Received: 8 October 2019 | Revised: 7 February 2020 | Accepted: 12 February 2020 DOI: 10.1002/ece3.6158 ORIGINAL RESEARCH Spatial variability of hosts, parasitoids and their interactions across a homogeneous landscape Anna Torné-Noguera1 | Xavier Arnan1 | Anselm Rodrigo1,2 | Jordi Bosch1 1 CREAF, Cerdanyola del Vallès, Spain 2 Abstract Universitat Autònoma de Barcelona, Cerdanyola del Vallès, Spain Species assemblages and their interactions vary through space, generating diversity patterns at different spatial scales. Here, we study the local-scale spatial variation of Correspondence Xavier Arnan, CREAF, Cerdanyola del Vallès, a cavity-nesting bee and wasp community (hosts), their nest associates (parasitoids), Spain. and the resulting antagonistic network over a continuous and homogeneous habi- Email: xavi.arnan@gmail.com tat. To obtain bee/wasp nests, we placed trap-nests at 25 sites over a 32 km2 area. Funding information We obtained 1,541 nests (4,954 cells) belonging to 40 host species and containing Spanish Ministry of Science and Innovation, Grant/Award Number: CGL2009-12646 27 parasitoid species. The most abundant host species tended to have higher para- and CSD2008-00040; FPI, Grant/Award sitism rate. Community composition dissimilarity was relatively high for both hosts Number: BES-2010-042520; RyC, Grant/ Award Number: RYC-2015-18448 and parasitoids, and the main component of this variability was species turnover, with a very minor contribution of ordered species loss (nestedness). That is, local species richness tended to be similar across the study area and community com- position tended to differ between sites. Interestingly, the spatial matching between host and parasitoid composition was low. Host β-diversity was weakly (positively) but significantly related to geographic distance. On the other hand, parasitoid and host-parasitoid interaction β-diversities were not significantly related to geographic distance. Interaction β-diversity was even higher than host and parasitoid β-diversity, and mostly due to species turnover. Interaction rewiring between plots and between local webs and the regional metaweb was very low. In sum, species composition was rather idiosyncratic to each site causing a relevant mismatch between hosts and par- asitoid composition. However, pairs of host and parasitoid species tended to interact similarly wherever they co-occurred. Our results additionally show that interaction β-diversity is better explained by parasitoid than by host β-diversity. We discuss the importance of identifying the sources of variation to understand the drivers of the observed heterogeneity. KEYWORDS beta-diversity, homogeneous habitat, host-parasitoid food web, local scale, spatial variation, species interactions, trap-nests This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited. © 2020 The Authors. Ecology and Evolution published by John Wiley & Sons Ltd. 3696 |  www.ecolevol.org Ecology and Evolution. 2020;10:3696–3705. TORNÉ-NOGUERA et al. | 3697 1 | I NTRO D U C TI O N Mouquet, and Gravel (2012) proposed a dissimilarity index to ex- plore differences between interaction networks across space. To ad- Diversity patterns we observe in nature are the outcome of multi- ditionally delve into the source of variability between networks, they ple biotic and abiotic factors and the interactions occurring among subdivided β-diversity of interactions into two components. The them. Understanding these patterns and their underlying processes first component includes dissimilarity due to species turnover and is one of the main goals of community ecology (Gaston, 1996). For the second dissimilarity due to interaction turnover (same species in- many decades, ecologists have studied species diversity patterns teracting differently, or rewiring). Poisot et al. (2012) also proposed at scales ranging from several meters to thousands of kilometers. the comparison of interactions of each local web to its counterpart Whittaker (1960) was the first to propose a partitioning of diversity in the regional web as a means to determine whether interactions across three different spatial scales: α-diversity, that is, species rich- found at the regional level are always found at the local level (which ness in a particular locality, γ-diversity, a measure of regional species would indicate that species interact similarly across the region). This richness, and β-diversity, a measure of species composition dissimi- approach provides a measure of the extent to which realized and po- larity between localities. β-diversity provides a measure of commu- tential interactions differ, that is, the intensity of interaction filtering. nity spatial variability, reflecting historical processes, and revealing Recently, some studies have empirically addressed β-diversity of information on population dynamics and species responses to habi- mutualistic (CaraDonna et al., 2017; Carstensen et al., 2014; Costa tat modifications such as environmental gradients and perturbations et al., 2018; Dáttilo et al., 2013; Dáttilo & Vasconcelos, 2019; Luna, (Cavender-Bares, Kozak, Fine, & Kembel, 2009; Graham & Fine, Peñaloza-Arellanes, Castillo-Meza, García-Chávez, & Dáttilo, 2018; 2008). For this reason, β-diversity has become a relevant measure Norfolk, Eichhorn, & Gilbert, 2015; Simanonok & Burkle, 2014; in biological conservation (Condit et al., 2002). β-diversity can be Trøjelsgaard et al., 2015) and antagonistic (Novotny, 2009; Poisot partitioned into two additive components: species turnover (dissim- et al., 2012, 2017) interactions. However, our understanding of the ilarity due to species replacement), and nestedness of assemblages mechanisms underlying the distribution of interactions is still frag- (dissimilarity due to ordered species loss; Baselga, 2010). Knowing mentary. For instance, there is no consensus on the relative contri- the relative importance of these two components is essential to un- bution of species turnover and rewiring to interaction dissimilarity. derstand the causes of observed spatial variability. Current evidence suggests that species turnover is more important Because species are not isolated but immersed in complex net- than rewiring across spatial gradients (Novotny, 2009, Poisot et al., works connecting them directly and indirectly with other species, 2012, Carstensen et al., 2014, Trøjelsgaard et al., 2015; but see spatial heterogeneity in species community structure is expected to Dáttilo & Vasconcelos, 2019), whereas rewiring is more important profoundly affect interaction network structure (Olesen & Jordano, across temporal gradients (Caradonna et al., 2017; Costa et al., 2018; 2002; Vázquez, Blüthgen, Cagnolo, & Chacoff, 2009). An increasing Luna et al., 2018). It