Reciprocity and Interaction Effectiveness in Generalised Mutualisms Among Free-living Species

132 | Ecology Letters. 2023;26:132–146.wileyonlinelibrary.com/journal/ele INTRODUCTION Mutualisms are ecological interactions entailing beneficial outcomes for the interacting partners. These benefits broadly emerge from interspecific encounters where there is an exchange of resources (Kiers et al., 2011). Despite recent interest in interspecific exchanges, especially focusing on strict and intimate interactions (Guimarães et al., 2007), much of the reciprocal effect between generalised, freeliving, mutualistic partners remains unexplored (Thompson,2009). Specieslevel analyses of complex interaction networks have revealed highly heterogeneous structures (i.e. high variance in number of interactions per species), weak levels of mutual dependence, and high asymmetry in interaction strength (Bascompte & Jordano,2014; Johnstone & Bshary,2008; Wootton & Stouffer,2016). Interaction asymmetry in complex networks of freeliving species (Bascompte et al.,2006), as well as energy flow asymmetry in food webs (Rooney et al.,2006), appear as quintessential characteristics of these complex systems, closely associated with their stability (Berlow, 1999). Yet, our understanding of the factors behind the emergence of asymmetric interactions is very limited; for example, if generalised mutualistic interactions between freeliving species entail exchanges of services, is there a “fair twoway transfer” of resources (Chomicki et al.,2020; Kiers et al.,2011), that is, is there reciprocity? Reciprocity, as defined herein, is the existence of a positive association in the rewards provided between mutualistic partners. We consider a mutualistic system to be reciprocal if the reward provided by one organism LETTER Reciprocity and interaction effectiveness in generalised mutualisms among freeliving species ElenaQuintero1 | FranciscoRodríguez- Sánchez1,2 | PedroJordano1,2 Received: 11 April 2022 | Revised: 12 September 2022 | Accepted: 24 October 2022 DOI: 10.1111/ele.14141 This is an open access article under the terms of the Creative Commons Attribution- NonCommercial License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited and is not used for commercial purposes. © 2022 The Authors. Ecology Letters published by John Wiley & Sons Ltd. Both Francisco Rodríguez- Sánchez and Pedro Jordano are senior authors. 1Integrative Ecology Group, Estación Biológica de Doñana, Sevilla, Spain 2Departamento de Biología Vegetal y Ecología, Facultad de Biología, Universidad de Sevilla, Sevilla, Spain Correspondence Elena Quintero, Integrative Ecology Group, Estación Biológica de Doñana, EBDCSIC, Av. Américo Vespucio 26, E- 41092 Sevilla, Spain. Email: [email protected] Funding information Agencia Estatal de Investigacion, Spain, Grant/Award Number: CGL2017- 82847- P; Universidad de Sevilla / Junta de Andalucía / FEDER EU, Grant/Award Number: US- 1381388; FEDER EU, Grant/Award Number: LIFEWATCH- 2019- 09- CSIC- 13; Fundación la Caixa, Grant/Award Number: LCF/BQ/DE18/11670007; VI Plan Propio de Investigación (Universidad de Sevilla), Grant/Award Number: VI PPITUS - 2018- IV.2 Abstract Mutualistic interactions among freeliving species generally involve lowfrequency interactions and highly asymmetric dependence among partners, yet our understanding of factors behind their emergence is still limited. Using individualbased interactions of a supergeneralist fleshyfruited plant with its frugivore assemblage, we estimated the Resource Provisioning Effectiveness (RPE) and Seed Dispersal Effectiveness (SDE) to assess the balance in the exchange of resources. Plants were highly dependent on a few frugivore species, while frugivoresinteracted with most individual plants, resulting in strong asymmetries of mutual dependence. Interaction effectiveness was mainly driven by interaction frequency. Despite highly asymmetric dependences, the strong reliance on quantity of fruit consumed determined high reciprocity in rewards between partners (i.e. higher energy provided by the plant, more seedlings recruited), which was not obscured by minor variations in the quality of animal or plant service. We anticipate reciprocity will emerge in lowintimacy mutualisms where the mutualistic outcome largely relies upon interaction frequency. KEYWORDS asymmetry, Bayesian, ecological networks, frugivory, Mediterranean, plant– animal interactions, reciprocity, resource provisioning effectiveness, seed dispersal, seed dispersal effectiveness Editor: Ferenc Jordan | 133 QUINTERO et al. (e.g. pollen grains or fruits offered by plants) matches the reward from its mutualistic partner (e.g. fertilised ovules or dispersed seeds). If an increase or decrease in reward does not return proportional changes in the reward by the other partner, so that both rewards keep balanced, those interactions would be less, or not reciprocal at all. Without an external reference, it is not possible to determine if the exchange in resources between partners is equal or fair. A population or community perspective will allow us to understand whether specific pairwise interactions are exchanging their resources at ‘fair’ cost, or at least at the cost set by the population or community. Aside from previous work on mycorrhizal symbioses, the reciprocity ofless intimate and ‘lagged’ mutualisms (i.e. with delayed responses beyond the interaction) has been rarely addressed. However, previous studies explore other definitions of reciprocity using different approximations that are related to the degree of partner's dependence and not to the balance in rewards exchanges (e.g. Burns, 2003; Guerra & Pizo, 2014; Herrera, 1984; Reid,1990). Partner dependence, that is how much a partner relies upon another partner for its services, is a betterexplored aspect of mutualistic interactions. Dependence can be estimated as the proportion of service obtained from a specific partner relative to the total service obtained from all partners. Dependence differs from reciprocity in that it examines the reliance from the perspective of the partner, and not the whole population. Estimating dependence also allows the calculation of interactions asymmetry by comparing the mutual dependence of both partners. Asymmetry emerges when a species/organism depends a lot on one partner but, in turn, the partner does not rely as much on that particular pairwise interaction (Bascompte et al.,2006; Jordano,1987; Vázquez et al.,2007). A generalised property of freeliving species networks is the high frequency of weak interactions (Jordano,1987) so that when other interactions are strong, their mudependence becomes highly asymmetric. This pattern in the mode of interaction between organisms is known as disassortativity, whereby organisms that establish many interactions tend to interact with less connected organisms (Barabási, 2016), and is often found in biological networks (Newman,2003). Weak links appear as a characteristic feature of complex systems which are made up of highly diversified components (Csérmely, 2009; Granovetter,1973). Weak links also provide support for network stability (McCann et al.,1998). Most previous analyses of network patterns in realworld ecosystems have considered specieslevel interactions. However, interaction asymmetries at the individuallevel remain largely unexplored, despite likely being the most appropriate level to address interaction outcomes (Clark et al.,2011). Actual ecological interactions that we can observe, sample and document, occur from interspecific encounters between individuals (Dupont et al., 2014; Jordano, 2016). One might therefore wonder if, when looking at a more refined level (e.g. from species to individuals), we could still expect asymmetry in mutual dependence. Few studies so far have analysed interaction asymmetry beyond variation in just interaction frequency or strength, further examining differences in interaction quality (González- Castro et al., 2022; Guerra & Pizo, 2014; Herrera, 1984; Jordano, 1987). Interaction outcomes may yield different results from those expected solely on the basis of interaction frequency (Janzen 1983; González- Castro et al.,2022), and so it is possible that infrequent interactions result in higher fitness than frequent interactions, affecting the reciprocity balance. A useful tool to measure the functional outcome (fitness) of mutualisms in terms of both interaction quantity and quality at the individual level is the effectiveness framework (Schupp,1993; Figure 1a). Consideration of individual variation and interaction outcomes expands our understanding of the potential consequences, for example, demographic or evolutionary, of ecological interactions. In this study, we calculate the twosided rewards for seed dispersal mutualistic interactions between plants and animal frugivores by means of the Resource Provisioning Effectiveness (RPE) and Seed Dispersal Effectiveness (SDE) frameworks (Schupp et al., 2017). We look at mutual reciprocity (i.e. the balance in the exchange of resources) from