Social-ecological filters drive the functional diversity of beetles in homegardens of campesinos and migrants in the southern Andes

Homegardens are coupled social-ecological systems that act as biodiversity reservoirs while contributing to local food sovereignty. These systems are characterized by their structural complexity, while involving management practices according to gardener’s cultural origin. Social–ecological processes in homegardens may act as filters of species’ functional traits, and thus influence the species richness-functional diversity relationship of critical agroecosystem components like beetles (Coleoptera). We tested the species richness-functional diversity relationship of beetle communities and examined whether habitat structure across different levels, sociodemographic profiles, and management practices act as filters in homegardens in a Global Biodiversity Hotspot, Chile. For 100 homegardens (50 campesino and 50 migrant), we sampled beetles and habitat attributes, and surveyed gardeners’ sociodemographic profiles and management practices. We recorded 85 beetle species and found a positive relationship between species richness and functional richness that saturated when functionally similar species co-occur more often than expected by chance, indicating functional redundancy in species-rich homegardens. Gardener origin (campesino/migrant), homegarden area (m2), structural complexity (index), and pest control strategy (natural, chemical, or none) were the most influential social–ecological filters that selectively remove beetle species according to their functional traits. We discuss opportunities in homegarden management for strengthening local functional diversity and resilience under social-environmental changes.

. Estimated association between species richness and functional richness for 50 campesino (blue dots) and 50 migrants (yellow dots) homegardens for 85 species in beetle communities in Andean temperate ecosystems, southern Chile. Graphs were generated using R software version 4.0.4 (R Core Team, 2021. R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. https:// www. Rproj ect. org/). Table 1. Ranking of models for species richness, relative abundance, and functional richness as a function of social-ecological filters. Season and locality were random terms in all tested models. Model structure in bold indicates the best models with equivalent support. a Number of parameters estimated; b Difference in AICc values between each model and the lowest AICc model; c AICc model weight; d Log likelihood. www.nature.com/scientificreports/ abundance (Table 1a); the latter was higher and positive in campesino homegardens (mean ± SD = 77.9 ± 78.6; β = 79.26) and smaller and negative in migrant homegardens (43.2 ± 60.1; β = − 38.32) (Fig. 2b). Beetle relative abundance was positively correlated with using a natural (mechanical by hand or using biopreparations) pest control strategy (β = 78.00) and negatively correlated with chemical control (β = − 44.63), while no control did not have an effect on beetle relative abundance. Structural complexity did not have an effect on beetle relative abundance (Fig. 2c).
Spatial projections of beetle diversity. The resulting projections of beetle diversity indicated, graphically, a zone of high values for beetle relative abundance to the east of the study area (Fig. 3b). The spatial projections for beetle species richness and functional richness did not reveal a clear pattern of areas with high values for these parameters. Anyhow, this analysis indicated a relative spatial mismatch between estimates of beetle species richness, relative abundance, and functional richness in the study area (Fig. 3).

Discussion
This study extends previous research on the relationship between biodiversity and ecosystem functioning, acknowledging that homegardens, as part of larger agroforests, are coupled social-ecological systems in which biodiversity has the potential to thrive. We found that several beetle species may be performing similar roles (i.e., are functionally redundant) in southern Andean homegardens with relatively high number of species. Thereby, if some go locally extinct (removed from a diverse homegarden) this will likely not produce substantial loss in agroecosystem function 47 . This result associates with the observed steep relationship between beetle species richness and functional richness, in relation to a random expectation, that started to saturate with relatively high beetle richness 48,49 . This finding suggests that homegardens with high functional redundancy will be more resilient to shifts in social-ecological filters 50-52 . Beetle species richness-functional diversity relationship. Our recorded total number of species is only a subset of the total species recorded or likely to occur in nearby temperate forest ecosystems [42][43][44][45]53 . However, remarkably, and contrary to our expectations, we found that beetle communities in southern Andean homegardens have a relatively high functional richness and functional redundancy. This result is not characteristic of systems generally considered as "species-poor" 36,37,54,55 . Andean temperate ecosystems are relatively impoverished in terms of faunal species richness in comparison to other tropical, subtropical, Mediterranean, and temperate ecosystem types 43 . During the Pleistocene (most recent period of repeated glaciations), immigration of species from tropical latitudes was not able to compensate for the extinction of local biota resulting from www.nature.com/scientificreports/ the contractions on the distribution of temperate forests 56 . Climatic change and geographic barriers, such as the Andes mountain range and the Atacama Desert, resulted