Modelling the impact of defoliation by the leaf beetle, Chrysophtharta bimaculata (Coleoptera: Chrysomelidae), on height growth of Eucalyptus regnans

Forest Ecology and Management, 54 ( 1992 ) 69-87 69 Elsevier Science Publishers B.V., Amsterdam

Modelling the impact of defoliation by the leaf beetle, Chrysophtharta bimaculata ( Coleoptera: Chrysomelidae), on height growth of Eucalyptus

regnans

Steven G. Candy, Humphrey J. Elliott, Richard Bashford and Anna Greener Forestry Commission, Tasmania, GPO Box 207B, Hobart, Tas. 7001, Australia

(Accepted 2 December 1991 )

ABSTRACT

Candy, S.G., Elliott, H.J., Bashford, R. and Greener, A., 1992. Modelling the impact of defoliation by the leaf beetle, Chrysophtharta bimaculata (Coleoptera: Chrysomelidae), on height growth of Eu- calyptus regnans. For. Ecol. Manage., 54: 69-87.

A model of the effect of defoliation by the Tasmanian eucalyptus leaf beetle, Chrysophtharta birnaculata Olivier, on the height growth of Eucalyptus regnans F. Muell. in young plantations is described. Two separate artificial defoliation trials were set up to provide data to calibrate and test the model. The treatments involved combinations of different defoliation levels and timing with and without a repeat of these treatments in the following year as well as a treatment involving regular removal of refoliating shoots (disbudding). The effect of disbudding was found to be greater on growth than was low levels of defoliation. The model gave an adequate fit to the data, but not all combinations of treatments were available to calibrate and test the model. A small simulation study employing the model to investigate the effect of defoliation and disbudding on plantation economics indicated a potentially high benefit to cost ratio for control of C. birnaculata, especially where refoliating shoots were grazed.

INTRODUCTION

A major factor in the economic viability of eucalypt plantations in Tas- mania, particularly those of Eucalyptus regnans F. Muell. and Eucalyptus ni- tens Maiden, is the impact of defoliation by the Tasmanian eucalyptus leaf beetle, Chrysophtharta bimaculata Olivier, on the growth and survival of eu- calypts, particularly at young ages. Elliott et al. ( 1992 ) describe an integrated pest management (IPM) strategy for control of C. bimaculata. The IPM in- volves regular monitoring of population levels to predict the degree of defoliation. In order to define unacceptable defoliation levels in the IPM strategy,

Correspondence to: S.G. Candy, Forestry Commission, Tasmania, GPO Box 207B, Hobart, Tas. 7001, Australia.

© 1992 Elsevier Science Publishers B.V. All rights reserved 0378-1127/92/$05.00

7 0 S.G. CANDY ET AL.

the impact of defoliation on yield and economic return from plantations must be determined. For this reason, artificial defoliation trials were set up to provide data to calibrate a mathematical model of the effect of defoliation on height growth of plantation-grown E. regnans.

Both artificial (manual leaf removal) and natural defoliation can be used to evaluate the effects of defoliation on tree growth and survival. Artificial defoliation can be imposed to reasonably precise levels and timed to simulate major defoliation peaks. However, this technique can only very approximately mimic the nature and degree of defoliation caused by leaf beetles. Nat- ural defoliation can indicate the effects of particular population levels but these levels vary considerably from year to year and within attacked stands, making experimental replication and adequate control of defoliation levels very difficult.

In trials reported here, artificial defoliation is used in an attempt to simulate the browsing of chrysomelids. Manual removal of foliage and the regular removal of refoliating shoots (hereafter referred to as disbudding) was carried out at the time of year and at levels, in terms of percentage crown reduc- tion, designed to simulate leaf beetle attack. Data from these trials are used to construct and test a model predicting the effect on height growth of different levels, timing and frequencies of defoliation, combined with and without disbudding. Finally, this model is combined with a single-tree growth model in a simulation study which gives estimates of stand volume, mean annual increment of stand volume and net present value of eucalypt plantations for a range of defoliation histories.

METHODS

Two trials were conducted using artificial defoliation techniques to simulate attack by C. bimaculata. The first trial (Trial A), begun in 1981, imposed a range of defoliation levels on young E. regnans at times corresponding to peak natural defoliation from populations of C. bimaculata. The second trial (Trial B), begun in 1988, was a development of the initial trial and imposed additional regular removal of new shoots (i.e. disbudding) following major defoliation in an attempt to more closely simulate the feeding habits of C. bimaculata, i.e. major defoliation by larval populations with continuing browsing of refoliating trees by adult beetles.

Trial A

The effect of different defoliation levels, timing of defoliation episodes, and repeated defoliation on the height and diameter growth of E. regnans was examined in Esperance Plantation in southern Tasmania (map reference: Huon 1:100 000 DN936052).

MODEL OF EFFECT OF BEETLE DEFOLIATION ON EUCALYPT GROWTH 71

Three hundred and eighty plantation-grown E. regnans aged 3 years and in the height range 175-225 cm were selected for treatment. Selected trees were free from weed competition, and had good crown condition, stem form and stable root systems. Nineteen treatment combinations covering different levels, timing and repetitions of artificial defoliation, including a control, were randomly allocated to the selected trees (20 trees per treatment) (Table 1 ).

