Habitat structure and egg distributions in the processionary caterpillar Ochrogaster lunifer: lessons for conservation and pest management

Authors


*Correspondence: Graham J. Floater, Departamento de Ecología Evolutiva, Instituto de Ecología, Universidad Nacional Autónoma de México, Apartado Postal 70–275, Ciudad Universitaria, 04510 México DF, México (fax Mexico City 616 1976).

Summary

1. The spatial and temporal distribution of eggs laid by herbivorous insects is a crucial component of herbivore population stability, as it influences overall mortality within the population. Thus an ecologist studying populations of an endangered butterfly can do little to increase its numbers through habitat management without knowledge of its egg-laying patterns across individual host-plants under different habitat management regimes. At the other end of the spectrum, a knowledge of egg-laying behaviour can do much to control pest outbreaks by disrupting egg distributions that lead to rapid population growth.

2. The distribution of egg batches of the processionary caterpillar Ochrogaster lunifer on acacia trees was monitored in 21 habitats during 2 years in coastal Australia. The presence of egg batches on acacias was affected by host-tree ‘quality’ (tree size and foliar chemistry that led to increased caterpillar survival) and host-tree ‘apparency’ (the amount of vegetation surrounding host-trees).

3. In open homogeneous habitats, more egg batches were laid on high-quality trees, increasing potential population growth. In diverse mixed-species habitats, more egg batches were laid on low-quality highly apparent trees, reducing population growth and so reducing the potential for unstable population dynamics. The aggregation of batches on small apparent trees in diverse habitats led to outbreaks on these trees year after year, even when population levels were low, while site-wide outbreaks were rare.

4. These results predict that diverse habitats with mixed plant species should increase insect aggregation and increase population stability. In contrast, in open disturbed habitats or in regular plantations, where egg batches are more evenly distributed across high-quality hosts, populations should be more unstable, with site-wide outbreaks and extinctions being more common.

5. Mixed planting should be used on habitat regeneration sites to increase the population stability of immigrating or reintroduced insect species. Mixed planting also increases the diversity of resources, leading to higher herbivore species richness. With regard to the conservation of single species, different practices of habitat management will need to be employed depending on whether a project is concerned with methods of rapidly increasing the abundance of an endangered insect or concerned with the maintenance of a stable, established insect population that is perhaps endemic to an area. Suggestions for habitat management in these different cases are discussed.

6. Finally, intercropping can be highly effective in reducing pest outbreaks, although the economic gains of reduced pest attack may be outweighed by reduced crop yields in mixed-crop systems.

Introduction

The relationship between the spatial and temporal distribution of eggs laid by herbivorous insects, and subsequent growth and survival of the larvae, is a crucial component of population stability, as it influences overall mortality within the population (Soberón 1986; Price 1991). Thus an ecologist studying populations of an endangered butterfly can do little to increase its numbers through habitat management without knowledge of its egg-laying patterns across individual host-plants in the habitat. At the other end of the spectrum, a knowledge of egg-laying behaviour can do much to control pest outbreaks by disrupting egg distributions that lead to rapid population growth. Butterflies, moths and other insect herbivores often discriminate between plants of the same species, depositing eggs on ‘high-quality’ plants that enhance offspring performance and consequently lead to higher potential population growth (Firempong & Zalucki 1990; Zalucki, Brower & Malcolm 1990; Zalucki & Brower 1992; Zalucki 1993). In contrast, other studies have found no evidence for higher egg densities on high-quality hosts (Karban & Courtney 1987) and show that host-plant ‘apparency’ (which influences the likelihood of females locating a potential host) has more effect on egg-laying patterns in a habitat than host-plant quality (Rausher 1979; Courtney 1981, 1982; Soberón et al. 1988). This being so, the interaction between spatial variation in host quality and host apparency in a habitat would be expected to have important effects on herbivore dynamics, including the frequency of pest outbreaks and local extinctions of rare species.

Although previous studies have shown the effects of either plant quality or plant apparency on insect egg-laying behaviour in a particular habitat at a particular time (Rausher 1979; Zalucki, Brower & Malcolm 1990), the relative frequencies of apparent and high-quality plants may differ between habitats and may change over time, as habitat structure and host-plant attributes change with management regimes and habitat succession. In the present study, different habitat types, ranging from young single-species stands of acacia to mature multispecies habitats, were used to analyse the effect of habitat structure on the spatial distribution of eggs of the processionary caterpillar Ochrogaster lunifer Herrich-Schäffer on acacias trees from one year to the next. The subsequent effects of different egg batch distributions on larval survival were then used to discuss the long-term implications for conservation and pest management.

The thaumetopoeid moth O. lunifer is a common univoltine species in coastal Australia, ranging from temperate to tropical regions (Floater 1996a). In natural mixed-species habitats, outbreaks of the processionary larvae are frequent on a small proportion of individual acacias, while the majority of trees suffer little damage. However, in acacia plantations on post-mining regeneration sites, outbreaks can occur across much larger areas, leaving a high proportion of trees defoliated. Not only does this affect the long-term stability of caterpillar populations, but it also reduces the overall diversity of insects that would otherwise utilize the trees’ resources. In general, regeneration sites support much lower levels of biodiversity than undisturbed habitats.

