Simplifying understory complexity in oil palm plantations is associated with a reduction in the density of a cleptoparasitic spider, Argyrodes miniaceus (Araneae: Theridiidae), in host (Araneae: Nephilinae) webs

Abstract Expansion of oil palm agriculture is currently one of the main drivers of habitat modification in Southeast Asia. Habitat modification can have significant effects on biodiversity, ecosystem function, and interactions between species by altering species abundances or the available resources in an ecosystem. Increasing complexity within modified habitats has the potential to maintain biodiversity and preserve species interactions. We investigated trophic interactions between Argyrodes miniaceus, a cleptoparasitic spider, and its Nephila spp. spider hosts in mature oil palm plantations in Sumatra, Indonesia. A. miniaceus co‐occupy the webs of Nephila spp. females and survive by stealing prey items caught in the web. We examined the effects of experimentally manipulated understory vegetation complexity on the density and abundance of A. miniaceus in Nephila spp. webs. Experimental understory treatments included enhanced complexity, standard complexity, and reduced complexity understory vegetation, which had been established as part of the ongoing Biodiversity and Ecosystem Function in Tropical Agriculture (BEFTA) Project. A. miniaceus density ranged from 14.4 to 31.4 spiders per square meter of web, with significantly lower densities found in reduced vegetation complexity treatments compared with both enhanced and standard treatment plots. A. miniaceus abundance per plot was also significantly lower in reduced complexity than in standard and enhanced complexity plots. Synthesis and applications: Maintenance of understory vegetation complexity contributes to the preservation of spider host–cleptoparasite relationships in oil palm plantations. Understory structural complexity in these simplified agroecosystems therefore helps to support abundant spider populations, a functionally important taxon in agricultural landscapes. In addition, management for more structurally complex agricultural habitats can support more complex trophic interactions in tropical agroecosystems.

Cleptoparasites such as Argyrodes spp. (Araneae: Theridiidae) spiders may be particularly sensitive to habitat change due to their direct reliance on host spider success (Sheffield et al., 2013). Argyrodes are obligate cleptoparasitic spiders that depend solely on food resources caught by their hosts and living space provided by their hosts (Agnarsson, 2003;Cangialosi, 1997;Vollrath, 1987b). Yet, Argyrodes can also have profound negative effects on host fitness (Elgar, 1989;McCrate & Uetz, 2010;Rittschof & Ruggles, 2010;Rypstra, 1981;Tanaka, 1984). Argyrodes reach densities of up to 40 individuals per square meter of their host's web, and even at much lower densities, can consume enough prey to significantly impact host growth, web tenure, web damage, and host mortality (Agnarsson, 2003(Agnarsson, , 2011Grostal & Walter, 1997;Koh & Li, 2002;Miyashita, 2001;Rypstra, 1981). Argyrodes may also indirectly reduce host reproductive success: Male Nephila spiders, which are substantially smaller than females and co-occupy female webs, also act as cleptoparasites, thereby directly competing with resident Argyrodes for access to food (Christenson, Brown, Wenzl, Hill, & Goist, 1985). Because of the sensitive balance of competition with and dependence on hosts, any change in the interactions between cleptoparasites and their hosts could be an early indicator of changing trophic interactions within modified ecosystems (Sheffield et al., 2013;Tylianakis et al., 2007Tylianakis et al., , 2008. In addition, owing to their role as generalist predators, any changes in the trophic interactions of spiders have the potential to impact pest control in agricultural systems (Cardinale, Harvey, Gross, & Ives, 2003;Denno et al., 2005;Jonsson, Wratten, Landis, & Gurr, 2008). While Argyrodes, because they consume insects already caught in host webs, are not likely to be critical pest control agents, their impacts on Nephila host health or behavior have the potential to alter host pest control potential. Although little is yet known about the role of spiders in oil palm specifically, they can be important pest control agents in other tropical (Hlivko & Rypstra, 2003;Kobayashi, 1975;Settle et al., 1996;Sigsgaard, 2000) and tree (Mansour, Rosen, & Shulov, 1980;Mathews, Bottrell, & Brown, 2004) crops.
Argyrodes abundance is known to be positively correlated with web size, host body size, host density, and prey availability (Agnarsson, 2003(Agnarsson, , 2011Cangialosi, 1990Cangialosi, , 1991Elgar, 1989;Grostal & Walter, 1997;Hénaut, Delme, Legal, & Williams, 2005;Koh & Li, 2002), but the overarching effects of habitat structure on density and abundance are less well understood (Agnarsson, 2011;Cangialosi, 1997;Miyashita, 2002;Rittschof & Ruggles, 2010). Here, we present the first study on the effects of habitat complexity in an oil palm agroecosystem on host-parasite interactions between a spider and a cleptoparasite. We examine the impact of oil palm understory vegetation complexity as well as host characteristics (host web size and male Nephila presence) on Argyrodes miniaceus (Doleschall, 1857) cleptoparasites in the webs of female Nephila spp. hosts (Koh & Li, 2002;Miyashita, 2002;Rypstra 1981). We make the following hypotheses about the effects of these environmental factors on A. miniaceus density and abundance: agricultural management, habitat complexity, host-parasite relationships, population density, trophic interactions 1. Greater vegetation complexity allows for greater cleptoparasite density and abundance.

