Spatial patterns and level of mortality
Strong spatial patterns in both whole-plant mortality and leader damage among survivors occurred across our experimental understorey-gap gradient. The lowest survival and the greatest leader damage occurred in the most exposed portions of the large gaps (centre and north plots). Several studies have noted significant mortality and/or spatial patterns of mortality in gaps (Howe 1990; Ashton et al. 1995; Gray & Spies 1996) but mortality has generally not been compared with measured microenvironmental patterns. The exception is Brown & Whitmore (1992), who found that mortality of three advance regenerant dipterocarp species increased across a 10–1500 m2 range in gap size, representing an eightfold range of mean daily photosynthetically active radiation and a 10–20% increase in maximum daily air and soil temperature. Higher mortality in the centre and on the north sides of our large gaps is consistent with the occurrence of dramatically higher radiation loads in these microsites, especially diurnal and seasonal durations of direct beam radiation (Table 1; Canham et al. 1990). Mean daily photosynthetic photon flux (PPF) in the seedling plots increased over 10-fold across the gradient, from 34 to 375 µmol m−2 s−1, including marked differences within both large and small gaps. The irradiance gradient was even broader on clear days, ranging from 50 to 612 µmol m−2 s−1. Furthermore, the mean daily PPF values do not reflect the very high irradiances experienced by the north and centre plots in large gaps for several hours on clear days in June and July (> 1600 µmol m−2 s−1; Sipe 1990).
We stress that our artificial, cleared gaps produced greater exposure of small seedlings to high irradiances and temperatures than would often be the case in most natural treefall gaps, where downed boles, crowns and upgrowth would partially ameliorate increases in ground-level irradiance and air and soil temperatures. Gradual gap formation due to progressive standing death would also provide more time for advance regenerated juveniles to adjust to more slowly increasing irradiance and temperatures (Krasny & Whitmore 1992). Conversely, our data may be applicable to canopy openings caused by logging, which often exposes juveniles to a greater degree than natural treefall gaps (e.g. Marquis 1982).
We are aware of no published data specifically focused on naturally occurring levels of shoot damage and regrowth for juvenile trees in north-eastern USA forests. Our observations in the Harvard Forest system over the last 16 years suggest that the majority of juveniles (perhaps > 60%) experience some leader loss in the understorey before they reach 0.5 m in height.
Agent of mortality
Although leader damage for survivors was clearly greater in the north half of large gaps (> 80%), there was still a surprising level of leader damage (50–60%) in all other microsites, including the understorey (Fig. 1). The reason for extensive leader death in our study was not obvious. Deer removed a few buds from three A. rubrum in one large gap centre plot later in the study after these plants had already experienced leader damage and were exhibiting extensive regrowth. There was no evidence of damage by small mammalian herbivores. Several researchers have measured gap-related patterns of insect herbivory or pathogen activity (Nuñez-Farfan & Dirzo 1988, 1991; Howe 1990; Khan & Tripathi 1991). Although we noted several types of insect damage sporadically on foliage and stems, none was clearly related to the leader damage. In most cases, the terminal meristems died in place with no removal or external signs of damage by any identifiable source.
A pathogen could have been partially responsible. We observed abrupt leaf wilting and necrosis in the centre and north plots of large gaps during a warm, wet period in July 1988, particularly by A. pensylvanicum and A. saccharum (Sipe 1990). Small gap north plots were also affected, though much less so, but there was no wilting in the understorey. Marquis (1982) also noted browning and some leaf loss by exposed seedlings of these three species during midsummer in a large clearcut in Pennsylvania. The spatial pattern of wilting generally corresponds to the plot-specific and species-specific data on whole-plant mortality and leader damage among survivors in large gaps (Fig. 1), suggesting a link, but does not account for the substantial levels of leader damage in the understorey.