Defoliation of acacias by processionary caterpillars towards the end of the wet season (April–May) is related to the number of egg batches deposited on the trunk of a host-tree at the beginning of the season (October–November). The number of batches on a tree represents the number of females that have deposited eggs, as each female produces only a single batch of 150–550 eggs in her lifetime (Floater 1996a). The distribution of egg batches is spatially clumped, and particular trees in natural habitats receive eggs year after year (and may be defoliated with high larval mortality from one year to the next), while neighbouring trees receive none. The spatial arrangement of eggs at the beginning of the season therefore has important consequences for long-term caterpillar dynamics, host-tree health and overall diversity of other insect herbivores in the habitat. For 2 years, the interaction between host-plant quality and host-plant apparency, and its effect on the spatial pattern of egg-laying in O. lunifer, was investigated over time in different habitat types, in order to investigate methods of manipulating egg distributions with different management strategies.

The study system

While O. lunifer will deposit eggs and feed on several species of phyllodinous acacia (Floater 1996a), most populations in subtropical south-east Queensland, Australia, are effectively monophagous on Acacia concurrens Pedley, which grows in exclusive stands in many parts of the region. The second most abundant host-tree in south-east Queensland is A. aulacocarpa Cunn. ex Bentham, and this species also grows exclusively in some areas (Floater 1996a). As acacia phyllodes perform the same function as leaves in other plant groups, phyllode and leaf are used interchangeably in the text. Both species of acacia begin to produce flush growth in October–November with the first rains of the season.

Adults of O. lunifer are on the wing in October–November. They possess no functional mouth-parts, and females live only for 2 or 3 days. After mating, the female flies to a host-tree and tests its suitability by touching outer leaves of the canopy in 10–12 places before walking down the trunk to deposit a single cluster of 150–550 eggs at the base, covering the eggs with a thick white mat of deciduous scales from her anal tuft. The scales provide protection for the eggs and early instar larvae (instar I–IV). The female dies shortly after depositing her single egg batch. The eggs give rise to a cohort of gregarious larvae, which initiate feeding in instar II in December (instar I larvae do not feed). Larvae are central patch foragers, moving up into the canopy to feed on the outer leaves for a few hours each day before returning to the base of the tree. In general, early stage larvae feed on flush leaves formed after the first rain in November, although they will feed and can survive on senescent leaves (Floater 1997) which may occur on acacias growing in poor conditions or in low-rainfall areas. The later instars feed at night. The larvae moult synchronously in a silk nest at the base of the tree, where the exuviae remain intact. The larvae grow geometrically, and the exuviae can be used to assess the growth and survivorship rates of larvae with a high level of accuracy (Floater 1996b). The larvae remain gregarious throughout the larval stage (instars I–VIII) and, unless the cohort defoliates its host-tree, the larvae remain at the base of the tree until the end of larval development. If the tree is defoliated, the larvae leave the tree in a single-file procession in search of a secondary host-tree. Larvae from different clutches usually amalgamate on the same tree to form one large cohort. Final instar larvae disperse in May, travelling up to 200 m from the host-tree, and undergo prepupal diapause underground during the dry season. As the distance travelled by dispersing caterpillars is generally several times greater than the distance between neighbouring host-trees, and females are strong flyers, consistent deposition of eggs on the same trees year after year is not due to females emerging next to the host-tree on which they fed as larvae. Pupation occurs a few weeks before the adults emerge.

Ochrogaster lunifer has been known as Teara contraria Walker and O. contraria (Walker) (Froggatt 1923; Mills 1950, 1951a,b; Jenkins 1962; Common 1970; McFarland 1979; Common 1990), and the species name is often used to describe the bag-shelter moth, which is probably a separate species (Floater 1996a). The present paper deals exclusively with the coastal ground-nesting O. lunifer. Ochrogaster lunifer is also related to the pine processionary caterpillar Thaumetopoea pityocampa (Den. & Schiff.), a serious pest of pine plantations in southern Europe with similar life-history habits (Breuer & Devkota 1990). More details of the biology and life history of O. lunifer can be found in Floater (1996a,b,c).

Methods

Field surveys

In November 1993, surveys of O. lunifer eggs were conducted at 21 localities in south-east Queensland. Between 26 and 150 trees were sampled at each locality; a map of localities is available elsewhere (Floater 1996a). At 19 localities the dominant host-plant present in the area sampled was A. concurrens. Stands of A. aulacocarpa were sampled at three localities (BC23, BC4 and N1). Localities were grouped into early, intermediate and late successional habitats, depending on vegetation structure and the approximate age of acacias at the site. At each locality, host-trees were tagged, and for each tree a record was made of the number and size of egg batches. At 18 localities various tree attributes were measured. These included tree size (measured as trunk diameter), exposure of the base of the trunk, and the amount of plant cover surrounding each tree. Trunk exposure was recorded as bare (no undergrowth present) or sheltered (undergrowth present). Two measures of plant cover were recorded: (i) the number of woody perennials within 3 m of the sample tree (index I), and (ii) the number of directions (NSEW) in which these plants were growing around the tree (index II); thus 0 = no plants and 4 = plants on all four sides of the sample tree. In November 1994, 14 sites were revisited and the number and size of egg batches were recorded for each tree.