Level of defoliation was expressed as percentage of crown volume removed and applied at three nominal levels. The levels of 33%, 66% and 100% were subjectively assessed in the field by visualizing an imaginary line around the outside of the crown and removing leaves to the required level.

The timing treatment consisted of 'early', 'late' and 'early and late' summer defoliation episodes where 'early' defoliation was imposed in January and 'late' imposed in early March to simulate the normal peaks of chrysomelid defoliation which usually occur in the area. The repeat treatment involved repeating the above treatments in the following year with defoliation levels referring to the current volume of the crown.

Defoliation was carried out by manually stripping leaves from the branches and twigs as described by Came et al. (1974). Defoliation commenced at the top of the tree and branch tips and continued towards the centre of the crown to the required level. This method simulates defoliation by C. bimaculata lar-

TABLE 1

Artificial defoliation treatments imposed in a 3-year-old E. regnans plantation (Trial A ) in Esperance Plantation, Tasmania, Australia, 1981

Number Code Treatment description

1 Control 2 33%, early 3 66%, early 4 100%, early 5 33%, early, repeat 6 66%, early, repeat 7 100%, early, repeat 8 33%, late 9 66%, late

10 100%, late 11 33%, late, repeat 12 66%, late, repeat 13 100%, late, repeat 14 33%, early and late

15 66%, early and late 16 100%, early and late 17 33%, early and late, repeat 18 66%, early and late, repeat 19 100%, early and late, repeat

No defoliation Nominal 33% of crown removed in early summer Nominal 66% of crown removed in early summer Complete defoliation of the whole crown in early summer (2) repeated in second year (3) repeated in second year (4) repeated in second year Nominal 33% of crown removed in late summer Nominal 66% of crown removed in late summer Complete defoliation of the whole crown in late summer (8) repeated in second year (9) repeated in second year (10) repeated in second year Nominal 33% of crown removed in early and then late summer As for (14) but with 66% nominal level As for (14) but with complete defoliation (14) repeated in second year ( 15 ) repeated in second year ( 16 ) repeated in second year

72 S.G. CANDY ET AL.

vae at peak population times. Defoliation was spread over 2 weeks to simulate the feeding period of third and fourth instar chrysomelid larvae. All trees in the trial were protected from C. bimaculata browsing by regular spraying with insecticide.

In order to assess the actual level of defoliation achieved in the field, 60 trees adjacent to the trial area and in the same height range as the treated trees were defoliated to the same nominal levels as those used in the trial (i.e. 33%, 66%, 100%; 20 trees per level). The mean dry weight of leaves removed in each of the 33% and 66% treatments was then expressed as a percentage of the mean dry weight of leaves removed from the 100% defoliated trees to establish the actual level of defoliation achieved.

Tree heights were measured to the highest living portion using a telescopic fibreglass height pole. During the first 18 months the tree heights were measured before each defoliation (January and March) and twice over the inter- vening winter (May and August). Following completion of all defoliation treatments, heights of selected trees were measured three times a year (Janu- ary, March and August) for 2 years. The untreated control trees were remea- sured monthly for the duration of the trial. The diameter of all trees was measured at the t ime of the final height measurement in June 1985.

Trial B

A trial to measure the effect of defoliation with and without disbudding (to simulate continuous grazing of refoliating trees by adult C. bimaculata) was conducted in a 6-year-old E. regnans stand within the same plantation as Trial A.

Two hundred and twenty trees with good crown condition and stem form were selected within a height range of 250-300 cm. One of 11 treatments (Table 2 ) was randomly allocated to each tree (20 trees per treatment ). Two nominal levels of defoliation, 50% and 100% of the current season's foliage, were used. The timing treatment was the same as in Trial A with the excep- tion that the combined early and late treatments were only applied at the 50% defoliation level. The repeat treatment was not used in this trial. The disbudding (removal of all new shoots < 4 cm in length) treatment in this trial was conducted (where required by treatment specification) at approximately 2 weekly intervals following the defoliation treatment until the end of the C. bimaculata season (April).

Tree heights were measured, using the same method as in Trial A, in early November 1988, prior to the early summer defoliation, and in February, March, April, June, September, October and November 1989.

To determine the actual defoliation level imposed by the 50% treatment, a further 20 trees in the 250-300 cm height range were sampled and dry weights of buds and leaves were obtained by progressively defoliating 50% and 100%


TABLE 2

Artificial defoliation treatments imposed in a 6-year-old E. regnans plantation (Trial B ) in Esperance Plantation, Tasmania, Australia, 1988

Number Code Treatment description

1 Control 2 50%, early 3 50%, early, disbud 4 100%, early 5 100%, early, disbud 6 50%, late 7 50%, late, disbud 8 100%, late 9 100%, late, disbud

10 50%, early and late 11 50%, early and late

No defoliation or disbudding Nominal 50% of new foliage removed in early summer As for (2) plus disbudding Complete removal of new foliage in early summer As for (4) plus disbudding Nominal 50% of new foliage removed in late summer As for (6) plus disbudding Complete removal of new foliage in late summer As for (8) plus disbudding Nominal 50% of new foliage removed in early summer As for (10) plus disbudding

of new foliage and 100% of all foliage. Leaves were separated into new and old foliage classes and the weight in each class was determined. The mean dry weight of buds and leaves making up new foliage obtained by defoliating to the nominal 50% level was used to determine the actual defoliation level achieved by conversion to the percentage of mean dry weight of 100% of new foliage as was done in Trial A.

