Ir C.M.M. Mols, Netherlands Institute of Ecology, Centre of Terrestrial Ecology, PO Box 40, 6666 ZG Heteren, the Netherlands (fax +31 26 4723227; e-mail email@example.com).
1The potential contribution of vertebrate predators to biological control in orchards has been largely overlooked to date. A few studies have shown that birds reduce numbers of pests, but data are scarce on the effects on the pattern or timing of damage. Consequently, the practical value of birds as biocontrol agents remains unclear.
2This study considered whether great tits Parus major can reduce caterpillar numbers and fruit damage by caterpillars, and increase biological yield, in an experimental orchard of apple trees with high caterpillar numbers. The outcome would depend on the coincidence of the period during which great tits forage and the period during which caterpillars cause damage. In the first experiment nets were put over trees at different times of the growing season, thus creating different periods during which great tits had access to the trees. In the second experiment caterpillars were removed from trees at different times in the growing season. In both experiments, the resultant caterpillar damage to apples was assessed in the autumn.
3The longer the period of foraging by great tits, from the start of egg incubation until fledging of young, the less the overall pest damage to fruit. Damage caused by caterpillars was greater the later they were removed, from the young apple stage onwards.
4The effect of great tits on caterpillar damage to apples was small (percentage damage was reduced from 13·8% to 11·2%) but significant (P < 0·05), and the yield of fruit increased significantly (from 4·7 to 7·8 kg apples per tree, P < 0·05). The only cost to the producer was that of erecting nest boxes (c. 2 ha−1) to encourage great tits to breed in the orchard. Depending on the great tits’ numeric response to insect densities, their relative impact may be greater at lower densities more typical of commercial orchards and, if so, the presence of breeding great tits may allow economic thresholds for other controls to be reduced. Furthermore, the contribution of natural predators to biological control of insect pests may be especially useful in orchards and in the future when a further reduction of pesticide use may be enforced.
There is increasing interest in biological control of pests in apple Malus domestica Borkh. orchard systems because adverse public attitudes to pesticides have intensified (Solomon et al. 2000), resistance of harmful insects to pesticides is an ongoing problem, and legislation increasingly restricts the use of pesticides (Anonymous 2001). The main focus in the search for pest control agents that can contribute to biological control has been on parasitoids (Cross et al. 1999) and predatory insects such as predatory mites, earwigs, lacewings, mirids and anthocorids (Solomon et al. 2000). The potential contribution of vertebrate predators such as birds is mostly overlooked. The main reason given for ignoring vertebrate predators is their presumed lack of a sufficient numerical response to outbreaks of pests. However, most of the studies on birds as biological pest control agents do show a reduction in the population size of the harmful insect species (Tables 1 and 2; Kirk, Evenden & Mineau 1996), thereby demonstrating their potential. However, because these studies have not investigated whether the insect reduction leads to actual damage reduction to the crop, it remains unclear whether they have practical value.
Table 1. Review of studies that investigated the effect of bird predation on insect populations in agricultural systems
Codling moth Cydia pomonella in apple orchards
Two species of woodpeckers Dendrocopos spp.
– Removal of more than 50% of overwintering codling moth larvae
Leaf-chewing insects in general in a forest dominated by white oak Quercus alba
Forest birds in general
– Exclusion of birds increased leaf damage
In this study we investigated whether great tits Parus major L. can reduce fruit damage inflicted by caterpillars in spring in apple orchards. Great tits are partly insectivorous birds with a preference for caterpillars, especially when they are feeding their nestlings (Betts 1955; Gibb & Betts 1963; Royama 1966; van Balen 1973; Gruys 1982; Naef-Daenzer, Naef-Daenzer & Nager 2000). The nestling rearing period, when the number of prey items caught is high, coincides with the time that caterpillars of winter moths Operophtera brumata L. and tortricid moths occur in orchards. These caterpillars are key pests in apple orchards (Cross et al. 1999; Solomon et al. 2000). The great tit is also a common species that breeds readily in nest boxes, and hence the local density of great tits can be increased easily by putting up nest boxes in orchards. Together, these features suggest that great tits have the potential to contribute to the control of orchard caterpillar pests.
