nt is the number of snags at time t, N0 is the significantly from each other, growing in the order, number of snags at beginning of the time interval, and r fastest to slowest, of tulip tree, white ash, sugar maple, is a constant indicating the rate of deterioration. The and beech (Table 1). Growth was consistent over most value for r was calculated as the slope of a line fitted to sampling intervals with relatively low values for 1991– the natural log of the percentage of snags still snags as a 1997. The model relating annual tree diameter growth in function of the number of years since the snags first beech and sugar maple to individual trees (n ¼ 348 trees), appeared. The line was constrained to include the y- species (n ¼ 2 species), sampling interval (n ¼ 5 intervals), intercept of 100% of snags at time 0. Snag half life equals dbh, and the interactions of species and dbh was r ln (0.5)/r. Annual snag survival equals e . Annual significant (P 0.001, R2 ¼ 0.49, df model ¼ 354, df snag deterioration equals 100% minus survival. Snag error ¼ 958, F ¼ 2.59). However, although tree, sampling formation rate (the number of new snags per year) can r be estimated using the calculated value for and the TABLE 1. Annual growth rates of point-centered quarter point following equation: trees by species and by time intervals. Z Z t t 0 rt Category N Growth (cm dbh/yr) NT ¼ Ntdt ¼ N0e dt 0 0 Species A where N is the observed total number of snags formed Liriodendron tulipifera 84 0.70 T Juglans nigra 22 0.62AB during the period between samples that were still snags Quercus rubra 26 0.54BC at the time of the second sample and N 0 is the number of Fraxinus americana 144 0.53BC 0 D new snags formed per year (Runkle 2005). N is thus the Acer saccharum 801 0.38 T Prunus serotina 13 0.37CDE sum of the number of snags that formed each year of the Fagus grandifolia 512 0.31E given time interval reduced by the deterioration of snags Year interval before the end of the time interval. Annual snag 1977–1991 315 0.41AC rt D formation rates, N0, equal (rNt)/(e 1), where Nt 1991–1997 282 0.32 BC is the number of snags observed to be t years old; for 1997–2001 328 0.37 2001–2005 338 0.42A annual rates t ¼ 1. Snag density (i.e., integrating the 2005–2009 339 0.42AB above equation over all ages) equals N0/r. Because snag formation rates varied between sampling intervals, I Notes: Values within a category followed by the same superscripted letter are not significantly different from each calculated formation rates and densities for each other (Tukey’s test, P 0.05). N is the number of growth sampling interval. readings, some coming from the same tree. May 2013 TEMPORAL CHANGES IN AN OLD-GROWTH FOREST 1169

