Introduction

In temperate deciduous forests, several species of understory trees are clonal and produce ramets by root suckering (Del Tredici 2001). The production of ramets results in the formation of clonal patches that can occupy dozens to hundreds of square meters (Gilbert 1966; Reinartz and Popp 1987; Douhovnikoff et al. 2005). Dense patches formed by understory clonal trees can suppress other plant species, including seedlings of canopy trees (Beckage et al. 2000; Frelich et al. 2003; Cole and Weltzin 2005; Baumer and Runkle 2010) and understory herbs (Sammul et al. 2004). Thus, it is important to understand how clonal understory trees develop patches as they could influence forest vegetation and dynamics.

Patches of clonal understory trees are, however, not static. In addition to expanding through asexual propagation, individual ramets die or are damaged by herbivores or disturbances at a small scale. Falling litter, such as dead branches, snapped trunks and uprooted trees are common agents of physical disturbances (e.g., Webb 1999; Whigham et al. 1999) and have been estimated to affect 2.5–20 % of the total area per year in many types of the forests (Clark and Clark 1989; McCarthy and Facelli 1990; Mack 1998; Drake and Pratt 2001; Gillman and Ogden 2001). Although the capacities of both asexual reproduction and regrowth by sprouting seem both to be adaptive in tolerating damage caused by falling litter (Lasso et al. 2009), few have investigated how clonal understory trees respond to this small-scale disturbance. This study tests which type of recovery is dominant in patches subjected to small-scale disturbance.

When some ramets are damaged and the others escape, the damaged ramets may re-grow by support from undamaged ones via clonal integration (Schmid et al. 1988; Chapman et al. 1992; Bach 2000; Liu et al. 2009; Xu et al. 2012), or using resources stored in underground organs (Wang et al. 2004). In the former case, the impact of disturbance would depend on the extent to which ramets are supported by other members of the same patch, and different effects of damaging source ramets or sink ramets can be expected (Xu et al. 2012). We hypothesize that if large ramets have a greater capacity for export of resources than small ramets, less biomass would be gained by sprouting when large ramets are damaged. We also test the hypothesis that if damaged ramets benefit from clonal integration, undamaged ramets may incur costs in terms of reduced growth. Alternatively, the amount of biomass gained after disturbance would not be sensitive to which ramets are damaged if connections between ramets function as stores of carbohydrate and are capable of supplying local ramets (Suzuki and Stuefer 1999; Wang et al. 2004; Esmaeili et al. 2009). Undamaged ramets may not be affected by small-scale disturbance on neighboring ramets if these resources are sufficient to support regrowth.

In this paper, we present results of a field experiment using the understory clonal tree, Asimina triloba (L.) Dunal, which is common in temperate forests of eastern North America. Ramets of different sizes were cut at their bases to simulate damage caused by falling litter. About half of the aboveground biomass was removed from each patch of A. triloba by cutting a single large ramet or many small ramets. Specific questions addressed are (1) Does disturbance change the production of new ramets? (2) Do damaged ramets re-grow or die after the disturbance event? (3) Do disturbed patches gain less biomass when large ramets are damaged? and (4) Do undamaged ramets grow less compared to control ramets of the same size classes?

Methods

Asimina triloba is a native, clonal understory tree that is widely distributed in eastern North America from northern Florida to Ontario (Kral 1960). Asimina triloba annually produces new ramets that arise on a horizontal root system of the parent plant (Karizumi 1979; Hosaka et al. 2005). Ramets are typically avoided by herbivores because of their toxic Annonaceous acetogenins (Slater and Anderson 2014), which can benefit the species compared to others in areas with large numbers of white-tailed deer (McGarvey et al. 2013). Based on the number of tree rings at ground level, ramets of A. triloba are able to survive more than 30 years (Hosaka N, personal observation). A series of ramet generations form a patch with hierarchical size structure, ranging from less than 0.3 m to more than 10 m in height (Larimore et al. 2003). Root connections between ramets are likely to remain intact for at least a decade. Preliminary studies showed that severing root connections reduce shoot growth and leaf size of offspring ramets (Stuefer JF, personal observation), indicating that the root system is physiologically integrated.

Field experiment

A field experiment was conducted in a temperate deciduous forest at the Smithsonian Environmental Research Center (SERC), in Maryland, USA (38°53′N, 76°33′W). Common canopy trees in the forest are Liriodendron tulipifera L., Carya spp., Quercus spp., Fagus grandifolia Ehrh., and Liquidambar styraciflua L., and the height of the forest canopy reaches 25–35 m (Parker et al. 1989). Numerous distinct patches of A. triloba occur in the forest and the patches contain a few to more than 800 ramets (Hosaka et al. 2005).