is also important to understand how the species number of studies are either theoretically or empirically address- turnover of each trophic level contributes to interaction turnover. ing such relationship (Burkle & Alarcón, 2011). For example, spatial This could help find proxies of species interactions and thus improve aggregation, plant identity, and animal mobility have a strong influ- sampling efficiency. So far, most studies have found turnover of the ence on interaction identity, strength, and distribution in plant–ani- lower trophic level to be the major contributor of interaction turn- mal mutualistic networks (Morales & Vázquez, 2008). Others show over (Carstensen et al., 2014; Norfolk et al., 2015; Novotny, 2009; that network structural patterns are largely shaped by relative Simanonok & Burkle, 2014). However, such relationships may be species abundance and spatiotemporal patterns of interacting spe- blurred by methodological artifacts and at least one study attributes cies (CaraDonna et al., 2017; Trøjelsgaard, Jordano, Carstensen, & a greater role to the higher trophic level (Trøjelsgaard et al., 2015). Olesen, 2015; Vázquez, Chacoff, & Cagnolo, 2009). Here, we work with a community of cavity-nesting bees and The decay of similarity in species composition with geographi- wasps (henceforth hosts) and their nest associates, and study the cal distance is well established (Carstensen, Sabatino, Trøjelsgaard, spatial variation of the two communities and of their interactions. & Morellato, 2014; Dáttilo, Guimarães, & Izzo, 2013; Trøjelsgaard Nest associates include species from different arthropod groups et al., 2015), and usually explained as a result of decreasing environ- (coleopterans, dipterans, hymenopterans, mites) and with differ- mental similarity and/or dispersal limitation (Nekola & White, 1999; ent parasitic habits (parasitoids, cleptoparasites, predators, and Soininen, McDonald, & Hillebrand, 2007). However, little is known scavengers). Because they perform a similar ecosystem function about how interaction similarity decays and how this decay is related by contributing to the mortality of the host offspring, nest as- to species similarity decay. Because interactions are influenced by sociates are usually treated as a single group in the cavity-nest- the variability of the two trophic levels plus their inherent variability ing bee/wasp literature, and are referred to as “parasitoids” for (i.e., same species interacting differently), interactions are expected convenience (see Staab, Pufal, Tscharntke, & Klein, 2018 for a to display more spatial variation than species (Poisot, Guéveneux- review). Host-parasitoid systems are suitable models in which to Julien, Fortin, Gravel, & Legendre, 2015; Poisot, Stouffer, & Gravel, analyze spatial variability in species composition and to test the 2017). Although β-diversity is widely used among ecologists to ex- ecological mechanisms accounting for spatial variation in inter- plore community spatial variation, it has been seldom used to ex- actions. First, parasitoids develop on or within the body of their plore the spatial variation of interactions. Poisot, Canard, Mouillot, hosts and therefore are intimately linked to them and experience 3698 | TORNÉ-NOGUERA et al. similar microenvironmental conditions during their development. reason, we also expect low levels of interaction filtering (species This is especially true for nest associates, which develop within interacting similarly across the study region). Fourth, interaction the nest of their hosts. For this reason, it is usually assumed that turnover is often dependent on variation of the lower trophic host-parasitoid interactions are more specialized than other kinds level (Carstensen et al., 2014; Norfolk et al., 2015; Novotny, 2009; of interactions such as predator–prey (Price, 1997; Schowalter, Simanonok & Burkle, 2014), thus we hypothesize (H4) that inter- 1996) and mutualistic (excluding symbionts) (Blüthgen, Menzel, action β-diversity will be better predicted by host β-diversity than Hovestadt, Fiala, & Blüthgen, 2007; Fontaine, Thébault, & Dajoz, by parasitoid β-diversity. 2009). Therefore, we should expect a strong match between the distribution of parasitoids and that of their hosts (Poisot et al., 2017; Poulin, Krasnov, & Mouillot, 2011). 2 | M ATE R I A L A N D M E TH O DS Previous studies working with cavity-nesting bees and wasps analyze composition differences between seasons, habitats, or en- 2.1 | Study area vironmental gradients (Albrecht, Duelli, Schmid, & Müller, 2007, Tylianakis, Tscharntke, & Lewis, 2007, Osorio, Arnan, Bassols, The study was conducted in the Garraf Natural Park (Barcelona, NE Vicens, & Bosch, 2015, Osorio-Canadas et al., 2018, Morris et al., Spain), a Mediterranean scrubland dominated by Quercus coccifera, 2014, Staab et al., 2016; review in Staab et al., 2018). Conversely, Pistacia lentiscus, Rosmarinus officinalis and Thymus vulgaris. Our our study was conducted across a habitat that can be considered study area encompasses a surface of 32 km2. We selected 25 plots uniform (at a landscape scale), with no physical barriers or strong distributed across the study area (Appendix S1). Distances between environmental gradients and dominated by a single vegetation nearest plots ranged from 585 to 1,354 m. As mentioned, the study type, and addresses spatial variability at a local scale (distance be- area is homogeneous at a landscape scale, being occupied by a single tween plots is ca. 1,000 m). Therefore, our aim is to study the in- type of vegetation with similar physiognomic features (Reverté et al., trinsic variability of the host-parasitoid network and its underlying 2019; Torné-Noguera et al., 2014). mechanisms, rather than to establish how different environmental factors may affect interaction identity and network structure. We have the following objectives: (a) To analyze the spatial 2.2 | Surveys variation of the host and parasitoid communities across a con- tinuous habitat; (b) To measure the β-diversity of host-parasitoid In each plot, we placed a trap-nesting station facing SE. Each sta- interactions; (c) To test whether species interact similarly across tion contained seven drilled wood blocks with inserted paper the study region; and (d) To analyze the relationship between tubes. Each wood block accommodated 25 tubes of one of the host, parasitoid and interaction β-diversity. We developed four following diameters (2, 3, 4, 5, 6, 7, and 8 mm), resulting in 175 hypotheses. First, because our study organisms are highly mo- nesting cavities per station. Paper tube length was 5 cm for the bile (Gathmann & Tscharntke, 2002, Klein, Steffan-Dewenter, & 2 and 3 mm diameters and 15 cm for the rest. Nesting stations