TABLE 2. Annualized mortality rates of point-centered quarter in 2001, and 7.1% in 2005 and 2009. The chi-square point trees and, in parentheses, gap border trees. analysis showed that beech and sugar maple mortality did not differ significantly for any time interval and size Mortality (% per year) Beginning combination except for stems 25–40 cm dbh and 40–60 year 1991 1997 2001 2005 2009 cm dbh from 1977–1991 and stems 40–60 cm dbh from 1977 1.3 1.3 1.4 1.3 1.3 1997–2001; in all three cases, beech mortality was higher 1991 1.3 1.3 1.3 1.2 than sugar maple mortality. 1997 1.5 (1.9) 1.1 (1.2) 1.2 (1.1) 2001 0.7 (0.6) 1.0 (0.9) Stem size distributions of beech and sugar maple 2005 1.2 (1.3) differed for each sample and over time (Fig. 3). Beech in 1977 had more stems 60–80 cm dbh than in either of the two smaller classes. Over time, high mortality of the larger stems plus fewer stems in the smaller size classes interval and dbh were significantly related to annual tree have resulted in decreases in stem density overall and in diameter growth (each with P 0.001), species and their the relative number of larger stems. Sugar maple, on the interaction with dbh were not. This result indicates that other hand, in 1977 had many stems 25–40 cm dbh. beech and sugar of a given size did not grow at These small stems have survived and grown to larger different rates. Combining all trees and sampling sizes (though not appreciably yet to stems .80 cm dbh). intervals for beech and sugar maple together resulted in the equation, annual growth ¼ 0.45 0.0020(dbh), Snag dynamics i.e., tree diameter growth was smaller for larger trees (P 2 Snags formed regularly throughout the study period. 0.001, R ¼ 0.014, df model ¼ 1, df error ¼ 1311, F ¼ 18.95). Snags deteriorating to stumps or fallen stems were Annual mortality rates of canopy trees were close to replaced by newly formed snags such that the total 1.3% for all time intervals for both point-centered number of snags remained at about 30–40 snags out of quarter trees and for trees bordering gaps (Table 2). 408 point-centered quarter point trees in my data set. Annual mortality rates by size class showed a consistent Those 408 trees correspond to about 160 trees/ha so pattern for the five time intervals with stems 25–40 cm snag densities were about 12–16/ha. Snag deterioration dbh dying at rates of 0.4–1.0% per year and larger stems approximated an exponential decay (Fig. 4). The dying at higher rates (3.6–5.4% for stems .80 cm dbh exponential decay value, r ¼ 0.16, was significant (P except for 2001–2005 when no large trees died; Fig. 2). 0.05). Snag half life equals 4.3 yr. Annual snag survival Only for longer time intervals (1977–1991, 1991–1997), equals 85% and annual snag deterioration thus equals with their higher sample size of dead stems, was the 15%. Based on that exponential decay value, 1–3 new 1 1 variation with size significant (P 0.001). Larger trees snagsha yr were added to the stand and total snag increased as a fraction of the stand because ingrowth densities were about 10–15 snags/ha (Table 3). Newly from smaller sizes more than compensated for their fallen trees (stumps ,1.5 m height or uproots) were higher mortality rates: stems 80 cm dbh made up 4.7% common (Table 3). However, the majority of these were of the sample in 1977, 5.1% in 1991, 4.7% in 1997, 5.4% observed to have come from snags (‘‘fallen snags’’ in

FIG. 2. Annual canopy tree mortality by sampling interval and size class. 1170 JAMES R. RUNKLE Ecology, Vol. 94, No. 5

