INTRODUCTION
In addition to factors such as the timing, frequency and intensity of herbivory, the responses of plants to herbivory are influenced by both the type of damage and the ontogenetic stage of the plant. For example, consumption of woody stems and branches by mammalian herbivores (hereafter referred to as pruning) and leaf tissue consumption by insects (hereafter referred to as defoliation) typically result in very different responses in re-growth rates, shoot and leaf size and foliar chemical composition (Bergstrom et al. Reference BERGSTROM, SKARPE and DANELL2000, Bryant et al. Reference BRYANT, HEITKÖNIG, KUROPAT and OWEN-SMITH1991, Hrabar et al. Reference HRABAR, HATTAS and DU TOIT2009, Lehtila et al. Reference LEHTILA, HAUKIOJA, KAITANIEMI and LAINE2000, Messina et al. Reference MESSINA, DURHAM, RICHARDS and MCARTHUR2002, Rooke & Bergstrom Reference ROOKE and BERGSTROM2007). Contrasting effects of pruning versus defoliation on plant architecture are expected as only the former typically involves developmental changes in response to removal of the apical meristem and shifts in the root/shoot ratio. Woody plants show responses following pruning including increased branch length and often larger leaf size (Bergstrom et al. Reference BERGSTROM, SKARPE and DANELL2000, Hrabar et al. Reference HRABAR, HATTAS and DU TOIT2009, Lehtila et al. Reference LEHTILA, HAUKIOJA, KAITANIEMI and LAINE2000, Messina et al. Reference MESSINA, DURHAM, RICHARDS and MCARTHUR2002, Rooke et al. Reference ROOKE, BERGSROM, SKARPE and DANELL2004), whereas defoliation often results in smaller branch and leaf size (Piene et al. Reference PIENE, MACLEAN and LANDRY2003, Rooke & Bergstrom Reference ROOKE and BERGSTROM2007).
The ontogenetic stage of the plant when damaged can also play a key role in determining both the pattern of damage and plant responses (Boege & Marquis Reference BOEGE and MARQUIS2005, Dirzo Reference DIRZO, Dirzo and Sarukhan1984). The patterns of herbivory and the resources available for plant responses vary significantly from the seedling to mature reproductive stages. In general, lower resource-acquisition capabilities (leaf and root surface area), limited resource storage, and resource-allocation constraints of juveniles compared with adults suggest that herbivory tolerance (re-growth capacity) of juveniles might be lower than adults. An accurate assessment of the selection imposed by herbivores on plants requires study of responses at all ontogenetic stages, but our understanding of plant responses to herbivory is currently limited as most studies involve experiments and observations at only the adult stage (Boege & Marquis Reference BOEGE and MARQUIS2005).
Woody species of tropical savannas and woodlands typically experience both leaf and stem tissue damage from diverse guilds of vertebrate and invertebrate herbivores. Herbivory at the seedling recruitment stage is crucially important in these species. In savanna ecosystems, factors influencing tree seedling growth and mortality, and demographic constraints at the juvenile recruitment phase of tree populations are critical to explaining tree population dynamics, the coexistence of trees and grasses and their relative abundance (Higgins et al. Reference HIGGINS, BOND and TROLLOPE2000, Moe et al. Reference MOE, RUTINA, HYTTEBORN and DU TOIT2009, Sankaran et al. Reference SANKARAN, RATNAM and HANAN2004). Despite this importance of seedling recruitment and juvenile growth and survivorship in savanna plant communities, we still know little about patterns and effects of herbivory at the seedling or young sapling stages of trees in these ecosystems.