Analysis of the presence/absence of eggs on host-trees

Logistic regression, a form of generalized linear modelling, was used to test the relationship between each of the four tree attributes measured in 1993 and the presence/absence of eggs (McCullagh & Nelder 1989; Sokal & Rohlf 1995). The model used was of the form:

image(eqn 1)

where p is the probability of a tree having eggs; αi is the parameter for continuous variable xi (including tree size and both indices of cover); βjk is the constant for the kth level of the nominal variable βj (including presence/absence of nest material and trunk exposure); and ε is the residual error. The observed data used to fit the model were the binary presence/absence counts of egg batches. G-values (log-likelihood ratio statistics approximately distributed as chi-squared; Sokal & Rohlf 1995) were used to test the significance of the effect of each tree attribute on the presence of eggs.

Nutrient analysis of host-trees

In November 1994, leaves were removed from 75 trees of A. aulacocarpa at Toowong, Brisbane, to determine the nitrogen and water content of tree foliage in the outer canopy. Females test the outer canopy before depositing eggs, and the early instar larvae feed in the outer canopy (regardless of whether the leaves are flush or not). Using long-handled shears, 10 shoots were removed from the outer canopy of each tree. Shoots were sampled from all sides of the tree and from the upper and lower canopy. From each shoot, the leaf nearest the apex was removed, sealed in a plastic bag and placed in a cool-box for storage. The pooled leaves from each tree were weighed on the day of collection, dried and reweighed to estimate the percentage water content of fresh leaves. A similar survey was conducted at Lytton, with 81 trees of A. concurrens sampled.

The percentage nitrogen content of dry leaves was measured using a Leco FP428 rapid nitrogen determinator. To test the variability of nitrogen content among leaves on the same tree, a test sample was taken from each of the 10 leaves from 10 trees. Variation in foliar nitrogen was lower within trees (mean coefficient of variation V = 9·7; n = 10) than across trees (V = 13·1; n = 10). For each remaining tree, all 10 leaves were ground together and a single sample taken to estimate the nitrogen content of leaves for that tree.

Analyses of covariance (ancova) were used to test differences in foliar chemistry between trees with and without eggs in 1994. For tests of water levels, percentage nitrogen content was used as the covariate. For tests of nitrogen levels, percentage water content was used as the covariate.

Nest manipulation

In order to test whether larval nest material from the previous generation influenced the likelihood of eggs being laid on individual host-trees, old nest material was manipulated at two sites (P1 and P3) on North Stradbroke Island in early October. Each site had a relatively large number of trees with nest material from the previous year. Four treatments were established. Trees with nest material from the previous year received one of two treatments: (i) nest material was removed from the base of the tree, which was left bare; (ii) nest material was removed and replaced with nest material from a tree in treatment (i). Trees without nest material from the previous year also received one of two treatments: (iii) nest material was introduced at the base of the tree; (iv) trees were left bare as a control. The number of trees in treatments (i), (ii), (iii) and (iv) was 8, 8, 14 and 18 at site P1, and 7, 8, 11 and 13 at P3. A similar experiment was conducted at a third site (BC4) in October 1995. Because the number of trees with nests was small at this site, treatment (ii) was not included. The number of trees in treatments (i), (iii) and (iv) was 12, 22 and 20, respectively. The presence/absence of eggs was assessed at each site in November 1994.

Larval performance

In November 1993, 15 egg batches were collected from a site on North Stradbroke Island. This site was not part of the main survey. Before each egg batch was removed from the tree, the size of the scale mass around the eggs was measured. The egg mass is roughly cone-shaped with an elliptical base on the trunk of the tree. Three measurements were taken: d1 (the major axis of the base), d2 (the minor axis of the base) and h (the height of the cone perpendicular to the base). After collection, the number of eggs in each batch was recorded. A plot of egg number, y, against the volume of the scale mass, x (calculated as the volume of a cone), gave the relationship y = 0·08x + 98·007 (r2 = 0·63; P = 0·0004).

The regression equation was used to estimate clutch size (i.e. initial cohort size) of batches measured at localities in the main survey conducted in November 1993. In January 1994, localities were revisited. Each tree that had eggs in November was searched, and the number of extinct and extant cohorts at the second larval instar was recorded. Cohorts that went extinct before instar II (when larvae start to feed), or shortly afterwards as a result of predation, were not included in the analysis of larval survival, as these mortalities were due to factors unrelated to host-plant quality (Floater 1996a).

In June 1994, a third survey was conducted. At 12 of the 22 localities, nest material, including frass, exuviae and cadavers, was collected from the base of each tree, placed into a plastic bag and removed for examination. In order to estimate the number of larvae entering the final instar in each nest, the number of instar VII exuviae in each nest was recorded (Floater 1996b). The proportion of larvae in a cohort surviving to the final larval instar was calculated as the number of instar VII exuviae divided by the number of eggs in the original batch. As the number of batches on a tree affects larval performance (authors’ unpublished data), only trees with a single egg batch were used in analyses.

Head capsule widths of instar VII exuviae were used as a measure of larval growth. The head capsules of all instar VII exuviae present in the nest were measured, and the mean head capsule width calculated. Head capsule width is not only a reliable measure of size within instars of a cohort, it is a useful measure of comparison between cohorts (Floater 1996b). The mean head size was used to compare larval growth on different trees.

