Using a process-based model to analyse compensatory growth in response to defoliation: Simulating herbivory by a biological control agent

Biological Control 43 (2007) 119–129 www.elsevier.com/locate/ybcon

Using a process-based model to analyse compensatory growth in response to defoliation: Simulating herbivory by a biological control agent M.S. Watt

a,*

, D. Whitehead b, D.J. Kriticos c, S.F. Gous c, B. Richardson b

a

a Ensis, P.O. Box 29237, Christchurch, New Zealand Landcare Research, P.O. Box 40, Lincoln 7640, New Zealand c Ensis, Private Bag 3020, Rotorua, New Zealand

Received 12 April 2007; accepted 29 June 2007 Available online 10 July 2007

Abstract The weevil Cleopus japonicus Wingelm} uller has been identified as a biological control agent for the highly invasive weed species Buddleia davidii. To study the potential effect of C. japonicus on the growth of this prolific plant, browsing was simulated on field grown plants over the course of a year, using four artificial defoliation levels (0, 33%, 66% and 100%). A simple process-based model was fitted to measurements to identify compensatory mechanisms induced by defoliation and to quantify their influence on above-ground plant biomass (Wp) and the ratio of leaf to total biomass (Wl/Wp). The method outlined in this paper provides a framework for quantifying the net growth impact of feeding by folivorous biological control agents on weeds. This method also provides a means of understanding critical levels of defoliation needed to achieve target levels of weed suppression. Results showed relative values of Wp for treatments in which 33%, 66% and 100% of leaf area had been removed, were 0.61, 0.44 and 0.08, respectively, compared to the undefoliated control. Defoliation intensity was positively related to light use efficiency (e), daily allocation of biomass to leaves (c) and specific leaf area, and negatively related to rates of natural leaf loss. Model results show that defoliation induced increases in e and c to be the most effective means of compensating for removed leaf area in the two treatments with highest levels of defoliation.  2007 Elsevier Inc. All rights reserved. Keywords: Allocation; Buddleja davidii; Cleopus japonicus; Failure time analysis; Leaf longevity; Leaf loss; Light use efficiency; Specific leaf area

1. Introduction Biological control provides a very useful means of managing many invasive weeds, as it is target specific, and usually requires minimal input after agents are safety tested and established in the field. Often, it is the only viable long-term solution for weed management where widespread use of herbicides is too costly or is unacceptable to the public (McFadyen, 1998), or if mechanical control is not feasible. Although numerous examples of complete or partial control of weed species by biological control *

Corresponding author. Fax: +64 3 364 2812. E-mail address: [email protected] (M.S. Watt).

1049-9644/$ - see front matter  2007 Elsevier Inc. All rights reserved. doi:10.1016/j.biocontrol.2007.06.011

agents have been reported, there are also many instances where control of the target weed has been negligible (McEvoy et al., 1991; Ooi, 1992; Hoffmann, 1995; McFadyen, 1998; Julien and Griffiths, 1999). The prime challenge for biological control practitioners, after ensuring agent safety, is to select agents that have a high probability of establishing, and if established, will have a significant negative impact on the target weed. The success of this endeavour depends partly on how well the effects of the agent on the growth and survival of the target weed species can be predicted in the country of release (McFadyen, 1998). A broad understanding of how attack by a biological control agent influences a weed’s growth and life-history traits is helpful for prioritis-

120

M.S. Watt et al. / Biological Control 43 (2007) 119–129

ing guilds of insects or pathogens for inclusion in biological control programmes and quantifying the level of control that can be expected from individual agents (Kriticos, 2003; Kriticos et al., 2003). Knowledge of the per capita impacts of putative agents, and relative ranges of their natural rate of increase, can provide practitioners with an indication of the likely relative impacts that agents with different modes of attack might have on the target plant (Raghu and Dhileepan, 2005). When this information is incorporated into a population dynamics model of the weed it can indicate the sort of impacts that the agent might have on those aspects of the plant that make it weedy, e.g., a reduction in the canopy cover proportion occupied by the plant (Kriticos et al., 1999). For folivorous biological control agents, accurate determination of their influence on plant growth, and how these interactions change across environmental gradients, requires an understanding of the mechanisms by which leaf area reductions influence growth processes. In many species, reductions in biomass are proportionately lower than reductions in leaf area (Langstrom and Hellqvist, 1991; Hoogesteger and Karlsson, 1992; Lavigne et al., 2001), as plants can respond to defoliation through compensatory growth (McNaughton, 1983; Strauss and Agrawal, 1999; Trumble et al., 1993). Compensatory responses which have been observed include increased biomass allocation to leaves (Pinkard and Beadle, 1998) and increases in photosynthetic activity (Heichel and Turner, 1983; von Caemmerer and Farquhar, 1984; Trumble et al., 1993). Leaf tissue removal has also been shown to either increase (Mabry and Wayne, 1997) or reduce (Dirzo, 1984; Mabry and Wayne, 1997) longevity of remaining leaves. Process-based models of plant growth provide a useful framework for determining how leaf removal by defoliating agents influences processes regulating growth. A relatively simple process-based model that provides robust estimates of growth is the light use efficiency model. This model, first identified by Monteith (1977), utilises the linear relationship between cumulatively intercepted radiation and dry matter production. The slope (light use efficiency, e) of this line is then reduced by physiologically based multipliers for various limiting factors to determine dry matter production. While the light use efficiency model has been widely applied to determine growth of crops (Gallager and Briscoe, 1978; Legg et al., 1979), trees (Cannell et al., 1988; Raison and Myers, 1992) and forests (Saldarriaga and Luxmoore, 1991; Linder, 1985), to our knowledge it has not been used to investigate how defoliation influences processes regulating growth. Buddleia davidii is a perennial, semi-deciduous, wind and water dispersed shrub that grows to more than four metres in height. In recent years it has been recognised as a plant with high invasive potential that forms large monocultures that can severely modify natural habitats. In Europe, B. davidii invades natural habitats and has been identified as the highest priority for a biological control programme (Sheppard et al., 2006). In Northern Ireland, it displaces