In June 2001, we selected 27 patches that were at least 3 m apart. Each patch covered ca 20 m2 and consisted of fewer than 100 ramets, with a single large ramet (16–55 mm in basal stem diameter) and many small ramets (8 mm in basal stem diameter on average). We made hand-excavations in about one third of the patches to confirm that ramets were part of the same interconnected root system. In July 2001, all ramets in each patch were tagged and measured for height (cm), basal stem diameter (mm) at 3-cm above the ground, and the number of current-year shoots.

The 27 patches were randomly assigned to three treatments (L-removal, S-removal, and Control) as described below (Fig. 1). The patches were ordered by basal stem diameter of their large ramets and grouped into three classes (nine patches in each class). The ranges of basal stem diameter were 16–31, 32–37, and 38–55 mm for S, M, and L classes, respectively. Three patches from each class were randomly chosen for one of the treatments. In August 2001, the large ramet in each of the nine L-removal patches was cut 5 cm above the ground, and all small ramets in each of the nine S-removal patches were cut in the same manner. No ramets were cut in the Control patches.

Fig. 1
figure 1

Schematic diagrams of Asimina triloba patches subjected to three treatments: a Control, b L-removal, and c S-removal. Dark- and light-filled regions with solid lines represent plant sizes at the start and the end of the experiment, respectively. Open regions with broken lines represent ramets that have removed in 2001. Initial aboveground biomass is calculated for each patch as a sum of the ramets of dark-filled and open regions. The light-filled regions represent biomass gain after 2001

Full size image

Ten and 35 months after cutting (in June 2002 and July 2004, respectively), ramets that had not been cut were measured for diameter, height, and number of leafing branches. For each of the damaged ramets that had been cut in 2001, we determined the presence or absence of stem sprouts and measured heights of sprouts that were present. New ramets that had arisen on the horizontal roots of the patches were verified by hand excavation, tagged, and measured for height and basal stem diameter. Ramets were scored as dead if all aboveground parts were missing or if the ramet had no sprouts. One patch in the S-removal treatment was excluded from the data analyses because the large ramet of the patch was severely damaged by a windstorm in 2003.

The large and small ramets that were cut in 2001 were used to estimate aboveground biomass of the patches. The ramets were individually weighed after oven-drying (60 °C) for at least 72 h. Based on field measurements (height, basal stem diameter, and number of branches with leaves) that had been made before cutting, separate stepwise multiple regression equations were developed for large and small ramets. Terms that represented a significant improvement in the model were included in the equations (P < 0.05). The regression equations were B = 1156.81 − 89.10D + 1.80D 2 + 3.39L (r 2 = 0.99; n = 9) for large ramets and B = 0.25D 2 − 0.11H + 2.42L (r 2 = 0.93; n = 192) for small ramets, where B, D, H, and L represented aboveground biomass, basal stem diameter, height, and the number of leafing branches, respectively. Averages of the total aboveground biomass and the number of ramets per patch were similar between treatments before the start of the experiment (Table 1). Large ramets accounted for approximately half of the total aboveground biomass of each patch. Consequently, similar amounts of biomass were removed from the patches of L-removal and S-removal treatments, respectively (Table 1). A separate equation was obtained to estimate the aboveground biomass of the stem sprout and new ramets: B = 0.00296H 2 (r 2 = 0.85; n = 192). Based on the three equations, absolute growth was calculated for individual ramets by subtracting the amount of biomass at the start of the census period from the amount of biomass at the end of the census period. The changes in biomass were divided by the time period (years) to calculate growth rate of individual large and small ramets.

Table 1 Summary characteristics of Asimina triloba patches before (2001) and after (2002, 2004) the initial cutting of ramets. Mean values for the measurements for each treatment are shown with the standard errors in parentheses
Full size table

Data analysis

All statistical tests were performed using R 3.2.0 (R Core Team 2015). We analyzed treatment and patch class effects at patch scale with generalized linear models (GLMs). Different models were used depending on the nature and distribution of the dependent variables: gamma models for the total aboveground biomass and total gained biomass per patch; quasipoisson models for the total number of ramets per patch. For the number of new ramets per patch, poisson models were used with the total number of ramets recorded in the previous census as offset variables. For survival rates of the small ramets, quasibinomial models were used with the numbers of live and dead ramets per patch as combined objects using cbind function. When an analysis of deviance indicated a factor effect with P < 0.05, the model was simplified by aggregating non-significant levels in stepwise a posterior procedure (Crawley 2015).

We also analyzed effects of the treatment and patch class at ramet-scale with GLMs. The models were simplified as described above. For individual new ramets, aboveground biomass estimated in their first census year (2002 or 2004) were analyzed using gamma models. For individual large and small ramets, growth rates were analyzed using binomial models with the aboveground biomass estimated in previous census as an offset variable.