To be effective as biological control agents against caterpillar damage, great tits must remove caterpillars before damage is inflicted on the fruit. Surprisingly, the time at which caterpillars actually inflict damage is not clearly known. It is also unclear when great tits start removing caterpillars in sufficient numbers to reduce fruit damage. It is presumed that their activity during the chick-rearing period is the most effective, but predation during their egg-laying and incubation period might also be important because caterpillars are then removed at an early larval stage. It is also unknown if removal of late larval stages causes a reduction in fruit damage. In this study, we examined whether great tits reduce caterpillar damage, by assessing the periods in which (i) caterpillars cause, and (ii) foraging great tits reduce, fruit damage. The impact great tits have as biological pest control agents depends on the overlap between these periods. This impact was quantified by determining the magnitude of damage reduction.
Materials and methods
Two experiments were carried out in 2000 in the experimental orchard ‘De Schuilenburg’ at Kesteren (51°70′N 5°31′E) in the Netherlands. The apple orchard consisted of 12 blocks of ‘small spindle bush’ trees that were planted in 1984 and 1988. The area contained 36 nest boxes within these blocks (three per block) and 15 boxes in an older part of the orchard on the border of the 1988 blocks. Nest boxes were inspected at least once a week to determine laying date, clutch size, hatching date and the number of young fledged great tits.
Because the majority of insectivorous birds in the orchard were great tits, it was assumed that great tits were the main predators of caterpillars. Other birds occasionally observed in the orchards included blue tits Parus caeruleus L., chaffinches Fringilla coelebs L., goldfinches Carduelis carduelis L., chiffchaffs Phylloscopus collybita Vieillot, willow warblers Phylloscopus trochilus L., magpies Pica pica L., jays Garrulus glandarius L., blackbirds Turdus merula L. and tree sparrows Passer montanus L. Of these birds, blackbirds, magpies and goldfinches only occasionally prey upon caterpillars. Chaffinches, jays, chiffchaffs and willow warblers include caterpillars in their diet but numbers stay below 20% of the total diet. Only tree sparrows and blue tits are known to prey on caterpillars on leaves (Glutz von Blotzheim & Bauer 1997), but less than great tits and foraging mainly at the edges of the orchard near their nesting sites.
Four blocks were selected in the experiments: two blocks planted in 1984, and two blocks planted in 1988. No control measures against caterpillars had been taken in these blocks for at least 4 years. Trees of the common apple variety Elstar were used for the experiment. A double row of this variety occurred twice in each block, while the other varieties (Cox's orange pippin, Belle de Boskoop, Jonagold, Alkmene and Summerred) had at least one double row per block. The blocks of approximately 0·4 ha were 21 or 22 rows wide and about 50 trees long. The distance between trees within a row and between rows was 1·25 m and 3 m, respectively.
In experiment 1 we prevented great tits from foraging on apple trees by covering the trees with polyethylene nets (mesh 25 by 25 mm) at different phases of the growing season and leaving them covered until the end of July. The earliest experimental treatments were made in the early pink bud stage (e2, classification on scales according to Fleckinger 1948) of the apple trees, and the latest treatments were made 2 weeks after the end of the great tit breeding season (26 April until 26 June 2000). The early pink bud stage of the apple trees is the first stage in which caterpillars become visible.
At the start of the experiment 44 trees were covered with nets (Julian date 116, i.e. 1 February is Julian date 32). In the following 8-week period (Julian date 118–174) every Monday, Wednesday and Friday four additional trees were covered, i.e. one per block. A control group of 56 trees was never covered with nets. Great tits had access to the control trees during the whole period (i.e. the control group had Julian date 177 in the analyses). At the end of July (Julian date 205–208) all nets were removed from the experimental trees to prevent reduced growth of, or damage to, the apples.
A similar experimental design was used in experiment 2 but instead of placing nets over trees, trees were searched and caterpillars were removed by hand. These data on caterpillar numbers were used to determine how caterpillar densities changed over time. After searching, trees were sprayed with insecticide (Condor®; 240 g L−1 active ingredient parathion-methyl; Agrevo Nederland BV, Haren, the Netherlands) at 0·1% to ensure the total removal of all caterpillars. The insecticide can also have an effect on aphids and mirids. The insecticide was sprayed with a hand lance until drip off (i.e. the trees were saturated), using approximately 1 litre tree−1. Caterpillars were never removed on a control group of 56 trees (i.e. the control group had Julian date 177 in the analyses). Thus both experiments had their own control group treated identically.