an individual perspective in a plant population using SDE and RPE as estimates for the reward obtained in the relationship (Figure1d). We explore whether mutualistic dependencies are still asymmetrical when looking from a plant individual perspective and when incorporating both interaction frequency and quality (Figure1e). We use as study organism the plant Pistacia lentiscus, a species defined as supergeneralist species because it interacts with a large part of the local diversity of partner species, being heavily connected to the rest of the community (García,2016; Jordano et al.,2002; Parejo- Farnés et al.,2020). Supergeneralists play a fundamental role in ecological networks because they provide great cohesion (Guimarães et al.,2011). A twosided study of such mutualism at the plant individual level informs about the diversity of individual rewards, the diversity of mutualistic partners and their effects, and the consequences on resource exchange between them. Here, we address three specific objectives: (1) characterise the effectiveness of the mutual beneficial service between individual plants and their frugivorous species, (2) test if the service provided between partners in terms of the amount of reward is reciprocal and (3) explore if there exists asymmetry in the mutual dependencies when looking at a plant individual level and considering interaction outcomes; that is, accounting for interaction quality beyond interaction frequency. 14610248, 2023, 1, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/ele.14141 by Universidad De Sevilla, Wiley Online Library on [13/09/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License 134 | RECIPROCITY AND EFFECTIVENESS OF MUTUALISMS FIGURE 1 Schematic representation of this study approach to characterise plantfrugivore seed dispersal mutualisms, showing: (A) the three main subcomponents present in the mutualism between any two nodes in the network: the interaction frequency or quantity component (QTC) and the twosided quality (QLC) of the service provided by the partners. For this example we present the animal frugivore in orange, and the plant individual in blue; plant's quality is the energetic yield per fruit (QLCP) and bird's quality is the probability of seedling recruitment per consumed fruit (QLCA). On the right we provide an example adjacency matrix with simulated numbers of quantity and quality data, with two animals (A1 and A2), and four plants (P1– P4). (B) The two subcomponents (quantity and quality) are combined to calculate the effectiveness of the interaction from the bird (Resource Provisioning Effectiveness, RPE) and the plant's perspective (Seed Dispersal Effectiveness, SDE). (C) Resulting calculations of RPE and SDE using the example matrix in (A). (D) Reciprocity (i.e. the balance in rewards exchange between partners) is assessed by the correlation between RPE and SDE values of all pairwise interactions. (E) Derivation of mutual dependence estimates and interaction asymmetry for plant and animal partners. Dependence values for animals (i.e. how much the animal depends on each particular plant, orange upperleft cells) are calculated based on RPE values, while dependence values for plants (blue lowerright cells) are based on SDE values. The asymmetry of each pairwise interaction is calculated as the standardised difference between the two dependence values in each interaction, and ranges between −1 and +1 (see Box1). QTC QLCP A1 A2 QTC P1 5 8 1 P2 4 6 1.2 P3 0 3 1.3 P4 1 0 0.9 QLCA0.5 1 SDE (Seed Dispersal E ectiveness) = No. of fruits consumed (QTC) Dispersal service (QLCA) RPE (Resource Provisioning E ectiveness) = No. of fruits consumed (QTC) Energetic yield (QLCP) A1 A2 P1 0.47 0.42 0.24 0.76 P2 0.45 0.38 0.25 0.75 P3 0 0.20 0 1 P4 0.08 0 1 0 A1 A2 P1 -0.49 0.45 P2 -0.44 0.50 P3 0.80 P4 0.92 DEPENDENCE ASYMMETRY A1 A2 P1 5 8 P2 4.8 7.2 P3 0 3.9 P4 0.9 0 A1 A2 P1 2.5 8 P2 2 6 P3 0 3 P4 0.5 0 -1 -0.5 0 0.5 1 ASYMMETRY Animal depends more on Plant Plant depends more on Animal SDE 0 2 4 6 8 RPE 02468 RECIPROCITY RPE SDE Dispersal service (recruits per fruit) Energetic yield (kJ per fruit) N of fruits consumed QLCP QLCA QTC RPE SDE (b) (a) (c) (d) (e) 14610248, 2023, 1, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/ele.14141 by Universidad De Sevilla, Wiley Online Library on [13/09/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License | 135 QUINTERO et al. METHODS Species and study site Pistacia lentiscus (Anacardiaceae) is a dioecious, windpollinated, animal seeddispersed shrub that