Three decades of butterfly--plant interaction turnover explained by climate and species loss

Three decades of butterfly--plant interaction turnover explained by climate and species loss | Authorea try { document.documentElement.classList.add('js'); } catch (e) { } var _gaq = _gaq || []; _gaq.push(['_setAccount', 'G-8VDV14Y67G']); _gaq.push(['_trackPageview']); (function() { var ga = document.createElement('script'); ga.type = 'text/javascript'; ga.async = true; ga.src = ('https:' == document.location.protocol ? 'https://ssl' : 'http://www') + '.google-analytics.com/ga.js'; var s = document.getElementsByTagName('script')[0]; s.parentNode.insertBefore(ga, s); })(); Skip to main content Preprints Collections Wiley Open Research IET Open Research Ecological Society of Japan All Collections About About Authorea FAQs Contact Us Quick Search anywhere Search for preprint articles, keywords, etc. Search Search ADVANCED SEARCH SCROLL Ecology Letters This is a preprint and has not been peer reviewed. Data may be preliminary. 16 October 2025 V1 Latest version Share on Three decades of butterfly--plant interaction turnover explained by climate and species loss Authors : Pau Colom 0000-0003-0309-8886 [email protected] , Constantí Stefanescu , Jordi Corbera 0000-0003-3583-3929 , and Amparo Lázaro Authors Info & Affiliations https://doi.org/10.22541/au.176062539.92924262/v1 Published Ecology Letters Version of record Peer review timeline 397 views 150 downloads Contents Abstract Information & Authors Metrics & Citations View Options References Figures Tables Media Share Abstract Understanding the mechanisms behind interaction turnover over long-term periods is essential to predict how ecological networks respond to global change. We used a high-resolution dataset of butterfly--plant interactions spanning 13--29 years in seven Mediterranean communities to assess how climate fluctuations and community shifts shape interaction turnover and its components---species turnover and rewiring. Early in the time series, rewiring explained most interaction turnover, but its influence declined as species loss reduced the pool of shared partners between years. Consequently, species turnover became increasingly dominant, even though communities shifted toward butterfly species with generalist traits that promote rewiring. Nevertheless, rewiring intensified in years with stronger temperature fluctuations, when populations experienced greater shifts in phenology and abundance and were more likely to rewire. In the context of biodiversity loss, species turnover increasingly governs interaction dynamics, while the short-term flexibility provided by rewiring may collapse as communities become impoverished. Three decades of butterfly–plant interaction turnover explained by climate and species loss Pau Colom 1,2* , Constantí Stefanescu 3 , Jordi Corbera 4 & Amparo Lázaro 5 1 Department of Evolutionary Biology, Ecology, and Environmental Sciences, University of Barcelona, Barcelona, Spain. 2 Biodiversity Research Institute (IRBio), Barcelona, Spain 3 Natural Sciences Museum of Granollers, Granollers, BiBio Research Group, Spain 4 Delegació de la Serralada Litoral Central, Institució Catalana d’Història Natural (ICHN), Mataró, Spain 5 Mediterranean Institute for Advanced Studies (UIB-CSIC), Global Change Research Group, Esporles, Balearic Islands, Spain *Corresponding author contact details: [email protected] ; [email protected] / +34 662219748. Type of article : Letter Running title: Temporal interaction turnover & species loss Email authors: [email protected] ; [email protected] ; [email protected] ; [email protected] Statement of authorship : PC and AL designed the study. CS and JC collected the data. PC performed the analyses and wrote the first draft of the manuscript. AL and CS contributed substantially to the interpretation of results and to manuscript revisions. Data accessibility statement : All data and code supporting the results of this study have been archived in Zenodo and are available at: https://doi.org/10.5281/zenodo.17358385 Number of words : Abstract – 148; Main text – 4815. Number of references: 74 Number of figures & tables: 5 figures, 0 tables. Keywords: biotic homogenization; interaction dynamics; mutualistic networks; phenology; pollinator decline; rewiring; species turnover; species traits. Abstract Understanding the mechanisms behind interaction turnover over long-term periods is essential to predict how ecological networks respond to global change. We used a high-resolution dataset of butterfly–plant interactions spanning 13–29 years in seven Mediterranean communities to assess how climate fluctuations and community shifts shape interaction turnover and its components—species turnover and rewiring. Early in the time series, rewiring explained most interaction turnover, but its influence declined as species loss reduced the pool of shared partners between years. Consequently, species turnover became increasingly dominant, even though communities shifted toward butterfly species with generalist traits that promote rewiring. Nevertheless, rewiring intensified in years with stronger temperature fluctuations, when populations experienced greater shifts in phenology and abundance and were more likely to rewire. In the context of biodiversity loss, species turnover increasingly governs interaction dynamics, while the short-term flexibility provided by rewiring may collapse as communities become impoverished. Introduction Ecological interaction networks are inherently dynamic, changing over space and time in response to environmental variability and biotic dynamics. Understanding their temporal dynamics is particularly relevant under global change, as environmental change and the consequent alteration of population dynamics will predictably reshape not only the structure but also the stability of interactions between species (Burkle et al. 2013; Burkle & Alarcón 2011; Hegland et al. 2009; Tylianakis & Morris 2017). The structure and composition of plant-pollinator networks vary significantly across years (CaraDonna et al. 2017; Olesen et al. 2011; Petanidou et al. 2008; Schwarz et al. 2020), being temporal turnover often as high as spatial turnover (Hervías-Parejo et al. 2023; Magrach et al. 2023; Resasco et al. 2021). By comparing pairs of consecutive temporal networks, current statistical methods allow quantifying temporal interaction turnover and decompose it into two components: species turnover, i.e. interaction changes due to gains and losses of species in the communities, and interaction rewiring, i.e. reassembly of interactions among species shared by the two consecutive temporal networks (Fründ 2021; Novotny 2009; Poisot et al. 2012). The species turnover component reflects the contribution of community composition shifts to interaction turnover, while interaction rewiring reflects plastic behavioural responses or generalist foraging strategies that can potentially buffer the effects of community degradation due to global change (CaraDonna et al. 2017; Lázaro & Gómez-Martínez 2022). Although this framework is increasingly applied to describe temporal variation in plant–pollinator networks (CaraDonna et al. 2021; Resasco et al. 2021; Schwarz et al. 2020), a mechanistic understanding of how climate fluctuations and shifts in community composition drive species turnover and rewiring is still