in a net loss of species during the Pleistocene, especially of faunal groups with tropical lineage 57 . While little is known about biogeographic distribution of beetles in the southern temperate ecoregion 42,58-60 , our study shows that small-scale patches of habitat, like homegardens, can be both taxonomically and functionally rich. Our results support the idea that functional diversity is not only correlated with the pool of species occurring in beetle communities (first objective of our research: species richness-functional richness relationship). Beetle functional diversity is also influenced by social-ecological filters, which are coupled human-nature factors that selectively remove species according to their functional traits, likely through shifting the intensity and magnitude of competition in biological communities 23,61,62 . In accordance with other studies, the observed relative spatial mismatch for diversity parameters in the study area (shown graphically in Fig. 3), challenge the use of any diversity component as a surrogate for other parameters in agroecology, land-use planning, and biodiversity conservation 37,63 . Gardener origin and beetle communities. We found that gardener cultural origin (indigenous and non-indigenous campesino vs. lifestyle migrant) might influence both the taxonomic and functional diversity of beetle communities in homegardens. Our result supports previous studies exploring the role of gardener origin on the composition, structure, and functioning of homegardens, as the latter usually reflect many aspects of the food system, tastes, and agricultural traditions of people co-occurring in an area 19,64 . For instance, differences in both crop species and intensity of management practices are correlated with the gardener origin in Vietnamese homegardens 64 . Number of management practices and homegarden area are different among migrant and non-migrant homegardens and both social-ecological filters differentially influence beetle functional groups in Indonesian homegardens 19 . While we acknowledge that homegarden attributes are likely influenced by several factors beyond gardener origin 7 , our study sheds light on some of the underlying social-ecological filters explaining variation in the taxonomic and functional diversity of beetles in campesino and migrant homegardens of the southern Andes.
Homegarden area, structural complexity, and management correlate with beetle taxonomic and functional diversity. We found support for our prediction that homegarden area leads to an increase in beetle species richness, relative abundance, and functional richness, a result in accordance with the few studies dealing with taxonomic and functional diversity of beetle communities in homegardens 20,21 . The long-standing Island Biogeography Theory 65 provides a framework for examining the underlying forces shaping community assembly and species loss in homegardens. For example, beetle communities shaped in coupled social-ecological systems like homegardens may be chiefly determined by local extinctions, with smaller homegardens likely exhibiting the highest extinction rates of species 41,66 .
Furthermore, the distribution of traits as a function of habitat area extends the Island Biogeography Theory beyond the traditional species-area relationship 67 . Social-ecological filters may perform as non-random processes that act on beetle species traits including the influence of local habitat conditions on species' fitness and ecological interactions, such as competition, mutualisms, and other trophic associations 23,38,39,68 . For example, larger and heavier species that require relatively large territories or species with limited dispersal ability will have a higher likelihood of local extinction in response to a shrinking homegarden area 69,70 . Therefore, only subgroups of species sharing akin functional traits (i.e., appearing functionally clustered) will be able to persist or outcompete other species on small habitats 67,68 . In our study, for example, relatively large species like Apterodorcus bacchus and Calosoma vagans were never recorded in homegardens with an area smaller than 150 m 2 . In the southern Andes, homegarden area is definitely a non-random process. While campesinos generally have properties that are still larger than migrant ones, historical and contemporary processes of encroachment into indigenous and non-indigenous campesino way of life and the land upon which they live has been correlated with changes in agricultural practices and a decreasing trend in the area of agroforestry systems, including homegardens 71 .
As shown, larger homegardens likely provide more resource opportunities and they should tend towards being more representative of the regional pool of species or if there is high habitat structural complexity 4,72 . Indeed, we found that homegarden structural complexity was positively correlated with both taxonomic and functional diversity parameters. Generally, homegardens are complex microenvironments composed of multiple strata that generate diversified niches for multiple species and, likely, functional traits to coexist 19 . Interestingly, homegarden structural complexity was correlated with the homegarden age (Spearman > 0.6), the latter measured as the number of years that the homegarden has been in the same spatial location. Therefore, the oldest homegardens are located in the farms that have the longest history of settlement in the study area. Older homegardens, managed by local campesinos who have lived longer in the area, will generally host more vegetation layers including annual crops and perennial trees than homegardens owned by migrants, and will thus resemble the complex surrounding forest ecosystems 7 .