Modelling the effects of defoliation on growth

The treatments employed in Trial A and B represent a small subset of possible combinations of tree age and possible treatments using level, timing, and repeats of defoliation combined with or without disbudding. To allow forest managers to assess the effect of any level of defoliation on height growth for a range of ages, a mathematical model was constructed using the data obtained in these trials.

Despite the differences in experimental methods and design between Trial A and Trial B, the results from both trials were combined to allow a more complete model of the effect of defoliation and disbudding on height growth than could be estimated from either trial alone. As a result, data from some treatment combinations were not available to calibrate and test the model (e.g. t im ingx repeat × disbudding). Actual defoliation level was used as a continuous effect in the model rather than discrete nominal levels. Defolia- tion level was expressed as a percentage of all foliage in Trial A but only as a percentage of current season's growth (new foliage) in Trial B. To allow the results from both trials to be combined, the defoliation level for Trial A was partit ioned into new and old foliage components based on the dry weights obtained in Trial B.


It is clear that the effect of the treatments on height growth would be multiplicative since, for example, the difference in effect between treatments '33%, early' and '33%, early, repeat' would not be the same as between '66%, early' and '66%, early, repeat'. For this reason the models tested involved a log- linear component and were fitted using non-linear regression techniques.

Simulation study on the effects of defoliation on plantation yield and economics

To demonstrate the economic impact of defoliation and disbudding in eucalypt plantations the model of the effect of defoliat ion/disbudding on height growth described later was incorporated in a simulation study which compared three defoliation histories. (i) No defoliation or disbudding because of control of C. bimaculata by application of insecticide at a cost of $30 ha - year- ~ and carried out under each of two regimes: (a) at ages 3, 5 and 7 years simulating the decision to control C. bimaculata based on monitoring in the IPM strategy of Elliott et al. ( 1992 ), and (b) at each summer between ages 2 and 10 simulating routine control of C. bimaculata. (ii) Three episodes of light defoliation with 50% of new foliage removed early in the summer at ages 3, 5 and 7 years. (iii) As for (ii) but with disbudding at each of the defoliation episodes. The simulation of a hypothetical stand of 110 stems on 0.1 ha (i.e. assuming an initial stocking of 1100 stems per hectare and no subse- quent mortality) used mean initial height at age 2 of 3.5 m, with individual heights distributed around this mean using a Weibull distribution with coef- ficient of variation (c.v.) of 30%. Trees were then grown at a mean rate of 1.75 m year- ~ to age 9 with individual growth rates varied again using a Wei- bull distribution with c.v. of 30% and assigned to trees in the same rank order as initial height. The above mean initial height, mean growth rate, and c.v.s were based on unpublished data (C.R.A. Turnbull, personal communication, 1991 ) on growth rates to age 6 from a trial planting ofE. regnans which had insect control to age 4, weed control to age 3, and was fertilized to age 4. Simulation of defoliation level for each tree was calculated using 20 binomial trials (with probability level set to 0.5 corresponding to the level of 50% set for (ii) and (iii) above). The sum of "successful trials' (i.e. the number of random numbers ranging from 0 to 1 that were below 0.5 ) was divided by 20 to give the proportion of new foliage removed. Disbudding was applied to every tree at the defoliation ages for (iii). The model of the effect of defolia- t ion/disbudding on height growth was applied at each defoliation/disbudding age to reduce height growth for the following year.

From age 10 onwards tree height was converted to diameter under bark at 1.3 m above ground (DBHUB) using an equation given by Goodwin and Candy (1986), derived from data on growth of a Eucalyptus globulus plantation in northern Tasmania, and trees 'grown on' with the DBHUB incre-


ment equation given by Goodwin and Candy (1986). At each year from age 10 to 20 the tree total height and then entire stem volume under bark (m 3) (ESV) were calculated from DBHUB using equations given by Goodwin and Candy ( 1986 ). Stand ESV mean annual increment (MAI) (m 3 ha - 1 year- 1 ) and net present value (NPV) were also calculated at each of the above ages. The simulations assumed either a 4% or 6% discount rate, $1500 ha- ~ estab- lishment cost, $20 ha - ~ year- ~ maintenance cost (excluding the cost of insect control ) and a $ 30 m-3 stumpage. Since each simulation is based on the same site quality, the growth rates from age 10 onwards for (ii) and (iii) were obtained by shifting the growth rate/age relationship for (i) to an earlier age so that the yield trajectories for (ii) and (iii) would eventually 'catch-up' to that for (i).

RESULTS

Trial A

Quantitative assessments of the defoliation levels employed in this trial showed that the subjectively assessed 33% and 66% defoliation levels corre- sponded to 17% and 49% leaf dry weight removed, respectively. For simplic- ity, future references to defoliation level will still refer to nominal percentages although actual levels were used in the model.

