Oecologia (2015) 177:423–430 DOI 10.1007/s00442-014-3103-7
POPULATION ECOLOGY - ORIGINAL RESEARCH
Weaker resource diffusion effect at coarser spatial scales observed for egg distribution of cabbage white butterflies Marc Hasenbank · Stephen Hartley
Received: 14 January 2014 / Accepted: 22 September 2014 / Published online: 7 October 2014 © Springer-Verlag Berlin Heidelberg 2014
Abstract Mobile organisms frequently forage for patchy resources; e.g. herbivorous insects searching for host plants. The resource diffusion hypothesis predicts that insect herbivores, such as Pieris rapae butterflies, are disproportionally attracted to more isolated, or ‘diffused’, host plants. Surprisingly little is known about how this response to variation in resource density manifests itself at different spatial scales. We measured the outcome of oviposition by P. rapae butterflies foraging among groups of host plants, with plant density experimentally varied to achieve comparability between three nested scales: fine (1 × 1 m), medium (6 × 6 m), and coarse (36 × 36 m). Hierarchical linear models were used to measure density-dependent responses in the number of eggs laid per plant, with plant density measured at nested spatial scales. At a fine scale, isolated plants received significantly more eggs, while at medium and coarse scales the differences were less pronounced, and tended towards a neutral distribution of eggs across plants. Larger plants also tended to receive more eggs. Since multiple processes, acting at multiple scales, are likely to be the rule rather than the exception in ecology, methods for detecting and characterising multi-scale responses are important to ensure a robust transfer of ecological models from one situation to another.
Communicated by Merijn Kant. Electronic supplementary material The online version of this article (doi:10.1007/s00442-014-3103-7) contains supplementary material, which is available to authorized users. M. Hasenbank (*) · S. Hartley Centre for Biodiversity and Restoration Biology, School of Biological Sciences, Victoria University of Wellington, PO Box 600, Wellington, New Zealand e-mail: [email protected]
Keywords Area-restricted search · Pieris rapae · Hierarchical models
Introduction In the natural environment, patchiness of resources may be discerned at many different spatial and temporal scales (Levin 1992). A common challenge for ecologists is knowing which is the appropriate scale (grain and extent) at which a particular process acts, and matching the scale of experiments and observations accordingly (Wiens 1989). A single-scale approach risks missing much of the rich complexity of ecological interactions as the relationships between pattern and process can change depending on the scale at which the pattern is observed (Levin 1992; Denny et al. 2004; Cadotte and Fukami 2005). A multi-scale approach is therefore needed to better understand these relationships and to highlight where potential contradictions may arise from scale-dependent results. In search for biologically meaningful scales many studies focus on the hierarchical levels of a study system, such as leaves within plants, plants within patches, and patches of plants within the wider landscape (e.g. Rabasa et al. 2005). Cabbage white butterflies have the tendency to lay a disproportionate number of eggs on isolated individuals of their Brassicaceae host plants (Root and Kareiva 1984; Thomas 1984). This ecological phenomenon is consistent with predictions made by the resource diffusion hypothesis (RDH) (Yamamura 1999), which assumes that insect herbivores are disproportionally attracted to more isolated, or “diffused”, host plants. A similar behaviour has also been demonstrated by the seed-head fly, Botanophila seneciella, utilising ragwort, Senecio jacobaea (Crawley and Pattrasudhi 1988). Most investigations of the RDH have been made
13
424
in the field, measuring density at a single scale. Difficulties in interpreting results of these studies occur because: (1) it is not clear what (if any) is the correct scale at which to analyse the patterns, (2) when density is not experimentally controlled the areas with high density of plants may be more or less attractive (or suitable for population growth) due to environmental reasons such as variations in soil moisture etc. Furthermore, in natural situations densities tend to be correlated across scales so it is difficult to tease apart independent influences of processes acting at different scales. Multi-scale manipulative experiments overcome these concerns (Sandel and Smith 2009); however, there are few examples in the literature. Some notable exceptions include Kunin (1999) who surveyed wild and experimental populations of S. jacobaea for insect herbivores, Roslin (2000) who investigated the distribution and movement patterns of Aphodius dung beetles, and Gunton and Kunin (2007, 2009) who examined herbivory, flowering and survival of Silene latifolia plants. In this paper we present results from two manipulative field experiments where we used heterogeneous patches of host plants to investigate the effects of patch density, varied independently at multiple spatial scales, upon the egg-laying of a single species, the small cabbage white butterfly, Pieris rapae.
Materials and methods Study species Pieris rapae To investigate the influence of different spatial scales on observed egg distribution patterns, we measured the number of eggs oviposited by females of the small cabbage white butterfly, Pieris rapae L., among different density groups of its host plants. P. rapae females oviposit on members of the family Brassicaceae, but also on nasturtium (Trapaeoleum spp.) in the family Tropaeolaceae (Bernays and Chapman 1994). P. rapae females usually lay one egg after landing on a suitable host plant before flying off to the next host plant. The flight and oviposition behaviour has been well studied by Jones (1977a), Jones et al. (1980) and Root and Kareiva (1984). A common observation is that isolated plants tend to receive more eggs per plant compared to plants growing in dense groups. This distinctive egg distribution pattern results from the flight behaviour of ovipositing P. rapae females, in which females tend to maintain a directional flight path even upon encountering dense groups of host plants. Although ovipositing females of P. rapae may land more frequently upon encountering dense groups of host plants this effect
13
Oecologia (2015) 177:423–430
does not outweigh the effect of directional flight paths. Root and Kareiva (1984) argued that such a behaviour is likely to have evolved as a form of bet-hedging: by distributing their eggs over a wider area female cabbage white butterflies increase the likelihood of at least some offspring surviving local patch extinctions or local predation pressure and also decrease intra-offspring competition for food resources. While no concrete information exists on P. rapae’s visual perception range, Yamamura (1999) observed that between consecutive oviposition events female cabbage white butterflies travelled about 1.3 m. In their experiments Root and Karieva (1984) measured flight paths at a lowest plant spacing of 1 m. For our experiments we therefore adopted 1 m to be the subplot dimension at which we measured fine-scale plant density. The medium- (6 × 6 m) and coarse- (36 × 36 m) scale measures were, however, based on an arbitrary sixfold multiplication, which allowed for the greatest spacing between different scales of measurement that could be achieved within given logistical constraints. Brassica oleracea All our cabbage plants, B. oleracea var. Sommercross, were raised in a greenhouse until they were 6 weeks old. For each experiment we randomly assigned each plant a location in the experimental design. In order to account for differences in plant sizes that might have affected the distribution of cabbage white eggs in our analysis, we measured for each plant the width (in centimetres) of the largest leaf and multiplied the value by the total number of leaves with width greater than 2 cm. One day prior to laying out the field experiment one small leaf was cut off each plant in order to standardise the possible effect of defensive chemicals induced by physical damage (Chen 2008). This precaution was taken because some plants might have been damaged during transport or as part of the planting process. Field sites and experimental design The experiment was carried out four times, utilising two different sites (Kaitoke and Levin) over a 3-year period. At both locations large, local populations of naturally occurring cabbage white butterflies were present, whose female butterflies would oviposit on the plants used within the experiments. The layout varied slightly between sites, as did the background ground cover and the number of butterflies in flight. The experimental layout for Kaitoke adopted a Latin–square design with intermitted areas of high and low resource density. While the Levin layout contained some elements of the Kaitoke experimental design, i.e. size and layout of resource patches (plants), the Levin layout followed a longitudinal design with higher coarse—scale
Oecologia (2015) 177:423–430
425
Fig. 2 Plant densities and related scales of measurement for Kaitoke experiments
Fig. 1 Layout of the Kaitoke experiments in 2004 and 2005; letters indicate 36 × 36-m blocks (coarse-scale variation in plant density), closed circles indicate cabbage plants. Block D consists of a single 6 × 6-m plot of 40 cabbages enclosing 161 × 1-m subplots each containing one, four or 16 cabbages (fine-scale variation in plant density). Block A contains 40 plants and exhibits medium-scale variation in plant density. Block C contained only four plants, at 18-m spacing (isolated at all scales considered)
resource density at one end, and a comparatively lower coarse—scale resource density at the other end. By adopting two different designs we were also able to explore whether the response to plant density was consistent across both plant layouts. While the Kaitoke Latin– square design resembled a compact resource patch at the experimental scale, the Levin design was longitudinal with a plant-density gradient from low to high. Kaitoke layout The first two trials were conducted at Kaitoke in the second half of January 2004 and 2005, respectively. We utilised four fields at the AgResearch farm near Kaitoke, New Zealand (41º04′S, 175º11′E; each field was about 1 ha in size).