This study had three objectives. The first was to compare the effects of two different common forms of damage, defoliation by caterpillars and pruning by vertebrate herbivores, on growth and morphology of juvenile Colophospermum mopane (Kirk ex Benth.) Kirk ex J. Léonard, an ecologically and economically important tree species over a large portion of southern Africa. Our second objective was to compare the morphology and growth responses to herbivory of C. mopane at the seedling stage with reported responses of adult trees to similar intensities of pruning and defoliation, to develop an understanding of the integrated selective impact of herbivory across ontogenetic stages of this important African tree species. Unlike adults which are primarily damaged by pruning and bark-stripping by elephants, juveniles are damaged by small- and medium-sized ungulate browsers (Moe et al. Reference MOE, RUTINA, HYTTEBORN and DU TOIT2009). Both adults and seedlings of C. mopane are defoliated by Imbrasia belina caterpillars. Our third objective was to assess the herbivory tolerance of C. mopane seedlings and determine whether impacts of defoliation on seedling growth were proportional to defoliation intensity (percentage of leaf area removed).
We tested four hypotheses: (1) C. mopane seedlings show lower herbivory tolerance than adults. (2) C. mopane seedling performance is reduced by defoliation and impacts of defoliation on seedling performance are proportional to defoliation intensity (amount of leaf area removed). (3) Pruning causes greater shifts in plant architecture than defoliation. (4) Effects of pruning on plant architecture differ significantly between seedlings and adults as browsers typically consume the main stem apical meristem in seedlings but axillary shoot meristems in adults.
METHODS
Study species
Colophospermum mopane (Fabaceae) (common name; mopane) is a semi-deciduous broad-leaved tree species. It is widely distributed in southern Africa from northern Botswana and South Africa northwards to Zambia and Malawi. It is typically found in dry lowland woodland on shallow sands overlying poorly drained, often alkaline soils (Venter & Venter Reference VENTER and VENTER2007). It is a relatively slow-growing tree species (Coates Palgrave Reference COATES PALGRAVE2000, Palmer & Pitman Reference PALMER and PITMAN1973), and is a host plant for both vertebrate and invertebrate herbivores. Both adults and seedlings are heavily defoliated by larvae of the emperor moths Imbrasia belina and Gynanisa maja (Lepidoptera: Saturniidae). Colophospermum mopane is the primary host plant for I. belina which inflicts high levels of damage, frequently causing 100% defoliation of large areas of mopane woodland. Imbrasia belina produces two generations per rainy season in Botswana, the first generation from October/November to December/January, and the second from February/March to April/May. Thus, C. mopane is often defoliated twice during each growing season. Larval growth, moth emergence and variation in rates of defoliation vary with rainfall (Ditlhogo Reference DITLHOGO1996); however, the distribution of defoliation rates across individual trees or across time has not been quantified. Imbrasia belina occurs in semi-arid savannas throughout southern and East Africa, also feeding on other tree species, including Sclerocarya birrea, Terminalia sericea, Brachystegia speciformis, Julbernardia globiflora, Ficus spp., Rhus spp. and Diospyros spp. (Oberprieler Reference OBERPRIELER1995, Pinhey Reference PINHEY1972). Ditlhogo (Reference DITLHOGO1996) showed that, in natural populations only 14% of adult C. mopane trees defoliated by I. belina caterpillars produce seeds, while 84% of non-defoliated trees produced seeds, and that larger trees are better able to withstand the negative impacts of defoliation than smaller ones. Economically and culturally, the association between C. mopane and I. belina represents the most important plant–insect interaction in southern Africa. Imbrasia belina caterpillars form the basis of a multi-million dollar trade in edible insects and play an important role in the livelihoods of rural people throughout the mopane range in southern Africa (Ditlhogo Reference DITLHOGO1996, Moruakgomo Reference MORUAKGOMO, Gashe and Mpuchane1996, Taylor Reference TAYLOR1982).
