Clonal regeneration of an arrow bamboo, Fargesia qinlingensis, following giant panda herbivory
- First Online:
- Cite this article as:
- Wang, W., Franklin, S.B. & Ouellette, J.R. Plant Ecol (2007) 192: 97. doi:10.1007/s11258-006-9229-x
- 141 Views
Characteristics of giant panda herbivory sites and clonal regeneration of an arrow bamboo Fargesia qinlingensis following giant panda grazing were studied in the Qinling Mountains of China. Three types of plots were located in a pandas’ summer habitat in 2002: herbivory (naturally grazed by giant pandas), clipped (simulated panda herbivory), and control. Average area of herbivory sites was 2.92 m2 and average distance from herbivory sites to the closest tree (dbh > 10 cm) was 1.0 m. Pandas avoided thin bamboo culms with basal diameters <5 mm. Average height of stumps of culms grazed by panda was 0.67 m and average density of grazed culms was 9.0 culms m−2. Annual culm mortality rate was significantly greater in herbivory and clipped plots than in control plots while annual recruitment rate was not significantly different among the three plot types in 2003. Neither recruitment rate nor mortality rate were significantly different among the three plot types in 2004. Annual recruitment rate was significantly greater than annual mortality rate only in control plots in both 2003 and 2004, suggesting static ramet dynamics in disturbed plots (herbivory and clipped). Density of new shoots was not significantly different, but basal diameter of new shoots was significantly less in herbivory plots compared to control plots in 2002. Differences of annual mortality rate and growth of new shoots found between control plots and herbivory plots suggest that clonal regeneration of F. qinlingensis culms was negatively affected by giant panda grazing. Therefore, no evidence of a clonal integration compensatory response to panda herbivory was found in F. qinlingensis.
KeywordsBasal diameterClonal integrationCulmFertilizer effectMortalityNew shootRecruitment
Dependence between endangered giant pandas, Ailuropoda melanoleuca David, and the bamboo forests that provide both habitat and 99% of the panda’s diet is well documented (Schaller et al. 1985; Liu 2001; Long et al. 2004). Giant pandas are herbivorous carnivores that evolved as obligate bamboo grazers (Wei et al. 1999). Thus, the life history of the giant panda is directly related to the life histories of the bamboos it feeds upon (Taylor and Qin 1988a), and the regeneration of bamboo is linked directly to the panda’s survival and conservation (Taylor and Qin 1988b, 1993). Researches have demonstrated effects of canopy openings (and other disturbances) on vegetative regeneration, and effects of environmental factors on seedling germination, survival, growth, density, and subsequent use of habitat by pandas (Taylor and Qin 1988a, b, 1989; Hu et al. 1990; Linderman et al. 2005). However, no attempt has been made to examine the effect of panda herbivory on clonal regeneration of bamboos, despite literature suggesting grazing can influence population-scale to landscape-scale structure and function of clonal plants (Gough et al. 2002; Tolvanen et al. 2002). Our present research examined the two-year recovery of a clonal bamboo, Fargesia qinlingensis Yi and Shao, following giant panda herbivory.
Fargesia qinlingensis is a perennial monocarpic species which is known for its long period of vegetative growth followed by a mast seeding every 50 years (Tian 1987). The response of culms to mammalian herbivory is through vegetative propagation during un-flowering periods. F. qinlingensis is a woody bamboo, and thus, could potentially respond to herbivory through release of suppressed buds along the shoot (e.g., Salix spp.) or from shoot proliferation from rhizomes. Effects of mammalian herbivory on clonal plant growth and population dynamics are complex due to various possible responses to clonal integration, the ability to transfer resources and information among ramets (Stuefer et al. 2002). Responses to grazing may be positive with increases in productivity, often a result of overcompensation (McNaughton 1984; Dyer et al. 1991). However, Bullock et al. (1994) find tiller densities of Agrostis stolonifera and Lolium perenne are decreased by winter grazing, and Rooney (1997) shows that forest herb Maianthemum canadense is negatively affected by herbivory, essentially surviving in ungrazed refugia. Response to herbivory, especially in woody clonal species, is often negative (Romme et al. 1995; Zeigenfuss et al. 2002).