The experimental fields were surrounded by a post and wire fence ~2 m high on the northern edge and ~1 m high on the other three sides. The fields had been grazed down by sheep before the beginning of the experiments. The farm itself was surrounded by scrub, indigenous vegetation and planted forest. A nearby field (500 m north–east) held a planting of Brassica oleracea var. acephala (kale). Four hundred potted cabbage plants were set up on the fields as resource patches in a spatially nested design, such that different numbers of plants occurred within a 1 × 1-m subplot, within a 6 × 6-m plot, within a 36 × 36-m block. At the top level of blocks there were four possible layouts (A, B, C, and D) containing 40, 16, four and 40 plants, respectively (the coarse-scale densities). To minimise the effects of large-scale environmental gradients blocks were replicated four times (once per field) in the overall design of a Latin square (Fig. 1). Plant density was measured by counting the number of plants within a square of a certain size (blocks, plots and subplots). The three scales of measurement and the corresponding plant-density treatments are shown in Fig. 2. Each scale of measurement contained plant densities comparable to those of the next higher scale, e.g. at fine scale plants within the different top-level layouts (A, B, etc.) could be grouped as one, four or 16 plants within a 1 × 1-m subplot, while the same plants could be re-grouped as plant densities of one, four or 16 or 40 at medium-scale measurement (i.e. within a 6 × 6 m-subplot). On the third day of each experiment we counted the number of P. rapae eggs on each cabbage plant. Levin layout During January 2006 we conducted two field experiments at the Woodhaven farm in Levin, New Zealand (40º37′S, 175º14′E). The experimental field was a freshly ploughed arable field, 400 × 300 m (12 ha), with the long side stretching roughly northeast–southwest. It was surrounded by a corn field on the northwestern side and a single row of trees to the south (Populus sp.) and east (Pinus sp.) with adjacent grazing paddocks. In the first 3 weeks of the experiments, rows of lettuce (Lactuca sativa: Asteraceae) were present on the far northwestern side (400 × 50 m). In the fourth week the harvest of the lettuce started.
13
426
Oecologia (2015) 177:423–430
were only comparable between the two experimental replicates. After 10 days a second trial was set up with fresh plants, and the layout of coarse-scale blocks was reversed so that block P was in the south–west corner of the field. Statistical methods
Fig. 3 Set up of 2006 experiments in Levin (L1, L2); letters indicate 36 × 36-m blocks, closed circles indicates locations of cabbage plants
Fig. 4 Plant densities and related scales of measurement for Levin experiments
Four hundred and twenty four cabbages were used for the experimental field in Levin. The plants were arranged in four blocks measuring 36 × 36 m and one block 18 × 18 m (blocks E and P, respectively, Fig. 3). In each corner of the 36 × 36-m blocks (E) a 6 × 6-m plot of one, four, 16 or 40 cabbages was placed. The 18 × 18-m block P was subdivided into nine 6 × 6-m plots. Alternating groups of four or 40 plants were placed into these plots (five groups of four and four groups of 40 for a total of 180 plants). The groups of 40 in all patches had a similar set up to block D in Kaitoke. For these, 1 × 1-m subplots containing one, four and 16 plants were placed in different spatial arrangements to increase the fine-scale spatial heterogeneity. The scales of measurement and the corresponding plant densities at each scale are shown in Fig. 4. Similar to the Kaitoke experimental design plant densities at fine and medium scale were chosen to be comparable between the two different scales of measurement. However, this experimental layout added elements of different coarse-scale plant densities that
13
Each plant belonged to a 1 × 1-m subplot of a particular fine-scale density, nested within a 6 × 6-m plot of a particular medium-scale density, nested within a block of a particular coarse-scale density. We analysed these data using a linear multi-level mixed effects model (LME) (Raudenbush and Bryk 2002), also known as hierarchical linear model, which specifically took into account the hierarchical structure of our study design (McMahon and Diez 2007). Our response variable for the LME was the number of eggs counted on each individual plant. We included four fixed effects in a full model as covariates: fine-scale plant density, medium-scale plant density, coarse-scale plant density and plant size. The LME model also included the nested areas of the study design as well as the experiment as random effects: experiment/block/plot/subplot. The LME model was fit using maximum-likelihood methods. Model fitting and graphs were done using R (3.1) (Ihaka and Gentleman 1996; R Development Core Team 2013). Linear mixed effects models were fitted using the lmer method from the lme4 package (1.1–7). For examples of how to build LMEs to analyse ecological data sets refer to McMahon and Diez (2007) and Quian et al. (2010), or for more general explanations to Gelman and Hill (2007). The egg-count data were square–root transformed prior to model fitting to improve normality of the residuals. One model was fitted incorporating the data from all four experiments. To assess the significance of coefficients we report the upper and lower bounds of the 95 % confidence interval (CI). The proportion of raw variance attributed to each level of the hierarchy in the LME was assessed by fitting an unconditional model that only included a fixed intercept and the random effects structure (McMahon and Diez 2007). At the time that the egg counts were taken no other insects, such as aphids (Brevicoryne brassicae) or diamondback moths (Plutella xylostella), were observed to feed off, or lay eggs on, any of the cabbage plants used during the experiments. Therefore analysis of the cabbage white egg distribution was performed on a single-species basis only.