lacking. Inter-annual climate variability is expected to strongly influence temporal dynamics in plant–pollinator networks (Dupont et al. 2009; Forrest 2015; Hegland et al. 2009), as the timings of both trophic levels (i.e. flowering and pollinator activity) are highly sensitive to climate cues (Donoso et al. 2016; Duchenne et al. 2020; Peng et al. 2025). However, long-term evidence for how such climatic fluctuations shape interaction dynamics remains scarce. We expect that stronger inter-annual climate variation could raise the contribution of rewiring to interaction turnover, because greater shifts in species’ phenology and abundance are expected to increase the probability of rewiring by altering temporal and numerical matching with potential partners. Ongoing pollinator declines driven by climate change and other anthropogenic pressures (Potts et al. 2010; Wagner et al. 2021) are likely to increase interaction turnover in plant–pollinator networks over time due to increased species turnover. Yet species’ responses to global change are heterogeneous—while many decline or become locally extinct, others persist or even thrive, reflecting a continuum shaped by their unique ecological and life-history traits (Bartomeus et al. 2013; Kammerer et al. 2021; Sunde et al. 2023). As a result, specialist species often become increasingly rare and community composition shifts toward dominance by generalists, which are preadapted to better cope with rapid environmental changes (Clavel et al. 2011; Ghisbain et al. 2021; Gossner et al. 2023). These generalist species typically have more flexible habitat requirements, broader diets, plastic life cycles (e.g. multivoltinism) and higher dispersal ability (Carnicer et al. 2013; Dapporto & Dennis 2013). Morphological traits can also contribute to this continuum. Body size may influence specialization, as larger species with higher energetic demands may forage more selectively than smaller ones, relying more on rewarding resources. Proboscis length, although often associated with higher specialization in bees and other pollinators (Shimizu et al. 2014; Xu et al. 2025; Zhao et al. 2022), may show the reverse pattern in butterflies, where reported evidence suggests that longer-proboscis species tend to be more generalist in their diet (Jain et al. 2016). The growing prevalence of generalist traits is likely to alter the relative contribution of species turnover and rewiring to overall temporal interaction turnover. We expect generalist traits to be associated with lower species turnover but higher frequencies of interaction rewiring (Lázaro & Gómez-Martínez 2022; Olesen et al. 2011; Resasco et al. 2021), reflecting a greater capacity to persist within the community and flexibly adjust interactions in response to environmental change. Consequently, as generalist species become more dominant, the relative role of rewiring is expected to increase over time. Here, we use long-term time series data (13–29 years) on butterfly–plant interactions, butterfly abundance and traits, collected through standardized monitoring in seven Mediterranean communities, to investigate the mechanisms driving inter-annual turnover in mutualistic networks. Specifically, we examine, for the first time, how interaction turnover and its components are shaped by inter-annual climate variability and butterfly community composition shifts. We specifically test three main hypotheses: I) Mediterranean butterfly communities lose species and become increasingly dominated by generalists over the years; II) Interaction turnover increase over the years, driven by both rising species turnover and rewiring, as generalist species may be more prone to rewire; and III) Rewiring increase with greater inter-annual climate variability, reflecting stronger changes in the abundance and phenology of species. By integrating environmental, abundance, interaction, and trait data over nearly three decades, our study reveals how inter-annual climate variability and species loss shape long-term interaction dynamics in plant–pollinator communities. Methods Study sites, butterfly abundance and butterfly-plant interactions Data were collected at seven sites in north-eastern Spain spanning an altitudinal gradient from sea level to 1100 m and covering a broad range of Mediterranean habitats and climates (Table S1, Supplementary Material). The sites include coastal wetlands in arid Mediterranean areas, woodlands, shrublands, riparian forests and agricultural mosaics in the humid Mediterranean zone, and mid-mountain grasslands and shrublands under more humid to subalpine conditions. The dataset covers 13–29 years of standardized monitoring (mean ± SD = 21.3 ± 7.6 years). Sampling was conducted with a temporal resolution of 30 weekly surveys per year (March–September), covering the entire flight period of most butterfly species in the study region (Vila et al. 2018). Each sampling involved counting all butterflies within 2.5 m on each side and 5 m in front of the recorder along fixed transects of 1.1 to 4.3 km (Pollard & Yates 1993), as part of the Catalan Butterfly Monitoring Scheme (www.catalanbms.org). Moreover, we recorded all interactions between flower-visiting butterfly species and their nectar plants, along with their frequencies. We only recorded flower visits when butterflies were visibly feeding on nectar with their proboscis clearly extended. This protocol yielded two separate datasets: butterfly abundance (all individuals recorded along transects) and interaction data (from flower visits). Sampling was conducted on sunny days with low wind and temperatures generally above 15°C to ensure suitable conditions for butterfly activity (Lang et al. 2025). This fixed sampling protocol, in which each site was always surveyed by the same recorder (Table S1, Supplementary Material), minimized sampling variability throughout the study period. Sampling completeness was consistently high across sites and years (Fig. S1, Supplementary Material). Details on UTM coordinates, transect length, annual averages of temperature and precipitation, and the number of butterfly and plant species (interacting with butterflies) recorded at the seven sites are provided in Table S1 (Supplementary Material). To minimise bias in interaction turnover estimates, we excluded from our dataset those populations with fewer than three flower visits per year on average (82 populations excluded) to avoid confounding low detectability with true dietary specialization. The final dataset retained 202 populations from 59 butterfly species (Table S2, Supplementary Material). Climate data We obtained high-resolution climate data of sites (i.e., at the centroid of each transect) using the ClimateDT tool (Marchi et al. 2024), which employs dynamic lapse-rate calculations to downscale climatic surfaces. This tool relies on CHELSA v2.1 (Karger et al. 2017), a global dataset with 30 arcsec spatial resolution (1 km to 500 m). By incorporating temperature variations with elevation and orographic effects on cloud cover and radiation, ClimateDT provides reliable climate estimates for topographically complex landscapes, such as some of our study sites in mountain regions. To characterize climatic conditions at each site and year, we calculated annual estimates of mean temperature and precipitation, from October of the previous year to September of the focal