Results

Consistent patterns of egg-laying in successive years

Clutch size ranged from 140 to 555 eggs across all sites and years. Mean clutch size per site ranged from 253·0 ± 71·4 SD at site TH12 to 485·2 ± 93·0 at site MC12. The mean clutch size per site (measured as the mean of means) was 333·1 in 1993 and 296·5 in 1994. In particular habitats, a strong positive association was found between the presence of eggs on a tree in 1993 and the presence of eggs on the same tree in 1994, confirming that eggs are consistently deposited on the same trees in successive years (Table 1). Of the 14 sites surveyed in both years, 13 (93%) showed a positive association, eight (57%) of which were significant.

Table 1.  Association of oviposition patterns in successive years. The sign of the phi (φ) coefficient indicates the direction of the association: a positive value indicates that oviposition is positively associated with oviposition in the previous year (Sokal & Rohlf 1995)
Number of trees
Sitesampledeggs (1993)eggs (1994)Total
(both years)
Number with
φ
Number with
χ2
Number with eggs
Significance
  • *

    P < 0·05;

  • **

    P < 0·01;

  • ***

    P < 0·001; NS, not significant.

BC2378680−0·1000·74NS
BC4261821170·5217·04**
D1506330·68423·40***
DH142151080·52011·20***
G65014310·0320·05NS
J15015320·2032·04NS
J252221180·3195·29*
L1231502931140·49316·71***
MN1449730·2412·57NS
N150101580·54614·88***
TF412545220·61420·35***
TH129681760·45319·66***
TH45091340·1981·94NS
TI12282218160·3383·19NS

Host-tree characteristics and the presence/absence of eggs

Cover around the tree was the most consistent factor governing the presence of egg batches in 1993. At least one index of cover was associated with the presence of eggs at six (54%) out of 11 sites at a 5% level of significance. Furthermore, a total of eight (73%) out of 11 sites showed a negative relationship between the amount of cover around hosts and the presence of eggs at the 10% level (Table 2). At no site was there a positive relationship at the 10% level. Tree size was associated with the presence of eggs at six (33%) of 18 sites at the 5% level, and eight (44%) of 18 sites at the 10% level. In all but one of these cases, trees of larger size had a greater probability of receiving eggs. At BC4, eggs were deposited on smaller trees, although these trees were larger than the largest trees at other localities. There were several very large senescent trees at BC4 that did not receive eggs. The presence of eggs was related to trunk exposure at only one (6%) out of 16 sites at the 5% level (Table 2).

Table 2.  Results of logistic regressions showing effects of different host-tree attributes (trunk exposure, cover surrounding the tree and tree size) on the presence/absence of egg batches. The relationship between the dependent variable and independent variables is shown to be positive or negative with a (+) or (–) sign, respectively. A missing value indicates an attribute that was not measured at a particular locality
G-values of different tree attributes
Cover (index 1)Cover (index 2)Tree sizeSite
Trunk exposure
  • P < 0·1;

  • *

    P < 0·05;

  • **

    P < 0·001;

  • ***

    P < 0·0001.

BC43·50(–)1·375·75**(–)
BR10·091·29
D10·280·896·10*(–)0·01
DH10·10002·73 (+)
G61·3131·81***(+)
J10·290·030·400·67
J22·1619·00***(+)
L1234·10*0·20
MC120·3111·73**(–)06·34*(+)
MN10·080·144·55*(–)0·20
N10·3510·83**(–)4·69*(–)0·84
P12·90 (–)5·96*(–)0·94
P23·14 1·61
P30·213·04 (–)0·610·06
TH121·517·12*(–)0·250·73
TI120·040·010·4321·25***(+)
WW10·053·64 (+)
Y11·144·32*(+)

Foliar chemistry and the presence/absence of eggs

At the Lytton site, the size of A. concurrens trees had no effect on foliar water levels (r2 = 0·03; P = 0·315) or nitrogen levels (r2 = 0·04; P = 0·247), while nitrogen was positively correlated with water content (ρ = 0·44; P < 0·001). In contrast, no correlation between foliar water and nitrogen was found for A. aulacocarpa at Toowong (ρ = −0·11; P = 0·342).

Foliar water content of trees ranged from 43·4% to 62·7% (mean = 51·9 ± 4·0%) for A. aulacocarpa at Toowong, and from 52·0% to 69·3% (mean = 59·4 ± 4·1%) for A. concurrens at Lytton. For both tree species, high foliar water levels significantly increased the likelihood of trees receiving eggs in 1994 (Toowong: ancovaF1,72 = 4·02; P = 0·049; Lytton: F1,71 = 4·39; P = 0·040). Leaf nitrogen concentration of trees ranged from 1·4% to 3·6% (mean = 2·08 ± 0·29) at Toowong and 1·8% to 3·5% (mean = 2·35 ± 2·9%) at Lytton. Foliar nitrogen of A. aulacocarpa at Toowong had no effect on the presence of eggs (ancovaF1,71 = 0·19; P = 0·663). At Lytton, nitrogen had a significant effect, with A. concurrens trees that were low in nitrogen receiving more eggs (F1,70 = 11·5; P = 0·001).