Results

Number of ramets

During the 3 years after the initial cutting of ramets in 2001, both types of disturbed patches increased the number of ramets by sprouting of damaged ramets and recruitment of new ramet (Table 1). No significant difference was found between treatments in the total number of ramets per patch either in 2002 or 2004 (Table 2). There was also no significant difference between patch classes in the total number of ramets (Table 2).

Table 2 Results of the analysis of deviance in 2002 and 2004 examining effects of the treatment and patch class factors on total number of ramets, survival of small ramets, and recruitment of new ramets of Asimina triloba patches
Full size table

All damaged ramets in the L-removal patches produced one or more stem sprout(s) which persisted until the end of the experiment (Fig. 2a, b). There was also no mortality of either undamaged large ramets in S-removal patches or large ramets in Control patches over the three-year study period (Fig. 2a, b).

Fig. 2
figure 2

Cumulative treatment means of the number of large, small, and new ramets of Asimina triloba patches in a 2002 and b 2004. Error bars represent standard errors of the total number of ramets of whole patches for each treatment. Cumulative treatment means of the component biomass gained by large, small, and new ramets in the c 2001–2002 and d 2002–2004 periods. Error bars represent standard errors of the total gained biomass of whole patches for each treatment

Full size image

For small ramets, an average of 82 % of the damaged ramets of the S-removal patches sprouted in the first year after the disturbance, while 96 % of small ramets of Control patches survived during the same period. The damaged small ramets had lower survival than small ramets in Control patches by the end of the experiment with significant effects of the treatment during 2001–2002 and 2002–2004 (Table 2). Undamaged small ramets of the L-removal patches showed equivalent survival to small ramets in Control patches in both 2001–2002 and 2002–2004 (Table 2). There were no significant differences in small ramet survival between size classes of the patches (Table 2).

New ramet recruitment occurred in the disturbed patches as well as the Control patches in both census periods (Fig. 2a, b). Although the number of recruits varied among the patches, there was no significant effect of the cutting treatment or patch class (Table 2). Recruitment ratio calculated as the number of new ramets divided by the total number of ramets in the previous census was similar among treatments with overall means of 0.50 and 0.75 for the periods between 2001 and 2002 and 2002–2004, respectively.

Biomass at patch level

Total aboveground biomass of disturbed patches recovered to pre-treatment levels by the end of the three-year study period, but biomass in neither type of disturbed patch did not reach the same levels as biomass in Control patches (Table 1). Effects of the treatment and size class were significant both in 2002 and 2004 (Table 3). Total aboveground biomass decreased with the order of patch class L, M, and S.

Table 3 Results of the analysis of deviance in 2002 and 2004 examining effects of the treatment and patch class factors on total patch biomass, total gained biomass, component biomass gained by large ramets, component biomass gained by small ramets, and component biomass gained by new ramets of Asimina triloba patches
Full size table

Disturbed patches gained similar amount of biomass (Table 1), with no statistically significant difference among the treatments in total gained biomass (Table 3). However, damaged large ramets of the L-removal patches gained less biomass than ramets of the Control patches, and so did damaged small ramets of the S-removal patches (Fig. 2c, d). Treatment effects on component biomass were significant and marginally significant for large ramets and small ramets, respectively (Table 3). No significant effect of the treatment was found in biomass gained by new ramets either in 2002 and 2004 (Table 3).

Size class of the patches also affected total biomass, total gained biomass, and component biomass, and these tended to decrease with the size class L, M, and S. Highly significant differences among the classes were found only in total patch biomass at patch level analysis (Table 2).

Biomass at ramet-level

Large ramets showed a significant difference in growth rates among treatments during the first year after the manipulation (Table 4). Damaged large ramets in the L-removal patches showed lower growth rates than those of Control and S-removal patches in 2002 (Fig. 3a). Two years later, however, there were no significant differences in growth rates among the treatments (Table 4). Undamaged large ramets of the S-removal patches grew as well as the large ramets in Control patches both in the 2001–2002 and 2002–2004 periods (Fig. 3a, b). Size class of the patch had no significant effect on the growth rate of its large ramet either in 2002 or 2004 (Table 2).

Table 4 Results of the analysis of deviance in 2002 and 2004 examining effects of the treatment and patch class factors on growth rate of large ramets, growth rate of small ramets, and absolute growth of new ramets of Asimina triloba patches. Letters in parentheses denote levels of the treatment factor for each census data as described in Table 2
Full size table
Fig. 3
figure 3

Treatment means and standard errors of growth rate of large ramets of Asimina triloba for each patch class in the a 2001–2002 and b 2002–2004 periods. Treatment means and standard errors of growth rates of small ramets for each patch class in the c 2001–2002 and d 2002–2004 periods. Treatment means and standard errors of absolute growth of new ramets for each patch class in e 2002 and f 2004

Full size image

Damaged small ramets of the S-removal patches also had lower growth rates than those of the Control patches during the first year after the manipulation (Fig. 3c; Table 4). Two years later, however, no significant difference was found between the treatments in the growth rates of the small ramets (Table 4). Effects of size class were always significant (Table 4) and growth rates of the small ramets were higher in the order of patch class L, M, and S in 2002 and 2004.

