Aridity could have driven the local extinction of a common and multivoltine butterfly

Biodiversity is currently undergoing a human-induced extinction crisis comparable to previous mass extinctions (Ceballos et al., 2015; Dirzo 2014; Pievani, 2014). Among the main drivers of loss, climate change has already impacted biodiversity from individual genotypes whole ecosystems (IPBES, 2019; Scheffers 2016). This resulted in species extinctions, local extirpations and changes distribution, phenology morphology (Bellard 2012; Parmesan & Yohe, 2003; Thomas 2004). Arthropods constitute half current animal biomass (Bar-On 2018), which insects are most diverse cosmopolitan group, with more than 1 million an estimated 80% yet be described (Chapman, 2009). Insects play essential role maintaining trophic networks (Losey Vaughan, 2006; Vanbergen Insect Pollinators Initiative, 2013), also considered good bioindicators due their ectothermy, short generation times because some show narrow resource dependency (McGeoch, 1998; Markert Markert, 2007). insect bioindicators, butterflies stand out as best known studied group Thomas, 2005; Hill 2021), monitored many places around world following BMS framework (Butterfly Monitoring System; Pollard Yates, 1994), multiple studies reporting responses effects (Hill 2021; 1999; Rödder 2021). The genus Coenonympha Hubner, 1819 (Nymphalidae, Satyrinae) constitutes relatively Holarctic 30 Palearctic 2 Nearctic (Bozano, 2002; García-Barros 2013; but see Wiemers 2018, 2020 who propose total 20 18 Europe). In Iberian Peninsula, there 4 species: dorus (Esper, 1782), arcania (Linnaeus, 1760), glycerion (Borkhaunsen, 1788), pamphilus 1758) (Figure Figure S1). All four classified Least Concern by International Union for Conservation Nature (IUCN) Europe, like Mediterranean (Numa study focused on C. pamphilus, Small Heath 1). widely distributed region (Boillat, Meneguz, 2012). Previously published distribution maps indicated that was Peninsula Balearic Islands (García-Barros 2004; Romo 2003), it common (Cancela 2020; Vila 2018). Initially, proposed areas no records were lack information (Romo 2003). However, 2004 suggested become spatially scarcer influence, Montagud García-Alamá (2010) indicate populations Valencian Community, eastern Spain, very dispersed localised. generalist grassland species, multivoltine biology Southern having two (bivoltine) or three (trivoltine) generations per year depending area wide variety larval host plants Poaceae family. Bivoltine cycles mountainous areas, while trivoltine characteristic warmer areas. first annual occurs April May, second July August, third September October (Vila Northern high latitude altitude, univoltine. Its preferred habitat pasture meadows grass such Festuca rubra L., Poa annua pratensis Anthoxanthum odoratum Brachypodium phoenicoides (L.) Roem. Schult., Cynosurus cristatus L. Dactylis glomerata serve nymphalid exhibits territorial behaviour, forming leks Wickman Rutowski, 1999). Technological advances, digital photography, Internet social media have greatly facilitated dissemination georeferenced occurrence data (Newman 2011; Nov Data derived collaboration between citizen academic science, hosted free databases, GBIF (Global Information Facility; www.gbif.org), helped advance our knowledge key distributions (Goula Jiménez-Valverde Sardón-Gutiérrez Shirey 2019), invasive (Comont Ashbrook, 2017; Sesma Gil-Tapetado, 2020), identification threatened species. objective this analyse patterns using available model climatically suitable potentially habitable identify climatic variables processes could limiting its distribution. addition, we evaluate importance obtained science globally relevance monitoring last decade (2010–2020) example butterfly science. We downloaded all stored Global Facility (GBIF) complete coordinates, type origin (GBIF.org, GBIF, 2021a; 669,852 records). further filtered these data, eliminating duplicates georeferencing errors, retaining 88,557 global records. areas: (restricted Western Palearctic) pamphilus; different periods: 1665 2010 both 2009 Peninsula. analysed source number 4019 2020. homogenised one presence 10 × km grid cell. cell (1963 records) Due special redundancy prior 1970 been subsequently re-observed (especially decade, 2b). represented historical (1665–2009) at resolution. Historical came al. (2004), (2003) (specific east Peninsula), alongside 2009. period representative platforms experienced substantial activity increase Spain during period. cells into those To estimate sampling effort each province Papilionoidea Biodiversidad Virtual platform—the biggest platform—between 2018 (116,525 records), GBIF. calculated records, richness, inventory completeness, accumulation curve slopes province. rational cut-off value 0.1 slope curve. These analyses performed package KnowBr (Lobo 2018) software RWizard 4.3 (Guisande 2014) R v 3.5.0 (R Development Core Team, 2015) programming environment. ensure biases, compared dorus. similar appearance S1), differentiated curved, proximally convex row ocelli underside hindwing wing 2013). Unlike only year, May altitude Both share genera (Festuca, Cynosurus, Poa, Brachypodium; García-Alamá, 2010), thus making ideal compare pamphilus. 