FIG. 3. Size class distribution of Fagus and Acer for the seven sampling dates.

Table 3), not living trees (‘‘new falls’’ in Table 3). It is DISCUSSION impossible to distinguish between a fallen tree that broke Growth and mortality rates of canopy trees ,1.5 m high or uprooted immediately upon dying and one that died as a snag and then afterward (but before I Values for annual tree diameter growth recorded here had sampled it) broke off or uprooted. However, the agree with literature values. Runkle (2000) found tree calculated rates of snag formation were very similar to diameter growth rates of about 2.5–3.5 mm/yr for the the calculated rates for total annual mortality, which southern Appalachians, 3.0–4.5 mm/yr for Hueston includes trees that died but did not produce snags Woods (a subset of the data used in the present paper), because they either uprooted or broke ,1.5 m high. and 2.5–3.5 mm/yr for the Tionesta sites in northwest Therefore, most of the new falls probably also came Pennsylvania. Sites from further north find smaller from snags. These results are based largely on beech, values, e.g., 1.7 mm/yr for sugar maple and 1.3 mm/yr which made up 49 of the 69 snags observed. for beech in the Adirondacks of New York (Bedison et al. 2007). My results indicate that tree diameter growth Gap regeneration does not change greatly with tree size. Although gaps in the original 1977 sample were of The stand characteristics measured, basal area and several ages since formation most were 6 years old density, changed very little over the short (4 years) and (Runkle 1982). The number of stems in the gap first medium (three decades) time scales of this study. Annual increased greatly in response to gap conditions but has mortality rates also varied little across the different time declined since 1981 to present low values of about one intervals sampled. These results support the idea that stem per 10 m2 (Fig. 5). The surviving stems have forests can approach a steady state in stand properties increased in size so that the total basal area of the for a meaningful time period. regeneration has increased to 4.6 m2/ha, substantially The present study and other recent studies support the less than that for the stand in general. conclusion of Runkle (2000) that canopy mortality rates Sugar maple constitutes about two-thirds of the total often average about 1% per year. Japanese studies found importance of the gap regeneration (Appendix B). Beech 1.0% for evergreen old-growth forests in southwest constitutes about one-sixth. Other species of some importance include white ash, tulip tree, black cherry, and elm (combining Ulmus rubra and U. americana). According to Pearson’s correlation analysis, sugar maple and tulip tree significantly (P 0.05) increased in importance while white ash and elm significantly decreased in importance over the sample interval. Gap regeneration is influenced by recent disturbances as well as the original disturbance (Appendix C). Total density and basal area of gap regeneration are signifi- cantly correlated to the total number of gap border trees, an index of gap size. Those parameters also are correlated with the incidence of more recent disturbance. Values for the 2009 sample are positively correlated with the death of border trees from 2001–2005 (Pearson coefficient) and from 2005–2009 (Spearman coefficient). FIG. 4. Snag deterioration curve combining all cohorts. May 2013 TEMPORAL CHANGES IN AN OLD-GROWTH FOREST 1171