The only defoliation treatments which resulted in significant mortality were those imposing 100% defoliation late in the summer for 2 consecutive years (treatments: '100%, late' or '100%, early and late, repeat') (Table 3). Only

TABLE 3

Percentage mortality by the final measurement of Trial A in Esperance Plantation, Tasmania, Aus- tralia, 1985

Defoliation Control Early Late Early and late level (%)

Repeat Repeat Repeat

0 0 (18)2

33 0 0 0 0 0 0 (20) (20) (19)1 (20) (20) (18)2

66 0 0 5 5 0 5 (20) (20) (19) (19) (19)1 (19)

100 0 5 0 60 0 85 (18)2 (17)2 (20) (8) (20) (3)

Numbers in parentheses are the numbers of live trees at the final measurement; the numbers outside the parentheses are the numbers of deaths attributed to windthrow or other non-treatment causes.


three of the 20 trees in the '100%, early and late, repeat' treatment combina- tion survived after four complete defoliations.

Significant differences in height growth of treated trees resulted from different levels, timing and repeats of defoliation (Table 4). It is not possible to determine the individual factor effects since the trial does not have a factorial design structure. Defoliation level of 33% had little effect on the height growth of the trees. In some cases these lightly defoliated trees grew slightly more than the controls but this difference was not statistically significant (Table 4).

Defoliation level of 66% resulted in reduced height growth in all cases. However, the effect was not significant unless the defoliation was repeated. The effect of 100% defoliation was severe and resulted in significant height loss in all cases (Table 4).

The growth pattern and recovery of defoliation treatments is compared with that of the control in Fig. 1. Figure 1 (a) shows the extremely good recovery of trees treated with a single defoliation early in the season. The total defoliation (' 100%, early' ) is the only single, early season treatment which resulted in a growth pattern significantly different from the control.

The effect of severe repeated defoliation can be seen most clearly in the case of the 'early and late, repeat' treatments in Fig. 1 (b). These multiple defoliation treatments resulted in significant loss of growth. In the most severe case of 100% defoliation height growth was effectively stopped from the first treatment (January 1982 ) until the summer of 1984-1985, 3 years later.

The diameter over bark at a height of 1.3 m (DBHOB) of all trees was measured once only at the final measurement (winter 1985) and the mean

TABLE 4

Mean height (cm) (standard error) at the final measurement of Trial A in Esperance Plantation, Tasmania, Australia, 1985



0 781.1 (33.8)

33 798.5 763.7 726.1 726.7 832.1 673.1" (30.3) (31.4) (28.3) (24.3) (23.9) (23.4)

66 745.3 665.3* 704.5 599.7** 687.1 548.2** (22.5) (24.3) (30.9) (24.5) (30.1) (22.6)

100 624.4** 468.1"* 621.5" 429.4" 519.8"* 381.7" (36.8) (23.2) (27.9) (21.5) (24.6) (48.1)

*P< 0.05, treatment differs from control; **P< 0.01, treatment differs from control.


gGO

V~o

500

,k00

3O0

200

100

(a)

O - - -- -- -~ 33 PERCENT T + . . . . . . . . -4- 66 PERCENT ~ H

. . . . _.~ 100 PERCENT /~I i j./ TH

I ~ INITIRL /

(b)

.. c . ; ...... ~ ........ . . - " I

~ IN IT I ! MEI~UREiNENT

Fig. 1. Mean height of control and defoliation treatments in Trial A, Esperanee Plantation, Tasmania: (a) early, ( b ) repeat early and late. Arrows indicate defoliation treatment. Standard error bars are shown and abscissa values have been jittered for clarity.

TABLE 5

Mean DBHOB (cm) (standard error) at final measurement of Trial A in Esperance Plantation, Tas- mania, Australia, 1985



0 8.7 (0.4)

33 9.3 7.9 8.3 7.9 9.3 6.8** (0.6) (0.4) (0.4) (0.5) (0.4) (0.2)

66 7.6* 6.6** 7.3* 5.8** 7.0** 4.9** (0.3) (0.3) (0.4) (0.3) (0.4) (0.4)

100 5.5** 3.5** 6.3** 3.4** 4.3** 3.2** (0.5) (0.3) (0.3) (0.4) (0.3) (0.9)

*P< 0.05, treatment differs from control; *P< 0.01, treatment differs from control.

DBHOB for each treatment is shown in Table 5. Treatment effects are similar to those for height.

To test if treatments had any effect on the relationship between height and diameter, linear regressions of height on DBHOB were calculated. Of the 18 treatments, five had a regression slope and /o r intercept which differed significantly (P< 0.05) from the control. There were no obvious trends in these


parameters with either level or t iming of defoliation, with only one of the five being a 100% defoliation. Examination of residuals indicated no obvious de- parture from a linear relationship.