Results The number of eggs laid by P. rapae females varied considerably between the four experiments. For each
Oecologia (2015) 177:423–430
427
Fig. 5 Violin plots for egg counts against plant densities for the different manipulative field experiments; width of violin is based on kernel density estimate of the number of occurrences of a particular square root-transformed egg count within a given plant-density group. The mean square root-transformed egg count for each plantdensity group is shown as closed circle together with a bootstrapped 95 % confidence interval (CI) Fig. 7 Effect of plant density and plant size on the number of eggs oviposited per plant (results of the linear multi-level mixed effects model). Negative values of the β-coefficients (i.e. slopes) indicate that plants situated in relatively low-density plantings receive a disproportionate number of eggs per plant. Overlap of the 95 % CI with zero indicates that the β-coefficient is not significantly different from zero and therefore does not contribute significantly to explaining the observed variance
Fig. 6 Scatter plots of square root-transformed egg count against plant size for the different manipulative field experiments; plant size measured in centimetres as width of largest leaves multiplied by the number of leaves >2 cm
experimental layout (Kaitoke and Levin), we gained a result based on a high and low number of total eggs. In the Kaitoke 2004 experiment a total of 2,061 eggs were oviposited, while in Kaitoke 2005 only 109 eggs were laid by P. rapae females (Table S1, Appendix). This translates into a mean number of eggs per plant of 5.15 and 0.27, respectively. At Levin, 173 eggs were laid in the first experiment (mean = 0.41) and 2,442 eggs in the second experiment (mean = 5.76). When the number of eggs laid overall was relatively high (i.e. Kaitoke 2004 and Levin 2) the average number of eggs counted per plant tended to decline with increasing density of neighbouring plants at all three scales
13
428
of analysis (Fig. 5), while showing a slight increase with increasing plant size (Fig. 6). Despite differences in the number of eggs oviposited between the four different experiments and the two different experimental layouts used, the statistical model revealed a consistent and strong resource diffusion effect when plant density was measured at the fine scale, i.e. a negative relationship between plant density in the local neighbourhood and number of eggs per plant (Fig. 7; Table S2). At the medium scale the coefficient was likely to be similar to zero due to the overlap of the 95 % CI. At the coarse scale, the predicted coefficient indicated that any effects of plant density on the egg count, although weaker than for fine-scale plant density, were likely to be similar to a diffusion or concentration pattern. Furthermore, the statistical model indicated that the plant size coefficient was likely to be positive (i.e. larger plants received more eggs). Based on the variance components of the unconditional linear mixed effects models, about 8 % of the total variance in the number of eggs was expressed at the fine-scale level. In contrast, the percentage of variance at the medium- and coarse-scale level was 16 and 0 %, respectively. The toplevel random effect for the different experiments accounted for 48 % of the total variance (Table S3).
Discussion In theory, opposite responses to host plant density may occur simultaneously in a single system, apparent when the system is analysed at different spatial scales (e.g. Barritt 2008). We did not find evidence for a reverse of the egg distribution pattern in our experiments (in other words β-coefficients did not switch from negative to positive with a change in scale); nevertheless, the distribution of eggs per plant did shift from a strong RDH response at fine scales to a much more neutral response to coarse-scale variations in density. This fine-scale pattern of egg distribution corresponds to results found by Root and Kareiva (1984), who described the distinct flight behaviour of P. rapae females during oviposition, as an ‘egg-spreading syndrome’, leading to comparatively higher egg numbers on more isolated plants (see also Cromatie 1975). The two main features of the egg-spreading syndrome are a relatively linear flight path and relatively larger steps between oviposition events when plants are sparse (Root and Kareiva 1984), and its benefit is seen as maximising offspring survival as part of a riskspreading strategy. In general, different species may exhibit different behaviours and may differ in their ability to move between resource patches, thus leading to different responses to resource density. Behavioural responses to resource density
13
Oecologia (2015) 177:423–430
may also reflect the requirements of the larval stages. For example, the egg distribution of Tyria jacobaeae, the cinnabar moth, follows a resource concentration pattern, i.e. fewer eggs per plant on isolated plants (Kunin 1999). During the final stages of the larval development the original host plant of T. jacobaeae often becomes completely consumed, and late-instar larvae are required to disperse to other plants within the close vicinity to fulfill their food requirements. Therefore, for T. jacobaeae, laying the majority of its eggs in a high-density patch of plants assures that enough food is available locally for its offspring. The larvae of P. rapae, though mobile to some extent, usually stay on the original host plant (Jones 1977b), hence avoidance of conspecifics is more important. Only when there are no other plants in the general vicinity is P. rapae forced to use the same plant more than once. At the individual plant level ovipositing female P. rapae tended to favour larger plants in some of our experiments. What remains unclear from our experiments is how host plant selection played out at the different spatial scales, and to what extent, if any, oviposition by female P. rapae was influenced by volatile (e.g. volatile chemicals from experimental plants), non-volatile or visual cues (e.g. ground cover of grass vs. bare soil). Other experimental work on P. rapae suggests that visual cues, in particular the colour green (Traynier 1979), are important during host plant search (Myers 1985; Bukovinszky et al. 2005), while ultimate host-plant selection and oviposition are influenced more by non-volatile plant chemicals (Renwick and Radke 1983). So it comes as no surprise that Dempster (1969) and Cromatie (1975) noted a trend of fewer P. rapae on cabbages in the ‘weedy’ plots of their experiments compared to those with a background of bare earth, where host plants may have been more visible to P. rapae females during oviposition. It is possible that in some cases larger plants were more easily perceived by female P. rapae in our experiments. The observed effect of plant size is, however, far from proportional, which may be a result of the tendency of P. rapae females to fly a certain distance between oviposition events, thus potentially skipping some larger plants nearby. On the other hand, P. rapae females have been shown to respond to volatile cues (Hern et al. 1996), but it remains unclear to what extent (if any) volatile cues play a part in the host-selection process. It may be that volatile cues from a large group of host plants, or larger plants in general, accumulate in an additive and more pronounced fashion and that these cues may be distributed over a wider area by air movement. In such a case we would have expected positive responses to plant density to be more noticeable at coarser scales. Such a response may have been present, but when mixed with a degree of random movement, which tends to lead to greater oviposition on isolated plants (Cain 1985), the two opposing forces may
Oecologia (2015) 177:423–430
have balanced each other out, thus explaining the shift from negative to neutral responses to plant density in number of eggs per plant as we change the scale of analysis from fine to coarser scales. Perhaps at yet coarser scales P. rapae eggs do become more common in dense patches and less common in isolated patches. Concluding remarks The resource patches (plants) used in this experiment varied in size; furthermore, by virtue of their spatial arrangement relative to each other, some patches (the larger, isolated plants) received predictably more eggs than others. This RDH consistent effect was strongest at the fine scale, inconsistent at the medium scale and only weakly supported at the coarse scale. These specific results obtained for P. rapae may be explained mechanistically by its reliance on visual cues, behavioural adaptations that favour risk-spreading and the impossibility of moving completely freely between an unlimited number of resource patches. An uneven distribution of eggs may be a general and virtually unavoidable consequence of searching for resources in a spatially explicit and heterogeneous world. Other species likely show different responses at some or all of the scales examined here, thus opening up the possibility for intraspecific aggregation and interspecific segregation across apparently equivalent resource patches, which is the basic premise of the aggregation model of coexistence (Hartley and Shorrocks 2002), with important implications for coexistence between competitors and/or predators and prey. Acknowledgments We thank Robert Madsen, AgResearch, and John Clarke, Woodhaven Gardens for access and logistical support at the study site; the Royal Society of New Zealand Marsden Fund (grant no. VUW305) and the Victoria University of Wellington for financial support; Cornelia Blaga, Heather Collie, Catherine Duthie, Yvonne Fabia, Jim Barritt for help in the field; Bill Kunin, Phil Lester, Heiko Wittmer, members of the Victoria University bug club, and Merijn Kant (handling editor) as well as four anonymous reviewers for commenting on the manuscript.