year. This period was chosen to align climate data with the adult butterfly season. These two climatic variables play a key role in butterfly population dynamics, influencing life cycles, habitat suitability, and food availability (Hill et al. 2021). Finally, to estimate inter-annual climate variation, we calculated the log-ratio between consecutive years for each variable (log (X t / X t-1 )), which reflects the relative change and reduces the impact of extreme values. We retained climatic variation in two forms, as the raw difference between consecutive years (i.e., directional variation, indicating warming or cooling) and the absolute variation (i.e., the magnitude of inter-annual variation). Butterfly species’ traits We used four traits—voltinism (number of generations per year), adult habitat specialization, nectar resource selectivity, and mobility—that are tightly linked to the specialist–generalist trait continuum described in Mediterranean butterflies (Carnicer et al. 2013). Along this continuum, butterfly species tend to range from multivoltine, habitat/trophic generalists with high mobility to univoltine, habitat/trophic specialists with low mobility (Carnicer et al. 2013; Dapporto & Dennis 2013). In addition, we included two morphological traits—wing length (as a proxy for body size) and proboscis length which may independently affect foraging strategies and the capacity for interaction rewiring. We included voltinism as a categorical variable distinguishing between univoltine species (one generation per year) and multivoltine species (multiple generations per year), based on the species life-history in the study region (Vila et al. 2018). Habitat specialization was quantified using the Habitat Specialisation Index (HSI), calculated as the coefficient of variation in population density across several biotopes in over 200 sites of the Catalan Butterfly Monitoring Scheme (Colom et al. 2019); higher HSI values indicate greater habitat specialisation (range: 0.46–1.97). Flower selectivity was measured using the specialization index d′ (Blüthgen et al. 2006), which quantifies how selectively a butterfly species interacts with available flowering plants, with values ranging from 0 (generalist) to 1 (specialist). We calculated d’ for each species’ population and year and then obtained an averaged value per population (i.e., each species at each site across years). Mobility was assessed based on expert criteria and treated as a categorical variable with four levels, following Stefanescu et al. (2011): (1) species with highly isolated populations, typically exhibiting a metapopulation structure, (2) species structured in metapopulations but with frequent dispersal and colonization of new areas, (3) species with highly open populations, and (4) migratory species. Wing length (from the apex to the insertion on the thorax) was sourced from García-Barros et al. (2013) (range = 10.65–39 mm). To measure proboscis length, we collected butterfly individuals, took them to the laboratory where we then uncoiled the proboscis and extended it over a board with a hooked pin, marked the start and end point, and measured the distance with a ruler (range = 5.28–18.2 mm). For each species, we used an average distance for at least three individuals. The full list of butterfly species (n = 59) and their assigned trait values is provided in Table S2 (Supplementary Material). Partitioning interaction turnover into species turnover and rewiring For each study site, we calculated interaction turnover as the overall β-diversity (Whittaker’s dissimilarity) in plant–butterfly interactions between surveys conducted in two consecutive years (βWN: Whole Network dissimilarity; Fründ 2021; Novotny 2009; Poisot et al. 2012). Interaction turnover was partitioned into two components (Fründ 2021; Novotny 2009; Poisot et al. 2012): species turnover (βST), i.e., interaction changes resulting from species gains and losses between pairs of consecutive temporal networks, and rewiring (βOS), i.e., the reassembly of interactions due to partner-switching among species shared between consecutive networks. To do this, we used the method described by Fründ (2021) and based on Novotny (2009), which allows separating the contribution of species turnover and rewiring respect to a common denominator representing the total number of interactions in paired networks of two consecutive years. We chose this method because it enables further decomposition of species turnover into the two trophic levels, i.e., butterflies (βSTb) and plants (βSTp), which was needed to assess their relative contribution to interaction turnover. We calculated interaction turnover and its components using betalinkr function (Fründ 2021) from bipartite package in R version 4.3.3 (Dormann et al. 2008). Butterfly species’ persistence in the network and rewiring frequency For each butterfly population (i.e., species–site combination), we quantified two complementary temporal metrics that describe its contribution to the interaction network. Persistence was defined as the proportion of years in which the species was recorded interacting with flowers relative to the total number of sampling years at that site. Rewiring probability was first defined as a binary variable for each population and pair of consecutive years (1 = species rewired at least one interaction; 0 = no rewiring). Rewiring frequency was then calculated as the proportion of year pairs in which rewiring occurred relative to the total number of pairs evaluated. To relate these metrics to inter-annual changes in butterfly abundance and phenology, we fitted counts across Julian days for each combination of species, site, and year, using generalized additive models (GAMs) with a Poisson error distribution and restricted maximum likelihood estimation (Colom et al. 2022). From each model, we predicted (1) the area under the phenological curve as a proxy of annual abundance; (2) the length of the phenological curve (i.e. number of days); and (3) the Julian day in which the peak of abundance occurred. The GAM approach is less sensitive to variation in sampling effort and species detectability, providing more accurate estimates of abundance and phenology than other methods (Moussus et al. 2010; Schmucki et al. 2016). Inter-annual variation of these variables was calculated as the log-ratio between consecutive years (log (X t / X t-1 )) for each species population and pair of consecutive years. For phenological variables (flight period length and peak date), we retained both the directional variation (indicating advancement or delay) and the absolute variation (i.e., the magnitude of phenological change). Statistical analyses To assess temporal trends in butterfly species richness (i.e., the number of species recorded per site–year combination in the interaction dataset), we fitted a generalized linear mixed model (GLMM) with year as a continuous predictor variable, and richness as the response variable. Site was included as a random intercept to account for baseline differences among sites. We used a Gaussian distribution with a log link because Poisson and negative binomial models showed overdispersion (Zuur et al. 2009). We also attempted a model with random slopes for year by site (Zuur et al. 2009), but it provided unstable estimates and was not retained. To further inspect heterogeneity in butterfly