Betula spp. and in the United States, B. davidii has been found to displace native willows and hinder the establishment of Douglas fir seedlings in plantations (Binggeli, 1998). B. davidii also invades natural habitats in parts of Hawaii, Australia (Binggeli, 1998) Canada and New Zealand (Smale, 1990). The removal of B. davidii from invaded areas is difficult. Mechanical removal by cutting results in resprouting, while plant extraction by hand can increase plant density as a result of the disturbance. Given that spraying with herbicides is generally inefficient and very expensive, use of a biological control agent is an attractive control option. In New Zealand the Chinese weevil Cleopus japonicus Wingelm} uller has recently been approved for release as a biological control agent for B. davidii. This study uses artificial defoliation to simulate the effects of C. japonicus on B. davidii as pre-release evaluation of the efficacy of C. japonicus was not possible in the restricted quarantine facilities available. Though artificial defoliation does not account for the interactions between the insect and the plant (Baldwin, 1990), this method has been previously shown to offer useful insights into the response of plants to herbivory (see Raghu and Dhileepan, 2005). Data used in this study were taken from a field experiment in which young open grown B. davidii plants were subject to a wide range of artificial defoliation intensities. By fitting a process-based model to measurements from this field experiment the objectives were to (i) identify compensatory mechanisms induced by defoliation and quantify their influence on biomass growth and allocation to leaves, (ii) examine how defoliation influences leaf longevity and leaf loss. The fitted model was then used to simulate how defoliation influences above-ground biomass growth on sites with contrasting temperature, representative of the temperature range over which B. davidii occurs in New Zealand. Using results from this study the utility of this model for assessing the likely impacts of candidate biological control agents is discussed. 2. Materials and methods 2.1. Modelling approach Above-ground biomass growth was modelled using the light use efficiency model described in detail in the Appendix. Briefly, this model determines on a daily basis the sum of utilisable intercepted radiation from canopy characteristics (leaf area index, crown diameter), radiation and temperature. Above-ground biomass growth is then determined as the product of utilisable radiation and light use efficiency, and a fraction is allocated to the leaves. Both estimated leaf and biomass growth are then added to the value for the previous day to obtain cumulative total values. Estimates of plant leaf area are then determined as the product of specific leaf area and cumulative leaf mass, from which estimates of radiation interceptance, and biomass growth are then made over the next time step.

M.S. Watt et al. / Biological Control 43 (2007) 119–129 Table 1 List of symbols and units Symbol

Description

Unit

Wl Wp Wl/Wp Af Lc Qi Qu b a e c g S

Leaf mass Plant mass Ratio of leaf to plant mass Plant leaf area Leaf area per unit crown area Intercepted radiation Utilisable radiation Allometric constant Scaling exponent Light use efficiency Daily allocation of biomass to leaves Fractional leaf loss Specific leaf area

g g g g1 m2 m2 m2 MJ plant1 day1 MJ plant1 day1 Dimensionless Dimensionless g MJ1 Dimensionless Dimensionless m2 kg1

Daily meteorological data required for the model includes total photosynthetically active radiation and mean temperature. The model also requires parameter values for the allometric constant (b), scaling exponent (a), fractional daily leaf loss (g), light use efficiency (e), optimum (Topt), minimum (Tmin), and maximum (Tmax) temperature for growth and functions describing changes in specific leaf area and crown diameter. An explanation of all symbols used throughout the paper is given in Table 1. 2.2. Site description and experimental design The experimental site was located adjacent to the Ensis nursery at Rotorua, New Zealand (lat. 38.2S, long. 176.3E, elevation 285 m above sea level), where a mean annual rainfall of 1 491 mm is evenly distributed throughout the year, mean annual temperature is 12.7 C, and the annual average raised pan evaporation is 1186 mm (Anonymous, 1983). The deep moderately fertile pumice soil (yellow brown Ngakuru loam), is well drained and has a high water holding capacity. B. davidii does not significantly respond to irrigation on this site (Richardson et al., 1996), indicating that water availability at the site does not limit growth of this species. Previous characterisation of the nutrient status of this site indicated that the soil was moderately acidic and, with the possible exception of magnesium, there were no significant nutrient limitations (Richardson et al., 1996). In mid-winter 2004 small B. davidii seedlings were transplanted into single row plots (3 m · 3 m) laid out in a randomised complete block design, with ten blocks, and a two row perimeter buffer. This spacing ensured that plants were not subject to competition from adjacent plants for light, water or other resources. The 40 plants within the experiment were randomly allocated to 10 blocks which included the following four treatments (i) undefoliated control (ii) removal of 33% leaf area (iii) removal of 66% leaf area and (iv) removal of 100% leaf area. For the defoliation treatments entire leaves were removed manually to simulate the effect of insect defoliation on a monthly basis from late-spring to late-summer, initially (November) on all

121

leaves present, and thereafter (December to February) on newly emerged leaves following the previous defoliation. For convenience treatments will hereafter be referred to as D0 (control), D33 (33% leaf area removal), D66 (66% leaf area removal) and D100 (100% leaf removal). 2.3. Measurements 2.3.1. Climate, plant dimensions and biomass Buddleia davidii basal diameter (d), height (h) and crown diameter (Cd) were measured at monthly intervals over the period of 1 year starting from 1 June 2004. Measurements of air temperature and photosynthetically active radiation were taken from a meteorological station at Rotorua airport, located 7.5 km from the study site. Destructive sampling was undertaken to determine values of leaf area (Af), leaf mass (Wl) and above-ground plant biomass (Wp) immediately prior to defoliation (November 2004) using six of the plants in the buffer rows. Additional destructive samples were undertaken at the end of the experiment (July 2005), on all 40 monitored plants (10 per treatment) within the randomised complete block. In all destructively sampled plants, basal diameter, height and crown diameter were measured before each plant was cut at ground level and divided into stem, branches and leaves. One-sided leaf area was measured on subsamples taken from each plant, immediately after harvesting, using a leaf area meter (model LI 3100, Li-Cor Inc, Lincoln, NE, USA). All components, including the subsamples were dried at 70 C until constant mass was reached then weighed. Specific leaf area was determined as the quotient of fresh leaf area and dry mass on the subsample. Total leaf area for each plant was determined as the product of total dry leaf mass and specific leaf area. Using prediction equations developed from the initial destructive sample, Af (m2) was determined prior to the first defoliation from measurements of d (mm) using Af = 0.1308 + 0.0198d (P < 0.001; r2 = 0.97) while Wp (g) was determined from h (m) using Wp = 144.1h2.179 (P < 0.001; r2 = 0.92). Values for Af, Wl and Wp at the experiment end were directly obtained from measurements as all monitored plants in the experiment were destructively harvested. 2.3.2. Leaf lifespan and leaf loss Phenological measurements were taken on all 40 plants (n = 10 per treatment) within the randomised complete block, to determine rates of leaf loss and leaf longevity. Measurements were taken from a randomly selected branch every month from early spring (1 October 2004) to mid-winter the following year (11 July 2005). On this branch, nodes were separately numbered and the two leaves which occur at each node were classified as either present, absent through experimental defoliation or absent through natural causes. If leaves were present, leaf area at each node, Al, (mm2 per leaf) was determined from measurements of leaf length, Ll, (mm per leaf) and width, Lw,