Tscharntke, 2004, Greenleaf, Williams, Winfree, & Kremen, 2007, were checked every 2 weeks and tubes containing occupied nests Guedot et al., 2009, Cronin & Reeve, 2005), we should expect low were pulled out, taken to the laboratory and replaced with empty spatial variation in community composition. Yet, recent studies in ones, so that there were nesting cavities of all diameters available the same study area reported high levels of spatial heterogeneity at all times. Nesting stations remained in the field from February in pollinator (Reverté et al., 2019) and bee species composition to October, in 2011 and 2013. Data of the two years are pooled (Torné-Noguera et al., 2014), despite similar levels of abundance together in the analyses. and richness. We, thus, hypothesize (H1) moderate to high spatial Nests were kept in a temperature chamber simulating monthly variation in species composition for both taxa, and we expect this ambient temperatures of the study site. Upon one year from col- variation to be mainly due to species turnover (as opposed to nest- lection, nests were dissected. The nests of solitary bees and wasps edness). Second, because parasitoids necessarily depend on their contain a variable number of cells, some of which may be parasitized. hosts and usually have relatively narrow host ranges (Hawkins, We recorded the contents of each cell (identity of the host or the 1994; Jeffs & Lewis, 2013; Poulin et al., 2011), we hypothesize (H2) parasitoid). a strong match in species composition across space and similar distance decay patterns for both groups. Third, we expect higher heterogeneity in interactions than in species composition (Poisot 2.3 | Statistical analysis et al., 2017,2015), and because we work in a spatial (as opposed to temporal) gradient, we hypothesize (H3) a higher contribution Sampling completeness was assessed by building a mean sample- of species turnover than rewiring to overall interaction dissimi- based host species accumulation curve from 1,000 curves based larity. Assuming that host-parasitoid associations are relatively on random sample additions and calculating the Ace Mean species narrow, we expect a similar distance decay pattern for interac- richness estimator (Estimates 8.0.0; Colwell, 2006). We did not build tion similarity than for host and parasitoid similarity. For the same accumulation curves for parasitoids because our sampling approach TORNÉ-NOGUERA et al. | 3699 is host-centered and parasitoids can only be detected if they have of interactions among local networks (βWN). Following Poisot et al. parasitized a host. (2012), this dissimilarity was subdivided into dissimilarity due to spe- cies turnover (βST ) and dissimilarity due to interaction rewiring (βOS). Sometimes, it was not possible to calculate βOS due to the lack of 2.3.1 | Host and parasitoid communities shared species between two plots. In these cases, all network vari- ability (βWN) was assigned to species turnover (βST ). In addition, and To characterize community structure at each nesting station, we again following Poisot et al. (2012), we calculated β’OS, a measure used the variables host abundance (number of host cells produced, of dissimilarity of interactions between local webs and their coun- including both non-parasitized and parasitized cells), host richness terparts in the regional web. β’OS measures the extent to which the (number of host species), parasitoid abundance (number of cells realized and potential interactions differ in each local community parasitized), parasitoid richness (number of parasitoid species), and (i.e., the intensity of interaction filtering across the region). All these parasitism rate (% cells parasitized). indexes were calculated with betalink package for R. We measured the correlation (Spearman's rho) between host To analyze distance decay of the different components of net- abundance and richness (for all hosts and for parasitized hosts sep- work β-diversity, we performed Mantel tests based on Spearman arately), between parasitoid abundance and richness, and between rho between geographical distance and interaction β-diversity (βWN), host (all hosts) and parasitoid abundance and richness. β-diversity due to species turnover (βST ) and β-diversity due to re- To analyze the spatial variation in host and parasitoid species wiring (βOS). composition and the relative contribution of species turnover and We were interested in determining whether interaction β-diver- species nestedness to this variation, we used the species β-di- sity (βWN) was better predicted by the spatial variability of the host versity partition framework proposed by Baselga (2010). To ob- trophic level or that of the parasitoid trophic level. To this aim, we tain a measure of diversity for each site-to-site comparison, we performed two GLMs in which the response variable was interac- calculated β-diversity based on Sørensen's dissimilarity index for tion β-diversity (βWN) and the explanatory variables were parasitoid the host (β H), parasitized host (β PH), and parasitoid (β P) commu- β-diversity (βP) and either host β-diversity (βH) or parasitized host nities. This index accounts for the total compositional variation β-diversity (βPH). Because explanatory variables could be correlated between assemblages, including both turnover and nestedness (see results), we calculated the variance inflation factor (VIF) to make components. β-diversity was then broken down into the compo- sure collinearity was low (VIF = 1.05 for the all hosts model, and 1.17 nent due to species turnover (β SIM) (species replacement), and the for the parasitized hosts model). Geographic distance was not in- component due to species nestedness (β SNE ) (species loss; Baselga, cluded in these models because our results showed it did not have a 2010). β SIM is the Simpson dissimilarity index and is invariant to strong effect on interaction β-diversity (see Results). The percentage richness differences. β SNE is calculated as the difference between of variance explained by the model was calculated with pmvd metric the Sørensen and Simpson indices. These pairwise metrics use in- using relaimpo package (Grömping, 2006) for R. formation on the number of species shared between two sites and the number of species unique to each site. These analyses were conducted with the betapart package (Baselga & Orme, 2012) for 3 | R E S U LT S R (R Core Team, 2014). Correlations between host β-diversity (all hosts (βH) and para- We obtained 1,541 nests (4,954 cells) from 40 host species. Seventeen sitized hosts (βPH) separately) and parasitoid β-diversity (βP) were of these species were bees (Megachilidae—15 sp., Colletidae—2 tested with Mantel tests based on Spearman's rho. To analyze dis- sp.), and 23 were wasps (Crabronidae—12 sp., Pompilidae—5 sp., tance decay of the different components of β-diversity, we again Vespidae—4 sp., Sphecidae—1 sp., Ampulicidae—1 sp.). Twenty-four used Mantel tests to calculate correlations between geographical host species had, at least, one individual parasitized. We found 27 distance and host β-diversity (βH), parasitized host β-diversity (βPH), parasitoid species (16 wasps, 6 flies, 3 beetles, 1 bee, and 1 mite) and and parasitoid β-diversity (βP). 