References Barritt J (2008) Simulation of the effects of movement patterns and resource density on the egg distribution of Pieris rapae (Lepidoptera) at multiple spatial scales. Masters thesis, Victoria University of Wellington, New Zealand Bernays EA, Chapman RF (1994) Host-plant selection by phytophagous insects. Springer, Berlin Bukovinszky T, Potting RPJ, Clough Y, van Lenteren JC, Vet LEM (2005) The role of pre- and post-alighting detection mechanisms in the responses to patch size by specialist herbivores. Oikos 109:435–446 Cadotte MW, Fukami T (2005) Dispersal, spatial scale, and species diversity in a hierarchically structured experimental landscape. Ecol Lett 8:548–557
429 Cain ML (1985) Random search by herbivorous insects: a simulation model. Ecology 66:876–888 Chen MS (2008) Inducible direct plant defense against insect herbivores: a review. Insect Sci 15:101–114 Crawley MJ, Pattrasudhi R (1988) Interspecific competition between insect herbivores: asymmetric competition between cinnabar moth and ragwort seed-head fly. Ecol Entomol 13:243–249 Cromatie WJ (1975) The effect of stand size and vegetational background on the colonization of cruciferous plants by herbivorous insects. J Appl Ecol 12:517–533 Dempster JP (1969) Some effects of weed control on the numbers of the small cabbage white (Pieris rapae L.) on Brussels sprouts. J Appl Ecol 6:339–346 Denny MW, Helmuth B, Leonard GH, Harley CDG, Hunt LJH, Nelson EK (2004) Quantifying scale in ecology: lessons from a wave-swept shore. Ecol Monogr 74:513–532 Gelman A, Hill J (2007) Data analysis using regression and multilevel/hierarchical models. Cambridge University Press, New York Gunton RM, Kunin WE (2007) Density effects at multiple scales in an experimental plant population. J Ecol 95:435–445 Gunton RM, Kunin WE (2009) Density-dependence at multiple scales in experimental and natural plant populations. J Ecol 97:567–580 Hartley S, Shorrocks B (2002) A general framework for the aggregation model of coexistence. J Anim Ecol 71:651–662 Hern A, McKinlay RG, Edwards J G (1996) Effect of host plant volatiles on the flight behaviour of Pieris rapae. In: Proceedings of Brighton Crop Protection Conference: Pests and Diseases. British Crop Protection Council, Bracknell pp. 431–432 Ihaka R, Gentleman R (1996) R: a language for data analysis and graphics. J Comput Gr Stat 5:299–314 Jones RE (1977a) Movement patterns and egg distribution in cabbage butterflies. J Anim Ecol 46:195–212 Jones RE (1977b) Search behavior: study of 3 caterpillar species. Behav 60:236–259 Jones RE, Gilbert N, Guppy M, Nealis V (1980) Long-distance movement of Pieris rapae. J Anim Ecol 49:629–642 Kunin WE (1999) Patterns of herbivore incidence on experimental arrays and field populations of ragwort, Senecio jacobaea. Oikos 84:515–525 Levin SA (1992) The problem of pattern and scale in ecology. Ecology 73:1943–1967 McMahon SM, Diez JM (2007) Scales of association: hierarchical linear models and the measurement of ecological systems. Ecol Lett 10:437–452 Myers JH (1985) Effect of physiological condition of the host plant on the ovipositional choice of the cabbage white butterfly, Pieris rapae. J Anim Ecol 54:193–204 Quian SS, Cuffney TF, Alameddine I, McMahon G, Reckhow KH (2010) On the application of multilevel modeling in environmental and ecological studies. Ecol 91:355–361 R Development Core Team (2013) R: a language and environment for statistical computing. R Foundation for Statistical Computing, Vienna. http://www.R-project.org Rabasa SG, Gutierrez D, Escudero A (2005) Egg laying by a butterfly on a fragmented host plant: a multi-level approach. Ecogr 28:629–639 Raudenbush SW, Bryk AS (2002) Hierarchical linear models: applications and data analysis methods. Advanced quantitative techniques in the social sciences. Sage, Thousand Oaks Renwick JAA, Radke CD (1983) Chemical recognition of host plants for oviposition by the cabbage butterfly, Pieris rapae (Lepidoptera, Pieridae). Environ Entomol 12:446–450 Root RB, Kareiva PM (1984) The search for resources by cabbage butterflies (Pieris rapae): ecological consequences and adaptive significance of Markovian movements in a patchy environment. Ecology 65:147–165
13
430 Roslin T (2000) Dung beetle movements at two spatial scales. Oikos 91:323–335 Sandel B, Smith AB (2009) Scale as a lurking factor: incorporating scale-dependence in experimental ecology. Oikos 118:1284–1291 Thomas CD (1984) Oviposition and egg load assessment by Anthocharis cardamines (L.) (Lepidoptera: Pieridae). Entomol Gaz 35:145–148
13
Oecologia (2015) 177:423–430 Traynier RMM (1979) Long-term changes in the oviposition behavior of the cabbage butterfly, Pieris rapae, induced by contact with plants. Physiol Entomol 4:87–96 Wiens JA (1989) Spatial scaling in ecology. Funct Ecol 3:385–397 Yamamura K (1999) Relation between plant density and arthropod density in cabbage fields. Res Popul Ecol 41:177–182
POPULATION ECOLOGY - ORIGINAL RESEARCH
Weaker resource diffusion effect at coarser spatial scales observed for egg distribution of cabbage white butterflies Marc Hasenbank · Stephen Hartley
Received: 14 January 2014 / Accepted: 22 September 2014 / Published online: 7 October 2014 © Springer-Verlag Berlin Heidelberg 2014
Abstract Mobile organisms frequently forage for patchy resources; e.g. herbivorous insects searching for host plants. The resource diffusion hypothesis predicts that insect herbivores, such as Pieris rapae butterflies, are disproportionally attracted to more isolated, or ‘diffused’, host plants. Surprisingly little is known about how this response to variation in resource density manifests itself at different spatial scales. We measured the outcome of oviposition by P. rapae butterflies foraging among groups of host plants, with plant density experimentally varied to achieve comparability between three nested scales: fine (1 × 1 m), medium (6 × 6 m), and coarse (36 × 36 m). Hierarchical linear models were used to measure density-dependent responses in the number of eggs laid per plant, with plant density measured at nested spatial scales. At a fine scale, isolated plants received significantly more eggs, while at medium and coarse scales the differences were less pronounced, and tended towards a neutral distribution of eggs across plants. Larger plants also tended to receive more eggs. Since multiple processes, acting at multiple scales, are likely to be the rule rather than the exception in ecology, methods for detecting and characterising multi-scale responses are important to ensure a robust transfer of ecological models from one situation to another.