2.
Cleptoparasite density is constant or greater in larger webs.

3.
As male Nephila also act as cleptoparasites in female webs, and so benefit from similar environmental conditions to Argyrodes, their presence is positively associated with Argyrodes density.
By quantifying the effect of understory complexity on a host-parasite relationship, this study will yield novel insights into the effect of habitat structure and management on food web complexity in a tropical agricultural landscape.

| Study site
This study was conducted as part of the Biodiversity and Ecosystem   (Marshall, Brown, Boatman, Lutman, & Squire, 2001), at least one study indicates little to no direct effect of herbicide application on leaf-litter invertebrate communities (Lindsay & French, 2004).

| Sampling protocol
To measure cleptoparasite density, we walked along every row of oil palm trees in each 150 m x 150 m plot in March 2016, two years after the experimental treatments had been implemented. Tree rows were planted approximately 8 m apart, with each tree in a row also approximately 8 m apart. We noted every adult Nephila spp. web within the plot that was less than 3 m above the observer's head and was over 10 cm in both length and width. Species of Nephila present in the study plots included N. pilipes (Fabricius, 1793) and N. kuhlii (Doleschal, 1859). Argyrodes miniaceus was the only species of Argyrodes present on Nephila spp. webs in the study plots. Webs consist of a large central orb surrounded by varying amounts of barrier webbing. To obtain a rough estimate of the size of the web, we measured the length and width of the central orb of every web. We did not measure the size of barrier webbing, and we did not measure distances between webs, although we observed no interconnected webs. We counted all A. miniaceus cleptoparasites that were within the central orb and on the surrounding barrier webbing, including any web attachment threads.
We also noted whether or not any male Nephila were present in the web of the larger female host.

| Analyses
All analyses were conducted using R version 3.3.0 (R-Core-Team 2016). We calculated cleptoparasite density per web by dividing the number of cleptoparasites by the web area in square meters (calculated as orb length × orb width). To determine what factors best predict cleptoparasite density, we created a linear mixed effects model with square root transformed cleptoparasite density data as the response variable, using the R package "lme4" (Bates, Maechler, Bolker, & Walker, 2015). We tested treatment, host web size (m 2 ), and the presence of a male in the host web as potential explanatory variables and included triplet as a random effect.
We used an automatic drop-in-deviance test to select the best-fit model. The automatic drop-in-deviance test used a series of likelihood ratio tests to determine which variables most improved the model's goodness of fit. We chose the variable with the lowest reported Akaike's Information Criterion (AIC) to add to the model and repeated the test, adding variables one at a time until no more factors were significant at the α = 0.05 level (Burnham & Anderson, 1998). We then used additional likelihood ratio tests to determine whether including interactions between any of the selected explanatory variables increased the model's goodness-of-fit. The final model was the model with the lowest AIC value.
To examine differences in mean cleptoparasite density per web and abundance per plot among treatments, we conducted ANOVA tests on square-root transformed density data and untransformed abundance data, with triplet as a random effect. We square root transformed the density data to avoid violating assumptions of normality of residuals and homoscedasticity. Because the effect of a treatment can be underestimated by p-values for small sample sizes (Gelman & Stern, 2006;Ioannidis, 2005), we considered results marginally significant for p-values below α = 0.1 for ANOVA tests on per-plot abundance (where n = 6 plots). After any significant or marginally significant results, additional ANOVAs were used to test pairwise differences in A. miniaceus density and abundance between treatments.