654 parasitized cells, so parasitism rate was 13.2%. Of the 27 parasi- toid species, 18 were true parasitoids, 7 were cleptoparasites, 1 was a predator, and 1 was a scavenger (Figure 1, Appendix S2). The host 2.3.2 | Interactions species accumulation curve approached asymptoticness (Appendix S3). The Ace Mean estimator predicted a total richness of 42 host We first conducted two general linear models (GLMs) to analyze the species, with 98% sampling completeness. relationship between parasitism rate and host abundance at the plot level, and between species abundance and parasitism rate at the species level. 3.1 | Host community We then built an interaction network for each plot (25 local webs), and a regional network (metaweb) pooling the data from all As expected, host abundance and richness showed a positive relation- plots. We explored local web dissimilarity by measuring β-diversity ship, both when considering all hosts (ρ = 0.59, p = .002, n = 25 plots) and 3700 | TORNÉ-NOGUERA et al. F I G U R E 1 Regional host-parasitoids web (metaweb of 25 plots). Species names can be found in Appendix S2. Only parasitized host individuals are included when considering only parasitized hosts (ρ = 0.57, p = .003; n = 25 plots). 3.4 | Distance decay in host and parasitoid All-host β-diversity was moderate to high (βH, mean ± SD = 0.57 ± 0.14), community β-diversity and mostly due to species turnover (βSIM-H = 0.47 ± 0.18, 82.5%), with a poor contribution of nestedness (βSNE-H, mean ± SD = 0.10 ± 0.08, 17.5%). β-diversity of all hosts (βH) (Mantel ρ = 0.29, p = .002) and β-diversity Parasitized host β-diversity was higher (βPH, mean ± SD = 0.68 ± 0.21) of parasitized hosts (βPH) (Mantel ρ = 0.20, p = .004) were signifi- than all-host β-diversity, again mostly due to species turnover (βSIM-PH, cantly but weakly correlated to geographic distance. On the other mean ± SD = 0.56 ± 0.30, 82.4%), with a weak contribution of nested- hand, β-diversity of parasitoids (βP) was not correlated to geographic ness (βSNE-PH, mean ± SD = 0.12 ± 0.16, 17.6%). Therefore, local host distance (Mantel ρ = 0.06, p = .167) (Figure 2). communities showed notable differences in species composition, and these differences were mostly due to species replacement rather than species loss. 3.5 | Interactions Parasitism rate was not explained by host abundance at the plot level 3.2 | Parasitoid community (GLM: p = .75, n = 25 plots). However, host species that were more abundant at the regional level had higher parasitism rates (GLM: As expected, parasitoid abundance and richness were positively cor- p = .02, pseudo-R 2 = 0.13, n = 24 species). related (ρ = 0.57, p = .003). Parasitoid β-diversity was similar to host We detected 73 specific interactions. Most of these (63.9%) β-diversity (βP, mean ± SD = 0.60 ± 0.18), and again mostly attribut- were recorded in just one plot, and 40.3% were recorded only once able to species turnover (βSIM-P, mean ± SD = 0.46 ± 0.26, 76.7%) (i.e., only one individual host was parasitized). Pairwise interaction rather than nestedness (βSNE-P, mean ± SD = 0.14 ± 0.16, 23.3%). dissimilarity between plots was high (βWN, mean ± SD: 0.83 ± 0.16). Most of this dissimilarity was due to species turnover (βST, mean ± SD: 0.62 ± 0.32, 74.7%), with a much lesser contribution of interaction 3.3 | Relationship between host and parasitoid rewiring (βOS, mean ± SD: 0.21 ± 0.31; 25.3%). β’OS values were low β-diversity (between 0 and 0.31; mean ± SD=0.17 ± 0.10). That is, species in- teracted similarly across local networks, indicating weak interaction The parasitoid community was highly dependent on the host commu- filtering across the region. nity: there was a significant positive correlation between parasitoid and host abundance (all hosts: ρ = 0.62, p = .001), and between para- sitoid and host richness (all hosts: ρ = 0.75, p < .0001). β-diversity of 3.6 | Distance decay of interaction β-diversity parasitoids (βP) was significantly but weakly correlated to β-diversity of all hosts (βH) (Mantel ρ = 0.19, p = .011) and to β-diversity of para- As with parasitoid β-diversity, network β-diversity (βWN: Mantel sitized hosts (βPH) (Mantel ρ = 0.38, p = .001). ρ = 0.08, p = .145) was not related to geographical distance. In fact, TORNÉ-NOGUERA et al. | 3701 F I G U R E 2 Distance decay of all- host β-diversity (βH), parasitized host β-diversity (βPH), parasitoid β-diversity (βP), interaction β-diversity (βWN), the component of interaction β-diversity due to species turnover (βST ), and the component of interaction β-diversity due to rewiring (βOS). (Statistical significance; **p ≤ .01) neither of the two components of network β-diversity, β-diversity within a uniform habitat, the data that we collected support our due to species turnover (βST: Mantel ρ = 0.09, p = .109) or β-diversity first hypothesis that both the host and parasitoid communities due to rewiring (βOS: Mantel ρ = −0.03, p = .668) (Figure 2) showed showed relatively high spatial variability at a local scale as indi- significant distance decay. cated by the intermediate to high β-diversity values. Moreover, this high spatial variability was mostly due to species turnover, whereas species loss was almost irrelevant. That is, local commu- 3.7 | Relationship between species and interaction nities did not differ in species richness, but tended to have species β-diversity compositions idiosyncratic to each site. Given that cavity-nest- ing bees and wasps are highly mobile organisms (e.g., Gathmann All-host (βH), parasitized host (βPH) and parasitoid β-diversity (βP) & Tscharntke, 2002; Greenleaf et al., 2007; Klein et al., 2004, were moderately good predictors of interaction β-diversity (βWN) Guedot et al., 2009) one might expect a more uniform host distri- (Table 1, Figure 3). The model including all hosts explained 43.9% bution. However, previous studies in the same study area found of the variance (33.6% by βP and only 10.3% by βH). On the other high spatial variability in local community composition of bees hand, the model including only parasitized hosts explained as much (both cavity- and ground-nesting) (Torné-Noguera et al., 2014) as 65.6% of the variance (24.4% by βP and 41.2% by βPH) (Table 1). and pollinators in general (Reverté et al., 2019). In our