A number of non-pathogenic physiological processes could be involved, including transplant shock, thermal load, photodestruction (e.g. UV-B exposure), dehydration, and extreme high or low temperatures. Some transplant shock must have occurred, especially since the dormant seedlings were bare-rooted when transplanted. However, seedling survival in the understorey during the 18 months prior to gap formation was > 95%, and survival in the understorey was consistent across pre-gap and post-gap years (Sipe & Bazzaz 1995), implying that transplant shock was not serious. Our diurnal microclimatic and gas-exchange data suggest that high irradiances, air temperatures and water stress could have acted synergistically to dehydrate and overheat developing meristem tissues in large gaps in late summer, particularly on the well-drained soils in our site (Sipe & Bazzaz 1994, 1995). While this may explain the high levels of leader damage in the most exposed microsites, it does not readily account for the substantial levels of leader damage in the understorey. Thus the explanation for extensive leader damage in our study system remains unclear and may involve several factors, such as an interaction between physical stress and pathogen activity.
Site differences in growth and recovery
None of these studies has distinguished performance by damaged vs. undamaged plants, but it is logical to predict that recovery of damaged plants should be greatest in places supporting fastest growth by intact plants. As a combined group, our three species recovered best in large gaps, the same sites that had the lowest survival, greatest leader damage and highest irradiances. Greater growth has also been linked to higher irradiance in gap microsites that experience higher mortality (Gray & Spies 1996) or herbivore damage (Khan & Tripathi 1991). In contrast, Canham et al. (1993) found that the effect of simulated browsing on seedling stem tips (including A. rubrum) was greater for seedlings growing in full sun compared with 8% full sun. They showed that, unlike the shaded plants, full sun plants were placing most of their stored carbohydrates in stems rather than roots, which resulted in a more detrimental impact on carbon balance when stems were removed.
Species differences in survival, damage and recovery
We found strong differences among species in whole-plant survival and leader damage (Figs 1, 2), and in the relationship between survival and damage frequency (Fig. 3). Recent work on the importance of juvenile tree growth and mortality in forest dynamics has emphasized larger saplings (e.g. Kobe et al. 1995) and few data exist on comparative mortality for smaller juveniles of these species. However, our results agree with Houle (1991), who found greater survival by A. saccharum than A. rubrum for young seedlings in the understorey, and with Marquis (1982), who documented better survival by A. pensylvanicum (58%) and A. rubrum (55%) than A. saccharum (37%) after release of advance regenerant, small (6–9 cm tall) seedlings in a clearcut. There are no published data on leader damage for Acer except for Bartlett et al. (1991), who documented various sources of injury prior to whole-plant mortality in 1-year-old A. saccharum seedlings along a cliff-edge microenvironmental gradient. Terminal bud loss was highest along the cliff edge and declined towards the forest interior, but the authors indicate that the loss was probably caused primarily by herbivory and not by exposure.
Ecophysiological and growth differences have been reported for these species from both controlled environment and field studies and are discussed in Sipe & Bazzaz (1994, 1995), but no studies have compared growth by damaged vs. undamaged stems for more than one species. However, documented differences in shoot development and architectural flexibility, sprouting and shoot regrowth ability, and physiological plasticity, may help explain why A. rubrum was more capable of recovery from shoot damage than the other species, especially in large gaps.
First, these species differ in shoot development and architecture in ways that could be relevant to damage recovery. Sakai (1990) has distinguished three main branching models for Japanese species of Acer: (i) sympodial-spread, determinate extension; (ii) monopodial-spread, indeterminate extension; and (iii) elongate, indeterminate extension. A. rubrum clearly shows indeterminate shoot extension, while A. saccharum and A. pensylvanicum are partially determinate with limited neoformation and heterophylly (Critchfield 1971; Steingraeber 1982a). Indeterminate growth offers greater flexibility in shoot development during the growing season and may help explain the more extensive branching and flushing of sizeable numbers of new leaves by damaged A. rubrum during the two post-gap years. A. pensylvanicum appears to be sympodial in the understorey (Lei & Lechowicz 1990) and if the sympodial form was always determinate, this would limit responses to altered environments or damage. A. pensylvanicum, however, showed a dramatic transformation to rapid, elongate growth in our small gaps, including the south sides where irradiance levels were only slightly higher than in the understorey, implying significant plasticity in growth form and perhaps a greater ability to respond to leader damage in suitable environments (Wilson & Fischer 1977).