Studies of clonal integration have shown clonal fragments in stressed zones can be supported with resources from unstressed fragments (Slade and Hutchings 1987; Jónsdóttir and Callaghan 1989; Alpert 1999; Peltzer 2002; Wilsey 2002). For example, Peltzer (2002) finds clonal integration tends to improve ramet survival and growth of Populus tremuloides. Tolvanen et al. (2002) find greater bud proliferation in Salix artica in muskox-grazed plots compared to ungrazed plots. If clonal integration of F. qinlingensis occurs, it is reasonable to assume that integration will compensate for negative effects of giant panda herbivory (Herms and Mattson 1992; Wilsey 2002).
Previous ecological studies of the impact of herbivory on plants have failed to incorporate a natural grazed treatment along with clipping treatments (Paige 1999). Using mechanical damage to mimic natural herbivory usually fails to adequately simulate natural herbivory, because (1) mechanical simulations may not match natural timing, amounts, or types of tissues eaten, (2) mechanical simulations may differ from herbivore damage in triggering damage cues and subsequent defensive responses, and (3) herbivores may manipulate a plant’s physiology through hormones present in saliva (Baldwin 1990; Rooke 2003). For example, no evidence of overcompensation is found in Ipomopsis aggregata populations when ungulate herbivory is simulated in experimental clipping treatments (Bergelson et al. 1996), while evidence for overcompensation is found in I. aggregata populations when plants are naturally grazed by ungulates (Paige 1999). Herbivory by giant pandas may alter abiotic factors in the immediate vicinity of the grazed site. Grazing by giant pandas usually adds nutrients through the excretion of feces or urine at feeding sites (Wei et al. 2000), and grazing in other Fargesia species tends to create small openings in the bamboo canopy (Schaller et al. 1985). Light and nutrients are major factors controlling clonal ramet dynamics (Reid et al. 1991; Taylor et al. 1991; Peterson and Squires 1995), so increasing these resources will likely increase reproduction of new shoots; a ‘fertilizer effect’. Indeed, herbivory of bamboo by giant panda has the potential to create a spatial pattern based on preferential feeding (Reid and Hu 1991). The ability to integrate the spatial preferential feeding into a genet-level response to herbivory (e.g., Schmid et al. 1990) is unknown.
The aims of this study were to investigate the characteristics of herbivory sites and clonal regeneration of F. qinlingensis following giant panda grazing. We specifically tested the hypothesis that herbivory would affect culm dynamics. To separate confounding factors of herbivory and abiotic alterations of a site due to herbivory, we set up clipped and control plots adjacent to naturally grazed sites. We predicted increased annual recruitment rate and decreased annual mortality rate in herbivory and clipped plots compared to control plots. Furthermore, we predicted annual recruitment rate would be greater than annual mortality rate in all plots. Our final prediction was that grazed sites would have the greatest density and greatest basal diameter of new shoots due to a ‘fertilizer effect’ subsequent to panda grazing.
This study was carried out in Foping National Natural Reserve (FNNR, 33°33′–33°46′N, 107°40′–107°55′E), which is located in the middle part of the southern slope of the Qinling Mountains, Shaanxi Province, China. The aim of FNNR is to conserve giant pandas and their habitat. Sixty-five pandas were confirmed after the survey in 1998, averaging one panda 5 km−2 (Liu 2001). Suitable panda habitat occurs in 67% of the reserve (Liu 2001).
FNNR is at the northern edge of a subtropical area with an average annual temperature 11.5°C; −0.3°C in January and 21.9°C in July. Precipitation averages 924 mm year−1 with 91% of annual rainfall occurring between April and October. Frost-free days average 220. There are three types of soil with a vertical distribution: yellow brown earth below 1,500 m; brown earth between 1,500 and 2,300 m; and dark brown earth above 2,300 m (Liu 2001).
The vegetation is diverse due to the coexistence of both northern and southern Chinese taxa (Pan et al. 1988). Forest communities cover about 80% of FNNR (Mackinnon et al. 1996), made up of generally four vegetation types along an elevation gradient: deciduous broad-leaved forests (<1,800 m), mixed coniferous and broad-leaved forests (1,700–2,800 m), coniferous forests (2,400–2,900 m), and interspersed shrublands and meadows (>2,900 m) (Ren 1998). Three bamboo species are distributed in FNNR: F. qinlingensis, Bashania fargesii (Camus) P.C. Keng & Yi, and Fargesia dracocephala Yi (Masman). Fargesia qinlingensis and B. fargesii are the main panda grazing species within pandas’ habitat in FNNR (Yong et al. 1994; Li 2002).