study, less than 30% of the variability in host composition can be attributed to foraging distance limitations (as evidenced by the weak cor- 4 | D I S CU S S I O N relation between host β-diversity and geographic distances); the rest of the spatial variability might be attributed to other factors We demonstrate that both our host and parasitoid communi- such as food and nesting resource distribution, immigration his- ties show relatively high spatial variation across a uniform (at tory, philopatry, and topoclimatic variation (Torné-Noguera et al., the landscape level) habitat. Interestingly, the spatial variation of 2014). Information on parasitoid mobility is scarce and controver- parasitoids did only weakly match that of hosts. Host dissimilar- sial (Jeffs & Lewis, 2013), but the general belief is that parasitoids ity increased with geographic distance, whereas parasitoid dis- are less mobile than their hosts (Cronin & Reeve, 2005; Jeffs & similarity did not show a distance decay pattern. Host-parasitoid Lewis, 2013). However, the ability of insects to disperse is usually interactions displayed greater spatial variability than the host and associated to body size (Shegelski, Evenden, & Sperling, 2019) and, parasitoid communities. Interaction dissimilarity did not show in our study, the body size of most parasitoids is similar to that of a distance decay pattern and was mostly explained by species their hosts (unpublished data). We can thus assume similar mobil- turnover with a low contribution of rewiring. In other words, in- ity for hosts and parasitoids. Accordingly, we found similarly high teractions were usually realized whenever the two partners co- spatial variability in parasitoid and in host composition. However, occurred locally, indicating that interaction filtering across space in contrast to hosts, and contrarily to our second hypothesis, par- was very low. Finally, we demonstrate that spatial variability in asitoids did not show any distance decay pattern. host-parasitoid interactions is mostly explained by variability of Ecological theory predicts a distance decay pattern in com- the parasitoid rather than the host community. munity similarity (Nekola & White, 1999; Soininen et al., 2007). Our first objective addressed the spatial variability of the host Distance decay of species composition of the lower trophic level and parasitoid communities. Although the study was conducted seems to be widespread (Carstensen et al., 2014; Dáttilo et al., 3702 | TORNÉ-NOGUERA et al. TA B L E 1 Statistical outputs of Variance GLMs analyzing the contribution of Estimate SE t-value p explained (%) all-host β-diversity (βH), parasitized host Model 1 37.3 β-diversity (βPH) and parasitoid β-diversity βH 0.298 0.049 6.04 <.0001 10.3 (βP) to interaction β-diversity (βWN). βH (Model 1) includes all hosts; βPH (Model 2) βP 0.506 0.040 12.6 <.0001 33.6 includes only parasitized hosts Model 2 62.8 βPH 0.447 0.029 15.7 <.0001 41.2 βP 0.338 0.034 10.0 <.0001 24.4 F I G U R E 3 Relationship between all-host β-diversity (βH), parasitized host β-diversity (βPH), parasitoid β-diversity (βP) (explanatory variables), and interaction β-diversity (βWN) (predicted variable). Model 1 (a, b) includes all hosts (βH). Model 2 (c, d) includes only parasitized hosts (βH). Partial regression plots obtained from linear models when the effect of the other explanatory variable is removed 2013; Novotny, 2009; Trøjelsgaard et al., 2015). On the other hand, specialization (Hawkins, 1994; Jeffs & Lewis, 2013; Poulin et al., composition dissimilarity of the higher trophic level, has been 2011), we might expect a high correspondence between host and found to either increase (Dáttilo et al., 2013; Trøjelsgaard et al., parasitoids communities. Thus, former studies conducted in con- 2015), decrease (Novotny, 2009) or not vary (Carstensen et al., trasted habitat types also showed a high dependency of the com- 2014) with geographical distance. Discrepancies among these munity structure of the higher trophic level on the lower trophic studies, including ours, may be due to the fact that these studies level (Albrecht et al., 2007; Ebeling, Klein, Weisser, & Tscharntke, have been conducted on different biological systems (from more 2012; Osorio et al., 2015; Weiner, Werner, Linsenmair, & Blüthgen, generalized plant–pollinator to more specialized host–parasite 2014). However, and again refuting our second hypothesis, we did systems), in contrasted habitat types (from rupestrian fields and not find a strong matching in species composition between hosts shrublands to dense rainforests), at different levels of community and parasitoids across space, indicating that host-parasitoid sys- isolation (inland versus islands), and at different geographical dis- tems can be less specialized than previously thought (Blüthgen tances (from 600 m to 500 km). Even then, it is surprising that par- et al., 2007; Fontaine et al., 2009; Price, 1997; Schowalter, 1996). asitoids and hosts do not follow similar geographic patterns in our Our “parasitoid” assemblage included mostly true parasitoids (18 study. In relation to this, our results show that the structure of the species), but also some cleptoparasites (7), predators (1), and scav- parasitoid community is highly dependent on the host community, engers (1). Although true parasitoids are often thought of as highly since abundance and richness of the two communities were highly specialized two of the most abundant true parasitoids in our study, correlated. Since parasitoids tend to have high levels of trophic the chalcid wasp Melittobia acasta and the bee-fly Anthrax anthrax TORNÉ-NOGUERA et al. | 3703 had very wide host ranges (13 and 10 host species, respectively, most abundant resource provides an insurance against situations of including both bees and wasps). By comparison, the predator strong species turnover (Canard et al., 2014; Gravel, Massol, Canard, Trichodes leucopsideus and scavenger Trogoderma sp. 1 interacted Mouillot, & Mouquet, 2011). On the other hand, overall parasitism with 6 and 3 host species, respectively, again including both bees rate was not explained by overall host abundance at the plot level. and wasps. At any rate, even in the absence of strong environmen- Other studies on cavity-nesting bees and wasps have found similar tal gradients or apparent dispersal barriers, our metacommunity results (Albrecht et al., 2007; Tylianakis, Tscharntke, & Klein, 2006), was composed of a mosaic of idiosyncratic local communities at a but others have found a positive correlation between parasitism rate small spatial scale. and host abundance (Steffan-Dewenter, 2003). Regarding our second objective, we found that the spatial vari- In relation to our third objective, interaction