The effect of previous larval attack on egg-laying

The presence of larval nest material on trees did not increase the likelihood of eggs being laid. Indeed, at the Toowong site, those trees from which nest material was removed received significantly more egg batches than expected, suggesting that those trees received more eggs for reasons other than the presence of nest material (χ2 = 19·29; d.f. = 5; P < 0·01). These results were supported at the experimental sites on Stradbroke Island (P1 and P3), where the presence of eggs was not significantly related to the presence or absence of nest material (Site P1: χ2 = 6·02; d.f. = 7; P > 0·05; Site P2: χ2 = 3·56; d.f. = 7; P > 0·05).

In order to test whether egg deposition was related to larval attack in the previous year, oviposition was compared between trees in three groups: (i) trees that did not receive eggs in the previous year; (ii) trees with eggs the previous year, but with no larvae because the eggs were eaten by predators; and (iii) trees with eggs and larvae in the previous year. At three of five localities with sufficient data, trees in group (ii), having eggs in 1993 that did not develop into larvae, received significantly more egg batches than trees in group (i), which had no eggs in 1993 (Table 3). A fourth locality (J2) gave a similar result that was almost significant at the 5% level (χ2 = 3·72; P = 0·054). In contrast, at no locality did trees with larvae in 1993 (group iii) receive significantly more egg batches than trees in group (ii) at the 5% level (Table 3). Both these results show that factors other than larval attack make these trees more likely to receive eggs over successive years.

Table 3.  Comparison of oviposition in 1994 on trees with different oviposition histories in 1993 at five localities in south-east Queensland. Numbers refer to the number of trees in each of three groups (with the number of trees receiving egg batches in 1994 shown in parentheses). If larval attack in the previous year increases the likelihood of oviposition, trees in groups (i) and (ii) should receive the same number of batches, while trees in group (iii) should receive more batches. In contrast, if particular trees receive eggs in successive years due to factors other than larval attack, trees in groups (ii) and (iii) should receive the same number of batches while trees in group (i) should receive significantly less
History of trees in 1993
Locality

χ2
(ii) Eggs no larvaeχ2(iii) Eggs and larvae (i) No eggs
  • *

    P < 0·05;

  • **

    P < 0·01;

  • NS

    NS, not significant.

BC48 (4)5·32*14 (13)0·84NS4 (4)
DH127 (2)6·93**6 (3)0·05NS9 (5)
J230 (3)3·72NS15 (5)0·19NS7 (3)
L123121 (17)7·59**5 (3)0·33NS24 (11)
TI126 (2)0·70NS11 (6)3·67NS11 (10)

At Toowong, larval attack in the first year had no effect on foliar nitrogen or water content of A. aulacocarpa in the following year (Table 4). At Lytton, larval attack had no effect on foliar water content of A. concurrens, although there was a significant relationship between previous larval attack and higher nitrogen levels.

Table 4.  Relationship between previous larval attack and water content (% fresh weight) and nitrogen content (% dry weight) of A. aulacocarpa at Toowong and A. concurrens at Lytton in spring 1994. To reduce confounding variables, only trees with eggs in 1993 were included in analyses (trees with eggs but without larval attack were those on which eggs were destroyed in the egg and first instar stages, before larval feeding began)
% water (1994)% nitrogen (1994)
nMeanSDP-valueMeanSDP-value
Acacia aulacocarpa
Trees with larval attack (1993)452·503·18} 0·7872·240·23} 0·724
Trees without larval attack (1993)2052·024·08 2·050·29 
Acacia concurrens
Trees with larval attack (1993)1958·904·14} 0·6642·330·40} 0·044
Trees without larval attack (1993)458·684·38 2·080·19 

Host-tree quality and larval performance

The effect of tree size on larval survival and development was analysed for (i) cohorts at the Lytton site (L123) where 150 trees were sampled, and (ii) cohorts across all sites, in order to investigate the consistency of relationships across spatial scales. At Lytton, the proportion of larvae surviving on a tree increased with tree size (G = 108·4; d.f. = 1; P < 0·0001). A similar result was found for cohorts across all sites (G = 778; d.f. = 1; P < 0·0001). The size of instar VII larvae increased with tree size both at Lytton (r2 = 0·53; P = 0·008) and across all sites (r2 = 0·43; P < 0·0001). Across all sites, the amount of cover around a tree had no effect on larval survival (r2 = 0·005; P = 0·821) or larval development (r2 = 0·08; P = 0·262).

No significant relationships were found between leaf quality and larval performance at Toowong or Lytton, although little data were available for analysis. Even when data were combined (n = 10), foliar nitrogen had no significant effect on larval survival (r2 = 0·15; P = 0·441) or larval size (r2 = 0·01; P = 0·795). Similarly, foliar water content had no effect on larval survival (r2 = 0·06; P = 0·649). A positive relationship approaching significance was present between leaf water content and larval growth (r2 = 0·36; P = 0·067), suggesting that increased foliar water content may enhance larval growth. There was no relationship between tree size and plant quality at Toowong (foliar nitrogen: r2 = 0·01; P = 0·622; foliar water: r2 = 0·03; P = 0·301) or Lytton (foliar nitrogen: r2 = 0·04; P = 0·247; foliar water: r2 = 0·03; P = 0·315).