can be considered as a ‘foundation species’ (Whitham et al.,2006) in lowland Mediterranean scrublands. Numerous resident and migrant frugivorous birds rely on P. lentiscus fruits as a nutritional resource (González- Varo et al.,2019) and act as its seed dispersers, with infrequent consumption by mammals (Perea et al.,2013). Fieldwork was conducted between the years 2019– 2020 at two study sites in Doñana National Park (Huelva, SW Spain): La Mancha del Rabicano in El Puntal site (EP) and Laguna de las Madroñas (LM). Both areas consist of Mediterranean sclerophyllous scrubland dominated by P. lentiscus coexisting with a total of 28 fleshyfruited species recorded in the area. We monitored 40 individual P. lentiscus plants per study site (Suppl. Mat. A). This sampling included all the female fruiting plants found in LM population. In EP site, we chose a representative sample of female plants scattered across the site and covering the fullsize gradient in the population. Interaction frequency: Quantity component (QTC) The interaction frequency of Pistacia lentiscus plants with avian frugivore species was assessed through DNA- barcoding and continuousmonitoring cameras (Quintero et al.2021) during the complete fruiting season, between September 2019 and March 2020. We placed seed traps beneath individual plants at both populations, where we collected a total of 2691 faecal and regurgitated seed samples (1913 for EP and 778 for LM). Visiting avian species were identified with DNA- barcoding analysis of collected samples. Animalorigin DNA present in the surface of the samples was extracted, amplified and then sequenced following protocols in González- Varo et al. (2014) with minor modifications (Suppl. Mat. B.1). Retrieved sequences were identified using BOLD Systems database or BLAST from the NCBI. More than 90% of the collected samples were analysed (n=2510) and the identification success was 94%. Using monitoring cameras, we recorded animal visitation and feeding events at focal plants in one of the sites (EP). All individual plants were monitored every fortnight along the fruiting season for a total of nine times, accumulating c.19 h observation per plant. Recordings lasted c. 2.2 h and started in the early morning (Suppl. Mat. B.2). We analysed the video recordings with the help of DeepMeerkat software (Weinstein, 2018). We obtained the feeding frequency of animals (i.e. fraction of visits with actual fruit consumption) and the number of fruits consumed per visit. Avian species identification was possible for 91.5% of the n=3970 visits recorded by cameras and 24% of the interaction events included feeding records. The total number of frugivorous bird species recorded was 27; 26 were recorded with cameras and 22 with DNA barcoding. Interaction accumulation curves (IAC) were used to determine sampling completeness (see Suppl. Mat. B.3; Colwell & Coddington, 1994; Jordano, 2016). Overall, sampling completeness (sensu Chacoff et al.,2012) was 93% for both methods; 95% for cameras and 96% for DNA- barcoding. To estimate the total number of fruits consumed by each bird species at each individual plant, we multiplied four sequential steps: (1) the total number of visits at each site, (2) the probability that a given bird species visited a particular plant, (3) the probability that a visit included a feeding event and (4) the number of fruits consumed per visit by each bird species. We estimated these quantities using Bayesian models fitted with Stan (Stan Development Team, 2022) and brms package (Bürkner, 2017) in R 4.1.2 (R Core Team, 2021). The Bayesian approach allowed us to combine information on bird visitation rates coming from DNA- barcoding and monitoring cameras, to estimate a few unobserved quantities, such as missing feeding rates for some species, by sharing information across taxa, and to obtain and propagate uncertainties along multiple effectiveness components (see Suppl. Mat. E.1 for details on each model). Since posterior distributions were often rightskewed, we report their median throughout. Interaction outcome for the animal: Quality Component (QLC— RPE) Plant quality was defined as the energetic reward provided per fruit consumed. Feeding behaviour was different among the avian species recorded: some birds consume the whole fruit regurgitating or defecating the seed intact (legitimate seed dispersers), others consume part of its pulp discarding the seed (pulp thieves); while other birds peel the fruit, break the seed coat and consume the embryo inside the seed (seedpredators or granivores). Because avian species consume different parts of the fruit, the energy obtained refers to the pulp for swallowers and pulp consumers, and to the seeds in the case of predators (see Table S.A.1 for frugivory type categories). We collected fruits from each plant (mean=31 fruits, range=17– 63, Suppl. Mat. C) and measured both pulp and seed fresh mass. Fresh mass was converted to dry mass using P. lentiscus % water content (Jordano,1984). To obtain the energy contained per fruit, we then multiplied the pulp and seed dry mass by their estimated energy yields: 25.25 kJ/g for pulp and 28.14 kJ/g for seed (see Suppl. Mat. E.2). 