similarity between ability of host-parasitoid interactions was high (mean βWN = 0.83), local networks and their counterparts in the regional network was and notably higher than that of the host (β H = 0.59) and parasitoid high, as reflected by the low β’OS (mean ± SD = 0.17 ± 0.1). In other (βH = 0.60) communities. Again, species turnover was the main fac- words, interactions in local networks were highly nested in the re- tor contributing to this variability, which agrees with our third hy- gional network, and when partners co-occurred locally, interactions pothesis. There is a vast body of literature on the causes of variation were usually realized. Interaction nestedness in our system resulted in species composition in ecological communities; however, less ef- from a combination of (a) species interacting similarly across plots, and fort has been invested in understanding how interactions between (b) large differences between plots in species composition. Because species vary. Interaction dissimilarity is primarily attributable to local communities were highly idiosyncratic, the contribution of each dissimilarity in species composition, but is also affected by inher- plot to the regional web was high. In a mammal–ectoparasite regional ent variability even when the two species co-occur. Such interac- network, Poisot et al. (2012) also found interactions to be highly id- tion-inherent variability may be due to neutral effects (interactions iosyncratic at the local level. However, the mechanisms explaining are more abundant where populations are more abundant, Canard their result were completely different. As opposed to our study, they et al., 2014), competition effects (Novotny, 2009; Poisot et al., obtained high values of β’OS indicating that many interactions only 2015) and/or environmental conditions differently filtering not only occurred in a reduced number of local webs. They conclude that in- species but also their interactions (Poisot et al., 2017). For these teractions are highly determined by local conditions (high interaction reasons, higher variability in interaction composition than in spe- filtering). Conversely, in our study the proportion of shared species cies composition is expected. Given the high level of specialization between local webs is low (high levels of community dissimilarity), and in antagonistic networks (Blüthgen et al., 2007), species turnover shared species interact similarly (very low levels of rewiring). As a con- would be expected to be the major driver of interaction dissimilarity sequence, even if many interactions only occur in a few local webs, (Olesen, Stefanescu, & Traveset, 2011), with a lesser contribution β’OS values are low. In other words, in our community interactions are of rewiring (shared species interacting differently). Our results (in- site-idiosyncratic not because interaction composition per se is deter- teraction dissimilarity due mostly to species turnover (~75%) rather mined by local conditions, but because species composition is deter- than rewiring (~25%)) agree with this pattern and with previous mined by local conditions. These results agree with a previous study in studies (Novotny, 2009, Norfolk et al., 2015, Simanonok & Burkle, the same area in which composition of the bee community was found 2014, Trøjelsgaard et al., 2015; but see Carstensen et al., 2014 for to be conditioned by local effects (Torné-Noguera et al., 2014). Our re- a similar contribution of both components of interaction diversity). sults are also congruent with a study on cavity-nesting bees and wasps These studies, including ours, cover a variety of interaction systems that found local factors to be much more important than landscape (plant–pollinator, plant–caterpillar, and host-parasitoid) and a range factors in determining community and network structure and compo- of geographical scales (from 400 m to 500 km). Thus, species turn- sition (Osorio et al., 2015). over appears to be the main driver of network β-diversity across As with the parasitoid community, interaction dissimilarity did biological systems and spatial scales of observation. not increase with geographic distance. So far, few studies have ex- In a continuous and homogeneous habitat, we would expect spa- plored interaction similarity decay and they do not provide consis- tial differences in the interaction patterns of species to be explained tent results. Some works found distance decay patterns (Carstensen mostly by neutrality (Canard et al., 2014) and competition, rather et al., 2014, Trøjelsgaard et al., 2015, Dáttilo & Vasconcelos, 2019), than by environmental filtering (Poisot et al., 2017). In relation to but others did not (Dáttilo et al., 2013; Simanonok & Burkle, 2014). this, we found that more abundant host species had higher rates of Although the low number of studies prevents the establishment of parasitism. Such a relationship may be explained by two mechanisms. robust conclusions, the studies that did find interaction similarity First, prey abundance may influence prey choice, as predators with decay patterns were performed at much larger spatial scales that the ability to prey on different hosts obtain greater returns by learn- the ones that did not. ing how to handle the most abundant prey and focusing on it (Ishii Our fourth objective was to establish whether interaction spa- & Shimada, 2012). Second, specialized antagonists, with a restricted tial variability was better explained by the spatial variability of the range of potential preys/hosts, may not be able to build stable pop- host or the parasitoid communities. Our predictive models show that ulations on locally rare hosts (Newman, 2002). Therefore, our meta- both host and parasitoid beta-diversity are important. However, par- community would correspond to systems in which focusing on the asitoids (or parasitized hosts) explain a much greater fraction of the 3704 | TORNÉ-NOGUERA et al. variability, which contradicts our fourth hypothesis that interaction ORCID β-diversity will be better predicted by host β-diversity than by par- Xavier Arnan https://orcid.org/0000-0002-9904-274X asitoid β-diversity. Our results are counter to various studies finding a greater contribution of the lower trophic level (Carstensen et al., REFERENCES 2014; Norfolk et al., 2015; Novotny, 2009; Simanonok & Burkle, Albrecht, M., Duelli, P., Schmid, B., & Müller, C. B. (2007). Interaction diver- 2014). However, these studies reported greater turnover of the lower sity within quantified insect food webs in restored and adjacent inten- sively managed meadows. Journal of Animal Ecology, 76, 1015–1025. trophic level, whereas in our study the host and parasitoid communi- Baselga, A. (2010). Partitioning the turnover