Discussion

Spatial and temporal patterns of egg-laying in O. lunifer that lead to particular patterns of larval attack are due to the interaction between host-tree ‘apparency’ (the relative conspicuousness of host-trees in the habitat to adult females) and host-tree ‘quality’ (host-tree characteristics such as size and foliar chemistry that influence larval performance and may be used by adult females to discriminate between potential host-trees). The most significant factors found to affect egg batch distributions within a habitat were variation in cover around host-trees, host-tree size and foliar water content. In order to have influenced egg-laying patterns, these host-tree attributes must either affect the apparency of host-trees (Feeny 1976; Wiklund 1984) and pre-alighting discrimination, or influence egg-laying decisions by the female once the host-tree has been located.

Vegetative cover around acacias and egg-laying patterns

The canopy exposure of host-trees had no effect on offspring survival or growth, and it appears that exposed trees receive more egg batches simply because there is a greater probability of females locating them (Cromartie 1975; Dempster & Hall 1980; Soberón et al. 1988). Wiklund (1984) showed that butterfly species that use visually non-apparent host-plants frequently alight on non-host-plants during their search, while those species that use visually apparent host-plants seldom alight on non-host-plants (Prokopy & Owens 1983; Firempong & Zalucki 1990). This suggests that insects such as many Lepidoptera, which use visual cues to locate potential host-plants, reduce host-plant searching time when host-plants (species and individuals) are apparent. Adult females of O. lunifer fly in the late afternoon and dusk, and probably use visual cues while searching for host-plants. Furthermore, the adults do not feed and live for just a few days. Consequently host-plant searching time may be an important factor affecting the deposition of eggs on apparent trees. Alternatively, by laying eggs on exposed trees, females may reduce the risk of their own mortality (e.g. by web-spinning spiders; cf. Rausher 1979) before the act of egg-laying. In either case, increased cover around host-trees reduces the probability of egg-laying by females, leading to high egg densities on a few exposed trees, while many high-quality host-trees evade herbivore attack altogether.

Acacia size and egg-laying patterns

Unlike tree cover, tree size was a component of host quality, having a significant effect both on larval survival and on larval growth. Leaf quality can differ significantly between plants of different age (Bowers & Stamp 1993; Stamp & Bowers 1994) or growing under different conditions (Floater 1997). However, no relationship between acacia size and leaf chemistry was found in the field. One possible reason for enhanced larval performance on larger trees is the quantity and spatial arrangement of leaves in the canopy. In all feeding instars (II–VIII), caterpillars feed in a tight aggregation, with individuals seldom losing physical contact with others in the cohort (Floater 1996a). Consequently the feeding area for the cohort at any one time is relatively small. As the density of leaves within the feeding area will affect the amount of time spent searching for leaves during a feeding sortie, the spatial distribution of leaves in the canopy probably has a major influence on larval development and survival.

Older host-trees support a greater density of shoots in the periphery of the canopy until they become senescent, such as the largest individuals at site BC4 (G.J. Floater & M.P. Zalucki, personal observation). Consequently, large trees can provide a greater density of food during a feeding bout. Furthermore, the feeding area includes a greater quantity of flush growth, which is important for enhanced growth and survival in the early instars (Floater 1997). As well as having a greater density of leaves at the edge of the canopy, larger trees have a greater overall quantity of leaves. This reduces the likelihood of the larval cohort defoliating the entire tree and having to disperse to a secondary host-tree, a process that carries a high risk of mortality (Floater 1996c). Other species of Lepidoptera tend to lay eggs on large host-plants (Ives 1978; Myers, Monro & Murray 1981; Zalucki & Kitching 1982; Soberón et al. 1988) but whether females discriminate between host-trees of different size, or simply locate larger host-trees more often due to their visual apparency, is difficult to deduce without detailed studies of female searching behaviour (Oyeyele & Zalucki 1990; Zangerl & Berenbaum 1992). Either way, the high host quality of large trees is not exploited to its full potential by O. lunifer when a large amount of vegetative cover exists around host-trees.

Foliar chemistry of acacias and egg-laying patterns

Along with tree size, foliar water levels of host-plants can affect egg-laying patterns and larval growth and survival. Early instar larvae of O. lunifer suffered strong reductions in survival and growth under greenhouse conditions when reared on A. concurrens saplings with comparatively low levels of foliar water (Floater 1997). Larval performance was also reduced when cohorts were reared on senescent rather than flush leaves. Foliar nitrogen, on the other hand, did not affect larval survival (Floater 1997). While our field results showed no effect of foliar water content on larval growth or survival among trees on which eggs were deposited, these trees had relatively high foliar water levels compared to trees without eggs, suggesting a link between host-plant discrimination by females and subsequent larval performance. Foliar water may also affect the apparency of host-plants through its association with chemical stimuli. Females were observed to approach potential host-trees from downwind, suggesting that chemical stimuli may be used to locate suitable host-trees. Approaching a shoot, the female touches the edge of a leaf with her legs and, still in hovering flight, gradually moves down its length. She repeats the process at 8–12 shoots and may then accept the host-tree for egg-laying or reject it and leave.