122

M.S. Watt et al. / Biological Control 43 (2007) 119–129

(mm per leaf) using the following equation developed from a subsample (200 leaves) taken across all treatments; Al = 0.6112Ll Lw, P < 0.001, r2 = 0.99. Leaf area was then summed across all present leaves on each shoot to obtain total leaf area. 2.4. Data analysis of measurements All analyses were undertaken using SAS (SAS Institute, 1996). The influence of defoliation on plant dimensions (d, h, Cd) over time was analysed using a mixed effects model, with time specified as a repeated variable (Littell et al., 2006). The influence of defoliation on specific leaf area and the ratio of leaf to total above-ground biomass (Wl/Wp) at the end of the experiment were tested using a mixed effects model which included block as a random effect. All multiple comparisons were undertaken using Tukey’s test. Interval censored failure time analysis was used to examine patterns of leaf longevity determined from phenological measurements. Definitions of minimum and maximum leaf lifespans used follow Dungan et al. (2003). Failure time analysis allows estimation of probability functions describing age-specific leaf mortality risk. The probability density function, f(t), describes the probability that a leaf will die in the interval t to t + Dt. The survival function, S(t), describes the probability that a leaf will live longer than t without dying. We tested the fit of six commonly used parametric distributions (normal, logistic, exponential, log-normal, Weibull and generalised Gamma). Of these distributions the extreme value Weibull distribution was selected as this had the maximum log likelihood, and a plot of the survival function against time showed that the model fitted the data well. Using this distribution the probability density function is described by, ht  l t  li f ðtÞ ¼ r1 e e ð1Þ r r where t is time after leaf initiation (leaf age) and l and r are location and scale parameters, respectively. To investigate variation in leaf lifespan between treatments we used failure time analysis to determine the value of the location parameter, l, (see Eq. (1)) for each branch. A general linear model was then used to test if the defoliation intensity significantly influenced the value of the location parameter. 2.5. Modelling of biomass growth Biomass growth was modelled at a daily time resolution, using treatment level plant inputs averaged over all 10 replicates. A CLIMEX model of B. davidii (Kriticos and Potter (2006)) was used to select appropriate parameter values for Tmin, Topt and Tmax (6, 23, and 28 C, respectively). Changes in specific leaf area and crown diameter were modelled over the course of the growing season using

a spline function fitted to measured data in SAS. Leaf loss was determined as a fraction, g, of net leaf mass. Monthly phenological measurements of leaf loss and net leaf area were converted to a mass basis, to determine g, by dividing area values by specific leaf area. Values of g were then determined as the quotient of daily leaf loss over the measurement period and net leaf mass at the end of the period. Seasonal changes in g were modelled using a spline function fitted to measured data. The model was fitted to each treatment using non-linear least squares regression. Parameter values for e, b, and a were determined by minimising the sums of squares between predicted and actual values. The model was initially fitted to measurements for the ratio of leaf mass to above-ground plant biomass (Wl/Wp), and then to above-ground plant biomass (Wp). As values for the allometric constant b represents daily allocation of biomass to leaves at very low values of biomass (1 g in this case) this was constrained within the model so it could not exceed a value of 0.9. The relative importance of defoliation induced changes in daily allocation of biomass to leaves, light use efficiency, specific leaf area and rates of leaf loss on both Wp and Wl/Wp was quantified. This was accomplished by substituting D0 values for each trait first individually and then simultaneously into the fitted model for each defoliated treatment. The magnitude of the alteration to cumulative biomass and average Wl/Wp in each treatment was then determined. The parameterised model was used to simulate how defoliation intensity influenced above-ground biomass growth across a temperature range. Simulations were undertaken using a base mean annual temperature of 12.6 C, and mean annual temperatures four degrees lower (8.6 C), and four degrees higher (16.6 C). This variation approximates the mean annual temperature range over which B. davidii is found in New Zealand (Kriticos and Potter, 2006). 3. Results 3.1. Measurements 3.1.1. Climate and plant dimensions During the measurement year, daily mean temperature was 12.2 C, ranging from 3.0 C to 22.8 C. Temperatures lower than the value used for Tmin (6 C) were recorded on 35 days. Incident photosynthetically active radiation totalled 2.5 GJ m2 year1 reaching a maximum daily value of 16.2 MJ m2 on December 27. Basal diameter, height and crown diameter of B. davidii were all significantly (P < 0.001) influenced by defoliation and age. As all dimensions exhibited divergence between treatments over time (Fig. 1) the interaction between age and defoliation was also highly significant for basal diameter, height and crown diameter (P < 0.001). Growth reductions in all characteristics were positively related to defoliation intensity. At the end of the experiment D100 was most affected by defoliation exhibiting significantly

M.S. Watt et al. / Biological Control 43 (2007) 119–129

123

ments in July, these values were not significantly different from the control at the 5% level (Table 2). Specific leaf area prior to defoliation was 13.3 m2 kg1 (Table 1). At the end of the experiment values for specific leaf area were 9.5 m2 kg1 in D0. Ending values for defoliated treatments were higher than D0 and exhibited a significant positive relationship with defoliation intensity (P < 0.001). Multiple comparison tests reveal this significant positive relationship was mainly attributable to significant defoliation induced gains in specific leaf area for D100, which exceeded values in D0 by 79% (Table 2). Although plant biomass was a marginally significant determinant of specific leaf area (P = 0.02; r = 0.35), when included as a covariate in the model with defoliation, defoliation was highly significant (P = 0.0036), but biomass was insignificant (P = 0.256). 3.1.3. Leaf loss and leaf longevity Daily leaf loss fluctuated in all treatments over spring, summer and autumn, before increasing considerably over winter (Fig. 2). When averaged across all months, daily rates of leaf loss were negatively related to defoliation intensity, with values averaging 0.48, 0.47, 0.40 and 0.36% day1 of net leaf mass, respectively, in D0, D33, D66 and D100. Failure time analysis showed that defoliation intensity did not influence leaf longevity as values of l did not significantly (P > 0.05) differ between treatments. The probability density function, for data combined from all treatments, exhibited an asymmetrical right skewed distribution with the highest rate of leaf mortality occurring 142 days after emergence (Fig. 3a). The skewed distribution of leaf longevity was also evident in the survival function. While the first 25 percentile of leaves took 107 days to fall, the last 25 percentile were lost within 33 days (Fig. 3b). 3.2. Modelling of biomass

Fig. 1. Seasonal changes in Buddleia davidii (a) height (b) basal diameter and (c) crown diameter for plants in treatments D0 (thick solid line), D33 (dotted line), D66 (dashed line) and D100 (thin solid line). Each point shown is the mean ± standard error of 10 sample plots. The arrows A to D indicate the times of defoliations.