14610248, 2023, 1, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/ele.14141 by Universidad De Sevilla, Wiley Online Library on [13/09/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License 136 | RECIPROCITY AND EFFECTIVENESS OF MUTUALISMS BOX 1 Reward, reciprocity and asymmetry calculations Reward estimation using the Effectiveness framework We estimated the rewards exchanged between mutualistic partners using the Resource Provisioning Effectiveness and Seed Dispersal Effectiveness framework (RPE and SDE; Schupp et al.,2017, Quintero et al.,2020). The effectiveness of the mutualistic interaction is estimated as the product of a quantity and quality component. The quantity component (QTC) was common for both RPE and SDE, that is, the total number of fruits consumed by a specific bird species on a given plant. Quality from the animal's perspective was the energy acquired per fruit consumed (referred to pulp or seed energy depending on bird's feeding behaviour). From plants' perspective, the quality component represents the probability that a consumed fruit becomes a seedling surviving its first summer. RPE therefore estimates the total energy provided by an individual plant to a bird species across the fruiting season, and SDE estimates the potential number of seedlings recruited coming from an individual plant by interacting with a bird species. Below we define the components for the rewards calculation: RPE ij: Total energy that bird species j obtained by feeding in plant i SDE ij: Number of plant i seedlings recruited through consumption by bird species j QTC ij: Number of fruits consumed by bird species j in plant i along the fruiting season QLC_RPE i: Energy contained per fruit pulp or seed of plant i QLC_SDE j: Probability that a fruit consumed by bird species j becomes a seedling surviving its 1st summer Reciprocity To estimate reciprocity between partners we used Pearson correlation coefficients between the logtransformed RPE and SDE values. We aggregated the RPE and SDE values for each individual plant, that is adding up the values for all bird species with which it interacted, resulting in the total energy provided by the plant and the number of seedlings recruited through interactions with its bird assemblage. We used 1000 samples from the posterior distribution of RPE and SDE calculated for each plant, to consider uncertainties in the estimation of reciprocity (Suppl. Mat. F.1). A high positive correlation indicates high reciprocity, meaning that plants providing more energy (RPE) also obtain a higher number of seedlings recruited (SDE). Calculating dependence and asymmetry between bird species and individual plants. We calculated two mutual dependence (d) values for each pairwise interaction, one for the plant ( d P i →A j ) and one for the animal species ( d A j →P i ; Suppl. Mat. F.2). d Pi→Aj= SDE ij ∑ n A=1 SDE i , for the dependence of P. lentiscus plant i on animal species j; and. d Aj→Pi = RPE ij ∑m P =1RPE j , for the dependence of animal species j on plant i, RPEij =QTCij ×QLC_RPEi SDEij =QTCij ×QLC_SDEj QTC ij = Total no. of visits in population ×Prob. that bird species jvisits plant i ×Prob. bird species j consumes fruit during visit × No. fruits bird species jconsumes per visit QLC_ RPEi =Fresh pulp or seed mass of plant i( g ) ×non−water%in pulp or seed ×energetic yield per gram of dry pulp or seed QLC_SDE j= Prob. seed escaping predation when manipulated by bird species j ×∑m=5 m=1[Prob. bird species jdeposits a seed at microhabitat m ×Prob. seed escaping post dispersal predation in microhabitat m ×Prob. seedling emerging and surviving its 1st summer in microhabitat m ] 14610248, 2023, 1, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/ele.14141 by Universidad De Sevilla, Wiley Online Library on [13/09/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License | 137 QUINTERO