In order to distinguish treatment effects from block effects, all treatments were distributed across all blocks. Further, in order to avoid confounding of treatment effects with the effects of rows, all treatments were similarly distributed across all rows. The treatments were further randomized within rows to balance for carry-over effects so that treatments were not systematically positioned relative to each other in space within blocks and rows.
All abscissed young fruits in June (June-drop) and all full-grown apples (in October) were collected from all trees included in the experiments. June-drop apples were collected to investigate whether damaged apples are more likely to be shed, thereby reducing the treatment effects among harvested apples. All apples collected (June-drop and full-grown) were inspected for caterpillar damage, i.e. early spring damage, as described by De Reede, Gruys & Vaal (1985). The other classes of caterpillar damage (July damage and late summer damage) can be distinguished easily from early spring damage, which is characterized by corked scar tissue not present in other classes of caterpillar damage. Other classes of caterpillar damage were not included in the analyses because they were caused after the time at which all experimental treatments had ended.
To prevent bias in damage evaluation, the observer was not aware of the treatment of the inspected apples. The same observer checked all apples. Besides damage levels, the number of apples per tree and biological yield (in kg) were determined. For yield all apples (damaged and undamaged) that were hanging on the experimental trees at the harvest day in October were taken into account. Any effect of treatment on yield can be due to either the number of apples per tree and/or the weight of each apple. As the number of abscissed young fruits might explain effects in the number of harvested apples, June-drop apples were not only checked for damage but were also counted.
To test our data on, respectively, caterpillar damage, loss of caterpillar-damaged apples in June-drop, yield, number of apples, weight per apple and number of apples lost in June-drop, we carried out six groups of statistical analyses.
For variation in caterpillar damage (individual apples either did or did not have caterpillar damage; binary data) we used generalized linear models (GLIM4) with binomial errors. We used the number of damaged apples per tree as the response variable, with the number of apples per tree as the binomial denominator (effectively weighting data points by the number of apples scored). Five trees had no apples and were therefore omitted from this analysis.
We tested date of treatment, expressed in Julian dates (i.e. 1 February is Julian date 32), as a continuous variable, and also date2 to test for non-linear effects. Block was included as a factor, both as a main effect and in interaction with date and date2 to control for possible differences in damage related to block. Non-significant terms were dropped from the model starting with the highest order interaction (stepwise backward procedure). Because the residual deviance was substantially larger than the residual degrees of freedom, Williams’ adjustment for overdispersion was applied (Crawley 1993), and hence the significance of terms in the model was assessed using an F-test (Crawley 1993).
Loss of caterpillar damaged apples in June-drop
The effect of treatment on the harvest might be reduced by June-drop if trees shed relatively more apples with caterpillar damage than undamaged apples. If this were the case the slope of the regression of the percentage of damage in June-drop on the percentage of damage in harvest would be significantly different from 1. Analyses were carried out on the combined data of June-drop of experiments 1 and 2. Individual trees were treated as data points in a generalized linear model, with percentage of damage in June-drop as the response variable and binomial errors and percentage of damage in the harvest as a continuous variable and block as a factor, to control for possible differences in damage related to block. The full model with its interaction was fitted and we tested whether the slope of the regression line differed from 1 (using the offset option in GLIM).
Yield, number of apples and weight per apple
To investigate the effect of treatment on the biological yield (damaged and undamaged apples) at harvest we tested yield (in kg), number of apples and weight per apple as response variables. For these variables we used the same model structure as above but with different error distributions and link functions. The logarithm of yield and the weight per apple were treated as normally distributed variables, and the number of apples per tree as a gamma distributed variable with a reciprocal link function. We included an additional covariate, the percentage of rosy aphid Dysaphis plantaginea Pass.-infested apples per tree, because apples on aphid-infested trees were smaller.