In addition to defoliation, C. mopane is damaged by several vertebrate herbivores that prune seedlings and/or adults. The elephant (Loxodonta africana Blumenbach) consumes its stems and branches and strips bark from larger trees. Both herbivory and non-herbivory effects of elephant (breakage of stems and branches) typically result in the loss of 50–75% of above-ground biomass (Smallie & O'Connor Reference SMALLIE and O'CONNOR2000). Young seedlings and small saplings are damaged or killed by impala (Aepyceros melampus) and other small antelopes and small- and medium-sized ungulate browsers that consume the branches and/or main stem of small individuals, often resulting in the loss of >50% of above-ground biomass (Moe et al. Reference MOE, RUTINA, HYTTEBORN and DU TOIT2009). Although a few previous studies have examined the effects of pruning and defoliation on adult C. mopane (Ditlhogo Reference DITLHOGO1996, Hrabar et al. Reference HRABAR, HATTAS and DU TOIT2009, Smallie & O'Connor Reference SMALLIE and O'CONNOR2000), effects of herbivory at the seedling stage have not been examined. Thus, C. mopane provides an excellent system to compare plant responses to two different guilds of herbivores feeding on the same host tree species and differential impacts of herbivory at the juvenile versus adult stage.
Experimental methods
Mature C. mopane seed pods were collected from four natural populations in northern Botswana in June 2009. One seed was sown in each of 160 square plastic pots (9 × 9 × 9 cm) that were filled with a mixture of compost, sand and loamy soil (1 : 2 : 4). Pots were maintained in a greenhouse under ambient light conditions and a 28°C/16°C day/night temperature regime. Plants were watered thrice weekly until germination and once weekly thereafter. Approximately 90% of the seeds germinated within 1 wk and 100% germinated within 2 wk. Resources from cotyledons were depleted and cotyledons completely senesced within 8–12 d after germination. When the seedlings reached a height of 15–20 cm, they were transplanted into larger round pots (18 cm diameter × 20 cm deep) containing the same growing medium. Pot positions on the greenhouse bench were re-randomized every 2 wk.
Growth rates of the seedlings were measured by counting leaves for a 4-mo period prior to implementation of the simulated herbivory treatments. After establishment, when seedlings were approximately 40 cm tall and had accumulated approximately 10 leaves, a small coloured wire ring was attached to the stem above the attachment point of the youngest fully expanded leaf. At 2-wk intervals, the number of leaves above the ring and the remaining live leaves below the ring were counted to quantify leaf emergence and mortality. After measurement of pre-treatment leaf population dynamics over a 4-mo period, plant height (soil surface to the apical meristem) and basal stem diameter were measured at 2-wk intervals for an additional 2 mo. The growth rates of C. mopane seedlings were also quantified by increases in total estimated shoot volume over time (shoot volume = (½πd 2 × ht)/2; where d = stem basal diameter and ht = shoot height).
Following the 6-mo period of pre-treatment growth measurements, the C. mopane were divided equally and randomly into four treatments including: (1) 100% defoliation: All leaves on the plant were removed. Colophospermum mopane has bifoliate compound leaves with a 20–40-mm petiole. To best simulate natural defoliation by I. belina caterpillars, for each leaf the lamina of each leaflet was removed, leaving the petiole attached. (2) 50% defoliation: every other leaf on the plant was removed as described above. (3) Pruned: 50% of the total length of the main stem and 50% of the length of each lateral branch (both leaf and stem tissue) were removed. (4) Control: no defoliation or stem damage. For each defoliated or pruned plant, all of the removed biomass was oven-dried at 60°C to a constant mass and weighed.
Our experiment was designed to simulate field conditions as closely as possible. The herbivory treatments closely matched patterns of damage observed in the field and used in other studies (Hrabar et al. Reference HRABAR, HATTAS and DU TOIT2009, Rooke et al. Reference ROOKE, BERGSROM, SKARPE and DANELL2004, Smallie & O'Connor Reference SMALLIE and O'CONNOR2000), and the plants were maintained in sandy soil, were not fertilized, and were watered once weekly. Plants were spaced 20 cm apart and experienced some competition for light, but because they were grown individually in pots, they did not experience below-ground competition.
After implementation of the treatments, the growth responses of all C. mopane plants were again measured over a 6-mo period. Measurements of leaf demography, plant height and basal diameter were continued at 2-wk intervals for 3 mo and then at 3-wk intervals for an additional 3 mo. Due to pre-treatment and post-treatment mortality of 11 plants, the final sample sizes for each treatment included 38 plants subjected to 50% defoliation, 38 plants subjected to 100% defoliation, 36 pruned plants and 37 undamaged control plants.