Structurally complex homegardens will not only increase the functional niche space filled by species in beetle communities and enhance beneficial organisms, such as pest-control predators, pollinators, and seed dispersers 13 , they will also be more important carbon sinks than those that are structurally simplified and lack trees 73 . In a complexity science context, this result suggests that these small-scale systems have a social-ecological memory in which older and structurally complex homegardens act as long-lived system entities whose presence continues to influence compositional, structural, and functional states of the system over time 51 .
Using a natural (mechanical by hand or using biopreparations) pest control strategy positively influenced beetle functional richness and relative abundance, while chemical pesticides negatively correlated with functional richness. These results should be viewed with caution because it may be interpreted that controlling insects using www.nature.com/scientificreports/ natural strategies can potentially increase phytophagous beetles. However, we have recorded that controlling beetles that damage crops by hand is a widespread strategy (mostly to control Epicauta pilme) which reduces damage while increasing the relative abundance of benefic beetles (pollinators like Cantharis variabilis and pest controllers like Eriopis connexa; J. T. Ibarra Unpublished Data). The systematic use of pesticides in agriculture over the past decades has negatively impacted insect populations 74 , a pattern also reported for homegardens 20 , with persistent negative effects on biodiversity and biological control potential 75 . In our study area, campesinos report a higher use of pesticides than migrants because the former have been provided for decades with agro-chemicals (fertilizers, pesticides, herbicides, and hybrid seeds) by extension agents from governmental programs 33 . However, campesinos and migrants are progressively dismissing the use of agro-chemicals as a result of an increasing adoption of agroecological practices not only limited to chemical-free agriculture but also as an alternative movement for the defense and re-signification of rural areas 32,33 .
Recommendations for gardening while sustaining beetle diversity. Beetles are globally declining, principally, because of habitat loss and conversion to intensive agriculture. Paradoxically, beetles comprise many predator, pollinator, and saprophytic species of outstanding importance for agroecosystem functioning. Homegardens, usually multifaceted, can be oriented towards building synergies between local food sovereignty or income generation depending on the concerns of the family and biodiversity. Our results highlight the importance of increasing the size of homegardens as much as possible and promoting the cultivation of a multi-layered arrangement of crops (e.g., combination of roots and tubers, small annual and perennial plants, shrubs, and trees) that will increase habitat structural complexity across years, and thus resources for a diversity of beetle species, that will resemble with surrounding forests. Agricultural and environmental governmental agencies charged with supporting small-scale agriculture should discourage the use of pesticides to control beetles and other insects, as these chemicals likely have negative effects on ecosystem functioning and biological control potential. Furthermore, our results highlight the importance of incorporating campesino (indigenous and non-indigenous) agroecological knowledge on biodiversity friendly agroforestry management in homegardens. These measures may contribute to maintain ecosystem functioning, local livelihoods, and the resilience of beetle communities in times of rapid social-environmental changes.

Study design.
All methods were carried out in accordance with relevant guidelines and regulations. The study was approved by Scientific Ethics Committee of the Pontificia Universidad Católica de Chile (Resolution #160415004). We conducted homergarden surveys and interviews to gardeners after obtaining prior informed consent from each one of them. Fieldwork was conducted in two field seasons during the summer season between December and February of 2016-2017 and 2017-2018. In total, we studied 100 homegardens (50 homegardens from Mapuche indigenous and non-indigenous campesinos were surveyed the first field season and 50 homegardens from lifestyle migrants were surveyed the second field season). Mapuche indigenous and non-indigenous campesinos were grouped together because the latter are people who were born, live, and work in the territory, often in close relationship with Mapuche families; their agriculture resembles and integrates the Mapuche traditional agricultural system 33 . For their part, lifestyle migrants are people who migrated during adulthood from an urban setting to the study area 32 . We used successive-referral sampling as our nonprobability recruiting method 76,77 . The criteria for selecting a homegarden for study was that its main purpose was family consumption and that it was at least two years old.