Communicated by Merijn Kant. Electronic supplementary material The online version of this article (doi:10.1007/s00442-014-3103-7) contains supplementary material, which is available to authorized users. M. Hasenbank (*) · S. Hartley Centre for Biodiversity and Restoration Biology, School of Biological Sciences, Victoria University of Wellington, PO Box 600, Wellington, New Zealand e-mail: [email protected]
Keywords Area-restricted search · Pieris rapae · Hierarchical models
Introduction In the natural environment, patchiness of resources may be discerned at many different spatial and temporal scales (Levin 1992). A common challenge for ecologists is knowing which is the appropriate scale (grain and extent) at which a particular process acts, and matching the scale of experiments and observations accordingly (Wiens 1989). A single-scale approach risks missing much of the rich complexity of ecological interactions as the relationships between pattern and process can change depending on the scale at which the pattern is observed (Levin 1992; Denny et al. 2004; Cadotte and Fukami 2005). A multi-scale approach is therefore needed to better understand these relationships and to highlight where potential contradictions may arise from scale-dependent results. In search for biologically meaningful scales many studies focus on the hierarchical levels of a study system, such as leaves within plants, plants within patches, and patches of plants within the wider landscape (e.g. Rabasa et al. 2005). Cabbage white butterflies have the tendency to lay a disproportionate number of eggs on isolated individuals of their Brassicaceae host plants (Root and Kareiva 1984; Thomas 1984). This ecological phenomenon is consistent with predictions made by the resource diffusion hypothesis (RDH) (Yamamura 1999), which assumes that insect herbivores are disproportionally attracted to more isolated, or “diffused”, host plants. A similar behaviour has also been demonstrated by the seed-head fly, Botanophila seneciella, utilising ragwort, Senecio jacobaea (Crawley and Pattrasudhi 1988). Most investigations of the RDH have been made
13
424
in the field, measuring density at a single scale. Difficulties in interpreting results of these studies occur because: (1) it is not clear what (if any) is the correct scale at which to analyse the patterns, (2) when density is not experimentally controlled the areas with high density of plants may be more or less attractive (or suitable for population growth) due to environmental reasons such as variations in soil moisture etc. Furthermore, in natural situations densities tend to be correlated across scales so it is difficult to tease apart independent influences of processes acting at different scales. Multi-scale manipulative experiments overcome these concerns (Sandel and Smith 2009); however, there are few examples in the literature. Some notable exceptions include Kunin (1999) who surveyed wild and experimental populations of S. jacobaea for insect herbivores, Roslin (2000) who investigated the distribution and movement patterns of Aphodius dung beetles, and Gunton and Kunin (2007, 2009) who examined herbivory, flowering and survival of Silene latifolia plants. In this paper we present results from two manipulative field experiments where we used heterogeneous patches of host plants to investigate the effects of patch density, varied independently at multiple spatial scales, upon the egg-laying of a single species, the small cabbage white butterfly, Pieris rapae.
Materials and methods Study species Pieris rapae To investigate the influence of different spatial scales on observed egg distribution patterns, we measured the number of eggs oviposited by females of the small cabbage white butterfly, Pieris rapae L., among different density groups of its host plants. P. rapae females oviposit on members of the family Brassicaceae, but also on nasturtium (Trapaeoleum spp.) in the family Tropaeolaceae (Bernays and Chapman 1994). P. rapae females usually lay one egg after landing on a suitable host plant before flying off to the next host plant. The flight and oviposition behaviour has been well studied by Jones (1977a), Jones et al. (1980) and Root and Kareiva (1984). A common observation is that isolated plants tend to receive more eggs per plant compared to plants growing in dense groups. This distinctive egg distribution pattern results from the flight behaviour of ovipositing P. rapae females, in which females tend to maintain a directional flight path even upon encountering dense groups of host plants. Although ovipositing females of P. rapae may land more frequently upon encountering dense groups of host plants this effect
13
Oecologia (2015) 177:423–430
does not outweigh the effect of directional flight paths. Root and Kareiva (1984) argued that such a behaviour is likely to have evolved as a form of bet-hedging: by distributing their eggs over a wider area female cabbage white butterflies increase the likelihood of at least some offspring surviving local patch extinctions or local predation pressure and also decrease intra-offspring competition for food resources. While no concrete information exists on P. rapae’s visual perception range, Yamamura (1999) observed that between consecutive oviposition events female cabbage white butterflies travelled about 1.3 m. In their experiments Root and Karieva (1984) measured flight paths at a lowest plant spacing of 1 m. For our experiments we therefore adopted 1 m to be the subplot dimension at which we measured fine-scale plant density. The medium- (6 × 6 m) and coarse- (36 × 36 m) scale measures were, however, based on an arbitrary sixfold multiplication, which allowed for the greatest spacing between different scales of measurement that could be achieved within given logistical constraints. Brassica oleracea All our cabbage plants, B. oleracea var. Sommercross, were raised in a greenhouse until they were 6 weeks old. For each experiment we randomly assigned each plant a location in the experimental design. In order to account for differences in plant sizes that might have affected the distribution of cabbage white eggs in our analysis, we measured for each plant the width (in centimetres) of the largest leaf and multiplied the value by the total number of leaves with width greater than 2 cm. One day prior to laying out the field experiment one small leaf was cut off each plant in order to standardise the possible effect of defensive chemicals induced by physical damage (Chen 2008). This precaution was taken because some plants might have been damaged during transport or as part of the planting process. Field sites and experimental design The experiment was carried out four times, utilising two different sites (Kaitoke and Levin) over a 3-year period. At both locations large, local populations of naturally occurring cabbage white butterflies were present, whose female butterflies would oviposit on the plants used within the experiments. The layout varied slightly between sites, as did the background ground cover and the number of butterflies in flight. The experimental layout for Kaitoke adopted a Latin–square design with intermitted areas of high and low resource density. While the Levin layout contained some elements of the Kaitoke experimental design, i.e. size and layout of resource patches (plants), the Levin layout followed a longitudinal design with higher coarse—scale
Oecologia (2015) 177:423–430
425
Fig. 2 Plant densities and related scales of measurement for Kaitoke experiments
Fig. 1 Layout of the Kaitoke experiments in 2004 and 2005; letters indicate 36 × 36-m blocks (coarse-scale variation in plant density), closed circles indicate cabbage plants. Block D consists of a single 6 × 6-m plot of 40 cabbages enclosing 161 × 1-m subplots each containing one, four or 16 cabbages (fine-scale variation in plant density). Block A contains 40 plants and exhibits medium-scale variation in plant density. Block C contained only four plants, at 18-m spacing (isolated at all scales considered)
resource density at one end, and a comparatively lower coarse—scale resource density at the other end. By adopting two different designs we were also able to explore whether the response to plant density was consistent across both plant layouts. While the Kaitoke Latin– square design resembled a compact resource patch at the experimental scale, the Levin design was longitudinal with a plant-density gradient from low to high. Kaitoke layout The first two trials were conducted at Kaitoke in the second half of January 2004 and 2005, respectively. We utilised four fields at the AgResearch farm near Kaitoke, New Zealand (41º04′S, 175º11′E; each field was about 1 ha in size).