At the end of the experiment (12 mo after sowing and 6 mo after implementation of the treatments), five leaves from each plant were randomly selected and harvested by cutting the lamina from the petiole. The area of each leaf was measured using a Decagon leaf area meter (Delta T Model RS232C). In addition, for each plant the final number of live leaves, plant height, and basal diameter, number of branches and length of each branch were recorded. Each plant was then harvested by clipping at soil level and leaves were separated from stems and branches. Leaves and stems were oven-dried to a constant weight at 60°C and weighed. At final harvest, mean individual leaf area, total leaf area, plant height, basal diameter, and stem, leaf and total above-ground leaf and stem biomass were measured.
Analysis of variance (ANOVA) was used to examine the effect of treatments on growth and morphological response variables. Final plant mass and total biomass production were examined in a one-way ANOVA with treatment as the factor in a completely randomized design with homogeneous variances. Leaf mass was examined in a one-way ANOVA with treatment as the factor in a completely randomized design with heterogeneous variances (one variance was estimated for each treatment). The number of leaves per plant fitted a gamma distribution and was analysed via separate pre- and post-treatment two-way ANOVAs with defoliation treatment and date in a repeated-measures design with a first-order autoregressive covariance structure for the pre-treatment ANOVA and a first-order autoregressive covariance structure varied by treatment group for the post-treatment ANOVA (Proc GLIMMIX, SAS). Shoot volume was investigated via two separate ANOVAs with defoliation treatment and date in a repeated-measures design (Proc MIXED, SAS). One ANOVA examined shoot volumes before treatment initiation and used a compound symmetry covariance structure, while the second ANOVA examined shoot volumes after treatment using an unstructured covariance structure. Date was the repeated measure in all analyses. Mean leaf size and mean total leaf area per plant were each analysed via one-way ANOVA assuming homogeneous variances among treatment levels (Proc GLM, SAS). Final main-stem height, mean branch length, total final branch length per plant, and leaf/stem ratios were each analysed with one-way ANOVA in a completely randomized design with homogeneous variances (except for main-stem length where a model with heterogeneous variances was used; Proc MIXED, SAS). The number of lateral branches per plant was investigated using a one-way ANOVA with the factor of defoliation treatment in a completely randomized design assuming a negative binomial distribution (Proc GLIMMIX, SAS). In all ANOVAs, date and treatment were considered fixed effects and, when necessary, denominator degrees of freedom were approximated using the Kenward–Roger's method. Pairwise comparisons were conducted on all one-way ANOVAs. Pairwise comparisons and sets of contrasts controlled the Type I error rate using Bonferroni's criteria. Prior to initiation of the treatments, there were no significant differences among treatment groups in any of the response variables measured.
RESULTS
Plant growth responses – shoot biomass and volume
All three simulated herbivory treatments resulted in a significant decrease in above-ground biomass compared with controls (F3,145 = 4.59, P < 0.01, Figure 1). Final mean biomass of control plants was approximately 13 g whereas final biomass of defoliated or pruned plants was approximately 9– 10 g. The proportion of final biomass in stem versus leaf tissue did not differ significantly among the three herbivory treatments (F2,145 = 2.52, P = 0.08); however, leaf mass and leaf/stem ratio were significantly lower in all three treatments compared with the control (F1,456 = 36.7, P < 0.01, Figure 1). When the off-take biomass (biomass removed by treatment implementation) was added to the total final biomass to estimate total production, plants subjected to 100% defoliation showed the highest compensatory growth. Total biomass production of 100% defoliation plants was 30% greater than both undamaged controls and partially defoliated plants, and significantly higher than all other treatments (F3,145 = 3.96, P < 0.01, Figure 1).