Number of apples lost in June-drop
To investigate whether the number of June-drop apples increased with higher levels of caterpillar damage, we tested the number of apples shed in June against the total percentage of apples with caterpillar damage. A generalized linear model with binomial errors was used with the number of apples per tree in June-drop as the response variable and total number of apples (i.e. harvest plus June-drop) as the binomial denominator. Data on number of June-drop apples in both experiments were combined. We tested total caterpillar damage (i.e. damage in June-drop and harvest) as a continuous variable with number of June-drop apples as a response variable. Block was included as a factor to control for possible differences between blocks. The full model with percentage of damage and block was fitted together with their interactions.
breeding pairs and caterpillar densities
During the experiment, nine pairs of great tits bred in the vicinity of the experimental blocks. Place of breeding could not be manipulated and was a result of the birds’ preferences for breeding places. The first block had one nest within and one nest in a non-experimental neighbouring block. The second block had two nests within and two nests in a non-experimental neighbouring block. The third and fourth blocks had no nests within but, respectively, two and one nest in a non-experimental neighbouring block. Additionally, three pairs of great tits and two pairs of tree sparrows bred on the edge of the orchard. The birds deserted one of the broods in the first experimental block and the chicks died at 7 days old. Two other broods lost one of the parents during the nestling phase but both remaining parents managed to fledge three chicks. Average clutch size of great tits was 8·5 eggs (SD 1·3, n= 12), out of which on average 7·6 (SD 1·4, n= 12) chicks hatched and 5·0 (SD 2·9, n= 12) chicks fledged. The hatching dates ranged from 4 to 18 May.
Caterpillar numbers were high at the start of the experiment (on average 28 caterpillars per tree; SD 25·3, n= 56) and decreased rapidly due to pupation and predation from 10 May onwards (Fig. 1). Almost no caterpillars were present after 20 June. Winter moths comprised 38·7% of the sampled caterpillars, followed by the tortricid species Spilonota ocellana Denis & Schiffermüller, 25·3%, Pandemis heparana Denis & Schiffermüller, 14·6%, and Recuvaria leucatella Clerck, 11·4%. All these caterpillars prefer feeding on leaves. Injury to apples is caused by accident when leaves are close to or connected with young fruits and flowers. Injuries appear as cork scars on the surface of the apple. There is no distinction between scars of different caterpillar species.
In experiment 1, we found decreasing caterpillar damage levels with an increasing period of foraging by great tits (experiment 1; Fig. 2a and Table 3). The reduction was linear in relation to the date of exclusion (F1,193 = 4·9, P < 0·05), with observed damage decreasing from 13·8% (SE 1·0%, n= 44), when great tits were excluded on Julian date 116, to 11·2% (SE 0·9%, n= 56), when great tits were never excluded (treatment date 177). The total percentage of caterpillar damage differed between blocks (F3,193 = 14·7, P < 0·001) but the effect of the experimental treatment did not (block × date interaction was not significant; Table 3).
Table 3. Logistic regression of the proportion of apples per tree with caterpillar damage. In experiment 1, great tits were excluded from foraging on trees, and in experiment 2 caterpillars were removed from trees, at different times during the growing season. Date in experiment 1 is the Julian date (i.e. 1 February is Julian date 32) on which great tits were excluded from trees. A later Julian date means a shorter period of great tit exclusion, thus implying a longer period of great tit predation and hence a lower exposure to caterpillars. Date in experiment 2 is the Julian date on which caterpillars were removed from the tree. A later Julian date means higher exposure to caterpillars
Increase in deviance
All interactions and variables that were not significant are not listed.
In experiment 2, caterpillar damage increased when caterpillars were left for a longer period on the trees. This increase in relation to the date of caterpillar destruction was approximately linear at first and flattened out after Julian date 153 (date, F1,191 = 15·8, P < 0·001; date2, F1,191 = 12·8, P < 0·001; Fig. 2b and Table 3). The percentage of caterpillar damage differed between the blocks (F3,191 = 10·8, P < 0·001), but the effect of treatment did not (block × date interaction was not significant; Table 3). The observed damage in the harvest increased from 2·6% (SE 0·5%, n= 42), when caterpillars were destroyed on date 116, to 11·4% (SE 1·2%, n = 56), when caterpillars were never destroyed (treatment date 177). The percentage damage for the control group of both experiments was almost equal (11·2% and 11·4%).
loss of caterpillar damaged apples in june-drop
There was no difference in the percentage of damaged apples in the June-drop and the percentage in the harvested apples (F1,372 = 0·73, P > 0·39). Thus the percentage of damaged apples among those harvested was not influenced by trees preferentially dropping damages apples in June.