and nestedness compo- ties displayed similar levels of turnover. In addition, hosts and para- nents of beta diversity. Global Ecology and Biogeography, 19, 134–143. sitoids in our study appear to be differently affected by geographic Baselga, A., & Orme, C. D. L. (2012). betapart : An R package for the distance. Our results imply that there is no easy short-cut to charac- study of beta diversity. Methods in Ecology and Evolution, 3, 808–812. Blüthgen, N., Menzel, F., Hovestadt, T., Fiala, B., & Blüthgen, N. (2007). terize the spatial distribution of host-parasitoid interactions. Even at Specialization, constraints, and conflicting interests in mutualistic the small spatial scale used in our study, a large fraction of the inter- networks. Current Biology, 17, 341–346. action variability is missed if only hosts (or parasitoids) are sampled. Burkle, L. A., & Alarcón, R. (2011). The future of plant-pollinator diver- Overall, our study demonstrates that communities of mobile or- sity: Understanding interaction networks across time, space, and global change. American Journal of Botany, 98, 528–538. ganisms and their interactions show a high level of variability even Canard, E. F., Mouquet, N., Mouillot, D., Stanko, M., Miklisova, D., & at local scale (~500 m) in a homogeneous habitat. Parasitoid spatial Gravel, D. (2014). Empirical evaluation of neutral interactions in turnover is the major driver of the observed spatial heterogeneity host-parasite networks. American Naturalist, 183, 468–479. in host-parasitoid interactions. Nonetheless, interactions are also CaraDonna, P. J., Petry, W. K., Brennan, R. M., Cunningham, J. L., subjected to their own intrinsic variability. Thus, it is important to Bronstein, J. L., Waser, N. M., & Sanders, N. J. (2017). Interaction rewiring and the rapid turnover of plant–pollinator networks. Ecology take into account both levels of variability (communities and in- Letters, 20, 385–394. teractions) to adequately characterize ecological function. Our Carstensen, D. W., Sabatino, M., Trøjelsgaard, K., & Morellato, L. P. C. results also show the need to sample a large number of plots to ad- (2014). Beta diversity of plant-pollinator networks and the spatial equately characterize a regional network even across a continuous turnover of pairwise interactions. PLoS ONE, 9, e112903. Cavender-Bares, J., Kozak, K. H., Fine, P. V. A., & Kembel, S. W. (2009). and uniform habitat. Distinguishing between the two components The merging of community ecology and phylogenetic biology. of interaction β-diversity is essential to understand the drivers of Ecology Letters, 12, 693–715. such diversity. When interaction β-diversity is mostly due to in- Colwell, R. K. (2006). EstimateS: Statistical Estimation of Species teraction rewiring, the resulting metaweb will reflect the regional, Richness and Shared Species from Samples. Version, 8, 2006. http:// purl.oclc.org/estim​ates but not the local behavior of interacting species. In contrast, when Condit, R., Pitman, N., Leigh, E. G., Chave, J., Terborgh, J., Foster, R. B., interaction β-diversity is mostly due to species turnover, the met- … Muller-Landau, H. C. (2002). Beta-diversity in tropical forest trees. aweb will reflect the local behavior of interacting species (Poisot Science, 295, 666–669. et al., 2012). Our system falls within the latter scenario, with high Costa, J. M., Ramos, J. A., Timóteo, S., da Silva, L. P., Ceia, R. S., & Heleno, R. H. (2018). Species activity promote the stability of levels of species turnover and low levels of rewiring. However, fruit-frugivore interactions across a five-year multilayer network. even then it is necessary to sample a sufficiently large number of BioRxiv, 421941. plots because species composition is highly dependent on local Cronin, J. T., & Reeve, J. D. (2005). Host-parasitoid spatial ecology: A factors and subjected to variation at very small scales. Our results plea for a landscape-level synthesis. Proceedings of the Royal Society B, 272, 2225–2235. have important implications for conservation policy. Even within a Dáttilo, W., Guimarães, P. R., & Izzo, T. J. (2013). Spatial structure of ant- homogeneous landscape it is necessary to preserve large areas of plant mutualistic networks. Oikos, 122, 1643–1648. land to cover the entirety of the regional taxonomic diversity along Dáttilo, W., & Vasconcelos, H. L. (2019). Macroecological patterns and with all potential interactions and ecological functions. correlates of ant-tree interaction networks in Neotropical savannas. Global Ecology and Biogeography, 28, 1283–1294. Ebeling, A., Klein, A.-M., Weisser, W. W., & Tscharntke, T. (2012). AC K N OW L E D G M E N T S Multitrophic effects of experimental changes in plant diversity on We are very grateful to H. Barril-Graells, M. Palamara, R. Novella, I. cavity-nesting bees, wasps, and their parasitoids. Oecologia, 169, Lobato, J. Ramoneda, M. Escolà, S. Reverté, A. López, and B. Cuadra for 453–465. their help in the field and in the laboratory. We also thank N. Blüthgen Fontaine, C., Thébault, E., & Dajoz, I. (2009). Are insect pollinators more generalist than insect herbivores? Proceedings of the Royal Society B: (Technical University of Darmstadt) for his analytical advice and J. Biological Sciences, USA, 276, 3027–3033. Mederos (Museum of Natural Sciences, Barcelona) and P. Bogusch Gaston, K. J. (1996). Biodiversity–latitudinal gradients. Progress in Physical (Univrsity of Hradec Králové) for their help with Diptera and Mutillidae Geography, 20, 466–476. identification, respectively. We are also thankful to Diputació de Gathmann, A., & Tscharntke, T. (2002). Foraging ranges of solitary bees. Journal of Animal Ecology, 71, 757–764. Barcelona for permission to work in the Natural Park of el Garraf. Graham, C. H., & Fine, P. V. A. (2008). Phylogenetic beta diversity: Linking ecological and evolutionary processes across space in time. Ecology DATA AVA I L A B I L I T Y S TAT E M E N T Letters, 11, 1265–1277. Host-parasitoid interactions data: Dryad https://doi.org/10.5061/ Gravel, D., Massol, F., Canard, E., Mouillot, D., & Mouquet, N. (2011). Trophic theory of island biogeography. Ecology Letters, 14, 1010–1016. dryad.kprr4​xh1t. TORNÉ-NOGUERA et al. | 3705 Greenleaf, S. S., Williams, N. M., Winfree, R., & Kremen, C. (2007). Bee across a continuous habitat: High heterogeneity in the face of appar- foraging ranges and their relationship to body size.Oecologia, 153, ent uniformity. Ecography, 42, 1558–1568. 