| RESULTS
We  (Table 1). No interactions between variables were significant. Model parameter coefficients indicate that, on average, larger webs have lower cleptoparasite densities, and male presence is associated with higher cleptoparasite densities (Table 1).
host webs in oil palm plantations. Understory vegetation complexity, host web size, and male Nephila presence were all significant predictors of A. miniaceus density in Nephila spp. host webs. Greater F I G U R E 1 Boxplots depicting medians and interquartile ranges of Argyrodes miniaceus cleptoparasite density per square meter of Nephila spp. web in BEFTA Project plots of enhanced complexity (n = 89 webs), standard complexity (n = 96 webs), and reduced complexity (n = 28 webs) understory management treatments. Letters (a or b) indicate significant differences in means: a different letter indicates a difference from the others F I G U R E 2 Boxplots depicting medians and interquartile ranges of Argyrodes miniaceus cleptoparasite abundance per plot in BEFTA Project plots of enhanced complexity (n = 6 plots), standard complexity (n = 6 plots), and reduced complexity (n = 6 plots) understory management treatments. Letters (a or b) indicate significant or marginally significant differences in means: a different letter indicates a difference from the others cleptoparasite density was associated with enhanced and standard levels of understory complexity, smaller webs, and the presence of a male Nephila spp. in the web. In addition, total cleptoparasite abundance differed across understory complexity treatments, with significantly fewer cleptoparasites in reduced understory complexity plots.

| Cleptoparasite density and vegetation complexity
Our results suggest that greater habitat complexity in oil palm plantations supports higher populations of cleptoparasitic spiders, both per host web and in terms of total abundance. This finding adds to a body of literature demonstrating that vegetation complexity and diversity support complex food webs, although these studies were primarily conducted in temperate ecosystems (Macfadyen, Gibson, Symondson, & Memmott, 2011;Memmott et al., 2007;Pocock, Evans, & Memmott, 2012). For example, a study by Goulson, Hughes, Derwent, and Stout (2002) found that an increase in floral resources in suburban and agricultural habitats in the United Kingdom increased both the abundance of native bumblebees and the abundance of their specialist parasites. Ebeling, Klein, Weisser, and Tscharntke (2012) also report that greater plant diversity in German grasslands increases not only the diversity of host bee and wasp species, but the diversity of their parasitoids as well. Our study indicates that a similar relationship exists in tropical agroecosystems.  Hansen, 2000;Landis, Wratten, & Gurr, 2000;Langellotto & Denno, 2004;Weibull, Östman, & Granqvist, 2003). Argyrodes populations are limited by competition for food resources-both with the host and with other cleptoparasites (Miyashita, 2001;Whitehouse, 1997)-and so an increased abundance of prey would allow each Nephila web to support a greater density of Argyrodes cleptoparasites (Cangialosi, 1991;Miyashita, 2001;Whitehouse, 1988Whitehouse, , 1997. Distance between host webs may also play a role in cleptoparasite density. Isolated habitat patches are less likely to be inhabited by any given species (Prugh, Hodges, Sinclair, & Brashares, 2008;Watling & Donnelly, 2006) and are expected to exhibit high extinction and low immigration rates (Brown & Kodric-Brown, 1977;Fahrig & Merriam, 1985;Hanski, 1999). Although we did not directly measure distances between host webs, the observed lower abundance and therefore density of Nephila host webs in reduced complexity plots suggests webs in these plots are more isolated than in the densely populated enhanced complexity plots. If we consider host webs as habitat patches, the isolation of host webs in reduced complexity plots may be an additional contributing factor to lower cleptoparasite density. However, previous research has indicated that host web isolation does not correlate with cleptoparasite abundance (Agnarsson, 2011). Future research might more closely examine the relationship between host web interdistance and cleptoparasite density.

