Study system and focal taxa
The yellow sac spider Cheiracanthium mildei (Miturgidae) was the studied exotic predator; this nocturnal, wandering hunter is a Mediterranean native that was first reported in North America in the late 1940s (Bryant 1951). We used vineyards in Napa County, California, as a test ecosystem, where up to 95% of the arthropod predators are spiders (Costello & Daane 1999). The grape leafhopper Erythroneura elegantula Osborn (Cicadellidae) was the dominant herbivore and the primary spider prey. In previous surveys, increased dominance of C. mildei was accompanied by reduced native spider abundance and diversity (Hogg, Gillespie & Daane 2010; Hogg & Daane in press), and we suspected that C. mildei may have played a role in driving this pattern.
We used a field experiment to assess interactions between C. mildei and two native spiders: the nocturnal wandering spider Anyphaena pacifica Banks (Anyphaenidae), the most abundant native ecological homologue of C. mildei; and the cobweb weaver Theridion melanurum Hahn (Theridiidae), which is a potential leafhopper predator (Costello & Daane 2003) but is a sit-and-wait forager and functionally dissimilar to the two wandering spider species. While T. melanurum was first described in Europe, initial records in North America were also quite early (Banks 1897) and it is distributed throughout the Holarctic; therefore, we consider it a native species, although we acknowledge the possibility of an early introduction. To assess the predatory ability of C. mildei and A. pacifica under more controlled conditions, a greenhouse experiment was conducted.
Field experiment: predator identity, niche partitioning or predator interference
The field experiment took place in a Cabernet Sauvignon cv. vineyard, planted in 1998 in Napa County, California. No pesticides were applied, except sulphur dust to control fungus. Although C. mildei dominated the vineyard’s spider community, a second exotic species, Cheiracanthium inclusum Hentz, was present in low numbers. Cheiracanthium inclusum and C. mildei are similar in size and appearance, and are likely to be close ecological homologues. In addition to the grape leafhopper, the variegated leafhopper Erythroneura variabilis Beamer and a sharpnosed leafhopper, Scaphytopius sp., were present at low densities. The experiment took place from August to September 2007, when spider and prey abundances in Napa vineyards are highest (Hogg & Daane 2010).
Preliminary tests using marked spiders indicated that enclosures would be necessary to reduce spider emigration and immigration. Cages enclosed one grape vine (∼1·5 m wide × 1 m tall) and were made of nylon mesh (∼2·75 × 3 m, 1 mm gauge), which was draped over vines and sealed on all open sides with staples. Holes on either side of the enclosure, which could be opened and resealed with clips, allowed access to vines.
Nine treatments were established in a complete factorial design: (i) no spiders added, (ii) C. mildei alone, (iii) A. pacifica alone, (iv) T. melanurum alone, (v) C. mildei + A. pacifica, (vi) C. mildei + T. melanurum, (vii) A. pacifica + T. melanurum, (viii) C. mildei + A. pacifica + T. melanurum (additive-series version) and (ix) C. mildei + A. pacifica + T. melanurum (replacement-series version). Five replicates were completed for all treatments except the A. pacifica + T. melanurum and the additive C. mildei + A. pacifica + T. melanurum treatments, where one cage from each of these treatments was excluded due to complications during the final collection of spiders and leafhoppers from cages. All multiple-species treatments except the replacement-series treatment followed an additive-series design, which elevates the total number of predators in the multiple-species treatments. This design is more appropriate for examining interspecific interactions among predators, as it holds intraspecific interactions constant across different levels of diversity (Schmitz 2007). Each of these replicates received nine individuals of all appropriate spiders, which is within the density range found in Napa vineyards for each species (B. Hogg & K. Daane, unpublished data). A replacement-series version of the three-species treatment was also included, a design that isolates the effects of intraspecific and interspecific predator interactions on lower trophic levels and is more appropriate for examining niche partitioning (Byrnes & Stachowicz 2009). This treatment received three individuals of each species, giving a total of nine spiders, equal to that in the single-species treatments. Unmanipulated caged and uncaged reference vines (five replicates of each) were included to assess cage effects and ensure that numbers of spiders and leafhoppers in treatment cages reflected field densities.
To control for possible edge effects, replicates were blocked by location along vineyard rows, in a complete randomized block design. Caged vines were at least five vines (∼10 m) apart along rows. All canes were pruned to a length of 3 m before applying netting to vines, and any foliage touching neighbouring vines was cut back. To provide a measure of vine size, the number of canes and average cane length per vine were measured and were not significantly different among treatments (anova, number of canes: F8,44 = 0·16, P = 0·99; cane length: F8,44 = 0·59, P = 0·78). Before tested spiders were added, resident spiders were cleared from vines between 13 and 20 July, by shaking and beating vine foliage for 60 s over a 1 m2 cloth funnel equipped at the bottom with a removable plastic bag, hereafter referred to as a ‘beat’ sample, as in Costello & Daane (1997).