Previous herbivore attack and subsequent egg-laying patterns

While other studies have shown an association between host-plant damage and subsequent herbivore attack, either positively (Danell & Huss-Danell 1985; Harrison & Karban 1986; Pullin 1986, 1987; Hunter 1987; Craig, Itami & Price 1990) or negatively (Bergelson, Fowler & Hartley 1986; Leather, Watt & Forrest 1987; Silkstone 1987; Tscharntke 1989; Kouki 1991), previous larval attack by O. lunifer does not appear to affect egg-laying decisions by adult females. Van Schagen, Hobbs & Majer (1992) showed that larvae of the bag-shelter moth (a sibling species of O. lunifer;Common 1990; Floater 1996a), feeding on shoots of A. acuminata in late summer, induced a significantly greater flush growth on those shoots in the following spring. However, in the present study no overall difference in the quality of spring growth (of either A. concurrens or A. aulacocarpa) was found between trees with previous larval feeding (some of which had been entirely defoliated) and trees without. Consequently, herbivore attack on the same individual trees year after year appears to be the result of the spatial interaction of host-tree quality and host-tree apparency that changes little over time, rather than a direct temporal link between annual herbivore attacks.

Habitat structure and spatial outbreaks

If the interaction between spatial variation in host-plant apparency and host-plant quality changes over time within a habitat, the management regime or successional stage of a habitat should have a significant effect on population growth and the probability of outbreaks at different spatial scales. At the level of the tree both high host quality and high host apparency increase the probability of eggs being laid, and each will therefore lead to higher egg loads. The interaction between high host quality and high apparency will further increase egg loads, with trees that are both high-quality and highly conspicuous receiving a disproportionately higher number of egg batches. The frequency of defoliated trees should therefore be higher in habitats that have high variation in tree quality and apparency. At the site level, high variation in host-tree quality and apparency will result in a small proportion of trees being heavily attacked, while a large proportion of trees will suffer no, or little, damage. Also, the high mortality of larvae on a small proportion of defoliated trees will significantly reduce population growth across generations. Furthermore, as the degree of variation in apparency increases, the number of eggs and larvae on high-quality host-trees will decrease, reducing the potential for site-wide outbreaks. In contrast, in habitats with little interspersed vegetation, and little variation in the size of trees, egg distributions and larval damage will be more even, and complete defoliation of individual trees will be rare until population levels are high enough to cause a site-wide outbreak with a very high proportion of trees in the habitat being heavily attacked.

In mature mixed-species habitats, habitat structure is heterogeneous but changes little from year to year, and the amount of cover surrounding a potential host-tree is the most important factor determining the likelihood of receiving eggs (Table 5). In these heterogeneous habitats, eggs were consistently deposited on exposed trees year after year, even though the size range of acacia hosts in these habitats was relatively high. In contrast, in relatively homogeneous early successional habitats, there is less overall cover, and a greater proportion of trees is relatively conspicuous. Consequently, moths have more opportunity to discriminate between trees of different size and foliar chemistry, leading to enhanced larval growth and survival and thereby increasing the probability of population establishment in these new habitats. In these relatively open homogeneous habitats, females deposited eggs on significantly larger trees, even though the range of tree size was relatively narrow compared with that in more mature habitats (Table 5).

Table 5.  Habitat attributes (host size and surrounding cover) that increase the likelihood of particular trees receiving eggs within and between years. A missing value indicates that a host attribute or oviposition was not measured at a particular locality
LocalitySuccessional
stage of habitat
(1993–94)
Tree attribute
affecting oviposition
(1993)
Consistency
of oviposition
(1993–94)
Size range
of trees
(mm)
  • P < 0·1;

  • *

    P < 0·05;

  • **

    P < 0·001;

  • ***

    P < 0·0001.

G6EarlySize***No62
J2EarlySize***Yes63
J1EarlyNeitherNo91
Y1EarlySize*111
WW1IntermediateSize165
DH1IntermediateSizeYes78
L123IntermediateYes295
MC12IntermediateCover** and size*197
TH12IntermediateCover*Yes253
BR1Intermediate272
N1LateCover**Yes126
P1LateCover*220
P2Late232
MN1LateCover*No249
TI12LateSize***Yes268
P3LateCover350
D1LateCover*Yes385
BC4LateCover and size*Yes520

While habitat structure and tree size are relatively dynamic in early successional habitats, at two localities (J2 and DH1) where females tended to lay eggs on large trees in 1993, the same trees received egg batches in the following year, suggesting that, at least in some young habitats, the size of host-trees relative to one another remains fairly constant across years. Foliar quality of individual acacias may also remain relatively unchanged. At Toowong, individuals of A. aulacocarpa with the highest levels of foliar water in 1994 were those that received eggs in 1993 and 1994, while trees with the lowest levels did not receive eggs in either year.

Although consistent variation in host-tree apparency and host-tree quality can both result in consistent egg distributions across generations, these patterns may have very different consequences for herbivore abundance. The proportion of performance-enhancing host-trees that receive eggs depends on the frequency of trees that are both high-quality and highly apparent. In young habitats, most trees are small, and therefore relatively poor for larval performance, but because most trees are conspicuous females can exercise a high degree of host-plant discrimination and so deposit eggs on the best host-plants for offspring performance. In older habitats, most trees are large, and therefore relatively good for larval performance. However, the closed structure of the habitat reduces the number of conspicuous host-trees amongst which females can choose. Consequently, females are forced to lay eggs on small trees with low foliar water levels and/or little flush growth (especially in very old habitats where trees are becoming senescent), so leading to reduced offspring performance and reduced population growth.