Trial B

Sampling the dry weight of new foliage excluding the weight of buds gave an actual defoliation level for new foliage of 63% for the 50% nominal level treatment calculated in the same way as described for Trial A. The total foliage removed for the 100% defoliated trees was made up of 65% old and 35% new foliage. As a result the nominal 33% and 66% levels from Trial A were partitioned into actual levels of 50% ( ~ 17/0.35) and 100% of new foliage and 0% and 21% ( = ( 4 9 - 3 5 ) / 0 . 6 5 ) of old foliage, respectively. The nominal 100% level from Trial A gives obviously 100% new and 100% old foliage removal. Old foliage removal can only be greater than 0% if new foliage defoliation is 100% because of the progressive nature of the artificial defoliation from the outer to inner part of the tree. Old foliage defoliation level was 0% for all Trial B treatments.

The mean heights at the final measurement for the 11 treatment combinations are presented in Table 6. Treatments which incorporated disbudding following the imposed defoliations produced a significantly (P< 0.05 ) lower final height than defoliation alone for each of the other five treatment combinations. Figure 2(a) shows the effect of early defoliation at the nominal 50% level on height growth for each of disbudding and no disbudding treatments. Figure 2 (b) similarly shows this effect including late as well as early 50% defoliation. From these two figures the strong additional negative effect on height growth of disbudding can be clearly seen. Only when disbudding

TABLE 6

Mean height (cm) (standard error) at the final measurement of Trial B in Esperance Plantation, Tasmania, Australia, 1989


Disbud Disbud Disbud

0 373.2 (13.6)

50 353.2 324.3* 370.5 331.1" (9.7) (11.6) (13.0) (7.2)

100 359.7 308.1"* 376.7 332.5* (10.5) (6.7) (8.7) (6.7)

357.9 (10.9)

294.9** (5.9)

*P< 0.05, treatment differs from control; **P< 0.01, treatment differs from control.


(Q) (b) 4OO r 4OO

| X X CONTROL | ® - - _ _ -(9 NO OISBUDDIN8 T

~o + . . . . . . . . + OIBBUOOING E H

o so loo 150 2OO 25O SO0 ~0 4OO ~ 260 o 50 10o 15o 2oo 25o sac as0 ~o *~o

OH/S FItOM INITIAL NERStJIflEMENT ORY$ FRON INI'rlRL HER~F~¢ENT

Fig. 2. Mean height of control and nominal 50% defoliation with and without disbudding treatments in Trial B, Esperance plantation, Tasmania: (a) early, (b) early and late. Arrows indicate defoliation treatment. Standard error bars are shown and abscissa values have been ji t tered for clarity.

TABLE 7

Parameter estimates (standard error I ) for Model 1

0/I 0/2 f l l f12 f13 ~1 ~2 ~)3 ~4

781.0 373.2 0.056 0.091 0.042 0.516 0.388 0.873 -0 .295 (30.0) (13.6) (0.010) (0.013) (0.007) (0.120) (0.065) (0.126) (0.139)

I Based on residual mean square of 0.0274 using pooled lack-of-fit and pure error sums of squares.

was included were the t rea tments significantly different f rom the control (Ta- ble 6 ).

There was very little morta l i ty observed in this trial. A single tree died in each of t rea tments 'late, 50%, disbud ' and 'late, 100°/o '.

Modelling of effects of C. bimaculata on growth

The best form of the mode l of those tested was found to be

H = ( a , T, +a2T2) (1)

{1 - (fllO"P fl2Po + fl3Pn) exp [ y,E + y2L + y3R + y4ER ] }

where H is height ( cm) at the final measurement , T1 and T2 are indicator variables with T~ = 1, 7"2 = 0 for Trial A and T~ = 0 , / '2 = 1 for Trial B, D indicates d isbudding ( D = 1, 0 otherwise) , Po and Pn are levels of defoliation for


o 0 NO DI5BUDDING ~ DISBUDOING

MODEL WITHOUT DI$BUDDING . . . . . MODEL WITH DISBUODING

(o) ,., b)

o 1,.o5

o ,

c °.95

o 0.9

I I I

-t

1.05 o I

o c

0.95

i o,9

0.85

Fig. 3. Mean height as a proportion of control mean height at final measurements for Trials A and B, Esperance Plantation, Tasmania, versus percentage defoliation of new foliage for (a) early and (b) late defoliation. Nominal 66% and 100% defoliation treatments in Trial A have been excluded because defoliation included a proportion of old foliage. Actual defoliation levels have been used and predictions from Model 1 are shown where disbudding is assumed to occur when percentage defoliation is greater than zero. Standard error bars are shown and the abscissa value for the trial B control has been jittered for clarity.

old and new foliage, respectively, expressed as proport ions, E indicates early ( E = 1, 0 otherwise) , L indicates late ( L = l, 0 otherwise) t iming of defoliation, R indicates repeat of the first year's t rea tment in the second year (R = 1 (Trial A only) , 0 otherwise) and a t , a2, ill, f12, f13, 7~, 72, 73 and 74 are parameters to be est imated.

The parameters a~ and a2 can be es t imated directly using the control mean heights of 781.0 and 373.2 for Trial A and Trial B, respectively. When these parameters were es t imated simultaneously with the remaining parameters using all the data there was no significant ( P > 0.10) improvemen t in fit compared with the fit where a ~ and a2 were es t imated using the control means.