Carapace width and body length of spiders were measured before experiments to the nearest 0·1 mm, and spiders were distributed equally among replicates and treatments according to size class (Supporting Information, Table S1). Spider sizes used in the experiment reflected ambient conditions (B. Hogg & K. Daane, unpublished data). To allow time for collection and measurement of spiders, additions of spiders to cages occurred in phases. On 3 August, six spiders per vine of all appropriate species were added for additive treatments and two spiders of each species were added for the replacement-series treatment. An additional three spiders per vine and one spider per vine (for additive and replacement treatments, respectively) were added on 10 August. Grape leafhoppers, collected on leaves from a nearby vineyard, were introduced before spiders were added at 200 nymphs per cage on 25 July, and 80 adults per cage on 27 July. Initial leafhopper numbers were visually sampled (10 leaves/cage) before introducing spiders. To ensure that spiders would not exhaust their food supply, leafhoppers were again added 13 and 28 days after the first spider addition at 100 nymphs per cage on 16 and 31 August. Total numbers of leafhoppers added to cages were far below the maximum leafhopper densities for this site.
The experiment ended immediately before harvest, when harvesting machinery would have destroyed the cages. All arthropods were collected at the end of the experiment using beats, which were performed from 5 to 6 September between 21 : 00 and 6 : 00 h, when lower night temperatures allowed better collection of the winged leafhopper adults. Specimens were stored in plastic bags at −4 °C until sorting. Treatment influences on final numbers of leafhoppers and non-focal spiders (C. inclusum + all other spider species) in additive treatments were analysed using 2 × 2 × 2 factorial mixed-model anova, with the addition of each of the three spider species and their interactions as fixed effects and block as a random effect. Variegated leafhoppers, sharpnosed leafhoppers and non-focal spiders were present in all treatments, and were therefore included in analyses. Numbers of all other taxa were too low to be analysed. Grape, variegated and sharpnosed leafhoppers were analysed separately. All dependent variables were (log10 + 1) transformed prior to analysis to meet assumptions of normality. The blocking factor was dropped when non-significant. Effects of spider species on each other in appropriate additive treatments (i.e. only additive treatments that initially received a particular species were included in analyses for that species) were assessed using 2 × 2 factorial anova with the addition of the other two spider species as factors. The number of grape and variegated leafhoppers remaining in the replacement-series version of the three-species treatment was compared with the best-performing single-species treatment and the mean of the single-species treatments using t-tests and the standard Bonferroni correction to compensate for Type I error. The former comparison tests for whether one predator species dominated effects on herbivores, while the latter tests for non-additive predatory effects; if the effects of predator species in combination diverges from the sum of their individual impacts, the number of herbivores remaining in the replacement-series treatment should differ from the average of the single-species treatments (Griffin et al. 2008). As cages were not entirely effective in excluding C. mildei from cages, and C. mildei was present in all treatments, the effects of C. mildei abundance on leafhoppers and focal spiders were also examined using regression analysis.
To assess effects of spider treatments on leaf damage caused by leafhoppers, five grape leaves per cage were randomly collected at the beginning and end of the experiment. Leaves were stored in plastic bags at −4 °C until they could be scanned and analysed using procedures modified from Skaloudova, Krivan & Zemek (2006). The proportion leaf damage was determined using NIH ImageJ (U.S. National Institutes of Health; http://rsb.info.nih.gov/nih-image/). Proportions were averaged across each cage and arcsine square-root transformed, prior to comparing leaf damage between treatments using a factorial anova.
To examine possible cage effects, (log10 + 1)-transformed numbers of spiders and leafhoppers and arcsine square-root transformed leaf damage proportions were compared between unmanipulated caged and uncaged reference vines using t-tests.
Greenhouse experiment: comparing predatory impacts of exotic and native spiders
In a greenhouse trial, we compared the impacts of the exotic C. mildei and the native A. pacifica on leafhoppers. Spiders were collected from vineyards and surrounding natural vegetation. Hunger levels were standardized ∼24 h before the experiment by providing two fruit flies Drosophila melanogaster Meigen to each spider. The test arenas were Chardonnay cv. grape vines (in 7·6 L pots) enclosed by ∼45 L nylon bags, with plant size (estimated by the number of leaves) and position on the greenhouse table set in a randomized block design before treatments were randomly assigned. Prior to the addition of spiders, each plant was infested with 100 leafhopper nymphs in two installments of 50 nymphs each, 5 days apart. To provide differently sized prey, first, second, third, and fourth instars were added in the ratio of 5 : 10 : 15 : 20 and 10 : 25 : 10 : 5 on the first and second inoculation, respectively. Smaller leafhoppers were deliberately overrepresented, and fifth instars were not used to minimize the number of leafhoppers developing into adults during the experiment.
Spiders were added to appropriate caged plants, 5 days after the second leafhopper installment, in three treatments: no spiders, two C. mildei added, two A. pacifica added. Spiders were allotted to treatments such that spiders within each cage and among all treatments in each block were of similar size (by weight). All spiders were immatures. After a 10-day period, all foliage was clipped (with bags still in place such that no arthropods escaped) and stored at −4 °C for 2 days to kill all arthropods. Afterwards, the numbers of leafhoppers and spiders were recorded. Twelve replicates of each treatment were initially included; if a spider was missing, that replicate was excluded from analysis, resulting in 12, 10 and 11 replicates of control, C. mildei alone and A. pacifica alone treatments, respectively. Effect of treatment on numbers of leafhoppers remaining at the end of the experiment was assessed using anova and a Tukey HSD test for multiple comparisons.