| Web size, male presence, and cleptoparasite density
Argyrodes miniaceus density was also correlated with host web size: Density was greater in smaller webs. Previous research suggests that cleptoparasite density should remain constant across host webs, due to the strong linear relationship between web size and cleptoparasite abundance (Agnarsson, 2011). In several studies, removal or addition of Argyrodes spp. individuals to host webs resulted in nearly immediate immigration to or emigration from the web, keeping density constant (Miyashita, 2001(Miyashita, , 2002Rypstra, 1985). Argyrodes often behave aggressively toward each other, with population density limited by competition for food and space (Miyashita, 2001(Miyashita, , 2002. The observed disparity in cleptoparasite density among host webs of different sizes is therefore surprising. We found no significant differences in web size across treatments (D.M. Spear unpublished data), and so the observed trend cannot be explained by disparities in average Nephila web size across our treatments. Female Nephila host body size may be a contributing factor to the observed differences in density. Large webs are typically occupied by larger or older hosts (Eberhard, 1972;Grostal & Walter, 1999;Kuntner, Gregorič, & Li, 2010;Moore, 1977;Witt, Reed, & Peakall, 2012), which can behave more aggressively toward cleptoparasites than younger "naïve" females, predating upon or chasing occupants from webs (Cangialosi, 1991;Vollrath, 1979Vollrath, , 1987bWhitehouse, 1988Whitehouse, , 1997. It is possible that these older hosts may more effectively limit the density of Argyrodes within their webs.
Further research is necessary to determine the potential interactive effects of host web size and habitat complexity on cleptoparasite density. Enhanced complexity treatment was used as the reference category during model construction, so was assigned a coefficient of zero.
The correlation between male Nephila presence and Argyrodes density may also reflect differences in food resources among treatments. Males frequently act as cleptoparasites in females' webs, and so benefit from similar factors to cleptoparasites, such as prey abundance (Christenson et al., 1985;Elgar et al., 2003;Vollrath, 1987b).
Webs that are preferable to males may be located in areas of high resource abundance, and so may be equally suitable for high densities of cleptoparasites. Grostal and Walter (1999), who similarly found a positive correlation between presence of Nephila plumipes host males and abundance of Argyrodes antipodianus cleptoparasites, also propose that males may distract the female host, thus decreasing the risk of predation by Nephila females for Argyrodes inhabitants and making male-occupied webs safer. An increased cleptoparasite load therefore may constitute an additional cost of reproduction for Nephila females.
However, high occurrence of both Nephila spp. and cleptoparasites in enhanced and standard complexity plots suggests that any detrimental effects of an increased cleptoparasite load are not severe enough to significantly limit Nephila population levels. Resource availability in enhanced complexity plots must therefore be great enough to support both high cleptoparasite abundance and high Nephila density, providing evidence that higher vegetation complexity increases the ability of oil palm ecosystems to support both more predators and more complex trophic interactions. The lower densities of cleptoparasites per web area in reduced complexity plots suggest that fewer resources were available to support either the hosts or the parasites. Because cleptoparasites rely on taking any remaining food once their host's energy requirements are met, cleptoparasites are likely to be more sensitive to habitat change than their hosts (Sheffield et al., 2013). Thus, changes to cleptoparasite abundance could be a first indicator of changing trophic interactions across modified ecosystems (Sheffield et al., 2013;Tylianakis et al., 2007Tylianakis et al., , 2008.

| CONCLUSIONS
While increased vegetation complexity and diversity provides beneficial resources to spiders in agricultural systems (Diehl, Mader, Wolters, & Birkhofer, 2013;Rypstra, Carter, Balfour, & Marshall, 1999;Schmidt, Roschewitz, Thies, & Tscharntke, 2005), this study shows that vegetation complexity also has the potential to increase cleptoparasite loads. The most likely reason for this is that maintaining understory vegetation in oil palm plantations provides additional resources that support a greater abundance and density of Nephila hosts, and also a greater number of cleptoparasites per web. This study demonstrates the potential of within-plantation management to increase the complexity of tropical food webs, increasing the abundance of both predators and the parasites they support, with potential impacts on the ecosystem services that predators provide.

CONFLICT OF INTEREST
None declared.

AUTHORS' CONTRIBUTIONS
DMS helped design the study, conducted the fieldwork, analyzed data, and drafted the manuscript; ECT and WAF coordinated and helped with all aspects of the study; SHL helped design the study, draft the manuscript, and analyzed the data; ADA helped design the study and conduct fieldwork; MN, JPC, JLS, ADA, and SP all helped design the study, and helped design, coordinate, and maintain BEFTA Project plots and treatments. All authors contributed critically to the drafts and gave final approval for publication.

DATA ACCESSIBILITY
The data used in this study will be archived in the Dryad Digital Repository.