The experimental fields were surrounded by a post and wire fence ~2 m high on the northern edge and ~1 m high on the other three sides. The fields had been grazed down by sheep before the beginning of the experiments. The farm itself was surrounded by scrub, indigenous vegetation and planted forest. A nearby field (500 m north–east) held a planting of Brassica oleracea var. acephala (kale). Four hundred potted cabbage plants were set up on the fields as resource patches in a spatially nested design, such that different numbers of plants occurred within a 1 × 1-m subplot, within a 6 × 6-m plot, within a 36 × 36-m block. At the top level of blocks there were four possible layouts (A, B, C, and D) containing 40, 16, four and 40 plants, respectively (the coarse-scale densities). To minimise the effects of large-scale environmental gradients blocks were replicated four times (once per field) in the overall design of a Latin square (Fig. 1). Plant density was measured by counting the number of plants within a square of a certain size (blocks, plots and subplots). The three scales of measurement and the corresponding plant-density treatments are shown in Fig. 2. Each scale of measurement contained plant densities comparable to those of the next higher scale, e.g. at fine scale plants within the different top-level layouts (A, B, etc.) could be grouped as one, four or 16 plants within a 1 × 1-m subplot, while the same plants could be re-grouped as plant densities of one, four or 16 or 40 at medium-scale measurement (i.e. within a 6 × 6 m-subplot). On the third day of each experiment we counted the number of P. rapae eggs on each cabbage plant. Levin layout During January 2006 we conducted two field experiments at the Woodhaven farm in Levin, New Zealand (40º37′S, 175º14′E). The experimental field was a freshly ploughed arable field, 400 × 300 m (12 ha), with the long side stretching roughly northeast–southwest. It was surrounded by a corn field on the northwestern side and a single row of trees to the south (Populus sp.) and east (Pinus sp.) with adjacent grazing paddocks. In the first 3 weeks of the experiments, rows of lettuce (Lactuca sativa: Asteraceae) were present on the far northwestern side (400 × 50 m). In the fourth week the harvest of the lettuce started.
13
426
Oecologia (2015) 177:423–430
were only comparable between the two experimental replicates. After 10 days a second trial was set up with fresh plants, and the layout of coarse-scale blocks was reversed so that block P was in the south–west corner of the field. Statistical methods
Fig. 3 Set up of 2006 experiments in Levin (L1, L2); letters indicate 36 × 36-m blocks, closed circles indicates locations of cabbage plants
Fig. 4 Plant densities and related scales of measurement for Levin experiments
Four hundred and twenty four cabbages were used for the experimental field in Levin. The plants were arranged in four blocks measuring 36 × 36 m and one block 18 × 18 m (blocks E and P, respectively, Fig. 3). In each corner of the 36 × 36-m blocks (E) a 6 × 6-m plot of one, four, 16 or 40 cabbages was placed. The 18 × 18-m block P was subdivided into nine 6 × 6-m plots. Alternating groups of four or 40 plants were placed into these plots (five groups of four and four groups of 40 for a total of 180 plants). The groups of 40 in all patches had a similar set up to block D in Kaitoke. For these, 1 × 1-m subplots containing one, four and 16 plants were placed in different spatial arrangements to increase the fine-scale spatial heterogeneity. The scales of measurement and the corresponding plant densities at each scale are shown in Fig. 4. Similar to the Kaitoke experimental design plant densities at fine and medium scale were chosen to be comparable between the two different scales of measurement. However, this experimental layout added elements of different coarse-scale plant densities that
13
Each plant belonged to a 1 × 1-m subplot of a particular fine-scale density, nested within a 6 × 6-m plot of a particular medium-scale density, nested within a block of a particular coarse-scale density. We analysed these data using a linear multi-level mixed effects model (LME) (Raudenbush and Bryk 2002), also known as hierarchical linear model, which specifically took into account the hierarchical structure of our study design (McMahon and Diez 2007). Our response variable for the LME was the number of eggs counted on each individual plant. We included four fixed effects in a full model as covariates: fine-scale plant density, medium-scale plant density, coarse-scale plant density and plant size. The LME model also included the nested areas of the study design as well as the experiment as random effects: experiment/block/plot/subplot. The LME model was fit using maximum-likelihood methods. Model fitting and graphs were done using R (3.1) (Ihaka and Gentleman 1996; R Development Core Team 2013). Linear mixed effects models were fitted using the lmer method from the lme4 package (1.1–7). For examples of how to build LMEs to analyse ecological data sets refer to McMahon and Diez (2007) and Quian et al. (2010), or for more general explanations to Gelman and Hill (2007). The egg-count data were square–root transformed prior to model fitting to improve normality of the residuals. One model was fitted incorporating the data from all four experiments. To assess the significance of coefficients we report the upper and lower bounds of the 95 % confidence interval (CI). The proportion of raw variance attributed to each level of the hierarchy in the LME was assessed by fitting an unconditional model that only included a fixed intercept and the random effects structure (McMahon and Diez 2007). At the time that the egg counts were taken no other insects, such as aphids (Brevicoryne brassicae) or diamondback moths (Plutella xylostella), were observed to feed off, or lay eggs on, any of the cabbage plants used during the experiments. Therefore analysis of the cabbage white egg distribution was performed on a single-species basis only.