et al. Interaction outcome for the plant: Quality Component (QLC— SDE) We estimated the quality of animals as seed dispersers according to the: (1) probability of seeds escaping predation by granivorous birds, (2) microhabitat use by each bird species, (3) probability of seeds escaping postdispersal predation and (4) probability of seedling emergence and early survival (past their first summer) in each microhabitat. We estimated these probabilities using Bayesian models as above (see Suppl. Mat. E.3 for details). The product of these four steps determined the probability of seedling recruitment resulting from the consumption of one fruit by a specific avian consumer. We found a few undamaged, depulped seeds (n=36) in the seed traps beneath plants that were attributed to seed predators through DNA- barcoding. Seeds dropped during fruit handling indicated sporadic dispersal events by seed predators, whose probability was estimated using the total number of preyedupon seeds (open seed endocarp halves) and the number of undamaged seeds found in seed traps attributed to granivores (Suppl. Mat. D.1 and E.3). The intensity of microhabitat use by the different bird species was inferred from the seed rain of P. lentiscus seeds collected at five microhabitats: under Pistacia lentiscus conspecifics (PL), under other fleshy fruited species (FR), under nonfleshy fruited species (NF), under pine trees (Pinus pinea; PP) and open ground areas (OA). At each microhabitat, we collected P. lentiscus dispersed seeds and identified the bird species through DNA- barcoding. For the PL microhabitat, we used the seed traps located beneath the focal individual plants (see above). To sample microhabitats FR, NF and PP we placed additional seed traps at 15 replicated points per microhabitat. For open areas (OA) we scanned 17, 1- m wide, transects during the season. The number of dispersed seeds collected at each microhabitat allowed us to estimate the probability of dispersal to each specific microhabitat by each bird species (Suppl. Mat. D.2 and E.3). Finally, we measured postdispersal seed predation, seedling emergence and survival at each microhabitat. To study postdispersal predation, we placed 10 seeds on a petri dish (six replicates per microhabitat) and monitored the rate at which seeds experienced predation (mainly by rodents, see Suppl. Mat. D.3 and E.3). We also installed germination stations (13 replicates per microhabitat, each containing 16 sown seeds) to estimate seedling emergence and survival (see Suppl. Mat. D.4). Effectiveness, reciprocity and asymmetry calculations We calculated RPE and SDE as the product of quantity and their respective quality components (i.e. the interaction frequency multiplied by its functional outcome; Figure1 and Figure S.E.1). We estimated the reciprocity in reward exchange between partners, their mutual dependence and asymmetry following the procedure presented in Box1. To assess the robustness of the observed asymmetry values to variations in our sampling design, we repeated the asymmetry calculations using replicated random subsets of 20, 40 and 60 plants, to examine potential effects of the number of focal plants on the distribution of asymmetry values. We also calculated the asymmetry in a randomised network of the same size (i.e. 80 focal plants) following Patefield and Vázquez null models to test if the observed and randomly expected asymmetry values differed (Suppl. Mat. H). RESULTS Plant individualbased interactions We estimated that birds consumed a total of 2.2 × 105 fruits from the 80 marked plants at both P. lentiscus populations (90% credibility interval: 1.5× 105– 6.6× 105). This consumption represents c. 20% of the total number of fruits produced by focal plants in the 2019– 2020 season (Suppl. Mat. G.1). We detected 27 bird species consuming P. lentiscus fruits, of which 12 are considered where d is the dependence of plant i on animal species j, or vice versa; SDE ij is the estimated number of seedlings recruited coming from plant i via frugivore species j; RPE ij is the amount of kilojoules plant i provided to frugivore species j; and n and m represent the total number of animal species and individual plants, respectively. Interaction asymmetry (AS) is defined as: AS values can range from −1 to 1, where 0 indicates total symmetry (both partners depend on each other with the same intensity), values approaching +1 indicate that the plant is more dependent on the animal than vice versa, and negative values indicate that the animal is more