In experiment 1, yield increased non-linearly in relation to the date of exclusion of great tits (Fig. 3a and Table 4). This increase was less marked at high levels of aphid infestation (interaction date2 × aphid; Table 4). Levels of aphid infestations were not affected by the period of great tit exclusion (F1,192 = 2·75, P= 0·10). The observed yield per tree increased from 4·65 kg (SE 0·59, n= 44), when great tits were excluded on Julian date 116, to 7·76 kg (SE 0·69, n= 56), when great tits were never excluded (treatment date 177).
Table 4. Regression of biological yield, weight per apple and number of apples per tree. In experiment 1 great tits were excluded from foraging on trees and in experiment 2 caterpillars were removed from trees at different times during the growing season. Date in experiment 1 is the Julian date (i.e. 1 February is Julian date 32) on which great tits were excluded from trees. A later Julian date means a shorter period of great tit exclusion, thus implying a longer period of great tit predation and hence a lower exposure to caterpillars. Date in experiment 2 is the Julian date on which caterpillars were removed from the tree. A later Julian date means higher exposure to caterpillars
Experiment 1 great tits excluded from foraging on trees
Experiment 2 caterpillars destroyed
Variables in the model
Increase in deviance
Variables in the model
Increase in deviance
All interactions and variables that were not significant are not listed.
In experiment 2, no difference in the yield per tree was found when caterpillars were left longer on the trees. None of the interactions or date2 and date had an effect on the yield per tree (Fig. 3b and Table 4). The average yield per tree was 8·08 kg (SE 0·42, n= 197). Only the percentage of aphid infestation had a significant effect on yield, with yield decreasing as the percentage of aphid infestation increased. Because the active ingredient, parathion-methyl, of Condor also affects aphids, the level of aphid infestations was higher in the treatments where caterpillars were removed from the tree later in the season (F1,193 = 5·42, P= 0·02).
average weight per apple and number of apples
We investigated whether the increase in yield was caused by an increase in the average weight of each apple or by an increase in the number of apples. The average weight of apples increased as the period during which great tits could forage on the trees increased, but this rate of increase varied between blocks (Table 4). Furthermore, the average weight per apple decreased with increasing levels of aphid infestation. The average apple weight ranged from 104·3 g (SE 8·4, n= 11) to 124·6 g (SE 7·4, n= 11) in the different blocks in the first treatment, and from 108·5 g (SE 6·1, n= 14) to 124·8 g (SE 4·8, n= 14) in the control group. In experiment 2, with varying periods of caterpillar exposure, the average weight per apple decreased when caterpillars were left longer on the trees. There was a significant interaction between experimental treatment and aphid infestation (both the interactions date × aphid and date2 × aphid were significant; Table 4): average apple weight decreased at higher levels of aphid infestation. Individual apples were larger in the first treatment (Julian date 116) than in the control (Julian date 177), on average the weight of each apple decreased from 119·1 g (SE 3·1, n= 42) to 117·7 g (SE 3·1, n= 56).
The number of apples harvested per tree increased when great tits could forage over a longer period on trees (Table 4). This increase differed between blocks. More apples were harvested when the percentage of aphid-infested apples was higher. The average number of apples ranged from 29 (SE 6·2, n= 11) to 57 (SE 17·2, n= 11) in the different blocks for the first treatment, and from 45 (SE 9·6, n= 14) to 101 (SE 21·6, n= 14) for the control group. In experiment 2, there was a marginally non-significant decrease in the number of apples with an increasing period of caterpillar exposure, depending on the degree of aphid infestation (date × aphid, F1,183 = 3·40, P < 0·07). The number of apples per tree varied significantly between blocks, ranging from 52 (SE 5·1, n= 50) to 83 (SE 8·3, n= 50).
number of apples lost in june-drop
In order to explain some of the variation in the number of apples per tree, we investigated whether the number of June-drop apples increased with increasing harvest damage levels. Trees with a high percentage of caterpillar damage at the time of harvest had also dropped a larger percentage of their apples in June (percentage of caterpillar damage, F1,372 = 6·81, P < 0·01). The effect of damage on June-drop was the same for all blocks (interaction, F3,369 = 0·64, P > 0·59) but the average level of June-drop differed between blocks (block, F3,372 = 9·34, P < 0·001). The effect of damage levels on June-drop was small. An increase of 30% in caterpillar damage led to an increase of about 8% in June-drop.