Results The number of eggs laid by P. rapae females varied considerably between the four experiments. For each
Oecologia (2015) 177:423–430
427
Fig. 5 Violin plots for egg counts against plant densities for the different manipulative field experiments; width of violin is based on kernel density estimate of the number of occurrences of a particular square root-transformed egg count within a given plant-density group. The mean square root-transformed egg count for each plantdensity group is shown as closed circle together with a bootstrapped 95 % confidence interval (CI) Fig. 7 Effect of plant density and plant size on the number of eggs oviposited per plant (results of the linear multi-level mixed effects model). Negative values of the β-coefficients (i.e. slopes) indicate that plants situated in relatively low-density plantings receive a disproportionate number of eggs per plant. Overlap of the 95 % CI with zero indicates that the β-coefficient is not significantly different from zero and therefore does not contribute significantly to explaining the observed variance
Fig. 6 Scatter plots of square root-transformed egg count against plant size for the different manipulative field experiments; plant size measured in centimetres as width of largest leaves multiplied by the number of leaves >2 cm
experimental layout (Kaitoke and Levin), we gained a result based on a high and low number of total eggs. In the Kaitoke 2004 experiment a total of 2,061 eggs were oviposited, while in Kaitoke 2005 only 109 eggs were laid by P. rapae females (Table S1, Appendix). This translates into a mean number of eggs per plant of 5.15 and 0.27, respectively. At Levin, 173 eggs were laid in the first experiment (mean = 0.41) and 2,442 eggs in the second experiment (mean = 5.76). When the number of eggs laid overall was relatively high (i.e. Kaitoke 2004 and Levin 2) the average number of eggs counted per plant tended to decline with increasing density of neighbouring plants at all three scales
13
428
of analysis (Fig. 5), while showing a slight increase with increasing plant size (Fig. 6). Despite differences in the number of eggs oviposited between the four different experiments and the two different experimental layouts used, the statistical model revealed a consistent and strong resource diffusion effect when plant density was measured at the fine scale, i.e. a negative relationship between plant density in the local neighbourhood and number of eggs per plant (Fig. 7; Table S2). At the medium scale the coefficient was likely to be similar to zero due to the overlap of the 95 % CI. At the coarse scale, the predicted coefficient indicated that any effects of plant density on the egg count, although weaker than for fine-scale plant density, were likely to be similar to a diffusion or concentration pattern. Furthermore, the statistical model indicated that the plant size coefficient was likely to be positive (i.e. larger plants received more eggs). Based on the variance components of the unconditional linear mixed effects models, about 8 % of the total variance in the number of eggs was expressed at the fine-scale level. In contrast, the percentage of variance at the medium- and coarse-scale level was 16 and 0 %, respectively. The toplevel random effect for the different experiments accounted for 48 % of the total variance (Table S3).
Discussion In theory, opposite responses to host plant density may occur simultaneously in a single system, apparent when the system is analysed at different spatial scales (e.g. Barritt 2008). We did not find evidence for a reverse of the egg distribution pattern in our experiments (in other words β-coefficients did not switch from negative to positive with a change in scale); nevertheless, the distribution of eggs per plant did shift from a strong RDH response at fine scales to a much more neutral response to coarse-scale variations in density. This fine-scale pattern of egg distribution corresponds to results found by Root and Kareiva (1984), who described the distinct flight behaviour of P. rapae females during oviposition, as an ‘egg-spreading syndrome’, leading to comparatively higher egg numbers on more isolated plants (see also Cromatie 1975). The two main features of the egg-spreading syndrome are a relatively linear flight path and relatively larger steps between oviposition events when plants are sparse (Root and Kareiva 1984), and its benefit is seen as maximising offspring survival as part of a riskspreading strategy. In general, different species may exhibit different behaviours and may differ in their ability to move between resource patches, thus leading to different responses to resource density. Behavioural responses to resource density
13
Oecologia (2015) 177:423–430
may also reflect the requirements of the larval stages. For example, the egg distribution of Tyria jacobaeae, the cinnabar moth, follows a resource concentration pattern, i.e. fewer eggs per plant on isolated plants (Kunin 1999). During the final stages of the larval development the original host plant of T. jacobaeae often becomes completely consumed, and late-instar larvae are required to disperse to other plants within the close vicinity to fulfill their food requirements. Therefore, for T. jacobaeae, laying the majority of its eggs in a high-density patch of plants assures that enough food is available locally for its offspring. The larvae of P. rapae, though mobile to some extent, usually stay on the original host plant (Jones 1977b), hence avoidance of conspecifics is more important. Only when there are no other plants in the general vicinity is P. rapae forced to use the same plant more than once. At the individual plant level ovipositing female P. rapae tended to favour larger plants in some of our experiments. What remains unclear from our experiments is how host plant selection played out at the different spatial scales, and to what extent, if any, oviposition by female P. rapae was influenced by volatile (e.g. volatile chemicals from experimental plants), non-volatile or visual cues (e.g. ground cover of grass vs. bare soil). Other experimental work on P. rapae suggests that visual cues, in particular the colour green (Traynier 1979), are important during host plant search (Myers 1985; Bukovinszky et al. 2005), while ultimate host-plant selection and oviposition are influenced more by non-volatile plant chemicals (Renwick and Radke 1983). So it comes as no surprise that Dempster (1969) and Cromatie (1975) noted a trend of fewer P. rapae on cabbages in the ‘weedy’ plots of their experiments compared to those with a background of bare earth, where host plants may have been more visible to P. rapae females during oviposition. It is possible that in some cases larger plants were more easily perceived by female P. rapae in our experiments. The observed effect of plant size is, however, far from proportional, which may be a result of the tendency of P. rapae females to fly a certain distance between oviposition events, thus potentially skipping some larger plants nearby. On the other hand, P. rapae females have been shown to respond to volatile cues (Hern et al. 1996), but it remains unclear to what extent (if any) volatile cues play a part in the host-selection process. It may be that volatile cues from a large group of host plants, or larger plants in general, accumulate in an additive and more pronounced fashion and that these cues may be distributed over a wider area by air movement. In such a case we would have expected positive responses to plant density to be more noticeable at coarser scales. Such a response may have been present, but when mixed with a degree of random movement, which tends to lead to greater oviposition on isolated plants (Cain 1985), the two opposing forces may
Oecologia (2015) 177:423–430
have balanced each other out, thus explaining the shift from negative to neutral responses to plant density in number of eggs per plant as we change the scale of analysis from fine to coarser scales. Perhaps at yet coarser scales P. rapae eggs do become more common in dense patches and less common in isolated patches. Concluding remarks The resource patches (plants) used in this experiment varied in size; furthermore, by virtue of their spatial arrangement relative to each other, some patches (the larger, isolated plants) received predictably more eggs than others. This RDH consistent effect was strongest at the fine scale, inconsistent at the medium scale and only weakly supported at the coarse scale. These specific results obtained for P. rapae may be explained mechanistically by its reliance on visual cues, behavioural adaptations that favour risk-spreading and the impossibility of moving completely freely between an unlimited number of resource patches. An uneven distribution of eggs may be a general and virtually unavoidable consequence of searching for resources in a spatially explicit and heterogeneous world. Other species likely show different responses at some or all of the scales examined here, thus opening up the possibility for intraspecific aggregation and interspecific segregation across apparently equivalent resource patches, which is the basic premise of the aggregation model of coexistence (Hartley and Shorrocks 2002), with important implications for coexistence between competitors and/or predators and prey. Acknowledgments We thank Robert Madsen, AgResearch, and John Clarke, Woodhaven Gardens for access and logistical support at the study site; the Royal Society of New Zealand Marsden Fund (grant no. VUW305) and the Victoria University of Wellington for financial support; Cornelia Blaga, Heather Collie, Catherine Duthie, Yvonne Fabia, Jim Barritt for help in the field; Bill Kunin, Phil Lester, Heiko Wittmer, members of the Victoria University bug club, and Merijn Kant (handling editor) as well as four anonymous reviewers for commenting on the manuscript.