Brown (1993) and Endler (1993) have documented striking differences in spectral quality of radiation, including red : far-red ratios, between the understorey and gaps of different sizes, and it is possible that A. pensylvanicum is using spectral cues to alter shoot development. A. saccharum shows some architectural plasticity in response to irradiance levels (Steingraeber 1982b; Bonser & Aarssen 1994) and more responsiveness to canopy gaps than its frequent codominant, Fagus grandifolia (Canham 1988), but it is widely regarded as less flexible architecturally than A. rubrum, which shows substantial shifts in allocation to branches and leaves in response to increases in irradiance (Wallace & Dunn 1980).
Secondly, some evidence suggests that A. rubrum is more capable of sprouting or recovering from simulated bud/stem loss than the other species. Powell & Tryon (1979) showed that A. rubrum is better than A. saccharum at producing seedling sprouts from stems < 2 cm in diameter. Cooper-Ellis et al. (1999) found that A. rubrum was more effective than A. saccharum at resprouting from damaged larger trees or saplings (> 30 cm tall but < 5 cm in diameter at breast height) in response to a simulated blowdown at another site in the Harvard Forest. Many of the A. rubrum seedlings that lost leaders in our study died back completely and resprouted from basal buds, whereas A. pensylvanicum and A. saccharum were less likely to recover if the entire stem died back. The ability to maintain populations of dormant buds is essential to recovery from damage (Tuomi et al. 1994). For shade-tolerant tree species, which tend to be slow-growing, large numbers of suppressed lateral buds may accumulate on stems that are no more than 20–30 cm tall. However, data on interspecific differences in dormant bud populations and the ease with which they can be induced to grow are generally lacking (Critchfield 1971). Nor is the degree of apical dominance well documented for these species. Our results strongly suggest that dominance declines in the order A. pensylvanicum > A. saccharum > A. rubrum for both damaged and intact plants across all sites. Remarkably, damaged A. rubrum sprouted effectively and grew more than intact individuals in our large gap centres. This is consistent with Canham et al. (1993), who found that A. rubrum seedlings from which terminal stem segments were removed in the winter showed more shoot growth than the undamaged control plants in full sun treatments the following summer.
Thirdly, these species differ in overall ecophysiological plasticity. A. rubrum is fairly plastic in photosynthetic biochemistry and rates, response to water stress, biomass and dimensional growth rates, root : shoot allocation and shoot architecture across a wide range of irradiance (Wallace & Dunn 1980; Bazzaz & Miao 1993). Although A. saccharum shows some physiological adjustment to higher irradiance, most researchers have concluded that it survives and grows better in small gaps than large gaps (e.g. Barden 1983; Runkle 1990), perhaps due to conservative responses to water stress (Ellsworth & Reich 1992a,b), and is thus relatively insensitive to gap size beyond fairly small canopy openings (Canham 1985; McClure & Lee 1993). A. pensylvanicum shows better photosynthetic acclimation to the understorey than A. saccharum (Lei & Lechowicz 1990; Sipe & Bazzaz 1994). In our study, A. rubrum had consistently higher photosynthetic rates per unit leaf area than A. pensylvanicum and A. saccharum for the north and south plots of all three site types. However, A. pensylvanicum showed greater total photosynthesis when scaled to the shoot level in all but the large gap north plots, and shoot photosynthesis was correlated with greater growth by this species across most of the gradient (Sipe & Bazzaz 1994). It has been shown that leaf biochemistry and photosynthesis may change significantly when simulated defoliation occurs on species such as A. rubrum and Quercus rubra (Heichel & Turner 1984). We have no such data for damaged individuals of our species since we measured gas-exchange only on intact plants. However, if the relative pattern among species for photosynthetic rates per unit leaf area were to remain the same for damaged plants as for intact plants, then damaged A. rubrum would be likely to show higher shoot-level photosynthesis than either of the other species in large gaps, and this would also help explain its recovery advantage in microsites with higher irradiance.