species richness trends among communities, we complemented the GLMM with site-specific linear models. To relate temporal and environmental gradients to trait-based changes in butterfly communities, we used the RLQ-fourth-corner framework (Dray et al. 2014). We assembled three matrices: L (species abundances by site–year), R (temporal/environmental variables per site–year), and Q (species traits; see section 2.3) from the interaction dataset. Quantitative traits (HSI, d′, wing length and proboscis length) were centred and scaled; categorical traits (voltinism and mobility) were treated as dummy variables. The RLQ approach (Dolédec et al. 1996), consisted in running a correspondence analysis of L to obtain row/column weights; a PCA of R with variables centred and scaled and weighted by L row weights; a Hill–Smith analysis (which accommodates both type of trait variables in a single ordination) of Q weighted by L column weights; and then combined them by RLQ to extract the main trait–environment co-structure (Dray et al. 2014). We fitted the RLQ including year, annual mean temperature and annual precipitation after partialling out the site effect from each R variable (residuals from models with site as a factor), thereby isolating temporal change within sites and avoiding bias from unequal time series among sites. Fourth-corner analysis was used to assess the statistical significance of trait–temporal/environmental associations (Brown et al. 2014). We used the max-test, which takes the larger of the p-values from permuting rows of L (breaking R–L) and columns of L (breaking L–Q), retaining only associations that survive both shuffles. We controlled the false discovery rate (Benjamini–Hochberg, α = 0.05) and assessed overall RLQ co-structure with a Monte-Carlo test on eigenvalues (randtest, 9,999 permutations) (Dray et al. 2014). We ran GLMMs to assess the role of inter-annual climate variation and time (years) on interaction turnover and its components. We built different models for the following response variables: total interaction turnover (βWN), total species turnover (βST), butterfly species turnover (βSTb), plant species turnover (βSTp) and rewiring (βOS). All models included temperature and precipitation variation (including separated variables for directional and absolute changes), as well as year (treated as a continuous variable), as predictors, and site as random intercept. In each model we used the error distributions that best fitted the data: βWN and βST were modelled using a Gaussian distribution with a log-link function, while βSTb and βSTp followed a Gamma distribution (log-link function), and βOS was fitted with a Gaussian model (identity-link function), as the response variable was approximately normally distributed and model diagnostics indicated that residuals satisfied the assumptions of normality. At the species level, we fitted GLMMs with persistence in the network and rewiring frequency as response variables, and species traits as predictors. Persistence was modelled with a beta distribution and rewiring frequency with a log-normal Gaussian distribution. Species and site were included as crossed random intercepts. Finally, we fitted a binomial GLMM to test the effect of abundance and phenology on the probability that a population rewired its interactions from year to year at a given site. In this model, rewiring probability (1 = rewiring; 0 = no rewiring) was the response variable, and the inter-annual variation in butterfly abundance, flight period length, and peak abundance date were the predictors—each included as both the directional and absolute variation. Species, site, and year pairs (e.g. 2000–2001) were included as random effects. We conducted all analyses using R (version 4.3.3). GLMMs were fitted with the glmmTMB package (Brooks et al. 2017). Model selection was performed using MuMIn retaining the most parsimonious model within the set of best-supported models (ΔAIC < 2) (Barton 2020). Model assumptions were evaluated with DHARMa (Hartig 2022), and normality of variable distributions and multicollinearity among predictors (all VIFs < 3) were assessed using the performance package (Lüdecke et al. 2021). RLQ and fourth-corner analyses were conducted using the ade4 package (Dray & Dufour 2007; Thioulouse et al. 2018). Results Temporal trends in butterfly richness and community trait composition We found a significant decline in butterfly species richness in the networks over time (χ² = 35.75, df = 1, p < 0.001), with richness decreasing at an average rate of 0.83% per year (Fig. 1). Site-specific linear models confirmed consistent negative trends across most sites, although the strength and significance of the declines varied (Fig. 1). Four sites showed strong declines, with annual losses ranging between 1.7% and 3.8%, whereas the remaining three sites did not exhibit significant temporal trends. See Table S3 (Supplementary Material) for site-level statistics. We found clear temporal and environmental patterns in butterfly community trait composition. The RLQ revealed a dominant first axis (RLQ1) contributing 82.3% to the co-structure (Fig. 2). Regarding temporal/environmental variables, RLQ1 captured a gradient of decreasing precipitation and increasing temperature and years (Fig. 2; see Supplementary Table S4 for variable scores). Regarding traits, RLQ1 was positively associated with multivoltine, extreme mobilities (low-dispersal or migratory) and morphologically larger species (including wing and proboscis length), and negatively associated with univoltine, medium or high dispersal and more habitat and flower-selective species (Fig. 2; see Supplementary Table S5 for trait scores). The fourth-corner analysis revealed a significant correlation between the temporal/environmental variables (R) and species traits (Q) structuring RLQ1 (r = 0.18, p = 0.006). This indicates that temporal community changes have been primarily driven by a shift towards multivoltine, habitat and flower generalists, low dispersal, migratory and morphologically larger species under warming and drying trends. RLQ2 accounted for 16.9% of the co-structure, but traits and temporal/environmental variables were not significantly correlated along this axis (r = 0.07, p > 0.05). Temporal and climate effects on interaction turnover and their components We found significant temporal trends in butterfly–plant interaction turnover and its components, with rewiring dominating early in the time series but species turnover progressively increasing its contribution over years (Fig. 3a). Inter-annual climate variation influenced only interaction rewiring (Fig. 3b). Total interaction turnover (βWN) increased significantly over time ( χ² = 4.77, df = 1, p = 0.029), primarily driven by increases in total species turnover (βST; χ² = 26.24, df = 1, p < 0.001). The decomposition of βST shows that both butterfly species turnover (βSTb) and plant species turnover (βSTp) increased over time (βSTb: χ² = 25.84, df = 1, p < 0.001; βSTp: χ² = 6.36, df = 1, p = 0.012), at similar annual rates (+1.85%/year, 95-CI: 1.14–2.56 for butterflies; +1.75%/year, 95CI: 0.59–2.92 for plants). In contrast, interaction rewiring (βOS) declined over time (χ² = 25.53, df = 1, p < 0.001) and only increased with the absolute temperature variation (χ² = 4.16, df = 1, p = 0.041; Fig. 3b). Alternative models (ΔAIC < 2) for