We show clearly that great tits can reduce caterpillar damage in apple orchards. Other studies have already shown that birds can reduce insect numbers and that this reduction is dependent upon prey densities (Tables 1 and 2; Kirk, Evenden & Mineau 1996). Some of these other studies also concluded that birds become ineffective when densities are low (Tinbergen 1960) or above a threshold insect density (Otvos 1979 and references therein) and that the relative impact of predation thus reaches the maximum at moderate caterpillar densities. It is unclear how the birds perceived the densities of caterpillars in the experimental orchard. Insect densities at the site have been building up over the last 10 years, during which time hardly any action was taken against caterpillars. Densities were six times higher than in commercial organic orchards in the same year, where mineral oil and the microbial insecticide, Bacillus thuringiensis, had been used for caterpillar control (C.M.M. Mols, unpublished data). Depending on the perception of caterpillar density by the birds, the damage reduction will either increase or decrease with increasing densities. On the other hand, if insect densities in the orchard become very low, great tits may forage more outside the orchard so that damage reduction by great tits would be reduced.
The second experiment, together with the data on caterpillar abundance (Figs 1 and 2), showed that caterpillars became harmful to the apples at the end of bloom (Julian date 120) and kept inflicting damage until they pupated (Julian date 156). Besides the fact that caterpillars start inflicting damage to apples at a very early stage of the growing season (before great tits exert any predation effect), the great tits’ impact as a biological pest control agent is further reduced by their preference for larger prey (Betts 1955; Tinbergen 1960; Gibb & Bett 1963). Naef-Daenzer, Naef-Daenzer & Nager (2000) found that great tits have a preference for caterpillars larger than 10–12 mg (9–10 mm). The caterpillar species occurring in the orchard differed in full-grown size as well as their time of appearance. At a given sample date a range of different caterpillar size classes was available to great tits, of which they prefered the relative larger caterpillars. A minimum amount of damage will therefore always occur as small caterpillars inflict damage prior to reaching 9–10 mm or larger, the preferred size of predatory great tits. However, as the relationship between time of removal by great tits and caterpillar damage is linear, few large caterpillars can inflict the same damage as many small ones.
In our experiment great tits reduced damage by 2·6% from 13·8% to 11·2%. This damage reduction was not influenced by differential loss of damaged apples in June-drop. The remaining damage of 11·2% is far above the economic thresholds of fruit growers (either with organic or integrated pest management). Solomon et al. (2000) concluded for several polyphagous arthropod predators that they alone are unlikely to prevent pest damage. Our results show that this is also the case for great tits. However, their benefits should be seen when used in concert with other control measures, where it is possible, under some circumstances, that economic thresholds could be reduced in the presence of great tits.
Besides reducing damage, great tits also improved the biological yield per tree. Yield increased due to an increase in the number of apples rather than the weight per apple. In experiment 2, where caterpillars were removed, the same trend was found: lower damage was associated with more apples per tree. However, care should be taken in interpreting these results because we experimentally excluded great tits using nets. Although mesh size was large enough to allow access by bees, the nets might have reduced pollination. However, since an effect on the number of apples was also found when caterpillars were destroyed, we are confident that the effect is due to the exclusion of great tits and not to side-effects of the nets. Moreover, June-drop appears to increase with damage, explaining part of the variation in the numbers of apples.
Although great tits on their own cannot reduce caterpillar damage within the present economic thresholds, they certainly contribute to biological control. Furthermore, the only cost to the fruit growers is that of putting up nest boxes (c. 2 ha−1) to allow great tits to breed in the orchards. With the tightening of regulations on the use of pesticides, resistance of harmful insects to pesticides and the adverse public attitudes to pesticides, great tits should be encouraged as a pest control agent for caterpillars in orchards.
We thank Fredy Vaal for his indispensable assistance with fieldwork, rearing and classifying caterpillars, and Leontien Witjes, Koen Pluis and Frank Mertens for assisting with the practical fieldwork. Without their help we could not have completed the study. We thank Kate Lessells, Arie van Noordwijk, Piet Drent, Leo Blommers and Herman Helsen for their comments on the manuscript and other advice. The Technology Foundation STW (code BBI. 3844), applied science division of NWO and the technology programme of the Ministry of Economic Affairs supported this research.