Like most bamboos in tropical and temperate regions (Janzen 1976; Gadgil and Prasad 1984), F. qinlingensis is a perennial monocarpic species which has a seeding cycle of ca. 50 years (Tian 1987). F. qinlingensis culm can grow up to 3.6 m in height and 1.1 cm in basal diameter. F. qinlingensis is a clump-forming bamboo with a pachymorph rhizome system. Rhizome growth is the only way to spread ramets during un-flowering period. New ramets regenerate from rhizomes in May and June (Li 2002).
Nine panda herbivory (natural grazed) sites were chosen from a pandas’ summer habitat at altitudes ranging from 2,600 to 2,642 m in 2002. When pandas graze, they often remove culm tops, leaving a stump behind, at a height based on preference. All herbivory sites were the same age and grazed in summer of 2001 (Gaodi Dang, research expert from FNNR, personal communication). The distance between any two herbivory sites was at least 50 m to avoid pseudoreplication. Each of the nine herbivory sites was enclosed by string to an elliptical shape. Length and width of each plot were measured, and the area was calculated by using an elliptical formula. Adjacent to each herbivory plot, in the ungrazed area, we set up two additional plots with the same shape, size, and approximate culm density as the herbivory plot: a clipped plot and a control plot. In the clipped plot, we simulated giant panda herbivory by clipping the same number of culms at the same height as those grazed by pandas in the herbivory plot. The clipped plots were clipped only once in 2002. The experiment was a block design; there were nine blocks, and each block had an herbivory plot, a clipped plot, and a control plot. The distance between any two plots in each block was 2–3 m to maintain same soil conditions, but none of the three plots were located within the same visible clump. The basal diameter of each new shoot and culm in plots was measured in October 2002, and was re-measured in September 2003 and July 2004. The height of stumps grazed by panda was also measured in October 2002. One site was destroyed by another mammal and one site was destroyed by a fallen tree in the winter of 2002. Thus, analyses involving 2003 and 2004 data had seven blocks. In this study, a new shoot was defined as a ramet of less than one year old (dark green, sheaths still attached to the nodes), and a culm (i.e., established shoot) was defined as a ramet of greater than one year old (light green, sheaths fallen out from the nodes).
Characteristics of herbivory sites
Characteristics of herbivory sites were recorded in October 2002, including distance between herbivory site and the nearest tree >10 cm diameter at breast height (dbh), percent tree canopy cover, maximum culm height, percent bamboo cover, percent herbaceous species cover, and dominant herbaceous species based on cover. Cover was visually estimated.
Grazing size selection of culms
Chi-square goodness-of-fit test was used to test for significant differences in size class distributions of bamboo culms in herbivory and control plots, and in culms that were grazed compared to total culms within herbivory plots. Size class 1 included all culms <3.0 mm basal diameter. All other size classes were 0.5 mm increments ranging from 3.0 to 13.5 mm in basal diameter (e.g., size class 2 included culms with basal diameter 3.0–3.5 mm).
We calculated annual mortality rate and annual recruitment rate in 2003 and 2004. Recruitment rate is expressed as the percentage of new shoots at the beginning of the growing season, and mortality rate is expressed as the percentage of culms dying over a 12-month-period (sensu Taylor and Qin 1987). Friedman test, a nonparametric test of treatments that blocks by plot sets, was used to compare mortality and recruitment rates among herbivory, clipped, and control plots in 2003 and 2004. Tukey multiple comparison was subsequently used if the difference was significant. Friedman test was also used to test the difference between recruitment and mortality rates in herbivory, clipped, and control plots in 2003 and 2004.
Assuming plots had already responded to herbivory (since plots were grazed in 2001), we tested differences in density and basal diameter of new shoots between herbivory and control plots with paired t-tests using the 2002 data. We further tested the effects of simulated herbivory by comparing the clipped and control plots using repeated measures ANOVA on density and basal diameter of new shoots. Factors were time (2002, 2003, and 2004) and treatment (clipping and control). All analyses were performed using SAS/STAT (SAS 2001). Differences were considered significant at P < 0.05 level.