dependent on the plant than the plant on the animal. AS PiAj= d Pi→Aj −d Aj→Pi max(d) 14610248, 2023, 1, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/ele.14141 by Universidad De Sevilla, Wiley Online Library on [13/09/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License 138 | RECIPROCITY AND EFFECTIVENESS OF MUTUALISMS residents, 9 summer or trans- Saharan, migrants and 6 winter migrants (Suppl. Mat. A). More than 85% of the consumed fruits were eaten by just three species: Curruca melanocephala, Erithacus rubecula and the seed predator Chloris chloris. These species behaved as supergeneralists, interacting with the majority of individual plants (Figure2). The next most significant consumers were Turdus merula and the winter migrant Sylvia atricapilla. Resource provisioning and seed dispersal effectiveness Pistacia lentiscus plants were highly variable in Resource Provisioning Effectiveness (RPE) provided to avian species (Figure3). On a perinteraction basis a frugivore species consumed a median of 101 fruits on each plant (90% CI: 2– 2680). We estimated that Curruca melanocephala and Erithacus rubecula ate more than 4000 fruits, and Chloris chloris predated more than 5500 seeds, at certain individual plants. This intensity of consumption represents, however, just a small proportion of the available crop offered: 90% of the plants had less than half their crop size removed by birds (Suppl. Mat. G.1). The quantity component accounted for almost all (93%) of the variation in RPE (Suppl. Mat. E.5). Regarding quality, we found up to sevenfold differences in the energetic content of fruits from individual plants. Birds exploited the full gradient of fruit sizes (Suppl. Mat. G.2), but in general, avian consumption was higher in plants with larger crops, canopy area and pulp content (Suppl. Mat. G.3). Seed Dispersal Effectiveness (SDE, Figure3) was also determined more by the quantity than the quality component, the latter varied little among bird species (variance partitioning: quantity = 69%, quality = 31%; Suppl. Mat. E.5). Except for seed predators, which had negligible contributions to recruitment (because they destroyed 99.9% of the seeds consumed), the probability of recruiting a seedling per consumed fruit was similar for all bird species (median=1.1× 10−4; 90% CI=1.0× 10−5– 9.6 × 10−4), with Curruca melanocephala emerging as the highest quality disperser, followed by other members of the Sylviidae family (Figure3; Figures E.3.4). Recruitment probabilities at the final stage were low; even the most effective pairwise interaction (involving C. melanocephala and plant ‘314’), would have resulted in SDE=0.53 seedlings (<1 seedling) surviving the first summer. Differences among frugivore species in dispersal quality result from their distinctive microhabitat use (Suppl. Mat. E.3.1) and existing tradeoffs between recruitment stages in different microhabitats (Suppl. Mat. E.3.2; E.3.3). For example, seeds falling under pine trees had the highest probability of surviving rodent predation (median probability = 0.023), followed by those arriving to open areas (median probability = 0.013). Seedling emergence and survival, on the other hand, was highest in open areas and lowest beneath pines (median probability=0.038 in OA versus 0.003 in PP). Overall, Open Area was the microhabitat with highest probability of recruitment, yet very few seeds arrived there, hence this microhabitat hardly contributed to recruitment. The high quality of C. melanocephala (median probability= 1.2 × 10−4) emerged from its preferential dispersal towards the most suitable microhabitats: beneath nonfleshy fruited plants and P. pinea. In contrast, heavy P. lentiscus fruit consumers like E. rubecula showed medium quality (median probability = 1 × 10−4) because it frequently deposits seeds under P. lentiscus plants, a microhabitat where the probability of escaping postdispersal seed predation and seedling survival were medium to low. Reciprocity We found a high correlation between RPE and SDE (mean Pearson r on log– log values=0.93; Suppl. Mat. F.1). High correlation indicated high reciprocity in the interactions between individual P. lentiscus plants and their bird consumers: more seedlings were recruited from plants supplying more energy (Figure4). In other words, the larger the reward provided by one interaction partner, the larger the reward contributed by the other partner. This high reciprocity stems from both RPE and SDE being mainly driven by the quantity component (intensity of