Figure 1. A comparison of the effects of pruning and defoliation treatments on cumulative total above-ground dry mass per plant in Colophospermum mopane. Treatments included: plants with 50% of leaves removed (50D); plants with all leaves removed (100D); pruned plants with 50% of the main-stem and lateral branch length removed (50P); undamaged control plants (CON). White bars = final stem mass. Diagonally hatched bars = final leaf mass. Double-hatched bars = biomass removed by implementation of the treatments. Different capital letters above bars indicate significant differences in total final biomass. Different lower-case letters above bars indicate significant differences in biomass production (Bonferroni pairwise comparisons).
Pruning significantly reduced estimated shoot volume compared with controls (F3,147 = 5.50, P < 0.01, Figure 2). Pruning immediately reduced the estimated shoot volume by the intended 50% (from 14.5 cm3 to 7.4 cm3). Pruned plants subsequently grew at a slightly faster rate than controls, such that they were only 27% smaller than controls by the end of the 6-mo period (Figure 2), although this treatment × date interaction was not quite significant (F321,289 = 1.54, P = 0.06, Figure 2). Final shoot volume (day 316) of pruned plants was significantly lower than controls (F1,147 = 10.7, P < 0.01). As expected, the leaf removal treatments caused no immediate change in total estimated stem volume. Plants subjected to 50% or 100% defoliation remained similar to controls in total estimated shoot volume over the entire 6-mo post-treatment period (Figure 2).
Figure 2. A comparison of the effects of pruning and defoliation treatments on mean (± 1 SE) estimated total shoot volume of Colophospermum mopane plants throughout the course of the study. Treatments included: plants with 50% of leaves removed (50D); plants with all leaves removed (100D); pruned plants with 50% of the main-stem and lateral branch length removed (50P); undamaged control plants (CON). Treatments were imposed on day 211.
Plant growth responses – leaf demography
During the pre-treatment measurement period, C. mopane seedlings showed a gradual and linear increase in number of leaves at a rate of approximately one new leaf every 15 d (Figure 3) and accumulated an average of 30 leaves by the end of the 6-mo period. Leaf death rates were minimal during early seedling growth and there were no pre-treatment differences in number of leaves per plant among treatment groups (F3,161 = 1.25, P = 0.29, Figure 3).
Figure 3. A comparison of the effects of pruning and defoliation treatments on mean (±1 SE) total number of live leaves per Colophospermum mopane plant over the course of the experiment (24 February 2010 – 5 January 2011). Treatments included: plants with 50% of leaves removed (50D); plants with all leaves removed (100D); pruned plants with 50% of the main-stem and lateral branch length removed (50P); undamaged control plants (CON). Treatments were initiated on 22 September 2010.
After defoliation or pruning, leaf population dynamics changed rapidly. Total leaves per plant varied significantly by treatment (F3,92 = 11.5, P < 0.01) and date (F8,545 = 1051.3, P < 0.01) and there was a significant treatment by date interaction (F24, 928 = 463.1, P < 0.001). After 50% leaf removal or pruning, leaf numbers increased from 15 to approximately 19 within a 2-wk period, and thereafter leaf population size remained very constant over the 4-mo post-treatment period (Figure 3). By contrast, plants subjected to 100% defoliation showed an extraordinarily rapid increase in leaf number over the first few weeks after defoliation, greatly surpassing control plants and surpassing pre-treatment leaf numbers within 3 wk (Figure 3). In completely defoliated plants, mean leaf numbers per plant increased from zero to 35 within 2 wk (rate of 2.5 leaves d−1). By the end of the measurement period (5 January 2011), plants that experienced 100% defoliation had almost double the number of leaves as that of the control and other treatment groups (F1,207 = 67.5, P < 0.001, Figure 3).
After showing consistent leaf population growth for the 6-mo pre-treatment period, control plants showed a gradual decline in total leaf number (leaf mortality > leaf natality) over the next 4-mo period (Figure 3). Thus, by the end of the experiment, there were no significant differences in leaf population sizes between the control and pruned plants (F1,107 = 3.47, P = 0.065) or between the control and 50% defoliated plants (F1,102 = 2.93, P = 0.09).