all response variables consistently retained year as a predictor variable and, in the case of βOS, also the absolute temperature variation, along with other non-significant variables (Table S6, Supplementary Material). Trait-based drivers of species’ persistence in the networks and rewiring frequency Persistence in the networks was explained by flower selectivity ( d’ ) (Fig. 4a), while rewiring frequency was explained by proboscis length (Fig. 4b) and voltinism (Fig. 4c). More selective species were less persistent over time than generalist ones ( χ ² = 5.85, df = 1, p = 0.016). Species with longer proboscises rewired interactions more frequently than those with shorter proboscises ( χ ² = 8.01, df = 1, p = 0.004), and multivoltine species more frequently than univoltine (χ² = 17.47, df = 1, p < 0.001). All alternative persistence models (ΔAIC < 2) included flower selectivity along with other non-significant traits. For rewiring frequency, all alternative models retained proboscis length and voltinism together with other non-significant traits (Table S7, Supplementary Material). Abundance and phenology effects on rewiring probability Phenological and abundance year-to-year shifts significantly explained variation in rewiring probability (Fig. 5). Rewiring probability increased with the increase in the absolute variation in peak flight day ( χ² = 7.39, df = 1, p = 0.007; Fig. 5a), flight period length ( χ² = 17.53, df = 1, p < 0.001; Fig. 5b), and abundance ( χ² = 4.33, df = 1, p = 0.037; Fig. 5c). Alternative models (ΔAIC < 2) consistently retained these three predictors in their absolute form but differed in the inclusion of additional non-significant predictors in their directional form (Table S8, Supplementary Material). Discussion Over the past three decades, most of the studied butterfly–plant networks experienced rapid losses of species, with butterfly community composition shifting toward generalist, multivoltine and larger-bodied species. For instance, El Puig—the richest network of the study system —lost about one-fifth of its butterfly richness over 29 years, illustrating the magnitude of these changes. In line with these compositional shifts, interaction turnover rose across years, driven by increasing species turnover. In contrast, the contribution of rewiring declined over time, even though communities became dominated by traits associated with greater rewiring capacity, particularly multivoltinism and longer proboscis length. Yet rewiring increased in years with stronger temperature fluctuations, when greater variation in butterfly abundance and phenology elevated the probability of partner switching. Our results are consistent with the strong declines reported for butterfly communities across the Mediterranean region in recent decades—largely attributed to increasing aridity and vegetation encroachment (Colom et al. 2022; Melero et al. 2016; Stefanescu et al. 2011a, b; Ubach et al. 2020). Notably, these declines are likely underestimated in our study, as rare species with few flower-visit records were not included in our dataset. The progressive increase in interaction turnover indicates that such long-term biodiversity erosion, rather than short-term climate variability, has been the dominant force shaping network dynamics across decades. Species turnover increases when disappearing species account for a larger proportion of the network’s total interactions (Minachilis et al. 2023; Zhang et al. 2023). As richness declines, each species loss removes a larger share of total interactions, reducing partner overlap between consecutive years and thereby amplifying species turnover. This pattern is consistent with theoretical expectations indicating that lower diversity is generally linked to higher temporal turnover (Shurin 2007). Rewiring was the dominant component of temporal interaction turnover, consistent with other studies at both intra-annual (CaraDonna et al. 2017; Hervías-Parejo et al. 2023) and inter-annual scales (Petanidou et al. 2008). However, our long-term approach shows that the contribution of rewiring vs. interaction turnover declines as plant–pollinator networks lose species, since a shrinking species pool constrains the possibility of reorganizing interactions among persisting taxa, thereby reducing the potential for rewiring (Zhang et al. 2023). A similar pattern was reported in a century-long resurvey of a prairie pollination network, where species loss led to a collapse of interaction richness with little compensatory rewiring (Burkle et al. 2013). Inter-annual climate variability did not alter overall interaction turnover, yet it influenced how much of that turnover was due to rewiring. Stronger temperature variations between years were associated with higher rewiring, and butterfly species that experienced greater shifts in abundance or phenology were more likely to rewire interactions. This pattern supports the idea that populations under fluctuating conditions adjust their interactions to persist, either because there are phenological or numerical mismatches with the preferred partners or because changes in phenologies and abundances generate new opportunities for interaction (Bartley et al. 2019; Hegland et al. 2009; Visser & Gienapp 2019). Such flexibility, although bounded by structural constraints (Carstensen et al. 2016), is often seen as a mechanism buffering communities against the destabilizing effects of demographic and phenological variability (CaraDonna et al. 2017; Fuzessy & Pizo 2025; Saavedra et al. 2016). Yet our results also suggest that this buffering capacity is fragile: as species richness declines, the pool of potential partners shrinks, and repeated or extreme climate fluctuations may eventually outpace species ability to establish effective new interactions, potentially triggering cascading effects on network stability (Pires et al. 2020). Butterfly traits played a clear role in shaping the species’ persistence in the networks and rewiring potential. Flower selectivity influenced persistence, with generalist-feeding butterflies persisting longer across years than specialists. This aligns with evidence that interaction generalists tend to occupy central, stable positions in pollination networks, whereas specialists are more exposed to partner loss and environmental change (Resasco et al. 2021; Zografou et al. 2020). Rewiring was more frequent in multivoltine than in univoltine species, likely because their extended activity windows across multiple generations increase the likelihood of overlapping with a broader range of flowering phenologies (Stefanescu & Traveset 2009). Butterflies with longer proboscises also contributed more to rewiring. In other pollinators, such as bees or bumblebees, long proboscises are often associated with dietary specialization and reduced partner flexibility (Shimizu et al. 2014; Xu et al. 2025; Zhao et al. 2022). In contrast, our results show that proboscis length in butterflies correlates negatively with flower selectivity (Fig. 2), a relationship we further confirmed with a mixed-model analysis accounting for site variation (χ² = 19.4, df = 1, p < 0.001). This pattern is consistent with evidence that butterflies with longer proboscises tend to be more generalist, visiting a wider diversity of flower types (Jain et al. 2016). Unlike many hymenopteran and dipteran pollinators, whose tongue