TABLE 3. Snag and treefall changes.

YR1 new Snag Mortality YR0 density YR1 snag density formation rate Snag density YR1 YR1 (treesha1 YR0–YR1 (trees/ha) new snags (snags/ha) (snagsha1yr1) (snags/ha) new falls fallen snags yr1) 1977–1991 171 35 14.7 2.6 16.5 32 2.2 1991–1997 159 13 5.1 1.3 8.3 18 10 2.1 1997–2001 160 18 7.1 2.4 15.1 6 21 2.4 2001–2005 153 6 2.3 0.8 4.8 6 8 1.1 2005–2009 158 15 5.8 2.0 12.4 6 24 1.9 Notes: Snag formation rates and densities calculated from deterioration rates. New snag density in YR1 was calculated as (YR0 tree density) 3 (YR1 new snags)/(total point-centered quarter point trees) (total ¼ 408). Annual mortality is calculated from rates in Table 2 multiplied by YR0 tree density. Treefalls (stumps ,1.5 m high or uproots) were recorded directly in the field. Values use the decay constant from Fig. 4 based on all snags from all years.

Japan (Miura et al. 2001), 1.3% for a Fagus crenata every 4–10 years for about 30 years altogether. They stand in northern Japan (Umeki and Kikuzawa 1999), found no clear trend in mortality with size for stems .30 and 1.3–4.3% for Japanese broad-leaved forest stands cm dbh. Low mortality values are often associated with (Masaki et al. 1999). Bellingham et al. (1999) found forests composed on long-lived species such as in annual mortality of 1.4% for stems 10 cm dbh in New temperate old-growth coniferous forests in Washington Zealand. Sugar maple mortality in Brownfield Woods, (Larson and Franklin 2010) and the northern sites listed Illinois, was 0.8–1.2% (Yi-ching and Augspurger 2006). above. However, some forests dominated by long-lived Mortality from western Massachusetts was lower, at species, such as Hueston Woods, may exhibit high 0.5% per year (D’Amato and Orwig 2008) and 0.9% per mortality rates because a large fraction of their canopy year (Wilson and McComb 2005). For trees of canopy trees are in larger size classes, which are characterized by size, mortality increases with stem size (Forrester and higher mortality rates. Runkle 2000, Runkle 2000, Lorimer et al. 2001, Miura et al. 2001). Busing (2005) found mortality in the southern Changes in canopy composition of beech–maple forests Appalachians to average 0.7% per year with values of unaffected by beech bark disease 1.3% for beech and 0.2% for sugar maple. Bedison et al. In contrast to the relative steady state observed for (2007) found the opposite species pattern in the stand basal area and density, species composition Adirondacks with 1.9% per year mortality for sugar changed markedly. The pattern observed in Hueston maple and 1.0% per year for beech despite the existence Woods of increasing sugar maple and decreasing beech of beech bark disease. In the old second-growth forest of fits their general pattern throughout most of the eastern Hubbard Brook, New Hampshire, the mean annual , though not in northern Pennsylvania and mortality of stems 10 cm dbh was 1.1% with values of New England (Runkle 2000). More recently, sugar 0.88% for sugar maple and 0.93% for beech (Siccama et maple was found to be increasing in Illinois (Yi-ching al. 2007). In Washington state, Larson and Franklin and Augspurger 2006, Lin and Augspurger 2008) and (2010) found average mortality rates to be 0.3–0.8% per Tennessee (Reid et al. 2008) while decreasing in Quebec year for stems .15 cm dbh in plots in 12 stands censused (Duchesne et al. 2005) and the Adirondacks (Bedison et