Morphological responses – leaf size
There were significant differences in post-treatment mean leaf size among treatments (Figure 4a). New leaves produced after 100% defoliation were 45% smaller in area than leaves present prior to defoliation and were significantly smaller than all other treatments (Figure 4a). In plants that were partially defoliated or pruned, the leaves that were produced following damage were not-significantly different in size compared with leaves of controls and compared with the residual leaves remaining on the defoliated plants. There was no significant difference in leaf size between plants subjected to 50% defoliation or pruning.
Figure 4. A comparison of the effects of pruning and defoliation treatments on mean (± 1 SE) leaf size (a) and mean total leaf area per plant (b) in Colophospermum mopane. Treatments included: plants with 50% of leaves removed (50D); plants with all leaves removed (100D); pruned plants with 50% of the main-stem and lateral branch length removed (50P); undamaged control plants (CON). F-statistics and P-values indicate whether each metric varied significantly by treatment. Different letters above bars indicate significant differences between means (Bonferroni pairwise comparisons).
The changes in leaf size and population dynamics following defoliation or pruning together resulted in significant effects on mean total leaf area per plant (Figure 4b). Pruned and partially defoliated plants both showed a significant reduction in total plant leaf area relative to the control 6 mo after treatment. However, plants subjected to 100% defoliation had significantly greater total leaf area than partially defoliated plants, and they showed no reduction in final total leaf area relative to controls (Figure 4b).
Morphological responses – branching and plant architecture
At final harvest, pruned plants showed significantly reduced height and lateral branch length, but no difference in the number of lateral branches, relative to control plants (Figure 5a–c). The net result was that the total summed length of the primary stem and lateral branches was significantly lower than that of the control (Figure 5d). By contrast, plants subjected to 50% defoliation showed no significant difference in mean lateral branch length, main stem height, or number of lateral branches relative to control plants (Figure 5a–c). The total summed length of primary stem and lateral branches was not significantly lower in partially defoliated plants compared with the control (Figure 5d).
Figure 5. A comparison of the effects of pruning and defoliation treatments on architectural traits of Colophospermum mopane. Treatments included: plants with 50% of leaves removed (50D); plants with all leaves removed (100D); pruned plants with 50% of the main-stem and lateral branch length removed (50P); undamaged control plants (CON). Traits measured included mean (± 1 SE) final main-stem length (a), number of lateral branches per plant (b), lateral branch length (c), and summed length of main stem plus all lateral branches (d). Diagonally hatched portion of bars indicates stem removed (off-take) by implementation of the pruning treatment. F-statistics and P-values indicate whether each metric varied significantly by treatment. Different letters above bars indicate significant differences between means (Bonferroni pairwise comparisons).
Plants subjected to 100% defoliation responded very differently from all other treatments. Complete defoliation resulted in a doubling of the number of lateral branches relative to control plants (Figure 5b), and a large and statistically significant decrease in lateral branch length (Figure 5c). These opposite responses in branch number and individual branch length resulted in a change in plant architecture but no difference in total stem and branch length between completely defoliated plants and controls (Figure 5d).