(P < 0.001) lower values in all characteristics, which ranged from 43% of D0 values for crown diameter to 53% of D0 values for both plant basal diameter and height (Fig. 1). 3.1.2. Ratio of leaf to total biomass and specific leaf area Prior to defoliation the ratio of leaf to total aboveground biomass, Wl/Wp, was 0.64. This ratio declined to a value of 0.28 in D0 by the experiment end (July 2005). Although Wl/Wp was slightly higher in the defoliated treat-

3.2.1. Allocation of biomass to leaves, light interception and light use efficiency The model fitted measurements of Wl/Wp well (see Fig. 4a and Table 2), with measured values of Wl/Wp at the end differing from modelled values no more than 1.5%. In D0 daily biomass allocation to leaves declined over the course of the experiment by 37% (Fig. 4b) from 0.51 to Table 2 Treatment variation in specific leaf area (S) and ratio of leaf mass to total biomass (Wl/Wp) for plants at the end of the experiment (11 July 2005) Treatment

S (m2 kg1)

Wl/Wp (g g1)

D0 D33 D66 D100

9.5 (0.46)a 10.7 (0.51)a 11.0 (0.31)a 17.0 (2.21)b

0.28 0.33 0.33 0.34

(0.01)a (0.03)a (0.02)a (0.04)a

Values shown represent the mean ± standard error for 10 plants per treatment. Treatment values with the same letter following them are not significantly different from each other at the 5% level.

124

M.S. Watt et al. / Biological Control 43 (2007) 119–129

Fig. 2. Seasonal variation in daily leaf loss, expressed as a percentage of net leaf mass, over time for treatments D0 (thick line), D33 (dotted line), D66 (dashed line) and D100 (thin line). The arrows A to D indicate the times of defoliations.

Fig. 4. (a) Modelled quotient of leaf mass (Wl) and above-ground plant biomass (Wp) over time. Also shown is daily biomass fraction allocated to leaves (b) over time and (c) against above-ground plant biomass for treatments D0 (thick line), D33 (dotted line), D66 (dashed line) and D100 (thin line). Also shown in 4a are measured values for D0 (open triangles), D33 (closed triangles) D66, (closed diamonds), and D100 (open diamonds) immediately prior to first defoliation (15 November) and at the experiment end. For figs. a and b the arrows A to D indicate the times of defoliations.

Fig. 3. (a) Probability density function, f(t), showing probability a leaf will die over the interval t to t + Dt, and (b) Survival function, S(t), showing the probability that a leaf will live longer than t without dying. In Figure (a) guide lines are drawn to the maximum rate f(t), from both axes, while for figure (b) guide lines are drawn to the 25% and 75% percentile of S(t) from both axes. The curves shown in both figures are from data pooled across all four treatments as treatment differences were not significant (P > 0.05).

0.37. Defoliation induced a sharp increase in daily biomass allocation to leaves, in all defoliated treatments. Ending values of daily biomass allocation to leaves in D33, D66 and D100 exceeded values in D0 by 35%, 41%, and 65%, respectively (Fig. 4b). Temporal variation in daily biomass allocation to leaves between treatments was partially attributable to treatment

M.S. Watt et al. / Biological Control 43 (2007) 119–129

125

variation in biomass (Fig. 4c), as evidenced by the similarity in values of b and a between treatments (Table 3). When modelled values were compared at the ending biomass for D100, values in D100 exceeded those in D0 by 20% (0. 61 vs. 0.51). Values of total utilisable radiation, Qu, (defined as QifT) in D33, D66 and D100 were, respectively, 68%, 39% and 4% of the value of 774 MJ recorded in D0 (Table 3). Light use efficiency in D0 was 1.45 g MJ1. Compared to this value, light use efficiency was 10% lower in D33 and 14%, and 133% higher in D66 and D100, respectively (Table 3). 3.2.2. Seasonal changes in biomass and leaf area Daily predictions of cumulative biomass using the model showed good correspondence with measured values (Fig. 5a). Total biomass for plants in the D0 treatment, including the leaves lost through natural causes was 1 119 g (Table 3). Cumulative biomass in the defoliated treatments as fractions of the control were 0.61, 0.44 and 0.08 for D33, D66, and D100, respectively. Defoliation strongly influenced the rate of leaf area development. In treatments D0, D33 and D66 plant leaf area increased rapidly from mid-spring (October) to early autumn (March), reaching respective, maximum values of 3.67, 2.54 and 1.77 m2 in late-autumn (May) before declining over winter (Fig. 5b). The highest leaf area for D100 (0.51 m2) was attained during early winter in June (Fig. 5b). Values of Lc averaged 1.72, 1.12, 1.23 and 0.47 m2 m2 in treatments D0, D33, D66, and D100, respectively. 3.2.3. Relative importance of compensation on biomass and Wl/Wp With the exception of biomass allocation to leaves, compensatory traits had little positive effect on biomass growth in D33 (Table 4). For D66, light use efficiency had the strongest influence on biomass growth. Analyses indicate defoliation induced gains in light use efficiency in this treatment increased relative biomass, expressed as a fraction of D0 values, from 0.33 to 0.44. Defoliation induced changes in all other traits, apart from fractional leaf loss, improved biomass growth in D66, and compensation from the combination of traits improved relative biomass growth from Table 3 Treatment variation in the allometric parameters b and a, total utilisable intercepted irradiance (Qu) total biomass growth, including mass of defoliated and naturally lost leaves (Wp) and light use efficiency (e) Treatment

b

a

Qu (MJ)

Wp (g)

e (g MJ1)

D0 D33 D66 D100

0.90 0.90 0.90 0.90

0.871 0.906 0.906 0.911

774 526 (0.68) 299 (0.39) 28 (0.04)

1 119 683 (0.61) 496 (0.44) 95 (0.08)

1.45 1.30 (0.90) 1.66 (1.14) 3.39 (2.33)

All values were determined at the plant level over the period of the simulation from 15 November 2004 (day prior to first defoliation) to 11 July 2005. Values in brackets represent the fractional value for each variable, in the defoliated treatments, relative to the control treatment (D0).