consumption) rather than by differences in plant and frugivores quality. As a result, more seedlings were recruited from plants which had more fruits consumed (mean slope of log SDE ~ log RPE = 0.83, SD = 0.06; Figure4), regardless of differences in the composition of their frugivore assemblages. Deviation from a slope of 1 indicates a ‘diminishing return’ effect, so that the number of seedlings recruited did not increase in the same proportion as the total energy provided by plants. This diminishing return was not caused by interactions with seed predators (mean slope of log SDE ~ log RPE excluding seed predators=0.85, SD=0.04); who damaged all plants in similar proportions. Additionally, plants presenting greater rewards had larger crop sizes and were consumed by a higher number of bird species (Figure4). Dependence and asymmetry Mutual dependencies on the partner were in general low (Figure5). Most pairwise interactions (96% from birds' perspective, and 76% from plants') showed dependencies below 0.25, indicating that most interactions contributed to the partner only a small fraction of the total reward obtained (i.e. energy income for birds or seedlings recruited for plants). There were, however, some strong, highly dependent interactions, namely 14610248, 2023, 1, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/ele.14141 by Universidad De Sevilla, Wiley Online Library on [13/09/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License | 139 QUINTERO et al. FIGURE 2 Interaction network between avian consumer species and individual Pistacia lentiscus plants, where the node and link width is proportional to the total number of fruits consumed on each plant. Nonlegitimate dispersers (n=7) are grouped at the end of the network. Cyanopica cooki Turdus philomelos Turdus viscivorus Erithacus rubecula Phoenicurus phoenicurus Luscina megarynchos Turdus merula Saxicola rubicola Lanius meridionalis Curruca melanocephala Hippolais polyglotta Curruca communis Curruca undata Sylvia atricapilla Curruca hortensis Ficedula hypoleuca Sturnus unicolor Sylvia borin Muscicapa striata Curruca cantillans Chloris chloris Coccothraustes coccothraustes Parus major Cyanistes caerulus Fringilla coelebs Columba palumbus Pyrrhula pyrrhula slaudividni tnalP 14610248, 2023, 1, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/ele.14141 by Universidad De Sevilla, Wiley Online Library on [13/09/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License 140 | RECIPROCITY AND EFFECTIVENESS OF MUTUALISMS FIGURE 3 Landscapes for Resource Provisioning Effectiveness (RPE) and Seed Dispersal Effectiveness (SDE). Each point represents an individual pairwise interaction between an individual plant and a frugivorous avian species. In both landscapes, the horizontal axis depicts the total number of fruits consumed by each bird species in each individual plant. Symbols represent feeding behaviour (circles for legitimate dispersers, and triangles for nonlegitimate dispersers). In the RPE landscape, the vertical axis represents the median energy (kJ) obtained from the pulp or seed from each individual plant. In the SDE plot, the vertical axis represents the posterior median probability of recruiting a seedling from a fruit ingested by each bird species. Hence, the product of the horizontal (Quantity) and vertical (Quality) axis gives the effectiveness of each birdplant pairwise interaction: the total energy (kJ) in the case of RPE, and total number of plant recruits for SDE. Different combinations of quantity and quality can produce equal effectiveness values, as shown by isolines. Note seed predators are not shown in the SDE landscape visualisation, because their dispersal quality is zero or close to zero and their inclusion distorts the graph (see Supp Mat E.4 for complete SDE landscape). 1.0 x 10-4 1.2 x 10-4 010002000 3000 4000 Total number of fruits consumed (Quantity) Prob. of seedling recruitment (Quality) Curruca melanocephala Erithacus rubecula Sylvia atricapilla Chloris chlorisTurdus merula Other species Resource Provisioning Effectiveness - RPE Seed Dispersal Effectiveness - SDE 0.02 0.420.30.21 0.14 0.09 129 325 580 901 1310 1849 2630 0.0 0.2 0.4 0.6 0.8 02000 4000 6000 Total number of fruits consumed (Quantity) kJ in pulp or seed per fruit (Quality) 0.9 x 10-4 1.1 x 10-4 Total kJ Total recruits 14610248, 2023, 1, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/ele.14141 by Universidad De Sevilla, Wiley Online Library on [13/09/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License