DISCUSSION
Both defoliation and pruning altered plant growth and architecture in juvenile C. mopane. Responses to pruning included reductions in total leaf area, plant height and final above-ground biomass, and architectural changes including shorter lateral branch length and a reduction in total main stem and branch length. The reductions in height and biomass were not proportionately as large as the amount of biomass removed (50%), indicating that seedlings show some compensatory re-growth capacity. However, comparison of our results with other studies on C. mopane indicates that seedlings are not as herbivory-tolerant as adult trees. Hrabar et al. Reference HRABAR, HATTAS and DU TOIT(2009) subjected C. mopane trees to a 50% pruning treatment that was very comparable to our treatment imposed on seedlings. They also subjected established C. mopane trees to a simulated defoliation (≥90% of leaves removed) that was very comparable to our treatment imposed on seedlings (100% of leaves removed). In a similar study, Rooke et al. Reference ROOKE, BERGSROM, SKARPE and DANELL(2004) subjected 3–5-y-old C. mopane to 50% stem/branch removal. Comparisons of these studies with our results indicate that pruning or defoliation has much greater impact on C. mopane at the seedling stage. For example, seedlings showed a 50% reduction in total branch length in response to pruning in our study, whereas saplings and adults showed a large increase in branch length (Hrabar et al. Reference HRABAR, HATTAS and DU TOIT2009, Rooke et al. Reference ROOKE, BERGSROM, SKARPE and DANELL2004). Similarly, pruning resulted in an increase in leaf size in adults (Hrabar et al. Reference HRABAR, HATTAS and DU TOIT2009, Rooke et al. Reference ROOKE, BERGSROM, SKARPE and DANELL2004) but no change (and perhaps a slight decrease) in leaf size in seedlings. Seedlings in our study showed nearly a 50% reduction in both leaf size and branch length in response to 100% defoliation whereas Hrabar et al. Reference HRABAR, HATTAS and DU TOIT(2009) reported no significant effect on these traits in similarly defoliated adults. Pruning reduced seedling shoot biomass in our study but increased shoot biomass in sapling or adults of C. mopane and other savanna tree species (Rooke et al. Reference ROOKE, BERGSROM, SKARPE and DANELL2004). Thus, our hypothesis that C. mopane seedlings would show lower herbivory tolerance than adults was supported.
Compensatory growth following pruning is a general pattern observed in African savanna trees (Rooke et al. Reference ROOKE, BERGSROM, SKARPE and DANELL2004), but our results suggest that seedlings are significantly less tolerant of damage than adults, possibly due to their lower resource acquisition capacity, lower stored reserves to support re-growth, and different resource allocation priorities for optimizing their resource-foraging balance compared with adults (Boege & Marquis Reference BOEGE and MARQUIS2005). The demographic implication of these results is that the impacts of herbivory may be particularly significant at juvenile stages, and seedling herbivory or predation may be an important constraint on C. mopane recruitment. The population persistence and the conservation of savanna and woodland trees are not only dependent on the survival of adults, but also critically dependent on successful recruitment through seedling establishment and growth (Belsky Reference BELSKY1984, Dublin et al. Reference DUBLIN, SINCLAIR and MCGLADE1990, Moe et al. Reference MOE, RUTINA, HYTTEBORN and DU TOIT2009, Ruess & Halter Reference RUESS and HALTER1990, Staver et al. Reference STAVER, BOND, STOCK, RENSBURG and WALDRAM2009).
Responses to defoliation varied with the proportion of leaf tissue removed but included both positive and negative effects. Partial defoliation resulted in reductions in leaf number, total leaf area and plant biomass, but no effect on plant height, number or length of lateral branches, or total biomass production. Surprisingly, complete defoliation did not result in responses of larger magnitude, as would be expected if plant responses were proportional to the amount of tissue removed. Rather, 100% defoliation resulted in some responses that were either not observed or were opposite of the response shown in partially defoliated plants. For example, 50% defoliation resulted in reductions in leaf number and total leaf area; but100% defoliation resulted in a large and rapid increase in leaf number and no effect on total leaf area (due to an off-setting reduction in leaf size). The number of lateral branches and biomass production were unaffected by partial defoliation but greatly increased by 100% defoliation. In addition to the surprisingly large positive responses and lack of large negative responses to 100% defoliation, the recovery of 100% defoliated plants was very rapid, particularly with respect to leaf population size. No such large and rapid recovery of leaf numbers and leaf area was observed in the partially defoliated plants. Thus our hypothesis that C. mopane seedlings show negative responses to defoliation that are proportional to defoliation intensity was not supported. Colophospermum mopane and I. belina have likely had a very long co-evolutionary history, and 100% or near 100% defoliation of the host plant is frequent in natural populations. Our results suggest that C. mopane has become adapted to these high levels of leaf loss (complete defoliation once or twice per growing season) such that severe defoliation triggers a large and rapid compensatory response. Why similar responses to low levels of defoliation have not evolved in C. mopane seedlings is not known. Mechanistically, 50% defoliation may be below a threshold that triggers an abiotic environmental cue or internal physiological change that elicits a strong compensatory response.