Fig. 5. Modelled (a) above-ground biomass and (b) leaf area for D0 (thick solid line), D33 (dotted line), D66(dashed line) and D100 (thin solid line). For both graphs measured values are shown for D0 (open triangles), D33 (closed triangles) D66, (closed diamonds), and D100 (open diamonds). The arrows A to D indicate the times of defoliations.

Table 4 Effects of substituting D0 values for light use efficiency (e), daily allocation of biomass to leaves (c), fractional leaf loss (g), specific leaf area (S) and all four traits (e, c, g, S) on relative total cumulative biomass and relative average Wl/Wp for the three defoliated treatments, over the period of the simulation (15 November 2004 to 11 July 2005) Change

Defoliated treatment D33

D66

D100

Relative total cumulative biomass No change 0.61 e 0.76 c 0.50 g 0.62 S 0.56 e, c, g, S 0.57

0.44 0.33 0.36 0.46 0.39 0.26

0.08 0.01 0.06 0.06 0.04 0.004

Relative average Wl/Wp No change e c g S e, c, g, S

0.90 0.89 0.75 0.91 0.89 0.77

0.41 0.14 0.33 0.36 0.31 0.08

1.07 1.06 0.92 1.09 1.07 0.93

The original relative value with no change in any trait is shown as reference for each treatment in bold.

126

M.S. Watt et al. / Biological Control 43 (2007) 119–129

Fig. 6. Simulated influence of defoliation intensity on normalised aboveground biomass, expressed as a fraction of control (D0) values, for mean annual temperatures of 8.6 C (open circles), 12.6 C (closed circles) and 16.6 C (closed squares). Linear lines of the form y = 1-cx drawn through each simulated temperature have values for parameter c (rate of decline in normalised biomass with increasing defoliation) of 0.0096 for 8.6 C (dashed line), 0.0092 for 12.6 C (dotted line) and 0.0089 for 16.6 C (solid line).

0.26 to 0.44. Similarly in D100 increased light use efficiency had the strongest influence on relative biomass, increasing values from 0.01 to 0.12. In this treatment, defoliation induced changes to all other traits had positive effects on relative biomass, and when included together all four traits increased relative biomass from 0.004 to 0.12. The ratio Wl/Wp was most sensitive to biomass allocation to leaves in D33, which increased values, relative to D0, from 0.92 to 1.07. Daily biomass allocation to leaves was also the compensatory trait with most influence on this ratio in D66, increasing relative Wl/Wp from 0.75 to 0.90. In contrast, increased light use efficiency had the largest influence on Wl/Wp in D100, increasing fractional values from 0.14 to 0.41. Defoliation induced alterations to all other compensatory traits increased this ratio in D100, and when all four traits were included together values for relative Wl/Wp increased from 0.08 to 0.41. 3.2.4. Model simulations For the temperatures tested normalised above-ground biomass growth declined linearly with increasing defoliation (Fig. 6). The rate of decline in normalised aboveground biomass with increasing defoliation decreased slightly with increasing temperature, ranging from 0.0096 at 8.6 C to 0.0092 at 12.6 C and 0.0089 at 16.6 C (Fig. 6). 4. Discussion Selection of biological control agents is very time consuming and costly (McFadyen, 1998). The model based approach outlined in this paper could provide a rapid cost effective solution for assessing the likely impacts of candidate biological control agents. Once parameterised for a particular weed species from field measurements, the model

can be used to examine how a large number of potential biological control agents, with a wide range of per capita defoliating intensities, influence growth of the target species. Given the sensitivity of net defoliation rates to agent abundance, and uncertainties around the population dynamics of exotic agents prior to their release and establishment in a new range, it is unlikely that a precise prediction of an individual agent’s success could be made using this model. However, this type of model could at least help compare the likely effects of folivores compared with agents from other guilds. Model simulations demonstrate the utility of such a model based approach in comparing the effects of defoliation on growth across a wide temperature range. These simulations suggest that high levels of defoliation will have the most adverse impact on B. davidii growth at sites with low temperatures. This approach could be refined by using seasonal estimates of insect defoliation, for different sites obtained from insect phenology studies. As the model is sensitive to the timing of defoliation, the effects of agents with different phenological patterns, or voltinism, on above-ground biomass growth could be readily determined. An additional advantage to using a process-based model is that the mechanisms by which the weed compensates for leaf loss can be identified and quantified, which increases understanding of how the target weed species responds to defoliation. In this paper the model highlighted the importance of compensation on above-ground biomass in the defoliated treatments D66 and D100 which increased values of above-ground biomass by 69% and 20-fold, respectively. An increase in light use efficiency was found to be the most effective mechanism by which B. davidii compensated for defoliation in these two treatments. Similarly, results show the importance of compensation mechanisms, in maintaining the ratio of Wl/Wp in defoliated plants, particularly for those plants in D100. For this treatment all compensation mechanisms evaluated were important and in combination these increased average relative values of Wl/Wp in D100 by fivefold (0.41 vs. 0.08). Despite the considerable leaf loss through defoliation, Wl/Wp in defoliated treatments exceeded values in D0 by the end of the experiment. Restoration of Wl/Wp to values exceeding those of the control is remarkable, particularly for D100, and demonstrates the ability of B. davidii to recover from complete leaf loss to a balanced allometric state in a relatively short period of time. It is likely that the development of new leaves in the completely defoliated D100 treatment was facilitated by translocation of high levels of stem nitrogen, which are typical of deciduous species (Bryant et al., 1983). Once leaves had grown, the recovery in D100 was largely attributable to defoliation induced alterations to compensatory traits, without which model results show ending values in D100 for Wl/Wp, considerably lower than actual recorded values (0.34 vs. 0.04). Defoliation induced increases in allocation to leaves have also been noted in Eucalyptus nitens (Pinkard and Beadle, 1998). This