FIG. 5. Gap regeneration. Changes in basal area (BA) and density over time. 1172 JAMES R. RUNKLE Ecology, Vol. 94, No. 5 al. 2007). In contrast, Gravel et al. (2010) found sugar for snags .1.5 cm dbh but only 18 snags/ha for snags maple to be increasing in southern Quebec, especially in .30.1 cm dbh. Densities based on stems of the same size sites of higher disturbance (higher percentage of stems range varied among sites with densities of larger snags showing release). being more consistent than densities of smaller snags. The reason for the change in species composition in For example, the density per hectare of snags 10 cm Hueston Woods is obscure. Beech was more successful dbh was 19 for Fagus stands in Italy, 34 for Acer–Tilia in the past (based on its dominance of larger size classes) stands in Minnesota, 73 for northern hardwoods–Tsuga and sugar maple more successful recently (based on its stands in Wisconsin and Minnesota, and 80–151 for dominance of smaller size classes). Similar long-term northern hardwoods in Maine. The density of snags of shifts in the relative success of beech and sugar maple canopy size (25–30 cm dbh) was more consistent, were found by Poulson and Platt (1996) who attributed ranging 10–40 snags/ha. Values for the present study them to shifts in gap frequency. Sugar maple grows also fell within that range. Fewer studies estimated the faster in the light and is favored by higher disturbance number of new snags appearing per year. That number periods; beech survives better in the shade and is favored also varied with the size of stems considered. The values by low disturbance periods (Canham 1988, Gravel et al. from the present study of about 0.8–2.6 new 2010). snagsha1yr1 agreed with literature values for similar One explanation for long-term shifts in species dbh classes. composition could be changes in climate that have These results indicate that snags are a dynamic but favored sugar maple regeneration over that of beech. predictable part of the forest. Some variation in snag However, I did not find July and annual temperatures to formation rates and densities are to be expected. For show a significant trend over time for the last century instance, the period 2001–2005 in the present study was (Appendix A). marked by lower disturbance rates in general and by I suggest that the shift is occurring as a slow response lower snag formation rates than other time intervals. to the fragmentation of the landscape. Landscape Forests need to be large enough that such periods do not changes (Lehva¨virta and Rita 2002) and interactions cause undue stress for species that require snags for food with forest edges (Kupfer and Runkle 2003) affect or shelter. species composition. Clearing of the forest around The rate of snag deterioration generally followed an Hueston Woods in the early 1800s and the development exponential decay curve, allowing snag population of the state park in 1940 have undoubtedly affected its parameters to be estimated. However, the fit to the interior climate and species regeneration. At present, the curve was not as good for younger snags (4–6 year total width from border road to lake is about 600 m. intervals) as for older snags. This result suggests that Edge influence can be substantial in a woodlot that size. snag decay includes a time delay with snags ,6orso Kupfer and Runkle (2003) found significant correlations years not falling as often as older snags, in accordance between the vegetation in gaps and distances to edges, with other reports in the literature (e.g., Garber et al. partly due to differences in the seed rain. Other studies [2005] found evidence for a lag time of ,2 years). have found edge influence on climate and vegetation to Immediately after death, trees may become snags, low extend 25–100 m or so into a forest. Chen et al. (1992) stumps, or uprooted stems. The relative importance of found edge influences up to 137 m for old-growth these fates varies with forest type (Tyrrell et al. 1998). Douglas-fir stands near clearcuts. In a review, Harper et However, the deterioration of snags to low stumps and al. (2005) found many primary and secondary responses uproots makes it difficult to quantify the frequency of of edge formation such as stand structure and species snags at time of tree death. For instance, in 1991, I composition to be influenced by the edge at distances of found that the trees in my 1977 sample had generated 35 50–100 m and even greater. These results indicate that snags and 32 low stumps and uproots, almost equal the size of the old-growth stand at Hueston Woods is numbers. However, factoring in average snag longevi- small enough that edge influences may have played a ties, it is clear that almost all of the low stumps and large role in the relative fates of beech and sugar maple uproots were from trees that had originally been snags. over the last two centuries. Gap regeneration Snag dynamics The observed pattern of gap regeneration (Fig. 5), The rate at which snags form and the total density of with peak densities of small saplings at ages 5–10 years snags depends on the location of the forest, its and slow growth of larger stems afterward, fits other composition and history, and the sizes of stems included studies. Runkle (1998) found regeneration in gaps of in the definition of ‘‘snag’’ (Appendix D). Snags have old-growth forests in the southern Appalachians to been defined to be of various minimum heights with a increase in density from 1976–1977 (gaps mostly ,6 height of 1.5 m, as used in the present study, being a years old) to 1983 and then to decrease by 1990–1991. common criterion. Snag density varies greatly with the The basal area of gap regeneration increased during minimum diameter or diameter range used. For each interval, reaching 8.1 m2/ha in the last sample. instance, Hura and Crow (2004) found 107 snags/ha Madsen and Hahn (2008) artificially created 12 gaps in a May 2013 TEMPORAL CHANGES IN AN OLD-GROWTH FOREST 1173