New ramets produced in both types of disturbed patches grew less than those of Control patches in the first year after the manipulation (Fig. 3e). Significant differences were detected between the Control and both the L-removal and S-removal treatments in 2002 (Table 4). Two years later, new ramets of the patches of the S-removal treatment showed equivalent growth to those of the Control patches (Fig. 3f), and no significant difference was found between Control and S-removal treatment in 2004. The effect of patch class was also observed in 2004 and growth of the new ramets was greater in the order of patch class L, M, and S.

Discussion

When clonal patches of understory trees are disturbed at a small scale by falling litter, the damage can vary in both the size and the number of directly affected ramets within a patch. The goal of our research was to determine the effects of disturbance to a single large ramet or many small ramets on recovery and development of patches in the clonal understory tree, Asimina triloba. Our experiment showed that small-scale disturbance did not change new ramet production and the recovery response of A. triloba was primarily achieved by sprouting of damaged ramets in either type of disturbance. Both types of disturbed patches gained similar amount of biomass and undamaged ramets grew as Control ramets of the same size classes. These results did not confirm our hypothesis that regrowth will be supported by resources from adjacent ramets, suggesting that sprouting of damaged ramets were mainly supported by resources stored in interconnected root systems.

Previous studies that employed large-scale disturbances to patches (e.g., clear-cutting or burning entire patches) of clonal trees showed increased ramet production (Cirne and Scarano 2001; Tappeiner et al. 2001). Since an increase in ramet production is often followed by high mortality of new ramets in the first growing season (Clark et al. 1982; Tappeiner et al. 2001) or within a few years (Gardiner and Helmig 1997; Vilà and Terradas 1995), this type of response to disturbance would require higher levels of initial cost to produce new ramets compared to the recovery that we found in this study. Regrowth of damaged ramets is, therefore, an energetically economic compensatory strategy that is favored under a closed forest canopy, where small disturbances are frequent and productivity of the habitat is low due to shading.

Despite considerable differences in size and number of the damaged ramets, the types of disturbances used in this study demonstrated that stem sprouting of damaged ramets were resulted in no cost in terms of reduced growth of undamaged ramets. These findings are surprising if one assumes that previously established large ramets mitigate damages to small ramets, as a number of experimental studies on clonal plants have shown (Schmid et al. 1988; Chapman et al. 1992; Bach 2000; Liu et al. 2009; Xu et al. 2012). Resources for sprouting of damaged ramets are not likely to be directly supplied by the undamaged ramets but probably stored in the horizontal root systems. This is in agreement with a common finding in woody plants, in which sprouting abilities positively correlate with carbohydrate and nutrient contents in belowground structures (Canadell and López-Soria 1998; Cruz et al. 2003; Kabeya and Sakai 2005; Luostarinen and Kauppi 2005).

One of advantages of storage in underground root systems for A. triloba seems to ensure resources for sprouting against aboveground disturbance. If resources for sprouting are supplied from aboveground parts of ramets, less should be available after disturbance, especially risk of which is high and equal for all ramets within a patch. Damage caused by falling litter is a typical consequence of such frequent unpredictable disturbance. When disturbance is predictable, however, loss of resources for sprouting is to be reduced by supplying from ramets with a lower risk of damage. Unlike disturbance caused by falling litter, grazing or trampling is more likely to damage small ramets of clonal trees. For example, an experimental study of Populus simonii showed that large ramets suffering little trampling could mitigate damage to interconnected small ramets (Xu et al. 2012).

Damaged ramets grew less than Control ramets, especially during the first year after the disturbance. New ramets also grew less in disturbed patches than in the Control, and the effects of disturbance were prolonged in the treatment where large ramets were damaged. In A. triloba, damage to large ramets may compromise sexual reproduction, which is limited to large ramets (Hosaka et al. 2008), and have negative impacts on new patch establishment via sexual reproduction. Besides such a direct effect of disturbance, storage for sprouting could have an indirect effect of disturbance on sexual reproduction, as storage implies a cost in terms of reduced current growth and performance of ramets (Suzuki and Stuefer 1999; Wiley and Helliker 2012). Such cost-benefit considerations between storage and current growth suggest that there should be a close fit between disturbance mode in different resource availability and active storage intensity in clonal plants.