Microsite-specific recovery and implications for forest regeneration
There are at least two ways in which microsite-specific differences in the occurrence of leader damage and regrowth among our species could influence future forest composition. First, juveniles of shade-tolerant trees may grow very slowly for decades in the shade before canopy gaps form overhead (Canham 1985; Runkle 1990; Clark & Clark 1992) and it is highly probable that some leader damage will occur during these long pre-gap periods (Clark & Clark 1989, 1991; Paciorek et al. 2000). Consequently, the ability to recover from damage in the shade will affect the status of juveniles and their ability to respond to gaps when the openings finally occur. Our data suggest that in the understorey A. pensylvanicum responds somewhat better following damage than A. rubrum, at least for the most important growth variables (total leaf area, net height change, net basal diameter change), and that A. saccharum fares poorly in this regard. Secondly, juvenile trees located in the centres of gaps have the greatest chance of capturing canopy space as the margins close in through lateral crown growth. Our data show that A. rubrum outperforms the other species in large gaps where it exhibits substantially higher survival (Figs 2 and 3) and growth (Figs 4 and 5) in spite of the greater frequency of leader damage. This is mostly due to its dramatic response in large gap centres (Sipe & Bazzaz 1995; also see McClure & Lee 1993). It may realize a competitive advantage in capturing canopy gap space over other less-resilient, advance-regenerating species when leader damage is extensive. Some studies have shown that released, advance-regenerant species may be overtopped by faster-growing pioneer species which can establish in large gaps, generally > 500 m2 (e.g. Bicknell 1982; Tuomela et al. 1996), and the gap centre advantage for advance regenerants may therefore apply only to medium-sized gaps (200–500 m2).
Our results may relate to the widely documented increase of A. rubrum in both the understorey and overstorey throughout the eastern USA (Abrams 1998). Palik & Pregitzer (1992) noted the importance of A. rubrum sprouts in achieving the overstorey in aspen-dominated stands in northern Michigan, while Abrams (1998) concluded that A. rubrum is a ‘super-generalist’ with a suite of traits that collectively give it an advantage in many sites and forest types in the absence of fire. To this list of traits we would add the ability to recover following leader damage caused by natural or anthropogenic disturbance.
Many forests in eastern North America are currently under heavy browse pressure due to high densities of white-tailed deer. Deer selectively remove terminal buds, may seriously suppress juvenile tree survival and growth, and may have substantial impacts on overstorey regeneration (e.g. Tilghman 1989; Gill 1992; Strole & Anderson 1992). Within forested landscapes, deer are often attracted to disturbed areas with greater densities of woody regrowth and available browse, such as canopy openings caused by logging. We are not aware of any published work comparing these three maple species in their susceptibility or response to deer browsing. We have observed browse damage on all three species in and around the Harvard Forest, but deer densities and browse pressure outside the fence in our study area were generally low compared to other areas in Massachusetts and the north-eastern USA.
Although the leader damage we documented was not due to deer, it does resemble deer browse in the selective death of meristems followed by sprouting from lateral or basal buds. If species-specific recovery following light deer browse (i.e. removal of one to several terminal buds) parallels the patterns we documented in our study, then deer pressure may favour A. pensylvanicum in the understorey and A. rubrum in larger gaps, at least for smaller juveniles (25–50 cm tall). The presence of deer is likely to increase the frequency of leader damage overall, placing even more significance on the role of damage and recovery processes and strongly suggesting that further research needs to be done on microsite- and species-specific responses to leader loss.
In summary, there were significant differences in spatial patterns of whole-plant survival and leader damage for advance-regenerated juvenile trees across the understorey-gap gradient, and these patterns differed for our three species of Acer. Damaged plants grew more in gaps than in understorey, particularly in large gaps, and the species differed significantly, with A. rubrum showing the greatest growth overall by damaged stems. Growth differences among species varied among sites for both intact and damaged plants, with large gaps producing the most pronounced species differences. Differences among species in recovery ability were offset by inverse patterns in the frequency of leader damage at the population level. Although A. rubrum was more resilient to damage, it suffered very high leader losses and was less successful as a population than A. pensylvanicum, which was much better at maintaining intact stems. Finally, some microsite-specific differences in growth recovery among these species may influence gap regeneration processes and forest composition, particularly through (i) cumulative damage and recovery episodes during prolonged pre-gap understorey periods, where A. pensylvanicum may have a modest advantage, and (ii) successful recovery in the centres of large gaps, where capture of canopy gap space is most likely to occur and where A. rubrum shows the most decisive advantage over its congeners.