PLATE 1. Large snags are an important component of the forest. Though individually ephemeral, formation of new snags may balance deterioration of existing snags to make them a reliable component of forest structure. Photo credit: J. R. Runkle.

Fagus-dominated stand in southern Denmark. They Kupfer and Runkle (2003) showed for Hueston Woods foundgapstoclosein3–4yr,basedonlight that gap size influenced gap species composition for 4–8 measurements at 1 m height and response of regenera- years after the initial sampling but that afterward other tion (stems ,6 m tall, mostly ,1 m tall). The present environmental factors mattered more. That gaps can study was of larger stems so peaked later than 3–4 years. influence relative abundance more than composition was Beaudet et al. (2007) found the height growth of stems also found by Goldblum (1997) for a Tsuga–northern 0.3–5 m tall to be greatest 3–4 years after a glaze storm hardwoods stand in New York, by Abe et al. (1995) for in a Acer–Fagus stand in Quebec. a Quercus–Fagus stand in Japan, and by Madsen and The overall composition of small stems in gaps stayed Hahn (2008) for a Fagus stand in Denmark. fairly constant though some species did change in Very few gaps are filled by trees newly germinating relative importance over the course of the study. In after gap formation (Ziegler 2002, Cole and Lorimer particular, the relative importance of sugar maple 2005). Most are closed by lateral branch growth of increased while that of beech varied little as the gaps canopy neighbors or by the height growth of preexisting aged. Gaps enhance the growth of existing stems with saplings. Because individual stems are usually over- occasional opportunities for some new individuals and a topped by several trees and because gaps are surrounded few opportunities for long-term gain, as for the shade- by several canopy trees, random mortality should keep intolerant tulip tree. Similar patterns have been found many gaps open after the original effects of one opening by others in old-growth temperature zone forests where had disappeared and should allow many tall saplings to the canopy dominants also can persist in the understory. be continually exposed or re-exposed to gap conditions 1174 JAMES R. RUNKLE Ecology, Vol. 94, No. 5 before dying an understory death. The mortality rates of Forrester, J. A., and J. R. Runkle. 2000. Mortality and gap border trees and canopy trees were similar in the replacement patterns of an old-growth Acer–Fagus woods in the Holden , Northeastern Ohio. American present study, indicating that chance disturbance is more Midland Naturalist 144:227–242. likely to reopen gaps than is a predilection for border Frelich, L. E. 2002. Forest dynamics and disturbance regimes: trees to die more frequently. The impacts of repeat studies from temperate evergreen–deciduous forest. Cam- disturbance in general are to boost the growth rates of bridge University Press, Cambridge, UK. stems already present. That repeat disturbances are Garber, S. M., J. P. Brown, D. S. Wilson, D. A. Maguire, and L. S. Heath. 2005. Snag longevity under alternate silvicultural frequent, even by chance, allows a shade tolerant regimes in mixed-species forests of central Maine. Canadian regeneration strategy to be successful. Journal of Forest Research 35:787–796. Garnas, J. R., M. P. Ayres, A. M. Liebhold, and C. Evans. ACKNOWLEDGMENTS 2011. Subcontinental impacts of an invasive tree disease on Many people have helped collect data in the many years forest structure and dynamics. Journal of Ecology 99:532– involved. Financial and logistic support for different parts of 541. the project came from the National Science Foundation and the Goldblum, D. 1997. The effects of treefall gaps on understory Ohio Biological Survey. Field assistance during the last round vegetation in New York State. Journal of Vegetation Science of sampling was provided by Holly Schlotman, Marilyn 8:125–132. Baumer, and Jenny Runkle. Gravel, D., M. Beaudet, and C. Messier. 2010. Large-scale synchrony of gap dynamics and the distribution of under- LITERATURE CITED story tree species in maple–beech forests. Oecologia 162:153– 161. Abe, S., T. Masaki, and T. Nakashizuka. 1995. Factors Griffin, J. M., G. M. Lovett, M. A. Arthur, and K. C. influencing sapling composition in canopy gaps of a Weathers. 2003. The distribution and severity of beech bark temperate deciduous forest. Vegetatio 120:21–32. disease in the Catskill Mountains, N.Y. Canadian Journal of Beaudet, J., J. Brisson, D. Gravel, and C. Messier. 