Homegarden habitat, sociodemographic profiles, and management practices. We identified all the crop species intentionally cultivated in each of the 100 homegardens and visually estimated the ground cover (%) of each crop vertical stratum through guided walks with gardeners ( Table 2; 78 ). We measured homegarden area (m 2 ) and used a handheld GPS to record the homegarden spatial location (geographic coordinates). We used Google Earth (Map data ©2021 Google, Maxar Technologies) images to measure the distance from the homegarden to the nearest native forest edge (normally seen as a clear-cut line between forest and a different land cover; e.g., pasture). We further conducted structured interviews with data on sociodemographic profiles and management practices, including gardener origin, age, gardening experience, homegarden age, and pest control strategies ( www.nature.com/scientificreports/ we constructed sample-based rarefaction accumulation curves for both sampling methods. We considered an adequate sampling effort when there was no longer an increase in species as individuals accumulated 81 . We randomly deployed four pitfall traps every 25 m 2 with a maximum of 16 traps (determined through accumulation curves) for three nights per homegarden 19 . We deployed traps between 8:00-11:00 am and were collected at the same time the fourth day. Each trap was buried 12 cm, had a diameter of 7.3 cm and was placed at the soil surface. Traps were filled to a third of their capacity with an ethylene glycol solution and covered by a suspended lid. For sweep netting, we performed one 10 m transect of 1.5 min every 25m 2 of homegarden with 3 m between transects and a maximum of nine transects per homegarden (determined through accumulation curves; Lister and Garcia 2018). We performed sweep netting transects from 12:00 to 16:00 on clear days with temperatures ranging from 15 °C to 25 °C. In total, we deployed 1.410 pitfall traps over 371 nights and conducted 371 sweep netting transects. We collected all beetle individuals and identified at the species level utilizing dichotomous keys in guides and the Coleoptera reference collection available at the Natural History Museum of Chile. Finally, we measured the length of a minimum of three individuals per species for functional trait analysis (below in section "Beetle traits and functional diversity").
Beetle traits and functional diversity. We used three traits of beetle species, including two categorical (foraging guild and habitat-use guild) and one continuous (body weight) measures (Table 3). These traits are correlated with resource use by species and are mechanistically linked to ecosystem functioning (e.g., quantity, type, and strategies for obtaining resources by each species; Table 3). For example, foraging guild has been used for linking resource production and disruption to beetle diversity 82,83 . Data on foraging guild and habitat-use guild were extracted from 34 bibliographic references (including 84-92 , among others). For its part, body weight has been utilized to show how environmental change has indirectly precipitated a bottom-up trophic cascade and consequent collapse of the food-web structures 93 . Body weight for each beetle species was calculated from measured body lengths using the function proposed by (Johnson and Strong 94 : According to their foraging guild, we classified each species as mainly beneficial (predator, pollinivorous, saprophagous, mycetophagous) or harmful (phytophagous, xylophagous) for homegarden production. Finally, we quantified functional diversity using the metric functional richness (FRic) 24 . FRic was calculated using the beetle traits (Table 3) and the presence/absence of each species per homegarden. We calculated FRic using the program R-FD 95 .
Data analysis. We used Generalized Linear Mixed-Effect models 96 , implemented in the packages lmer 97 and AICcmodavg packages 98 in R software version 4.0.4 99 (R Development Core Team, 2021). We first tested the species richness-functional diversity relationship by regressing species richness against FRic. Then, we examine the association between a dependent variable and independent variables (fixed effects; social-ecological filters; Table 2) collected in grouped units at different levels (random effects; season and locality). We first assessed collinearity to reduce the number of independent social-ecological filters presented in Table 2. With strongly correlated social-ecological filters (Spearman's r > 0.6), we kept for analysis either the one considered to be most ecologically influential for the studied taxa or the most feasible to implement in management practices (Table 2). We examined the fixed effect of homegarden area, crop richness, structural complexity, distance to forest, homegarden age, gardener origin, and pest control strategy on the following dependent variables: beetle species richness, relative abundance, and FRic. To find the best models for our dependent variables, we generated a candidate set of models based on model weights (w i ) and the precision of the estimated coefficients, using Akaike's Information Criterion (AIC; 100 . We considered models with a ΔAIC < 2 of the top model as the competitive set of bestsupported models. For easier interpretation of our results and for categorizing taxonomically and functionally ln weight = ln(b0) + b1 * ln length Table 2. Social-ecological filters used to evaluate homegarden associations of beetles (Arthropoda: Coleoptera) in Andean temperate ecosystems, southern Chile. a Social-ecological filters retained for tests of homegarden associations of beetles after reducing collinearity.