In almost all response variables measured, completely defoliated seedlings showed greater performance than partially defoliated seedlings. Only individual leaf size was smaller in fully defoliated seedlings compared with partially defoliated seedlings. Hrabar et al. Reference HRABAR, HATTAS and DU TOIT(2009) reported that adults of C. mopane also show reduced leaf size after defoliation by I. belina, and other studies have reported reduced size of re-growth leaves produced after defoliation in woody plants (Gadd et al. Reference GADD, YOUNG and PALMER2001, Piene et al. Reference PIENE, MACLEAN and LANDRY2003, Rooke & Bergstrom Reference ROOKE and BERGSTROM2007). The significance of reduced leaf size as a response to herbivory has not been well studied. Severe defoliation may create light and temperature conditions in the canopy that favour the production of smaller leaves after a defoliation event. The production of a large number of small leaves may allow more rapid recovery of canopy leaf area than growing out leaves to a larger size. Also, more rapid maturity of smaller leaves may allow earlier establishment of physical and chemical defensive traits that increase resistance to subsequent herbivory. Effects of herbivory on leaf size are complex because, in contrast to defoliation, pruning of stems and branches results in an increase rather than a decrease in leaf size in C. mopane (Hrabar et al. Reference HRABAR, HATTAS and DU TOIT2009, Rooke et al. Reference ROOKE, BERGSROM, SKARPE and DANELL2004).
Some of the changes in plant architecture in C. mopane seedlings in response to herbivory also were unexpected. In particular, pruning caused only a small increase in branching whereas defoliation resulted in a very large stimulation of lateral branching. Defoliated plants also showed greater length of lateral branches than did browsed plants. Seedlings of C. mopane have a prominent main stem with little if any lateral branching. When browsed by vertebrates, the main stem apical meristem is removed, which typically releases apical dominance and stimulates the outgrowth of branches from lateral buds. Consumption of only the leaf lamina has no such effect on shoot meristems. Thus we predicted that pruning would elicit changes in plant architecture and defoliation would not, but the opposite occurred. Thus our hypothesis that pruning would stimulate greater changes in seedling architecture than defoliation was not supported.
Overall, our results indicate that, at the seedling stage, pruning has more negative effects on C. mopane than does defoliation. When intensity of herbivory was equivalent (50% removal), pruned seedlings showed lower growth performance as measured by final biomass, final shoot size (volume), and shoot height. Greater tolerance to defoliation than to pruning may be a result of the long history of selection imposed by I. belina, which typically imposes severe damage (e.g. up to two complete defoliations within a single growing season) compared with the lesser and perhaps more variable damage imposed by vertebrate herbivores. However, the patterns and relative intensities of vertebrate and invertebrate herbivory in C. mopane woodlands are changing and may not be reflective of historical selection pressures. Also, C. mopane growth responses alone may not be indicative of effects on plant fitness, and Ditlhogo (Reference DITLHOGO1996) showed that defoliation by I. belina causes large effects on flowering and seed production in C. mopane.
Our results suggest that both defoliating insects and browsing vertebrates, and particularly the latter, may impose a significant constraint on seedling recruitment and establishment in C. mopane that may influence its population persistence and abundance. Severe defoliation triggered a large and rapid compensatory response, indicating that C. mopane has become adapted to the high rates of defoliation inflicted by I. belina caterpillars. Future studies of the role of herbivores in the population dynamics of savanna trees should focus on the impact of herbivory at early ontogenetic stages as well as effects on adult trees.
ACKNOWLEDGEMENTS
The Department of Biological Sciences, University of Botswana provided technical support and Kansas State University provided travel support to D. Hartnett and J. Ott. We thank two anonymous reviewers who provided very helpful comments on the manuscript.