2007. Effect Forest Research 33:1754–1760. of a major canopy disturbance on the coexistence of Acer Harmon, M. E., et al. 1986. Ecology of coarse woody debris in saccharum and Fagus grandifolia in the understorey of an old- temperate ecosystems. Advances in Ecological Research growth forest. Journal of Ecology 95:458–467. 15:133–302. Bedison, J. E., A. H. Johnson, S. A. Willig, S. L. Richter, and Harper, K. A., S. E. MacDonald, P. J. Burton, J. Chen, K. D. A. Moyer. 2007. Two decades of change in vegetation in Brosofske, S. C. Saunders, E. S. Euskirchen, D. Roberts, Adirondack –fir, northern hardwood and -domi- M. S. Jaiteh, and P.-A. Esseen. 2005. Edge influence on forest nated forests. Journal of the Torrey Botanical Society structure and composition in fragmented landscapes. Con- 134:238–252. servation Biology 19:768–782. Bellingham, P. J., G. H. Stewart, and R. B. Allen. 1999. Tree Hura, C. E., and T. R. Crow. 2004. Woody debris as a species richness and turnover throughout New Zealand component of ecological diversity in thinned and unthinned forests. Journal of Vegetation Science 10:825–832. northern hardwood forests. Natural Areas Journal 24:57–64. Bormann, F. H., and G. E. Likens. 1979. Pattern and process in Kashian, D. M., and J. A. Witter. 2011. Assessing the potential a forested ecosystem. Springer-Verlag, New York, New for ash canopy tree replacement via current regeneration York, USA. following emerald ash borer-caused mortality in southeastern Braun, E. L. 1950. Deciduous forests of eastern North America. Michigan. Forest Ecology and Management 261:480–488. Blakisten, Philadelphia, Pennsylvania, USA. Reprinted 1972. Kupfer, J. A., and J. R. Runkle. 1996. Early gap successional Hafner, New York, New York, USA. pathways in a Fagus–Acer forest preserve: pattern and Busing, R. T. 2005. Tree mortality, canopy turnover, and determinants. Journal of Vegetation Science 7:247–256. woody detritus in old cove forests of the southern Appala- Kupfer, J. A., and J. R. Runkle. 2003. Edge-mediated effects on chians. Ecology 86:73–84. stand dynamic processes in forest interiors: a coupled field Canham, C. 1988. Growth and canopy architecture of shade- and simulation approach. Oikos 101:135–146. tolerant trees: response to canopy gaps. Ecology 69:786–795. Larson, A. J., and J. F. Franklin. 2010. The tree mortality Chen, J., J. R. Franklin, and T. A. Spies. 1992. Vegetation regime in temperate old-growth coniferous forest: the role of responses to edge environments in old-growth Douglas-fir physical damage. Canadian Journal of Forest Research forests. Ecological Applications 2:387–396. 40:2091–2103. Clements, F. E. 1936. Nature and structure of the climax. Lehva¨virta, S., and H. Rita. 2002. Natural regeneration of trees Journal of Ecology 24:252–284. in urban woodlands. Journal of Vegetation Science 13:57–66. Cole, W. G., and C. G. Lorimer. 2005. Probabilities of small- Lerch, N. K., P. E. Davis, L. A. Tornes, E. H. Hayhurst, and gap capture by sugar maple saplings based on height and N. A. McLoda. 1969. Soil survey of Preble County, Ohio. crown growth data from felled trees. Canadian Journal of U.S. Department of Agriculture Soil Conservation Service, Forest Research 3:643–655. Washington, D.C., USA. Cottam, G., and J. T. Curtis. 1956. The use of distance Lin, Y., and C. K. Augspurger. 2008. Impact of spatial measures in phytosociological sampling. Ecology 37:451– heterogeneity of neighborhoods on long-term population 460. dynamics of sugar maple (Acer saccharum). Forest Ecology D’Amato, A. W., and D. A. Orwig. 2008. Stand and landscape- and Management 252:3589–3596. level disturbance dynamics in old-growth forests in western Lorimer, C. G., S. E. Dahir, and E. V. Nordheim. 2001. Tree Massachusetts. Ecological Monographs 78:507–522. mortality rates and longevity in mature and old-growth Davis, M. D., editor. 1996. Eastern old-growth forests: hemlock–hardwood forests. Journal of Ecology 89:960–971. prospects for rediscovery and recovery. Island Press, Madsen, P., and K. Hahn. 2008. Natural regeneration in a Washington, D.C., USA. beech-dominated forest managed by close-to-nature princi- Duchesne, L., R. Ouimet, J.-D. Moore, and R. Paquin. 2005. ples—a gap cutting based experiment. Canadian Journal of Changes in structure and composition of maple–beech stands Forest Research 38:1716–1729. following sugar maple decline in Que´bec, Canada. Forest Mark, A. F., and A. E. Esler. 1970. An assessment of the point- Ecology and Management 208:223–236. centred quarter method of plotless sampling in some New May 2013 TEMPORAL CHANGES IN AN OLD-GROWTH FOREST 1175