2021b; 1044 them (652 cells). map. determine if plant driving Anthos platform (http://www.anthos.es) past bioclimatic environmental (10 arcminutes) (30 arcseconds) resolution, respectively. Our consisted 19 WorldClim 2.1 layers (see Table 1), (Fick Hijmans, 2017), aridity index (aI, Trabucco Zomer, 2018; low values aridity/low humidity vice versa). previously used Settele (2008), addressed modelling European including case variables, usual correlation among high, so useful models should not different. summarised mean, standard deviation, maximum minimum variable across time series (1665–2020) (2010–2020). obtain simple coverage model, aI (Gil-Tapetado Lobo 2010). had within (i.e., habitable) Areas least outside range unsuitable uninhabitable), restricting allowed limit influence developed (SDMs; Elith Leathwick, 2009), presences avoid correlation, selected subset methodology employed Polidori (2021) (2021). hierarchical cluster analysis, resulting dissimilarity dendrogram (Dormann analysis segregates clusters, considering distance them. Ward's clustering method, based Pearson's matrix (Harrell, 2015). established 0.3 threshold selection less 70% correlation). below threshold, prioritising: (1) restrict model; (2) important (following Schweiger (3) prioritising seasonal over variables). Variables unrealistic area, mean temperature wettest quarter (bio8) driest (bio9) (Fajardo 2014), discarded. Finally, variance inflation factor (VIF; Lin 2011) discarding overestimated contributed redundant (VIF >5; Miles, 2005). ensemble SDMs six algorithms: generalised linear (GLM; McCullagh Nelder, 1989), additive (GAM; Hastie Tibshirani, 1990), artificial neural network (ANN; Ripley, 1996), classification tree (CTA; Breiman 1984), entropy (MaxEnt; 2006) random forest (RF; Breiman, 2001). Background pseudo-absences generated points randomly (suitable areas), pseudo-absence (unsuitable uninhabitable areas) External evaluation dividing training dataset (75%) testing (25%) calculating under (AUC) Receiver Operating Characteristic (ROC) (Jiménez-Valverde, Each evaluated same procedure, subdivision 85% (training) 15% (testing). ran 600 (100 replicates algorithms) kept internal AUC >0.7, excellent performance (Thuiller, fulfilled condition calculate average consensus upper lower bounds models. suitability (=0.70) model. Therefore, higher 0.70 suitable, unsuitable. ANOVA test whether significantly suitability. Variable selection, VIF statistical environment 2015), RStudio 1.1.453 (RStudio biomod2 (Thuiller 2019). point done ArcGIS 10.4.1 (ESRI, throughout Europe 2), clustered central southern Scandinavia, Great Britain, northern Italy western There sparse Eastern reaching Russia Kyrgyzstan, south Italy, north Africa, (Valencian Community). Most schemes belonging (e.g., United Kingdom provided 44.57% 298,565, www.ukbms.org) platforms, Observation (www.observation.org) 2021a). pattern when 92% (82,044) come 3% (2675) scientific surveys preserved specimens 3a). Regarding 3b), oldest date 1665, increasing slowly 20th century exponentially 21st (dashed black line 3b). periods 2014 2015 roughly (44,278, marked 3b dashed green line). (around 2806), specifically (35.38%, 1422 Observation.org (26.5%, 1066 3c). started 3b, blue dotted line), trend data. showed notable decrease globally, probably combined COVID-19 pandemic lag upload sources. absent Spanish south-eastern Autonomous Community Murcia Almeria since 2010. latest latter 2008, localities Sinarcas (39.7322, −1.231) El Regajo (38.8815, −0.9153) 4a). easternmost 2002 locality Elche (38.2692, −0.6961) Many plants, P. annua, pratensis, B. D. can found and, specifically, Community. cristatus, F. A. missing area. None aside glomerata, present Almeria. S2) territories. 2000 3000 territories: (southeastern Spain). observed richness highest sub-plateau. completeness southeast (>85%), (<0.02), indicating adequate complementary sharing territories uniquely limits 4b). indicates effort, where absent. summary (Table 1) 2) priori little restrictions. (aI) diurnal (bio2) 7.57% 6.45% Palearctic, 2, southernmost especially S3 shows (including Kingdom, Iceland, islands), except mentioned Nevertheless, must acknowledge seem unreliable, temperatures 0°C. final set included isothermality (bio3), (bio7), coldest (bio11), precipitation (bio17), (aI). (bio16) they collinear rest S4). index, fluctuations absence atmosphere. SDM validation 0.957 sensitivity 88.8% specificity 90.57%. external validation, remaining 25% 1, optimal S5). peninsula highly (>0.70 value), southeastern inland 5). 