Characteristics of herbivory sites
Characteristics of giant panda herbivory sites in Foping National Nature Reserve, Qinling Mountains, Shaanxi, China
Herbivory site area (m2)
Distance to tree with dbh > 10 cm (m)a
Tree canopy cover (%)b
Maximum culm height (m)
Bamboo cover (%)
Herbaceous cover (%)
Dominant herbaceous species
Mean ± SE
2.92 ± 0.81
0.99 ± 0.28
20.78 ± 9.09
2.44 ± 0.05
73.33 ± 7.82
17.67 ± 7.29
Grazing size selection of culms
In 2003, annual recruitment and mortality rates were not significantly different in herbivory (n = 9, F = 1.35, P = 0.29) and clipping plots (n = 9, F = 0.13, P = 0.74), but recruitment rate was significantly greater than mortality rate in control plots (n = 9, F = 8.40, P = 0.027) (Fig. 3a). Recruitment rate was greater than mortality rate in 2004, but only significantly greater for herbivory (n = 7, F = 36.00, P = 0.001) and control plots (n = 7, F = infinity, P < 0.0001); not for clipped plots (n = 7, F = 0.13, P = 0.74) (Fig. 3b). During the 2-year study, herbivory plots averaged recruitment of 6 new shoots m−2, clipped plots 5 new shoots m−2, and control plots 14 new shoots m−2.
In 2002, new shoot density was not significantly different (n = 9, t = 1.618, P = 0.144) in herbivory plots (6.72 ± 2.25 culms m−2) compared to control plots (4.77 ± 1.49 culms m−2), but mean basal diameter of new shoots was significantly less (n = 9, t = −4.931, P = 0.001) in herbivory plots (7.37 ± 0.28 mm) than in control plots (8.14 ± 0.34 mm).
The aims of this study were to investigate the characteristics of herbivory sites and clonal regeneration of bamboo F. qinlingensis following giant panda grazing. Herbivory sites averaged ca. 3 m2 in size, with panda grazing an average of 9 culms m−2 (∼19% of culms in plots). Comparing to Schaller et al. (1985) who study pandas in Wolong Reserve in the Qionglai Mountains of China, where approximately 80% of herbivory sites have three or less culms eaten, sites in this study were heavily grazed. Density of culms in this study, averaging 48 culms m−2, was approximately half of 90 culms m−2 found in Fargesia sites in Wolong Reserve (Reid and Hu 1991). Trees were sparse in pandas’ summer grazing habitat (320 ± 80 trees ha−1), while bamboo cover were high (>70%). Taylor and Qin (1988b) demonstrate that shrub density and woody plant species richness are less in stands with high bamboo cover in Wolong Reserve because bamboos impede tree regeneration. Giant pandas potentially prefer these high bamboo density sites for grazing (Schaller et al. 1985; Reid and Hu 1991). Average distance between herbivory sites and the closest tree >10 cm dbh was only 1 m. Close distances between herbivory sites and trees are also found in Yele Natural Reserve in Sichuan Province of China (Wei et al. 2000). Yong (1989) suggests that adjacency facilitates giant panda escape by climbing trees when frightened by humans or mammalian predators.
While giant pandas are specialist grazers on bamboo, which generates 99% of their diet (Schaller et al. 1985; Liu 2001; Long et al. 2004), they are also selective herbivores. Selectivity has been shown in some mammalian herbivores, including choice among conspecifics (Swihart and Picone 1998), choice among habitats where the plant grows (Danell et al. 1991), and choice of where browsing occurs on a plant (Rounds 1979). Our data showed that grazing was biased toward thick culms between 5.47 and 11.81 mm in basal diameter and biased away from thin culms, regardless of the size class distribution of the herbivory plots. Pandas in Wolong Reserve show the same preference for thick culms and avoidance of thin culms (Schaller et al. 1985; Reid and Hu 1991). The reason is likely due to handling efficiency being greater for thicker culms compared to thin ones. In addition, pandas often remove culm tops, leaving a stump behind, at a height based on preference. Schaller et al. (1985) find winter grazing stumps tend to be higher than summer grazing stumps, and stump height is positively correlated with age. Stumps from our study averaged 67 cm, higher than in the study of Schaller et al. (1985), where stumps average 15–40 cm for summer grazing.
Annual recruitment and mortality rates in control plots averaged 14% and 7%, respectively, in 2003, and 13% and 2%, respectively, in 2004. Thus, there was an overall increase in culm density during this study: 14 new shoots m−2. Average mortality rate was greater in the herbivory and clipped plots compared to the control plots (significantly greater in 2003), opposite our original prediction. Mortality rate in 2003 fell within the 10–20% mortality range expected if death rate is constant and longevity of culms is 5–10 years (McClure 1966; Campbell et al. 1983), but averaged only 5% in 2004. We have no explanation for this annual variation, but it is typical in clonal plant populations (Slade and Hutchings 1987).