Zealand forests. Proceedings of the New Zealand Ecological Runkle, J. R. 2000. Canopy tree turnover in old-growth mesic Society 17:106–110. forests of eastern North America. Ecology 81:554–567. Masaki, T., H. Tanaka, H. Tanouchi, T. Sakai, and T. Runkle, J. R. 2005. Twenty-four years of change in an old Nakashizuka. 1999. Structure, dynamics and disturbance Tsuga canadensis woods affected by beech bark disease. regime of temperate broad-leaved forests in Japan. Journal of Journal of the Torrey Botanical Society 132:483–491. Vegetation Science 10:805–814. SAS Institute. 2002–2008. SAS 9. SAS Institute, Cary, North Miura, M., T. Manabe, N. Nishimura, and S. Yamamoto. Carolina, USA. 2001. Forest canopy and community dynamics in a temperate Siccama, T. G., T. J. Fahey, C. E. Johnson, T. W. Sherry, E. G. old-growth evergreen broad-leaved forest, south-western Denny, E. B. Girdler, G. E. Likens, and P. A. Schwarz. 2007. Japan: a 7-year study of a 4-ha plot. Journal of Ecology Population and biomass dynamics of trees in a northern 89:841–849. hardwood forest at Hubbard Brook. Canadian Journal of Moore, M. R., and J. L. Vankat. 1986. Responses of the herb Forestry Research 37:737–749. layer to the gap dynamics of a mature beech–maple forest. Small, J. J., C. J. Small, and G. D. Dreyer. 2005. Changes in a American Midland Naturalist 115:336–347. hemlock-dominated forest following woolly adelgid infesta- Oliver, C. D., and B. C. Larson. 1996. Forest stand dynamics. tion in southern New England. Journal of the Torrey Update edition. John Wiley, New York, New York, USA. Botanical Society 132:458–470. Perry, D. A., R. Oren, and S. C. Hart. 2008. Forest ecosystems. Tyrrell, L. E., G. J. Nowacki, T. R. Crow, D. S. Buckley, E. A. John Hopkins University Press, Baltimore, Maryland, USA. Nauertz, J. N. Niese, J. L. Rollinger, and J. C. Zasada. 1998. Peters, R., and T. L. Poulson. 1994. Stem growth and canopy Information about old growth for selected forest type groups dynamics in a worldwide range of Fagus forests. Journal of in the eastern United States. General Technical Report NC- Vegetation Science 5:421–432. 197. U.S. Department of Agriculture, Forest Service, North Pickett, S. T. A., M. L. Cadenasso, and S. J. Meiners. 2009. Central Forest Experiment Station, St. Paul, Minnesota, Ever since Clements: from succession to vegetation dynamics USA. and understanding to intervention. Applied Vegetation Umeki, K., and K. Kikuzawa. 1999. Long-term growth Science 12:9–21. dynamics of natural forests in Hokkaido, northern Japan. Poulson, T. L., and W. J. Platt. 1996. Replacement patterns of beech and sugar maple in Warren Woods, Michigan. Ecology Journal of Vegetation Science 10:815–824. 77:1234–1253. Vankat, J. L., W. H. Blackwell, and W. E. Hopkins. 1975. The Reid, J. L., J. P. Evans, J. K. Hiers, and J. B. C. Harris. 2008. dynamics of Hueston Woods and a review of the question of Ten years of forest change in two adjacent communities on the successional status of the southern beech-maple forest. the southern Cumberland Plateau, U.S.A. Journal of the Castanea 40:290–308. Torrey Botanical Society 135:224–235. Watmough, S. A. 2010. Assessment of the potential role of Runkle, J. R. 1981. Gap regeneration in some old growth metals in sugar maple (Acer saccharum Marsh) decline in forests of the eastern United States. Ecology 62:1041–1051. Ontario, Canada. Plant and Soil 332:463–474. Runkle, J. R. 1982. Patterns of disturbance in some old growth Whittaker, R. H. 1975. Communities and ecosystems. Second mesic forests of eastern North America. Ecology 63:1533– edition. Macmillan, New York, New York, USA. 1546. Wilson, B. F., and B. C. McComb. 2005. Dynamics of dead Runkle, J. R. 1990. Gap dynamics in an Ohio Acer–Fagus wood over 20 years in a New England forest. Canadian forest and speculations on the geography of disturbance. Journal of Forest Research 35:682–692. Canadian Journal of Forest Research 20:632–641. Yi-ching, L., and C. K. Augspurger. 2006. A long-term study of Runkle, J. R. 1996. Central mesophytic forests. Pages 161–177 neighbor-regulated demography during a decline in forest in Mary B. Davis, editor. Eastern old-growth forests. Island species diversity. Journal of Vegetation Science 17:93–102. Press, Washington, D.C., USA. Ziegler, S. S. 2002. Disturbance regimes of hemlock-dominated Runkle, J. R. 1998. Changes in southern Appalachian canopy old-growth forests in northern New York, U.S.A. Canadian tree gaps sampled thrice. Ecology 79:1768–1780. Journal of Forest Research 32:2106–2115.

SUPPLEMENTAL MATERIAL

Appendix A Changes in temperatures in the region from 1893 to 2011 (Ecological Archives E094-104-A1).

Appendix B Changes in the importance values of the most important species of the gap regeneration (Ecological Archives E094-104-A2).

Appendix C Correlations between gap regeneration in 2009 (total number, total basal area) and border tree mortality (Ecological Archives E094-104-A3).

Appendix D Snag densities and formation rates for old-growth forests from the literature (Ecological Archives E094-104-A4).