M.S. Watt et al. / Biological Control 43 (2007) 119–129

response is thought to occur as defoliation induces a strong sink demand for leaves in source limited plants (Waring et al., 1968; Ryle and Powell, 1975; Whigham, 1990). Failure time analysis provided a useful means for interpreting the negative scaling observed between defoliation intensity and rates of leaf loss after the final defoliation. Two possible explanations for this negative scaling were that defoliation induced greater leaf longevity or reduced the average age of leaves in the canopy. Results from failure time analysis show that leaf longevity was not significantly influenced by treatment. It therefore seems likely that lower rates of leaf loss for defoliated plants were attributable to negative scaling between defoliation intensity and the proportion of older leaves in the overall population. Given that older leaves have an increased risk of dying, treatments with a lower proportion of old cohorts will have an overall lower rate of leaf loss. This is most certainly the case for the completely defoliated treatment, D100. Following the last defoliation in February, leaf age could only reach a maximum of 150 days before the experiment ended. In contrast over the same period leaf age in D0 reached values as high as 221 days. The average leaf age for D33 and D66 fell between these two extremes. Both treatments included older cohorts of undefoliated leaves, which declined as a percentage of the overall population as defoliation intensity increased. Allometric analysis showed the substantial treatment differences in the daily fraction of biomass allocated to leaves, c, at the experiment end were partially attributable to size dependent shifts in allocation patterns. In the most extreme contrast, the 65% gains in D100 over D0 for c observed at the end of the experiment were reduced to 20% when c was compared between D0 and D100 at the ending biomass for D100. This reduction in treatment differences was attributable to the frequently observed decline in allocation to leaves with increasing biomass (Cannell, 1985). These results reinforce the findings of several other researchers (Gebauer et al., 1996; Oso´rio et al., 1998; Watt et al., 2003) that demonstrate that accurate determination of treatment effects on allocation requires separation of size-dependent changes in allocation from functional adjustments made in response to treatment. Low to moderate levels of defoliation had little influence on light use efficiency. We speculate that the significant gains in light use efficiency, over the control, induced by high levels of defoliation may be related to upward regulation of photosynthesis. Previous research shows that photosynthetic rates often scale positively with defoliation intensity (Helms, 1964; Heichel and Turner, 1983), a response that is thought to be attributable to reductions in the source:sink ratio which occur as defoliation levels increase (Neales and Incoll, 1968; Lavigne et al., 2001). Comparison of e values in this study with previous research on fast growing woody plants where intercepted radiation is related to above-ground biomass suggests that undefoliated B. davidii has a high light use efficiency. In a 1-year study using Salix viminalis and Populus trichocarpa

127

respective values for e of 1.58 and 1.50 g MJ1 were found (Cannell et al., 1988). For Pinus radiata values ranging from 1.14 to 1.34 g MJ1 have been reported for plants growing under optimal levels of nutrition and water availability (Grace et al., 1987; Raison and Myers, 1992). The ability of B. davidii to rapidly colonise sites, and initially outcompete other fast growing pioneer species (Richardson et al., 1996; Bellingham et al., 2005) may be partly attributable to its high light use efficiency. Mechanical defoliation experiments have been found to be useful for accurately assessing plant responses to various levels of defoliation (Strauss, 1988; Hja¨lte´n, 2004; Inouye and Tiffin, 2003; Raghu and Dhileepan, 2005; Raghu et al., 2006; Wirf, 2006), though they may not accurately reflect the full range of effects of herbivores (Lehtila¨ and Boalt, 2004). Artificial and real herbivory each have their respective strengths and weaknesses. Artificial herbivory can be a very useful tool for understanding plant response to herbivory (Raghu and Dhileepan, 2005), can be precisely applied and does not involve any biosecurity considerations, though it may not accurately reflect the process of interest. Conversely, real herbivory may be a more direct application of the treatment effect, but it may be difficult to achieve or measure treatment levels or covariates. Ideally, both artificial and real herbivory effects should be measured in order to draw on the strengths of each approach (Lehtila¨ and Boalt, 2004; Wirf, 2006). Whilst the use of mechanical leaf area reductions may not accurately reflect the effects of damage due to any specific herbivore, there is nonetheless a need to make timely informed prioritisations of agents, or at least agent guilds. The method outlined in this paper could be parameterised to help rapidly compare biological control agents for other invasive weed species. In our case, the use of real herbivory damage was not feasible as the putative insect was still being held in containment pending government approval for release into the New Zealand environment. However, there is nothing stopping the growth impact model from being parameterised with data derived from real herbivory if such data can be gathered prior to agent release. In conclusion, the use of a process-based model to explore how defoliation influences processes affecting growth shows considerable potential as a useful means of assessing potential biological control agents. To fully realise this potential however, it will be necessary to couple this form of growth model to a model of the population dynamics of the plant so as to compare the effects of other agent guilds (e.g., seed feeders) on the population dynamics of the plant, and to identify any other compensatory effects such as density-dependence that could reduce the net effect of the biological control programme. Acknowledgments We are grateful for field measurements undertaken by Natalie Watkins and acknowledge the help of Margaret Richardson who provided useful suggestions on the manu-

128

M.S. Watt et al. / Biological Control 43 (2007) 119–129

script. We also thank the two anonymous referees for useful comments on the final draft of the paper. This project was funded by the New Zealand Foundation for Research Science and Technology under Contract No. C04X0202. Appendix A Following Jackson and Palmer (1979, 1981) the daily integral of radiation intercepted by plants with a discontinuous canopy (Qi) is determined as, Qi ¼ Qo sð1  ekLc Þ

ðA1Þ

where Qo is available incident radiation, k is the light extinction coefficient (assumed to be 0.5 for a spherical leaf angle distribution). The term Lc is the leaf area per unit crown projected area, determined as, Lc = Af/s, where Af denotes plant leaf area, and s is the ground area covered by ellipsoid crowns described by, p ðA2Þ s ¼ ab 4 where a and b are the average horizontal crown diameters in the north–south and east–west directions respectively. Following Monteith (1977), increases in above-ground plant biomass, Wp, over the course of a day (Wpt+Dt) in an environment limited by temperature, but not limited by nutrition or water was related to intercepted incident radiation by W ptþDt ¼ eQi fT  W s

ðA3Þ

where e is light use efficiency, fT is a unitless modifier for air temperature and Ws is the mass of shed leaves, assumed to be lost at the end rather than throughout the course of the day. Daily leaf loss was determined as a fraction, g, of the net leaf mass on the previous day, Wlt-Dt, by, W s ¼ gW ltDt