589–596. Schowalter, T. D. (1996). Insect ecology. An ecosystem approach. London, Grömping, U. (2006). Relative importance for linear regression in R: The UK: Academic Press. package relaimpo. Journal of Statistical Software, 17, 139–147. Shegelski, V. A., Evenden, M. L., & Sperling, F. A. (2019). Morphological Guedot, C., Bosch, J., & Kemp, W. P. (2009). Relationship between variation associated with dispersal capacity in a tree-killing bark body size and homing ability in the genus Osmia (Hymenoptera; beetle Dendroctnous ponderosae Hopkins. Agricultural and Forest Megachilidae). Ecological Entomology, 34, 158–161. Entomology, 21, 79–87. Hawkins, B. A. (1994). Pattern and process in host-parasitoid interactions. Simanonok, M. P., & Burkle, L. A. (2014). Partitioning interaction turnover Cambriadge, UK: Cambridge University Press. among alpine pollination networks: Spatial, temporal, and environ- Ishii, Y., & Shimada, M. (2012). Learning predator promotes coexistence mental patterns. Ecosphere, 5, 1–17. of prey species in host – parasitoid systems. Proceedings of the Soininen, J., McDonald, R., & Hillebrand, H. (2007). The distance decay of National Academy of Sciences, USA, 109, 5119. similarity in ecological communities. Ecography (Cop.), 30, 3–12. Jeffs, C. T., & Lewis, O. T. (2013). Effects of climate warming on host-par- Staab, M., Bruelheide, H., Durka, W., Michalski, S., Purschke, O., Zhu, C.- asitoid interactions. Ecological Entomology, 38, 209–218. D., & Klein, A.-M. (2016). Tree phylogenetic diversity promotes host– Klein, A. M., Steffan-Dewenter, I., & Tscharntke, T. (2004). Foraging trip parasitoid interactions. Proceedings of the Royal Society B: Biological duration and density of megachilid bees, eumenid wasps and pompilid Sciences, 283, 20160275. wasps in tropical agroforestry systems. Journal of Animal Ecology, 73, Staab, M., Pufal, G., Tscharntke, T., & Klein, A.-M. (2018). Trap nests for 517–525. bees and wasps to analyse trophic interactions in changing environ- Luna, P., Peñaloza-Arellanes, Y., Castillo-Meza, A. L., García-Chávez, J. ments – a systematic overview and user guide. Methods in Ecology H., & Dáttilo, W. (2018). Beta diversity of ant-plant interactions over and Evolution, 9, 2226–2239. day-night periods and plant physiognomies in a semiarid environ- Steffan-Dewenter, I. (2003). Importance of habitat area and landscape ment. Journal of Arid Environments, 156, 69–76. context for species richness of bees and wasps in fragmented or- Morales, J. M., & Vázquez, D. P. (2008). The effect of space in plant-an- chard meadows. Conservation Biology, 17, 1036–1044. imal mutualistic networks: Insights from a simulation study. Oikos, Torné-Noguera, A., Rodrigo, A., Arnan, X., Osorio, S., Barril-Graells, H., 117, 1362–1370. da Rocha-Filho, L. C., & Bosch, J. (2014). Determinants of spatial dis- Morris, R. J., Gripenberg, S., Lewis, O. T., & Roslin, T. (2014). Antagonistic tribution in a bee community: Nesting resources, flower resources, interaction networks are structured independently of latitude and and body size. PLoS ONE, 9, e97255. host guild. Ecology Letters, 17, 340–349. Trøjelsgaard, K., Jordano, P., Carstensen, D. W., & Olesen, J. M. (2015). Nekola, J. C., & White, P. S. (1999). The distance decay of similarity in Geographical variation in mutualistic networks: Similarity, turnover biogeography and ecology. Journal of Biogeography, 26, 867–878. and partner fidelity. Proceedings of the Royal Society B-Biological Newman, M. E. J. (2002). Spread of epidemic disease on networks. Sciences, 282, 20142925. Physical Review E, 66, 016128. Tylianakis, J. M., Tscharntke, T., & Klein, A. M. (2006). Diversity, ecosys- Norfolk, O., Eichhorn, M. P., & Gilbert, F. S. (2015). Contrasting pat- tem function, and stability of parasitoid-host interactions across a terns of turnover between plants, pollinators and their interactions. tropical habitat gradient. Ecology, 87, 3047–3057. Diversity and Distribution, 21, 405–415. Tylianakis, J. M., Tscharntke, T., & Lewis, O. T. (2007). Habitat modifi- Novotny, V. (2009). Beta diversity of plant-insect food webs in tropical cation alters the structure of tropical host-parasitoid food webs. forests: A conceptual framework. Insect Conservation and Diversity, Nature, 445, 202–205. 2, 5–9. Vázquez, D. P., Blüthgen, N., Cagnolo, L., & Chacoff, N. P. (2009). Uniting Olesen, J. M., & Jordano, P. (2002). Geographic patterns in plant-pollina- pattern and process in plant-animal mutualistic networks: A review. tor mutualistic networks. Ecology, 83, 2416–2424. Annals of Botany, 103, 1445–1457. Olesen, J. M., Stefanescu, C., & Traveset, A. (2011). Strong, long-term Vázquez, D. P., Chacoff, N. P., & Cagnolo, L. (2009). Evaluating multiple temporal dynamics of an ecological network. PLoS ONE, 6, 2–6. determinants of the structure of plant-animal mutualistic networks. Osorio, S., Arnan, X., Bassols, E., Vicens, N., & Bosch, J. (2015). Local and Ecology, 90, 2039–2046. landscape effects in a host-parasitoid interaction network along a Weiner, C. N., Werner, M., Linsenmair, K. E., & Blüthgen, N. (2014). Land- forest-cropland gradient. Ecological Applications, 25, 1869–1879. use impacts on plant-pollinator networks: Interaction strength and Osorio-Canadas, S., Arnan, X., Bassols, E., Vicens, N., & Bosch, J. (2018). specialization predict pollinator declines. Ecology, 95, 466–474. Seasonal dynamics in a cavity-nesting bee-wasp community: Shifts Whittaker, R. H. (1960). Vegetation of the Siskiyou mountains, Oregon in composition, functional diversity and host-parasitoid network and California. Ecological Monographs, 30, 279–338. structure. PLoS ONE, 13, e0205854. Poisot, T., Canard, E., Mouillot, D., Mouquet, N., & Gravel, D. (2012). The dissimilarity of species interaction networks. Ecology Letters, 15, S U P P O R T I N G I N FO R M AT I O N 1353–1361. Additional supporting information may be found online in the Poisot, T., Guéveneux-Julien, C., Fortin, M. J., Gravel, D., & Legendre, P. Supporting Information section. (2017). Hosts, parasites and their interactions respond to different climatic variables. Global Ecology and Biogeography, 26, 942–951. Poisot, T., Stouffer, D. B., & Gravel, D. (2015). Beyond species: Why eco- logical interaction networks vary through. Oikos, 124, 243–251. How to cite this article: Torné-Noguera A, Arnan X, Rodrigo A, Poulin, R., Krasnov, B. R., & Mouillot, D. (2011). Host specificity in phy- Bosch J. Spatial variability of hosts, parasitoids and their logenetic and geographic space. Trends in Parasitology, 27, 355–361. interactions across a homogeneous landscape. Ecol Evol. Price, P. W. (1997). Insect ecology. New York, NY: Wiley-Blackwell. 2020;10:3696–3705. https://doi.org/10.1002/ece3.6158 R Core Team (2014). R: A language and environment for statistical comput- ing. Vienna, Austria: Foundation for Statistical Computing. Reverté, S., Bosch, J., Arnan, X., Roslin, T., Stefanescu, C., Calleja, J. A., … Rodrigo, A. (2019). Spatial variability in a plant-pollinator community