length often co-evolved with particular flower morphologies (Anderson 2015), lepidopteran species may face weaker constraints to specialize, potentially because they rely exclusively on nectar and not on pollen provisioning. However, the exact mechanisms behind this contrasting relationship remain unclear and deserve further study. Over time, community composition has shifted mainly towards multivoltine, migratory species with longer proboscises, indicating that species persisting under environmental change are those most capable of adjusting their interactions. Yet, this trait filtering, while favoring species able to rewire, has not prevented a gradual reduction in the relative contribution of rewiring at the community level. Rewiring is an inherently diversity-dependent process: as species are lost, the pool of potential partners and temporal overlap diminishes, constraining the ability of flexible species to reorganize interactions (CaraDonna et al. 2017; Fuzessy & Pizo 2025). Thus, trait composition can modulate but not compensate for the reduction of interaction opportunities that accompanies species loss. Overall, our findings highlight general mechanisms by which global change reshapes the dynamics of ecological networks. Long-term biodiversity loss amplifies the role of species turnover and erodes opportunities for interaction reorganization, while trait filtering favors species with grater rewiring potential over time, but cannot fully buffer against the consequences of species losses. Species’ ability to rewire interactions can buffer plant–pollinator mismatches under short-term climate variability, yet this flexibility may collapse as biodiversity loss reduces the pool of potential partners. Acknowledgements This work was supported by the Spanish Ministry of Science, Innovation and Universities and the Spanish Research Agency through the project FUNLAND (PID2023-150295OB-I00, Proyectos de Generación de Conocimiento 2023, BDNS: 732796). We also thank the Departament d’Acció Climàtica, Alimentació i Agenda Rural de la Generalitat de Catalunya and the Diputació de Barcelona for funding support. We are grateful to Moisès Guardiola for his valuable assistance with plant identifications. References Anderson, B. (2015). Coevolution in mutualisms. In: Mutualism (ed. Bronstein, J.L.). Oxford University Press, Oxford, pp. 107–130.Bartley, T.J., McCann, K.S., Bieg, C., Cazelles, K., Granados, M., Guzzo, M.M., et al. (2019). Food web rewiring in a changing world. Nat Ecol Evol, 3, 345–354.Bartomeus, I., Ascher, J.S., Gibbs, J., Danforth, B.N., Wagner, D.L., Hedtke, S.M., et al. (2013). Historical changes in northeastern US bee pollinators related to shared ecological traits. Proc Nat Acad Sci USA , 110, 4656–4660.Barton, K. (2020). MuMIn: multi-model inference. R package version 1.43.17.Blüthgen, N., Menzel, F. & Blüthgen, N. (2006). Measuring specialization in species interaction networks. BMC Ecol. , 6, 9.Brooks, M.E., Kristensen, K., van Benthem, K.J., Magnusson, A., Berg, C.W., Nielsen, A., et al. (2017). glmmTMB Balances Speed and Flexibility Among Packages for Zero-inflated Generalized Linear Mixed Modeling. R J , 9, 378–400.Brown, A.M., Warton, D.I., Andrew, N.R., Binns, M., Cassis, G. & Gibb, H. (2014). The fourth-corner solution - using predictive models to understand how species traits interact with the environment. Methods Ecol Evol , 5, 344–352.Burkle, L.A. & Alarcón, R. (2011). The future of plant-pollinator diversity: Understanding interaction networks acrosss time, space, and global change. Am J Bot , 98, 528–538.Burkle, L.A., Marlin, J.C. & Knight, T.M. (2013). Plant-pollinator interactions over 120 years: Loss of species, co-occurrence, and function. Science (1979), 340, 1611–1615.CaraDonna, P.J., Burkle, L.A., Schwarz, B., Resasco, J., Knight, T.M., Benadi, G., et al. (2021). Seeing through the static: the temporal dimension of plant–animal mutualistic interactions. Ecol Lett , 24, 149–161.CaraDonna, P.J., Petry, W.K., Brennan, R.M., Cunningham, J.L., Bronstein, J.L., Waser, N.M., et al. (2017). Interaction rewiring and the rapid turnover of plant–pollinator networks. Ecol Lett , 20, 385–394.Carnicer, J., Stefanescu, C., Vila, R., Dincǎ, V., Font, X. & Peñuelas, J. (2013). A unified framework for diversity gradients: The adaptive trait continuum. Glob Ecol Biogeogr , 22, 6–18.Carstensen, D.W., Sabatino, M. & Morellato, L.P.C. (2016). Modularity, pollination systems, and interaction turnover in plant-pollinator networks across space. Ecology , 97, 1298–1306.Clavel, J., Julliard, R. & Devictor, V. (2011). Worldwide decline of specialist species: Toward a global functional homogenization? Front Ecol Environ , 9, 222–228.Colom, P., Carreras, D. & Stefanescu, C. (2019). Long-term monitoring of Menorcan butterfly populations reveals widespread insular biogeographical patterns and negative trends. Biodivers Conserv , 28, 1837–1851.Colom, P., Ninyerola, M., Pons, X., Traveset, A. & Stefanescu, C. (2022). Phenological sensitivity and seasonal variability explain climate-driven trends in Mediterranean butterflies. Proc R Soc B , 289(1973), 20220251Dapporto, L. & Dennis, R.L.H. (2013). The generalist-specialist continuum: Testing predictions for distribution and trends in British butterflies. Biol Conserv , 157, 229–236.Dolédec, S., Chessel, D., Ter Braak, C.J.F. & Champely, S. (1996). Matching species traits to environmental variables: a new three-table ordination method. Environ Ecol Stat , 3, 143–166.Donoso, I., Stefanescu, C., Martínez-Abraín, A. & Traveset, A. (2016). Phenological asynchrony in plant–butterfly interactions associated with climate: a community-wide perspective. Oikos , 125, 1434–1444.Dormann, C.F., Gruber, B. & Fründ, J. (2008). Introducing the bipartite Package: Analysing Ecological Networks. R News , 8, 8–11.Dray, S., Choler, P., Dolédec, S., Peres-Neto, P.R., Thuiller, W., Pavoine, S., et al. (2014). Combining the fourth-corner and the RLQ methods for assessing trait responses to environmental variation. Ecology , 95, 14–21.Dray, S., & Dufour, A. B. (2007). The ade4 package: implementing the duality diagram for ecologists. J Stat Softw , 22, 1-20.Duchenne, F., Thébault, E., Michez, D., Elias, M., Drake, M., Persson, M., et al. (2020). Phenological shifts alter the seasonal structure of pollinator assemblages in Europe. Nat Ecol Evol , 4, 115–121.Dupont, Y.L., Padrón, B., Olesen, J.M. & Petanidou, T. (2009). Spatio-temporal variation in the structure of pollination networks. Oikos , 118, 1261–1269.Forrest, J.R.K. (2015). Plant-pollinator interactions and phenological change: What can we learn about climate impacts from experiments and observations? Oikos , 124, 4–13.Fründ, J. (2021). Dissimilarity of species interaction networks: how to partition rewiring and species turnover components. Ecosphere , 12, e03653.Fuzessy, L.F. & Pizo, M.A. (2025). Navigating a changing world: on the significance of rewiring for mutualistic interactions, caveats and future directions. Oikos , 2025, e11230.García-Barros, E., Munguira, M.L., Stefanescu, C. & Vives-Moreno, A. (2013). Lepidoptera: Papilionoidea . Museo Nacional de Ciencias Naturales-CSIC, Madrid.Ghisbain, G., Gérard, M., Wood, T.J., Hines, H.M. & Michez, D. (2021). Expanding insect pollinators in the Anthropocene. Biol Rev , 96, 2755–2770.Gossner, M.M., Menzel, F. & Simons, N.K. (2023). Less overall, but more