6 algorithms 3). comparing suitability, obtaining significant differences (p < 0.00001), versa 6). results rare, scarce, (Figures 4a 5), (2010). contrasts categorization abundant transects distributed, without threats S3). Previous affirm does negatively affected anthropic stressors (Montagud will substantially (Settele 2008). did find any Even unpublished inaccessible existed, would still rare publicly 3d). Murcia, Almeria, S2). absences true efforts (2003). It possible might impact Toledo (central 5 6b). ‘previous’ before 2010, dating 2002, years passed record Although need confirm gone locally extinct, highlights possibility depth (Mackenzie confirmation findings, however, transect (Alicante, Community) 2019 situation may scarce. Zerynthia Association (www.asociacion-zerynthia.org) reports País Vasco available. Catalan (https://www.catalanbms.org/), organisation, either. assumed (contributed users platforms) differ notably, consider Southeastern devoted nature 3a, c). undoubtedly provides community immense biological ecological processes, (Abolafya Champion 2020). use Big risk-free accumulate biases errors (Beck 2014)—as 0°C (Tables 2)—platforms, tools biogeography (Robertson photographs even though identified visually, be. For example, assume 10% visually photographs, percentage amount 10,000 identifiable If 1%, Applying 1% Hexapoda 360 ~36,000 (Ortuño Martínez-Pérez, identifiable. Considering is, fact, easily accessible tool biodiversity. recorded tend traits bioindicator conspicuous, recognisable, (Ribera Foster, 1997). increased 3B 3D) By combining standardised programmes, distributions, facilitating endangered. Here, applied scheme showing disappearing severe change. 5) related 3 Climate caused decades (Andrade Corte-Real, 2016; Paniagua 2019) driven disappearance contrary 2008. causes explain increased. Extreme made territory development (Brook 2008; Koh Los Monegros desert (centroid: 41.7752, −0.3458) Zaragoza arid Another potential cause direct negative effect larvae, vulnerable stage basin meridional long-term multiproxy demonstrated unprecedented heat waves drought impacts several (Briffa 2009; Della-Marta 2007; Luterbacher 2004), mainly larvae. hypotheses supported other S6) dorus, displaced although Catalonia parts peninsula, narrower diet breadth Community), certain hypothesised subspecies lyllus 2013) preference (distributed Europe), support suggestion. entire S6), southeast. As geographic barriers isolate interior unlikely individuals dietary preferences located west. interior, A lies species' bivoltinism, leading evolutionary trap (Schlaepfer Van Dyck bivoltine Catalonia, BMS, 2022 temperatures, Forrest, 2016 phenologies general Kingsolver, 2014, closer specific cases, Pieris rapae Lasiommata megera 1767), respectively). Apparently, territory, faster life cycle enters ‘obligate’ ‘forced’ bivoltinism. Together rising causing wither summer, nymphalid. Consequently, develop resources August facultatively, sedentary behaviour (Öckinger 2006), hindering recolonization displacing towards west suggest favouring dispersal (Cormont 2011). Ultimately, likely optimised speed (Roy 2001) resources. break synchronicity larvae generations, prematurely. Probably, univoltine, adult hibernate until next avoiding withering end season corroborate findings Melero (2016), linked multivoltinism steeper declines butterflies. Warmer univoltine add extra 1980 (Altermatt, While increment aid population growth, always positive effect, experience extreme droughts adverse conditions MacGregor escape (as Europe) accelerate cycle, taking advantage favourable spring food availability caterpillars, explored indirect Increasing events associated disruptive webs (Carnicer Whether large-scale web disruptions produced risk open question warrants research. sense, recommend conducting samplings establishing itineraries extirpation. Aridity frequency severity Carnicer contract centre changes. results, threat category revised Spain. what happening endangered complex conservation problems face future regarding Diego G. T. José Manuel S. M. conceived designed work; materials, experiments, data; T., Carmen S., Jose G., Francisco J. materials/analysis tools; wrote paper. thank ‘Asociación española de Entomología’(Spanish Entomology Association) awarding draft manuscript ‘VI Premios Nacionales Entomología para Jóvenes Investigadores’ (VI National Awards Young Researchers) granted DGT. observers scientists given time, article possible, friend Daniel Padilla. authors declare conflicts interest. openly FigShare https://doi.org/10.6084/m9.figshare.17290052.v1, reference 17290052.v1. S1. Photographs (left) 1782) (right). Moranas Fernando Camuñas. S2. Number estimation (Slope) (Lepidoptera, Papilionoidea) (www.biodiversidadvirtual.org). S3. (a) www.gbif.org) (b) S4. Dissimilarity red marks (0.3). boxed blue. after (variance factor; <5), discarded red. names Tables 2. S5. 5. S6. Occurrence nutrient (http://www.anthos.es). Please note: publisher responsible content functionality supporting supplied authors. Any queries (other content) directed corresponding author article.