Also contrary to our original prediction, annual recruitment rate did not significantly differ among plots, and thus, a nearly static ramet population was witnessed in 2003, with approximately equal recruitment and mortality rates in herbivory and clipped plots. Further, data did not support our prediction that recruitment rate would be greater than mortality rate in all plots, a pattern that occurred only in control plots for both years. Both Campbell et al. (1983) and Schaller et al. (1985) report static or declining Fargesia ramet dynamics. The data from 2004 suggest slightly different dynamics. Recruitment rate was significantly greater than mortality rate in both herbivory and control plots, due to a substantial decline in mortality rate compared to the data from 2003. Mortality rates were greatest in clipped plots for both years, suggesting clipping treatment had the greatest negative effect on culm densities.
The recruitment rate was significantly greater than the mortality rate in herbivory plots, but not in clipped plots, suggesting the importance of including natural grazed plots in studies of herbivory effects on plant response (Paige 1999). Data did support our hypothesis that herbivory would affect culm dynamics, although the herbivory plots showed no clear differences in new shoot recruitment compared to clipped plots. Thus, we found no evidence of a ‘fertilizer effect’ following panda herbivory, although nutrients and especially light are known to strongly affect ramet dynamics (Schaller et al. 1985; Reid and Hu 1991; Taylor et al. 1991). Indeed, herbivory plots also had less mean basal diameters of new shoots than control plots in 2002. Although insignificant, trends from the data suggest that clipping had a negative affect on new shoot recruitment; new shoot density increased from 2002 to 2004 in control plots, but decreased in clipped plots. New shoots during the course of the study were at least two times greater in control plots (14 new shoots m−2) compared to herbivory plots (6 new shoots m−2) and clipped plots (5 new shoots m−2).
The negative response to herbivory was different from other clonal perennials that respond positively (i.e., abundance increases following increased grazing intensity) or show no response to grazing (Sackville Hamilton et al. 1987; Solangaraachi and Harper 1987). Results appeared to mimic typical negative responses of woody clonal species, as opposed to clonal grasses (Zeigenfuss et al. 2002). This may be one reason why pandas avoid previously grazed sites (Reid and Hu 1991). We saw no sign that giant pandas had returned to this area for grazing in 2003 or 2004.
Our data showed no evidence of clonal integration. If F. qinlingensis was compensating for herbivory at the genet level, we should witness no differences in response to herbivory in the control and treatment plots. Wilsey (2002) concludes that integration among herbaceous grass ramets is probably unimportant in the ability of patches to regrow following large mammal herbivory on the Serengeti. However, physiological integration is evident in two bamboo species. Li et al. (2000) report that more new shoots of Phyllostachys pubescens develop in unfertilized patches (surrounded by fertilized area) compared to uniformly unfertilized patches because of an export of resources from the fertilized patches to the unfertilized patches through belowground rhizomes. Saitoh et al. (2002) demonstrate that clonal parts of dwarf bamboo Sasa palmate in shade are supported by translocation from connected clone parts in open areas, and such physiological integration can enhance the persistence of S. palmate in a heterogeneous resource environment such as the gap-understory continuum of temperate forests. The lack of physiological integration in F. qinlingensis may be explained by a ‘cut off point’, suggested by Jónsdóttir and Callaghan (1989). The translocation of resources between stressed (e.g., grazing) and unstressed modules has a “cut off point” beyond which subsidies are not provided. This is observed in Carex bigelowii, which cuts off carbon allocation between young and older tillers under heavy grazing pressure (Jónsdóttir and Callaghan 1989). An alternative strategy of clonal plants is an individual ramet response (Haukioja 1991). Plants may regulate, via growth regulators (hormones), the degree of integration, possibly compartmentalizing stressed zones. This alternative strategy would explain the lack of regeneration in response to increased light and an expected increase in nutrients on herbivory sites, and deserves further scrutiny.
We are grateful to Dr. Arnold van der Valk and two anonymous reviewers whose comments greatly improved the manuscript. We would like to thank the Foping National Nature Reserve for logistic support of our fieldwork in the reserve. We also thank Gaodi Dang, Leigang Zhao, Xichun Du, and Haining Li for their assistance in the field. This study was supported by the Memphis Zoo.