Parameter estimates for Model 1, with a l and a2 est imated directly using the control mean heights, are given in Table 7. The early × repeat (ER) interaction was the only significant interact ion ( P < 0.025 ) as de te rmined by com- parison of the corresponding parameter estimate, 94, with its s tandard error (Table 7 ). Statistical tests were carried out using the pure error componen t of the residual sum of squares which was available because of the repeat observations (the surviving trees) at each design point (Draper and Smith, 1981 ). The logari thmic t ransformat ion was applied to both sides of eqn. ( 1 )

M O D E L OF EFFECT OF BEETLE D E F O L I A T I O N ON EUCALYPT G R O W T H 81

U 16 / O~

1 / / .+

) 12 ~ . ' " +" -

. + e¢" . ÷ - '

tO .+ . •

. + . X X ( I ) CONTROL

8 4 - . + - O - - - - - - e ( I f ) LIGHT DEFOLIflTION

. .+. . + . . . . . . . . . + ( I l l ) DISBUDDING WITH ( I f )

8 4-- +.

Fig. 4. Simulation results for mean annual volume increment for stand (i) control, (ii) light (50%, early), and (iii) light defoliation combined with disbudding at ages 3, 5 and 7.

to stabilize the variance. This was partially successful with a Cochran's C test (Cochran, 1937) for homogeneity of variance giving a probability level of 0.04. The remaining degree of heterogeneity of variance was not considered sufficient to seriously affect estimation or hypothesis testing. The model given by eqn. ( 1 ) overall gave a non-significant lack-of-fit statistic ( P - 0.1 ). To test if the effect of defoliation level was non-linear, two extra parameters were added to Model 1. These parameters were used as powers of Pc and P,. When this model was fitted it was found that neither of these parameters differed significantly (P> 0.1 ) from 1, indicating that the linear effect of defoliation level in Model 1 is adequate given the data available here. The two trials differ in the age at which treatments were applied, so to test any residual effect of age, the mean of the residuals from the model for the two trials were compared and found to be not significantly different (P> 0.1 ).

The effect of disbudding and defoliation in Model 1 is to proportionally reduce the height expected if these effects are absent. The effect of early, late and repeat treatments are multiplicative on each other (via the exponential operator) and on defoliation level and disbudding effects. The significant negative ER interaction indicates a greater effect of repeat of late defoliation (i.e. 0.388 + 0.873 ) than the effect of consecutive late defoliations considered independently (i.e. 2 X 0.388 ). For the case of early defoliation the additive repeat effect gives a result (i.e. 0.516 + 0 . 8 7 3 - 0.295 ) much closer to two independently considered early defoliations (i. e. 2 X 0.516 ).

The term ( f l lD+ fl2Po + fl3Pn ) exp [ 7~E+ ~)2 L -t" 73R + 7~ER ] in the model de- fines the proportion of height loss from defoliation. Although mathematically

82 S.G. CANDY ET AL

2

T 1.5

N -I

J L 0.5: E

1

If1-0.5

(a)

x / . + / + . .

e l . + • " i

+-

+

.+

+.

+

-I X X (I) CONTROL O - - - - - - ' - ~ ( l i t LIGHT DEFDLIBTIDN + . . . . . . . + ( l i l t DISBUDDIN8 WITH (liT

RGE

2

Rp 1.5

U 0.5

o o o

) - 0 . 5

(b)

/ , m /

÷

÷ + .

.+ +

+.

3,5

3 T

2.5

2 T

L l .S

o ~ 0

o,)

1 o

-0.5

. f

/~ / +.+~÷+ / . + +-

/ /

d +

.+.

.÷

+-

Fig. 5. Simulation results for net present value for stand (i) control, (ii) light (50%, early), and (iii) light defoliation combined with disbudding at ages 3, 5 and 7. (a) Insecticide applied at ages 3, 5 and 7 and a 6% discount rate. (b) Insecticide applied at each age between 2 and 10 and a 6% discount rate. (c) As for (a) but with a 4% discount rate. Costs, stumpage and other assumptions are explained in the text.

this term is no t constra ined to be less than 1, g iven the parameter est imates in Table 7 and the most severe treatments o f D, Po, Pn, E, L, R = 1, this term takes a value o f 0 .83 (i .e. an 83% loss in height compared to contro l s ) . S ince the m o d e l can be expressed as a proport ion o f the height o f trees o f the same age which have not be defo l ia ted or d i sbudded , the m o d e l can be appl ied to

MODEL OF EFFECT OF BEETLE DEFOLIATION ON EUCALYPT GROWTH 8 3

trees of ages other than those in the trials reported here. However, there are no data available at this time to test the model for other ages.

Disbudding has a stronger effect on height loss as seen by the estimate of ill compared to defoliation effects f12 and f13 which are scaled by the proportion of defoliation. Only at relatively high defoliation levels is the effect of defoliation as strong as that of disbudding. This strong effect of disbudding reflects that displayed in the treatment means in Table 6 mentioned earlier. It is not appropriate to compare the relative magnitudes of/~2 and/~3 since, as mentioned earlier, the effect of Po cannot be assessed independently of Pn since defoliation of old foliage only occurs in the trials after all new foliage is removed.

Figure 3 compares the fit of the model to observed means expressed as a proportion of control mean height.