ðA4Þ

Following Sands and Landsberg (2002), the temperature modifier was described as, (   ðT max T opt Þ=ðT opt T min Þ ) T a  T min T max  T a fT ¼ max 0; T opt  T min T max  T opt ðA5Þ

where Ta is the measured daily mean temperature, and Topt, Tmin, and Tmax optimum, minimum, and maximum temperatures for growth. Increases in plant leaf mass, Wl, over the course of the day (Wlt+Dt) was determined as an allometric fraction of above-ground biomass growth, unadjusted for leaf loss, from which daily loss in leaf mass was deducted using W ltþDt ¼ cW ptþDt  W s

ðA6Þ

where c describes leaf mass as a fraction of cumulative aboveground plant biomass modelled using c ¼ ðbW p a Þ=W p

ðA7Þ

where b is the allometric constant (value of y at x = 1) and a is the scaling exponent (Landsberg and Waring, 1997). Both estimated leaf and biomass growth were added to the value for the previous day to obtain cumulative total values. Estimates of plant leaf area were then determined as the product of specific leaf area and cumulative leaf mass. These values were then used in Eq. (A1) to estimate intercepted irradiance, and Eqs. (A2)–(A7) were then used to determine biomass and leaf mass increment over the next timestep. References Anonymous, 1983. Summaries of Climatological Observations to 1980. New Zealand Meteorological Service Misc. Pub. 177, Wellington, New Zealand. Baldwin, I.T., 1990. Herbivory simulations in ecological research. Trends Ecol. Evol. 5, 91–93. Bellingham, P.J., Peltzer, D.A., Walker, L.R., 2005. Contrasting impacts of a native and an invasive exotic shrub on flood-plain succession. J. Veg. Sci. 16, 135–142. Binggeli, P., 1998. An overview of invasive woody plants in the tropics. School of Agricultural and Forest Sciences Publication No. 13, University of Wales, Bangor. Bryant, J.P., Chapin, J.S., Klein, D.R., 1983. Carbon/nutrient balance of boreal plants in relation to vertebrate herbivory. Oikos 40, 357–368. Cannell, M.G.K., 1985. Dry matter partitioning in tree crops. Attributes of trees as crop plants. In: Cannell, M.G.K., Jackson, J.E. (Eds.). Attributes of trees as crop plants. Institute of terrestrial ecology, Nature Environment Research Council, Cambria, Great Britain, pp 160–193. Cannell, M.G.R., Sheppard, L.J., Milne, R., 1988. Light use efficiency and woody biomass production of poplar and willow. Forestry 61, 125–136. Dirzo, R., 1984. Herbivory: a phytocentric overview. In: Dirzo, R, Sarukhan, J. (Eds.). Perspectives on plant population ecology. Sinauer, Sunderland, pp.141–165. Dungan, R.J., Duncan, R.P., Whitehead, D., 2003. Investigating leaf lifespans with interval-censored failure time analysis. New Phyt. 158, 593–600. Gallager, J.N., Briscoe, P.V., 1978. Radiation absorption, growth and yield of cereals. Agricultural Science (Cambridge) 91, 47–60. Gebauer, R.L., Reynolds, J.F., Strain, B.R., 1996. Allometric relations and growth in Pinus taeda: the effect of elevated CO2 and changing N availability. New Phyt. 134, 85–93. Grace, J.C., Jarvis, P.G., Norman, J.M., 1987. Modelling the interception of solar radiant energy in intensively managed stands. N. Z. J. For. Sci. 17, 193–209. Heichel, G.H., Turner, N.C., 1983. CO2 assimilation of primary and regrowth foliage of red maple (Acer rubrum L.) and red oak (Quercus rubra L.): responses to defoliation. Oecologia 57, 14–19. Helms, J.A., 1964. Apparent photosynthesis of Douglas-fir in relation to silvicultural treatment. For. Sci. 10, 432–442. Hja¨lte´n, J., 2004. Simulating herbivory: problems and possibilities. In: Weisser, W.W., Siemann, E. (Eds.), Insects and Ecosystem Function. Springer, Heidelberg, pp. 244–255. Hoffmann, J.H., 1995. Biological control of weeds: the way forward, a South African perspective. Proc. BCPC Symp.: Weeds in a Changing world 64, 77–89. Hoogesteger, J., Karlsson, P.S., 1992. Effect of defoliation on radial stem growth and photosynthesis in mountain birch (Betula pubescens spp. tortuosa). Func. Ecol. 6, 317–323. Inouye, B.D., Tiffin, P., 2003. Measuring tolerance to herbivory with natural or imposed damage: a reply to Lehtila. Evolution 57, 681–682. Jackson, J.E., Palmer, J.W., 1979. A simple model of light transmission and interception by discontinuous canopies. Ann. Bot. 44, 381–383.