Consequences for the management of ecological reserves and forestry plantations

An understanding of the interaction between host-plant quality and host-plant apparency on insect egg-laying behaviour can be used effectively in strategies for managing natural habitats and plantations for conservation and pest control. Site-wide outbreaks of processionary caterpillars are relatively frequent on post-mining regeneration plantations of acacia in south-east Queensland. This can be explained by the regular structure of the plantations and its effect on the host quality–host apparency interaction. In these artificial habitats, all trees are relatively similar in size and grow in regular, single-species rows, so that the amount of cover afforded by other acacias is similar for all trees. Consequently, the distribution of eggs is more even (G.J. Floater, unpublished data), leading to few defoliated trees in most years. However, as the population increases unchecked, the likelihood of a large-scale outbreak is much greater. This type of problem is often as acute in forestry plantations as in regeneration plantations. In the former, the costs to timber production from large-scale pest outbreaks can be high, while in the latter, the process of regenerating high levels of biodiversity can be slowed considerably, or prevented altogether, by the unstable population dynamics of immigrating or reintroduced species.

In both situations, the plantation, acting as a uniform habitat, does little to disrupt the host-searching behaviour of insects, leading to egg distributions that enhance population growth and can result in outbreaks. In order to reduce the destabilizing effects of high insect population growth rates, mixed planting can be used to disrupt the host-searching efficiency of potential insect pests. Although the spatial mechanism for outbreaks in monocultures and diverse habitats is poorly understood, it is well known in some areas of agriculture that mixed planting can reduce the density of pest species and insect damage on crops (Kenny & Chapman 1988; Dissemond & Hindorf 1990; Karel 1991; Skovgaard & Paets 1997). However, the economic gains from reduced insect damage may be offset by lower yields from mixed crops. Gold, Altieri & Bellotti (1990) have suggested that intercropping may reduce insect attack by reducing the size of cassava host-plants rather than interfering with the dispersal of pests, although detailed studies of the spatial distribution of pests and its relationship with the quality–appareny interaction have not been conducted for this system. Whatever the particular interaction for a particular crop system, economic gains and losses from intercropping must always be weighed. More experiments to investigate the effect of different levels of crop diversity on pest outbreaks are urgently required, particularly for forestry plantations, to test economic advantages and disadvantages of intercropping.

While mixed planting may have associated negative effects for crop yield, the use of mixed planting in areas of habitat regeneration has many associated benefits. Not only should mixed planting reduce the instability of populations that make up the emerging animal community, but a larger variety of host-plants will increase the number of herbivore and plant-associated species in the community. If the plant species chosen for regeneration are naturally occurring in the area, mixed planting should accelerate the regeneration process. Furthermore, rather than planting similar-sized plants that lead to greater uniformity of resources, individuals and organizations involved in the regeneration process should be encouraged to vary the size of introduced plants so as to increase host-plant heterogeneity and consequently the spatial heterogeneity of herbivores. Different host-plant sizes in the habitat also has the advantage of creating a variety of different microhabitats and resource quality; some herbivore species tend to feed on young plants while others develop more readily on older plants with different chemical and physical attributes. While planting large trees on regeneration sites is generally unfeasible, even a modest range of seedling and sapling sizes during initial planting should accelerate the process of regenerating animal communities.

With regard to the conservation of single species, different practices of habitat management will need to be employed depending on the goals of a particular project. For example, if a project is concerned with saving a declining population of butterflies from extinction, the overall project goal will be to increase population numbers as rapidly as possible. This being so, the project team could increase the conspicuousness of host-plants by introducing more plants into the habitat and placing them in areas of the habitat where females are most likely to locate them. A knowledge of the dispersal behaviour of adult butterflies would help in this regard. While increasing overall plant apparency, overall plant quality can also be increased by manipulating plant characteristics that enhance larval survival. Such characteristics may include plant size, foliar chemistry and the amount of shelter that the plant provides against predators, parasitoids and adverse weather conditions. Host-plants may be fertilized, allowed to grow without competition from other plant species, and provided with open netting that allows eggs to be laid but reduces predation.

On the other hand, if a project is concerned with the conservation of a stable, established population of butterflies that is perhaps endemic to an area, or the numbers of which have been increased using the management techniques outlined above, the overall project goal will be to maintain population stability rather than increase numbers rapidly, which may lead to unstable dynamics, resource depletion and a high probability of extinction. In this case, if the plant community is sufficiently diverse to maintain stable population dynamics, little habitat management may be required. However, as in all cases of conservation management, constant monitoring of the distribution and abundance of the species within the reserve should be encouraged.

Acknowledgements

We would like to thank Kelli Gray and Anthony O’Toole for their help with nitrogen analyses, and Joan Hendrikz for her advice on statistical procedures. Thanks also to David Walter, Gimme Walter and Miguel Franco, who made valuable comments on an earlier draft. The research was funded by a scholarship awarded to G. J. Floater by the Association of Commonwealth Universities under the Commonwealth Scholarship and Fellowship Plan (CSFP).

Received 30 December 1998; revision received 21 September 1999

Ancillary