Simulation study on the effects o f defoliation on plantation economics

The MAI and NPV results from the simulation study for defoliation/disbudding histories (i), (ii) and (iii) are shown in Figs. 4 and 5. The NPV results are the mean of 20 simulations for each history using a 6% discount rate combined with each of the insecticide regimes (a) and (b) within (i). The simulation was repeated using a 4% discount rate for each history but only using regime (a) within (i). The MAI results are the mean of all simulations carried out for each of (i), (ii) and (iii).

DISCUSSION

The method of artificial defoliation employed during the trials produced a reasonable approximation of natural chrysomelid defoliation in terms of the appearance of trees at time of defoliation and during the recovery period. In Trial A the levels of 33% and 66% of crown volume removal were used because they were relatively easy to assess in the field by visualizing an imaginary line around the outside of the crown. However, the smaller leaves and lower leaf density at the top of a rapidly expanding crown meant that these levels were overestimates of the actual leaf weight removed ( 17% and 49%, respectively ). Although this meant that no defoliation level between 50% and 100% leaf weight removed was imposed, the two defoliation levels actually achieved are more relevant to the field situation, as previous observation and research by the authors have shown that natural defoliation of young euca- lypts in forest areas of Tasmania is normally less than 20% of total crown leaf area and only occasionally exceeds 50%. Mortality at these levels of defoliation is uncommon, and, as can be seen from the trials reported here, defoliation needs to be of the order of 100% of all foliage and repeated in consecutive growing seasons before significant mortality occurs. A model of mortality was


not constructed and no adjustments for mortality were made in the simulation study for this reason. Apart from the fact that defoliation of this severity rarely occurs in plantations not managed under the IPM strategy, the objec- tive of this strategy is to prevent much less severe defoliations than those which produce mortality.

The level of defoliation and loss of growth increment of E. regnans in this study were linearly related. A similar relationship has been shown to exist for a wide range of deciduous and evergreen species in many studies reviewed by Kulman ( 1971 ). The effects of artificial defoliation on other eucalypt species are not well known although Came et al. (1974) reported that only very high levels of artificial defoliation of Eucalyptus grand& ( > 65%) produced significant loss of growth increment.

Graphs of the growth of trees in individual treatments show the importance of using repeated defoliation within a season andin consecutive years to gain more information on the likely effect of C bimaculata on growth of E. reg- naris. The recuperative ability ofE. regnans under repeated light and medium defoliations is illustrated by trees in these treatments regaining the same growth rates as untreated trees within 2 years of defoliation, although the original growth loss was not recovered. Imposition of further defoliations in consecutive years would probably have a far greater effect on growth and this appears to occur in E. regnans stands subjected to high levels of C bimaculata over many years. In Trial B, the fact that the '100% defoliation/no disbudding' treatments had, at the final measurement, a mean height not significantly different from the control is in contrast to Trial A and is probably because there was insufficient time between late defoliation in early March and the last measurement in November for the defoliation effect to be expressed. It would be expected that the diversion of resources by the tree into refolia- tion would impact height growth in the following growing seasons.

Adverse conditions during recovery may produce effects more severe and longer lasting than those exhibited by the vigorous Esperance Plantation re- generation. Observations by the authors in the Florentine Valley indicate that trees which have a defoliated upper crown when cold conditions occur in winter can suffer extensive dieback and death of naked shoots.

The imposition of regular removal of ref01iating shoots (i.e. disbudding) in some treatments in Trial B shows the severe effect that repeated browsing of such shoots has on height increment despite the short measurement period over which growth was measured. Adult C bimaculata have been observed by the authors to repeatedly remove refoliating shoots, but the effects of this feeding habit have not been accurately quantified before this trial. The results support those of Cremer ( 1972 ) who found, in a small trial using 3-year-old E. regnans, that a single removal of axillary buds as well as leaves from the top 40 cm of the crown had a more severe effect on growth than either defoliation or disbudding alone.


The incorporation of disbudding into the model allows a reasonably close mimic of the total effect on growth to be presented. The model in its current form does not allow repeat defoliations to have a different level of defoliation to that in the first year. Also, even though the effect of repeated defoliation and disbudding can be predicted from Model 1, there was no treatment com- bining repeats of disbudding; consequently the interaction between disbudding and repeat treatments could not be estimated. Based on the significant early × repeat interaction obtained in Trial A, our inability to estimate a prob- able disbud × repeat interaction could result in the model being conservative in predicting the impact on growth of consecutive defoliations combined with disbudding. Even where the repeat treatments were carded out in Trial A, only 1 further year of application of the treatment was carried out, so the effect on height growth of repeated year-to-year defoliation cannot be estimated. The model has been defined independently of age by expressing height loss as a result of defoliation/disbudding as a proportion of control height. The height of trees protected continuously after planting from defoliation/ disbudding, for a given site quality, will be largely a function of age, but for a variable history of defoliation/disbudding age will not necessarily be a good predictor of height growth. This can be seen in this study in the large difference between Trial A and Trial B in the mean height of controls, even though these trials were carried out in the same plantation and at final measurement were both approximately age 7. Although no significant difference in mean residual for the two trials was detected, a greater range of ages is required before the model, which excludes any explicit effect of age, can be considered sufficiently general.