of the same: drivers of insect population trends lead to community homogenization. Biol Lett , 19, 20230057Hartig, F. (2022). DHARMa: ResidDiagnostics for Hierarchical (Multi-Level / Mixed) Regression Models. R package version 0.4.6. Hegland, S.J., Nielsen, A., Lázaro, A., Bjerknes, A.L. & Totland, Ø. (2009). How does climate warming affect plant-pollinator interactions? Ecol Lett , 12, 184–195.Hervías-Parejo, S., Colom, P., Beltran, R., Serra, P., Pons, S., Mesquida, V., et al. (2023). Spatio-temporal variation in plant-pollinator interactions: a multilayer network approach. Oikos , 2023, e09661.Hill, G.M., Kawahara, A.Y., Daniels, J.C., Bateman, C.C. & Scheffers, B.R. (2021). Climate change effects on animal ecology: butterflies and moths as a case study. Biol Rev , 96, 2113–2126.Jain, A., Kunte, K. & Webb, E.L. (2016). Flower specialization of butterflies and impacts of non-native flower use in a transformed tropical landscape. Biol Conserv , 201, 184–191.Kammerer, M., Goslee, S.C., Douglas, M.R., Tooker, J.F. & Grozinger, C.M. (2021). Wild bees as winners and losers: Relative impacts of landscape composition, quality, and climate. Glob Chang Biol , 27, 1250–1265.Karger, D.N., Conrad, O., Böhner, J., Kawohl, T., Kreft, H., Soria-Auza, R.W., et al. (2017). Climatologies at high resolution for the earth’s land surface areas. Sci Data , 4.Lang, A., Kallhardt, F., Lee, M.S., Loos, J., Molander, M.A., Pettersson, L.B., et al. (2025). Do standard weather conditions and flower density affect the results of butterfly monitoring schemes? A field test in three bio-geographic regions in Europe. J Insect Conserv , 29, 1–11.Lázaro, A. & Gómez-Martínez, C. (2022). Habitat loss increases seasonal interaction rewiring in plant–pollinator networks. Funct Ecol , 36, 2673–2684.Lüdecke, D., Ben-Shachar, M., Patil, I., Waggoner, P. & Makowski, D. (2021). performance: An R Package for Assessment, Comparison and Testing of Statistical Models. J Open Source Softw , 6, 3139.Magrach, A., Artamendi, M., Lapido, P.D., Parejo, C. & Rubio, E. (2023). Indirect interactions between pollinators drive interaction rewiring through space. Ecosphere , 14, 1–15.Marchi, M., Bucci, G., Iovieno, P. & Ray, D. (2024). ClimateDT: A Global Scale-Free Dynamic Downscaling Portal for Historic and Future Climate Data. Environments , 11(4), 82.Melero, Y., Stefanescu, C. & Pino, J. (2016). General declines in Mediterranean butterflies over the last two decades are modulated by species traits. Biol Conserv , 201, 336–342.Minachilis, K., Kantsa, A., Devalez, J., Vujic, A., Pauly, A. & Petanidou, T. (2023). High species turnover and unique plant–pollinator interactions make a hyperdiverse mountain. J Anim Ecol , 92, 1001–1015.Moussus, J.-P., Julliard, R. & Jiguet, F. (2010). Featuring 10 phenological estimators using simulated data. Methods Ecol Evol , 1, 140–150.Novotny, V. (2009). Beta diversity of plant-insect food webs in tropical forests: A conceptual framework. Insect Conserv Divers , 2, 5–9.Olesen, J.M., Stefanescu, C. & Traveset, A. (2011). Strong, Long-Term Temporal Dynamics of an Ecological Network. PLoS One , 6, e26455.Peng, S., Ellison, A.M. & Davis, C.C. (2025). Climate change intensifies plant–pollinator mismatch and increases secondary extinction risk for plants in northern latitudes. Proc Natl Acad Sci USA , 122, e2506265122.Petanidou, T., Kallimanis, A.S., Tzanopoulos, J., Sgardelis, S.P. & Pantis, J.D. (2008). Long-term observation of a pollination network: Fluctuation in species and interactions, relative invariance of network structure and implications for estimates of specialization. Ecol Lett , 11, 564–575.Pires, M.M., O’Donnell, J.L., Burkle, L.A., Díaz-Castelazo, C., Hembry, D.H., Yeakel, J.D., et al. (2020). The indirect paths to cascading effects of extinctions in mutualistic networks. Ecology , 101, e03080.Poisot, T., Canard, E., Mouillot, D., Mouquet, N. & Gravel, D. (2012). The dissimilarity of species interaction networks. Ecol Lett , 15, 1353–1361.Pollard, E. & Yates, T.J. (1993). Monitoring butterflies for ecology and conservation: the British butterfly monitoring scheme . Springer Science & Business Media.Potts, S.G., Biesmeijer, J.C., Kremen, C., Neumann, P., Schweiger, O. & Kunin, W.E. (2010). Global pollinator declines: Trends, impacts and drivers. Trends Ecol Evol , 25, 345–353.Resasco, J., Chacoff, N.P. & Vázquez, D.P. (2021). Plant–pollinator interactions between generalists persist over time and space. Ecology , 102, e03367.Saavedra, S., Rohr, R.P., Fortuna, M.A., Selva, N. & Bascompte, J. (2016). Seasonal species interactions minimize the impact of species turnover on the likelihood of community persistence. Ecology , 97, 865–873.Schmucki, R., Pe’er, G., Roy, D.B., Stefanescu, C., Swaay, C.A.M. van, Oliver, T.H., et al. (2016). A regionally informed abundance index for supporting integrative analyses across butterfly monitoring schemes. J Appl Ecol , 53, 501–510.Schwarz, B., Vázquez, D.P., CaraDonna, P.J., Knight, T.M., Benadi, G., Dormann, C.F., et al. (2020). Temporal scale-dependence of plant–pollinator networks. Oikos , 129, 1289–1302.Shimizu, A., Dohzono, I., Nakaji, M., Roff, D.A., Miller, D.G., Osato, S., et al. (2014). Fine-tuned bee-flower Coevolutionary state hidden within multiple pollination interactions. Sci Rep , 4.Shurin, J.B. (2007). How is diversity related to species turnover through time? Oikos , 116, 957–965.Stefanescu, C., Carnicer, J. & Peñuelas, J. (2011a). Determinants of species richness in generalist and specialist Mediterranean butterflies: The negative synergistic forces of climate and habitat change. Ecography , 34, 353–363.Stefanescu, C., Torre, I., Jubany, J. & Páramo, F. (2011b). Recent trends in butterfly populations from north-east Spain and Andorra in the light of habitat and climate change. J Insect Conserv , 15, 83–93.Stefanescu, C. & Traveset, A. (2009). Factors influencing the degree of generalization in flower use by Mediterranean butterflies. Oikos , 118, 1109–1117.Sunde, J., Franzén, M., Betzholtz, P.E., Francioli, Y., Pettersson, L.B., Pöyry, J., et al. (2023). Century-long butterfly range expansions in northern Europe depend on climate, land use and species traits. Commun Biol , 6.Thioulouse, D., Dray, S., Dufour, A., Siberchicot, A., Jombart, T. & Pavoine, S. (2018). Multivariate Analysis of Ecological Data with ade4 . Springer, New York.Tylianakis, J.M. & Morris, R.J. (2017). Ecological Networks Across Environmental Gradients. Annu Rev Ecol Evol Syst , 48, 34.Ubach, A., Páramo, F., Gutiérrez, C. & Stefanescu, C. (2020). Vegetation encroachment drives changes in the composition of butterfly assemblages and species loss in Mediterranean ecosystems. Insect Conserv Divers , 13, 151–161.Vila, R., Stefanescu, C. & Sesma, J.M. (2018). Guia de les papallones diürnes de Catalunya . Lynx Edició. Barcelona.Visser, M.E. & Gienapp, P. (2019). Evolutionary and demographic consequences of phenological mismatches. Nat Ecol Evol , 3, 879–885.Wagner, D.L., Grames, E.M., Forister, M.L., Berenbaum, M.R. & Stopak, D. (2021). Insect decline in the Anthropocene: Death by a thousand cuts. Proc Natl Acad Sci USA , 118, 1–10.Xu, X., Liang, H., Ren, Z.X., Maruyama, P.K., Rodrigo Rech, A., Trunschke, J., et al. (2025). Generalised bumblebee–flower interactions demonstrate weak floral niche partitioning despite a high bee diversity. Ecography , 