M.S. Watt et al. / Biological Control 43 (2007) 119–129 Jackson, J.E., Palmer, J.W., 1981. Light distribution in discontinuous canopies: calculation of leaf areas and canopy volumes above defined ‘irradiance contours’ for use in productivity modeling. Ann. Bot. 47, 561–565. Julien, M.H., Griffiths, M.W., 1999. Biological Control of Weeds. A world catalogue of Agents and their Target Weeds, fourth ed. Commonwealth Agricultural Bureaux International, Wallingford, United Kingdom. Kriticos, D.J., 2003. The roles of ecological models in evaluating weed biological control agents and projects. In: Spafford-Jacob, H.S., Briese, D.T. (Eds.). Improving the selection, testing and evaluation of weed biological control agents. Proceedings of the CRC for Australian Weed Management Biological Control of Weeds Symposium and Workshop. CRC for Australian Weed Management, Adelaide, pp. 69–74. Kriticos, D.J., Potter, K.P., 2006. Something’s bugging butterfly bush in Tasmania. Proceedings of the Fifteenth Australian Weeds Conference. Weed Science Society of Victoria, Melbourne, Australia. Kriticos, D.J., Brown, J.R., Radford, I.D., Nicholas, M., 1999. Plant population ecology and biological control: Acacia nilotica as a case study. Biol. Control 16 (2), 230–239. Kriticos, D.J., Brown, J.R., Maywald, G.F., Radford, I.D., Nicholas, D.M., Sutherst, R.W., Adkins, S.A., 2003. SPAnDX: a process-based population dynamics model to explore management and climate change impacts on an invasive alien plant, Acacia nilotica. Ecol. Model. 163, 187–208. Landsberg, J.J., Waring, R.H., 1997. A generalised model of forest productivity using simplified concepts of radiation-use efficiency, carbon balance and partitioning. For. Ecol. Manage. 95, 209–228. Langstrom, B., Hellqvist, C., 1991. Effects of different pruning regimes on growth and sapwood area of Scots pine. For. Ecol. Manage. 44, 239–254. Lavigne, M.B., Little, C.H.A., Major, J.E., 2001. Increasing the sink:source balance enhances photosynthetic rate of 1-year-old balsam fir foliage by increasing allocation of mineral nutrients. Tree Phys. 21, 417–426. Legg, B.J., Day, W., Lawlor, D.W., Parkinson, K.J., 1979. The effects of drought on barley growth: models and measurements show the relative importance of leaf area and photosynthetic rate. J. Agric. Sci. 92, 703– 716. Lehtila¨, K., Boalt, E., 2004. The use and usefulness of artificial herbivory in plant–herbivore studies. In: Weisser, W.W., Siemann, E. (Eds.), Insects and Ecosystem Function. Springer, Heidelberg, pp. 258–275. Linder, S., 1985. Potential and actual production in Australian forest stands. In: Landsberg, J.J., Parsons, W. (Eds.), Research for Forest Management. CSIRO, Canberra, Australia, pp. 11–35. Littell, R.C., Milliken, G.A., Stroup, W.W., Wolfinger, R.D., Schabenburger, O., 2006. SAS for Mixed Models, second ed. SAS Institute Inc., Cary, NC. Mabry, C.M., Wayne, P.W., 1997. Defoliation of the annual herb Abutilon theophrasti: mechanisms underlying reproductive compensation. Oecologia 111, 225–232. McEvoy, P.B., Cox, C.S., Coombs, E.M., 1991. Successful biological control of ragwort. Ecol. Appl. 1, 430–432. McFadyen, R.E.C., 1998. Biological control of weeds. Ann. Rev. Ent. 43, 369–393. McNaughton, S.J., 1983. Compensatory plant growth as a response to herbivory. Oikos 40, 329–336. Monteith, J.L., 1977. Climate and the efficiency of crop production in Britain. Philosophical Transactions of the Royal Society of London B281, 277–294. Neales, T.F., Incoll, L.D., 1968. The control of leaf photosynthesis rate by the level of assimilate concentration in the leaf: a review of the hypothesis. Bot. Rev. 34, 107–123.

129

Ooi, P.A.C., 1992. Biological control of weeds in Malaysian plantations. Proc. 1st Int. Weed Control. Cong., Melbourne: Weed Sci. Soc. Victoria. Oso´rio, J., Oso´rio, M.L., Chaves, M.M., Pereira, J.S., 1998. Water deficits are more important in delaying growth than in changing patterns of carbon allocation in Eucalyptus globulus. Tree Phys. 18, 363–373. Pinkard, E.A., Beadle, C.L., 1998. Above ground biomass partitioning and crown architecture of Eucalyptus nitens following green pruning. Can. J. For. Res. 28, 1419–1428. Raghu, S., Dhileepan, K., 2005. The value of simulating herbivory in selecting effective weed biological control agents. Biol. Cont. 34, 265– 273. Raghu, S., Dhileepan, K., Tervin˜o, M., 2006. Response of an invasive liana to simulated herbivory: implications for its biological control. Acta Oecologia 29, 335–345. Raison, R.J., Myers, B.J., 1992. The Biology of Forest Growth experiment : linking water and nitrogen availability to the growth of Pinus radiata. For. Ecol. Manage. 52, 279–308. Richardson, B., Vanner, A., Ray, J., Davenhill, N., Coker, G., 1996. Mechanisms of Pinus radiata growth suppression by some common forest weed species. N.Z.J. For. Sci. 26, 421–437. Ryle, G.C., Powell, C.E., 1975. Defoliation and regrowth in the graminaceous plant: the role of current assimilate. Ann. Bot. 39, 297–310. Saldarriaga, J.G., Luxmoore, R.J., 1991. Solar energy conversion efficiencies during succession of a tropical rainforest in Amazonia. J. Trop. Ecol. 7, 233–242. Sands, P.J., Landsberg, J.J., 2002. Parameterisation of 3-PG for plantation grown Eucalyptus globulus. For. Ecol. Manage. 163, 273–292. SAS Institute., 1996. SAS/STAT Software : Changes and enhancements through release 6.11. Cary, NC : SAS Institute Inc. Sheppard, A.W., Shaw, R.H., Sforza, R., 2006. Top 20 environmental weeds for classical biological control in Europe: a review of opportunities, regulations and other barriers to adoption. Weed Res. 46, 93– 117. Smale, M.C., 1990. Ecological role of buddleja (Buddleja davidii) in streambeds in Te Urewera National Park. N.Z.J. Ecol. 14, 1–6. Strauss, S.Y., 1988. Determining the effects of herbivory using naturally damaged plants. Ecology 69, 1628–1630. Strauss, S.Y., Agrawal, A.A., 1999. The ecology and evolution of plant tolerance to herbivory. Trends Ecol. Evol. 14, 179–185. Trumble, J.T., Kolodny-Hirsch, D.M., Ting, I.P., 1993. Plant compensation for arthropod herbivory. Ann. Rev. Ent. 38, 93– 119. von Caemmerer, S., Farquhar, G.D., 1984. Effects of partial defoliation, changes of irradiance during growth, short-term water stress and growth at enhanced p(CO2) on the photosynthetic capacity of leaves of Phaseolus vulgaris L. Planta 160, 320–329. Waring, P.F., Khalifa, M.M., Treharne, K.J., 1968. Rate-limiting processes in photosynthesis at saturating light intensities. Nature 220, 453–458. Watt, M.S., Whitehead, D., Mason, E.G., Richardson, B., Kimberley, M., 2003. The influence of weeds on allocation and above-ground biomass growth of young Pinus radiata growing at a dryland site. For. Ecol. Manage. 183, 363–376. Whigham, D.F., 1990. The effect of experimental defoliation on the growth and reproduction of a woodland orchid, Tipularia discolour. Can. J. Bot. 68, 1912–1916. Wirf, L.A., 2006. The effect of manual defoliation and Macaria pallidata (Geometridae) herbivory on Mimosa pigra: Implications for biological control. Biological Control 37, 346–353.