The simulation study has a number of limitations in that it considered only a limited range of defoliation/disbudding episodes, with a single level of early summer defoliation and did not include sensitivity analyses of assumptions such as costs, discount rate and stumpage. Also, in the absence of a growth model for plantation-grown E. regnans, a single-tree growth model developed for an E. globulus plantation was used. As well as differences between these two species, the rate at which the yield from the defoliated stands would catch up to that of the undefoliated stand, which are all assumed to be of the same site quality, is based on conjecture rather than experimental data. To a degree, the use of an E. globulus growth model may be relevant to the situation here of estimating growth for stand history (i) since this species is fairly re- sistant to insect browsing. The assumption implicit in the simulation that defoliation/disbudding does not affect the height versus breast-height diameter relationship was found to be reasonable for Trial A at age 7. Whether this remains true for later ages is not known. Despite these limitations, this simulation study is instructive in demonstrating the relative effect of controlling versus not controlling C. bimaculata defoliation and disbudding on plantation economics. Even under a conservative regime of light defoliation, assum-


ing no year-to-year carryover effects, the benefit of controlling C. bimaculata using the IPM system is obvious (Fig. 5). The effect on NPV of applying insecticide each year compared with application only in the years when defoliation occurred was only slight, and reducing the discount rate from 6% to 4% had little effect on the relative difference in NPVs between histories. The results presented from this simulation study are only meant as a rough guide because of the limitations mentioned above, but they do indicate that the strong negative effect of disbudding on growth at early ages in the rotation, observed in the results from Trial B and quantified by Model 1, has a serious impact on plantation yields and economics by the end of the rotation. The defoliation/disbudding histories used in the simulation study are considered substantially less severe than those typically encountered as a result of C. bimaculata browsing in Tasmanian eucalypt plantations. As a result the economic benefit of controlling C. bimaculata could be substantially greater than that indicated by the simulation study.

Combined with the monitoring of C. bimaculata populations as part of the IPM strategy and the regression models relating population size to the level of defoliation described by Elliott et al. (1992), the model described here can be used to give the forest manager information on the cost-benefit of employing the IPM strategy before significant damage occurs. The regression model of Elliott et al. does not allow prediction of the occurrence of disbudding which has been shown in this study to have a large impact on growth and thus plantation economics. Also, the estimation of the effect of disbudding on growth in this study assumed an 'all-or-nothing' disbudding effect. If only a proportion of buds are removed the effect on growth will obviously be less. Obser- vation by the authors suggest that at high levels of defoliation total disbudding is fairly common, but further work is required to determine the incidence and intensity of disbudding as related to population size and phenology.

Finally, the information presented here applies to specific stands of E. regnans and, although extrapolation to other stands is reasonable, factors such as different site qualities, age and species must be taken into account. The authors recognize that the model presented here has shortcomings, but con- sider it to be a useful management tool until such time that additional factors can be addressed experimentally.

ACKNOWLEDGEMENTS

We thank Sue Jennings, Leigh Edwards, Matthew Long, Bill Brown, Joe Harries, Neil McCormick and Gerard Conlon (Forestry Commission) for their assistance with field trials, and Tim Wardlaw (Forestry Commission) for helpful comments on the manuscript. The assistance of Dr. Cliff Ohmart (CSIRO) in the development and design of these trials and the provision of unpublished data on E. regnans growth rates by Charles TurnbuU (CSIRO)


is gratefully acknowledged . M u c h o f the funding for these studies was pro- v ided by the T a s m a n i a n Forest Research Counci l .

REFERENCES

Came, P.B., Greaves, R.T.G. and Mclnnes, R.S., 1974. Insect damage to plantation grown eu- calypts in north coastal New South Wales, with particular reference to Christmas beetles (Coleptera: Scarabaeidae). J. Aust. Entomol. Soc., 13:189-206.

Cochran, W.G., 1937. Problems arising in the analysis of a series of similar experiments. J. R. Star. Soc. Suppl., 4:102-118.

Creme r, K.W., 1972. Effects of partial defoliation and disbudding on height growth of Eucalyp- tus regnans saplings. Aust. For. Res., 6: 41-42.

Draper, N.R. and Smith, H., 1981. Applied Regression Analysis, 2nd edn. Wiley, New York, 709 pp.

Elliott, H.J., Bashford, R., Greener, A. and Candy, S.G.. 1992. Integrated pest management of the Tasmanian Eucalyptus beetle Chrysophtharta bimaculata (Olivier) (Coleoptera: Chry- somelidae). For. Ecol. Manage., 53: 29-38.

Goodwin, A.N. and Candy, S.G., 1986. Single-tree and stand growth models for a plantation of Eucalyptus globulus Labill. in Northern Tasmania. Aust. For. Res., 16:131-144.

Kulman, H.M., 1971. Effect of insect defoliation on growth and mortality of trees. Annu. Rev. Entomol., 16: 289-324.

Modelling the impact of defoliation by the leaf beetle, Chrysophtharta bimaculata (Coleoptera: Chrysomelidae), on height growth of Eucalyptus regnans

Documents