48, e07643.Zhang, X., Dalsgaard, B., Staab, M., Zhu, C., Zhao, Y., Gonçalves, F., et al. (2023). Habitat fragmentation increases specialization of multi-trophic interactions by high species turnover. Proc R Soc B Biol Sci , 290, 20230454.Zhao, Y.H., Lázaro, A., Li, H.D., Tao, Z. Bin, Liang, H., Zhou, W., et al. (2022). Morphological trait-matching in plant–Hymenoptera and plant–Diptera mutualisms across an elevational gradient. J Anim Ecol , 91, 196–209.Zografou, K., Swartz, M.T., Tilden, V.P., McKinney, E.N., Eckenrode, J.A. & Sewall, B.J. (2020). Stable generalist species anchor a dynamic pollination network. Ecosphere , 11, e03066.Zuur, A.F., Ieno, E.N., Walker, N.J., Saveliev, A.A. & Smith, G.M. (2009). Mixed Effects Models and Extensions in Ecology with R . Springer, New York. List of figures Figure 1. Figure 2. Figure 3. Figure 4. Figure 5. List of legends Figure 1. Temporal trends in butterfly species richness across sites. Coloured lines show site-specific linear trends with 95% confidence intervals, and points represent annual observed richness per site and year. The overall trend across all sites is shown in black with its 95% confidence interval. Figure 2. RLQ biplot showing the joint structure of temporal/environmental variables (blue) and butterfly species traits (orange) across the first two RLQ axes. Categorical trait levels are represented as centroids (triangles for voltinism, circles for mobility). Significant trait–temporal/environmental relationships were detected along RLQ1 but not along RLQ2. Figure 3. Butterfly–plant interaction turnover and its components across years and in relation to temperature variation. (a) Temporal trend for total interaction turnover (βWN), total species turnover (βST), butterfly species turnover (βSTb), plant species turnover (βSTp) and interaction rewiring (βOS). (b) Positive relationship between βOS and the absolute temperature variation. The relationship was significant when temperature variation was included in absolute values, but real values (positive and negative) are shown in the plot for illustrative purposes. In both panels, each point represents a site-by-year observation, with fitted regression lines and 95% confidence intervals shown. Figure 4. Butterfly traits shaping persistence and rewiring. Relationships between persistence in the networks and (a) flower selectivity, and between rewiring frequency and (b) proboscis length and (c) voltinism. Panels (a) and (b) show the predicted trends (black line with 95% CI in blue shading), and panel (c) shows predicted means (points ±95% CI) for univoltine and multivoltine species. Predictions were back-transformed from the log scale. In both panels, points represent raw data of species–site combinations. Figure 5. Rewiring probability and changes in abundance and phenology. Rewiring probability in butterfly populations across years in relation to the absolute variation in the (a) peak flight day, (b) flight period length, and (c) annual abundance. Predictions are based on the most parsimonious binomial GLMM, with 95% confidence intervals shaded. Note that both negative and positive values are displayed for illustrative purposes, although the model included the absolute variation for the three variables. Points along the top and bottom margins represent observed values of the binary response variable (rewiring = 1, no rewiring = 0) for each butterfly population in a pair of consecutive years. Information & Authors Information Version history V1 Version 1 16 October 2025 Peer review timeline Published Ecology Letters Version of Record 20 Mar 2026 Published Copyright This work is licensed under a Non Exclusive No Reuse License. Collection Ecology Letters Keywords biotic homogenization interaction dynamics mutualistic networks phenology pollinator decline rewiring species traits species turnover Authors Affiliations Pau Colom 0000-0003-0309-8886 [email protected] Universitat de Barcelona View all articles by this author Constantí Stefanescu Natural Sciences Museum of Granollers View all articles by this author Jordi Corbera 0000-0003-3583-3929 ICHN View all articles by this author Amparo Lázaro Mediterranean Institute of Advanced Studies View all articles by this author Metrics & Citations Metrics Article Usage 397 views 150 downloads .FvxKWukQNSOunydq8rnd { width: 100px; } Citations Download citation Pau Colom, Constantí Stefanescu, Jordi Corbera, et al. Three decades of butterfly--plant interaction turnover explained by climate and species loss. Authorea . 16 October 2025. DOI: https://doi.org/10.22541/au.176062539.92924262/v1 If you have the appropriate software installed, you can download article citation data to the citation manager of your choice. Simply select your manager software from the list below and click Download. For more information or tips please see 'Downloading to a citation manager' in the Help menu . Format Please select one from the list RIS (ProCite, Reference Manager) EndNote BibTex Medlars RefWorks Direct import Tips for downloading citations document.getElementById('citMgrHelpLink').addEventListener('click', function() { popupHelp(this.href); return false; }); $(".js__slcInclude").on("change", function(e){ if ($(this).val() == 'refworks') $('#direct').prop("checked", false); $('#direct').prop("disabled", ($(this).val() == 'refworks')); }); View Options View options PDF View PDF Figures Tables Media Share Share Share article link Copy Link Copied! Copying failed. Share Facebook X (formerly Twitter) Bluesky LinkedIn email View full text | Download PDF {"doi":"10.22541/au.176062539.92924262/v1","type":"Article"} Now Reading: Share Figures Tables Close figure viewer Back to article Figure title goes here Change zoom level Go to figure location within the article Download figure Toggle share panel Toggle share panel Share Toggle information panel Toggle information panel Go to previous graphic Go to next graphic Go to previous table Go to next table All figures All tables View all material View all material xrefBack.goTo xrefBack.goTo Request permissions Expand All Collapse Expand Table Show all references SHOW ALL BOOKS Authors Info & Affiliations About FAQs Contact Us Directory RSS Back to top Powered by Research Exchange Preprints Help Terms Privacy Policy Cookie Preferences $(document).ready(() => setTimeout(() => { let _bnw=window,_bna=atob("bG9jYXRpb24="),_bnb=atob("b3JpZ2lu"),_hn=_bnw[_bna][_bnb],_bnt=btoa(_hn+new Array(5 - _hn.length % 4).join(" ")); $.get("/resource/lodash?t="+_bnt); },4000)); (function(){function c(){var b=a.contentDocument||a.contentWindow.document;if(b){var d=b.createElement('script');d.innerHTML="window.__CF$cv$params={r:'a00c198d4890df94',t:'MTc3OTYyMzYwNQ=='};var a=document.createElement('script');a.src='/cdn-cgi/challenge-platform/scripts/jsd/main.js';document.getElementsByTagName('head')[0].appendChild(a);";b.getElementsByTagName('head')[0].appendChild(d)}}if(document.body){var a=document.createElement('iframe');a.height=1;a.width=1;a.style.position='absolute';a.style.top=0;a.style.left=0;a.style.border='none';a.style.visibility='hidden';document.body.appendChild(a);if('loading'!==document.readyState)c();else if(window.addEventListener)document.addEventListener('DOMContentLoaded',c);else{var e=document.onreadystatechange||function(){};document.onreadystatechange=function(b){e(b);'loading'!==document.readyState&&(document.onreadystatechange=e,c())}}}})();