References Barritt J (2008) Simulation of the effects of movement patterns and resource density on the egg distribution of Pieris rapae (Lepidoptera) at multiple spatial scales. Masters thesis, Victoria University of Wellington, New Zealand Bernays EA, Chapman RF (1994) Host-plant selection by phytophagous insects. Springer, Berlin Bukovinszky T, Potting RPJ, Clough Y, van Lenteren JC, Vet LEM (2005) The role of pre- and post-alighting detection mechanisms in the responses to patch size by specialist herbivores. Oikos 109:435–446 Cadotte MW, Fukami T (2005) Dispersal, spatial scale, and species diversity in a hierarchically structured experimental landscape. Ecol Lett 8:548–557
429 Cain ML (1985) Random search by herbivorous insects: a simulation model. Ecology 66:876–888 Chen MS (2008) Inducible direct plant defense against insect herbivores: a review. Insect Sci 15:101–114 Crawley MJ, Pattrasudhi R (1988) Interspecific competition between insect herbivores: asymmetric competition between cinnabar moth and ragwort seed-head fly. Ecol Entomol 13:243–249 Cromatie WJ (1975) The effect of stand size and vegetational background on the colonization of cruciferous plants by herbivorous insects. J Appl Ecol 12:517–533 Dempster JP (1969) Some effects of weed control on the numbers of the small cabbage white (Pieris rapae L.) on Brussels sprouts. J Appl Ecol 6:339–346 Denny MW, Helmuth B, Leonard GH, Harley CDG, Hunt LJH, Nelson EK (2004) Quantifying scale in ecology: lessons from a wave-swept shore. Ecol Monogr 74:513–532 Gelman A, Hill J (2007) Data analysis using regression and multilevel/hierarchical models. Cambridge University Press, New York Gunton RM, Kunin WE (2007) Density effects at multiple scales in an experimental plant population. J Ecol 95:435–445 Gunton RM, Kunin WE (2009) Density-dependence at multiple scales in experimental and natural plant populations. J Ecol 97:567–580 Hartley S, Shorrocks B (2002) A general framework for the aggregation model of coexistence. J Anim Ecol 71:651–662 Hern A, McKinlay RG, Edwards J G (1996) Effect of host plant volatiles on the flight behaviour of Pieris rapae. In: Proceedings of Brighton Crop Protection Conference: Pests and Diseases. British Crop Protection Council, Bracknell pp. 431–432 Ihaka R, Gentleman R (1996) R: a language for data analysis and graphics. J Comput Gr Stat 5:299–314 Jones RE (1977a) Movement patterns and egg distribution in cabbage butterflies. J Anim Ecol 46:195–212 Jones RE (1977b) Search behavior: study of 3 caterpillar species. Behav 60:236–259 Jones RE, Gilbert N, Guppy M, Nealis V (1980) Long-distance movement of Pieris rapae. J Anim Ecol 49:629–642 Kunin WE (1999) Patterns of herbivore incidence on experimental arrays and field populations of ragwort, Senecio jacobaea. Oikos 84:515–525 Levin SA (1992) The problem of pattern and scale in ecology. Ecology 73:1943–1967 McMahon SM, Diez JM (2007) Scales of association: hierarchical linear models and the measurement of ecological systems. Ecol Lett 10:437–452 Myers JH (1985) Effect of physiological condition of the host plant on the ovipositional choice of the cabbage white butterfly, Pieris rapae. J Anim Ecol 54:193–204 Quian SS, Cuffney TF, Alameddine I, McMahon G, Reckhow KH (2010) On the application of multilevel modeling in environmental and ecological studies. Ecol 91:355–361 R Development Core Team (2013) R: a language and environment for statistical computing. R Foundation for Statistical Computing, Vienna. http://www.R-project.org Rabasa SG, Gutierrez D, Escudero A (2005) Egg laying by a butterfly on a fragmented host plant: a multi-level approach. Ecogr 28:629–639 Raudenbush SW, Bryk AS (2002) Hierarchical linear models: applications and data analysis methods. Advanced quantitative techniques in the social sciences. Sage, Thousand Oaks Renwick JAA, Radke CD (1983) Chemical recognition of host plants for oviposition by the cabbage butterfly, Pieris rapae (Lepidoptera, Pieridae). Environ Entomol 12:446–450 Root RB, Kareiva PM (1984) The search for resources by cabbage butterflies (Pieris rapae): ecological consequences and adaptive significance of Markovian movements in a patchy environment. Ecology 65:147–165
13
430 Roslin T (2000) Dung beetle movements at two spatial scales. Oikos 91:323–335 Sandel B, Smith AB (2009) Scale as a lurking factor: incorporating scale-dependence in experimental ecology. Oikos 118:1284–1291 Thomas CD (1984) Oviposition and egg load assessment by Anthocharis cardamines (L.) (Lepidoptera: Pieridae). Entomol Gaz 35:145–148
13
Oecologia (2015) 177:423–430 Traynier RMM (1979) Long-term changes in the oviposition behavior of the cabbage butterfly, Pieris rapae, induced by contact with plants. Physiol Entomol 4:87–96 Wiens JA (1989) Spatial scaling in ecology. Funct Ecol 3:385–397 Yamamura K (1999) Relation between plant density and arthropod density in cabbage fields. Res Popul Ecol 41:177–182
