Detecting early‐warning signals of concern in plant populations with a Citizen Science network. Are threatened and other priority species for conservation performing worse?

We are facing unprecedented rates of biodiversity loss and extinction at planetary scale (IPBES, 2019). Species’ begins with the local populations that result from habitat destruction or demographic decline, often promoted by stochastic deterministic factors (e.g. genetic drift land-use changes). However, extinctions rare events pass unnoticed for most species, which hinders our ability to adopt timely management strategies. Small population sizes, negative highly fluctuating trends precede extinctions, therefore constitute early-warning signals collapse impairment. For reason, these variables considered important criteria identifying species concern in IUCN Red list process (IUCN Standards Petitions Committee, This type information can only be gathered through large-scale, long-term monitoring programmes (Pereira & Cooper, 2006), is not available populations. Citizen Science (CS) play an invaluable role obtaining on conservation status species. The contribution volunteers has greatly increased last decades now a powerful tool addressing major challenges, providing up 70% records (Chandler et al., 2017). Volunteers also key maintaining large longstanding (McKinley 2017) used estimate global indices trends, such as ‘Living planet index’. index combines data repeated censuses thousands vertebrate globally (WWF, 2020) repeatedly indicated alarming declines. Empirical evidence however, equally different taxonomic groups because largely biased, 83% CS projects focusing animals (Buckland Johnston, 2017; Burns 2018; Chandler Meanwhile, other abundant like plants systematically underrepresented. greater difficulty taxonomical identification plant (of there many more than vertebrates) their lack movement might have made them less attractive programmes. Plants, structural elements habitats overall ecosystem services, making particularly suitable indicators 2006). In addition, part lists threatened, high-conservation value they must periodically assessed regional, national supranational level. Article 17 Habitats Directive, example, requests Member States European Union report every 6 years EU Community interest. present here working protocol participatory programme aimed characterizing status, assessing potential risk, wide variety plants. It was designed detect poor performance high vulnerability, based three sources information: (a) size (small face higher risk; Matthies 2004), (b) temporal changes abundance (decline rate environmental stochasticity intrinsically bound Menges, 1992) (c) observed threats (directly linked declines). developed ‘Adopt plant’ (AP; https://biodiversidadipe.csic.es/ciencia_ciudadana.html) which, best knowledge, unique terms careful scientific design used, detailed obtained non-professional people. Launched 2010, its objective track dynamics diversity, emphasis threatened ‘rare plants’ endemics. Following scientifically robust sampling designs, rangers monitor across range NE Iberian Peninsula (Aragón region), semi-desert valleys Mediterranean mountains alpine summits. Our aim threefold: provide guidelines participants set field designs gather structured heterogeneous group plants; demonstrate testing if size, variability priority lower compared non-priority assess threat frequency effects believe extensive ours help scientists validate ecological biology paradigms, improve planning managers policymakers. To this end, we discuss expand globally. Three interest were defined. first identified following international official lists. includes (critically endangered, vulnerable sensitive perturbation), well those listed Annexes II IV Directive (full Appendix S1 Supporting Information). Since Institution holds herbarium, distributions region. Common co-occurring ones monitored when possible, control possible deterioration affecting all study area irrespective contribute richness. A second target selected according singularity regional rarity. These included endemics Pyrenean small Peninsula, occurring distribution limit within area. All but concern. Finally, third indicator non-forested (Annex I) cannot remote sensing, endorheic lakes, bogs, cliffs, shrublands subalpine belt gypsum soil lowlands, diverse (sub)alpine grasslands. Because specificity, good climatic change very dependent humidity) under slow transformation). Specific sites (‘monitoring units’, hereafter MUs) expert particular some specific threat), terrain accessibility, cases, participants’ preferences protected area, proximity residence). MUs (N = 213 total), 2 4 km2, typically contain one two (range: 1–12 plants), distributed along altitudinal gradient (139–2,837 m a.s.l.) located Natura 2000 network (69%; Figure 1). Table 1 shows general rules select appropriate estimation methods (counts, presence cover) permanent units. Such units go macro-plots hundreds m2 micro-plots 20 × cm evenly transects (usually 10–50 meters long), 10 miniquadrats grids. cliff-dwelling use ‘virtual plots’ where individual counted, drawing fixed areas good-quality digital photos surveyed year eye aided binoculars (see bottom left picture uncountable forming population, gain occupancy comparing pictures time. count-based surveys medium (~74% populations, 1), add until minimum 300 counted. More 1,000 usually (~21% populations), cover (4%) replicated transects. minimize observation error, adjust density, include 30 per plot possible. surveys, number distance between adjusted total fall (20%–80%) presences so detected years. Sampling (number plots space, transects, etc.) conducted situ experienced scientist agreement team rangers. guided principles: being simple effective will able complete own just day, period robust, produce trustable gathered. Nevertheless, new established sometimes participants, who decide suggestions 1, assistance. Anyway, final method (increasing reducing effort) year. Each thus each MU units, etc., adjustment. examples provided García, Sanz, al. (2019). Data analysed look weakness bad performance: trends. Given differ primary hypotheses prone show them, suffer threats. vary interested comparative results, cleaned dataset prior analysis. removed 3 and/or extremely low sizes (<20 individuals), severely affected stochasticity. excluded counting flowering individuals (flowering does necessarily correlates whole population) annual (their require series, yet available). analysis fluctuations contained 242 150 taxa, 942 transitions (annual consecutive years; range: 3–10 MU; median: 4). association contingency table Chi-squared test S4 same using random samples size). whether decline fluctuate ones, fitted linear models PAch PAchsd response variables, separately. Residuals both showed non-constant spread variance, indicating violation homogeneity assumption (Appendix S5 Figures S2). By plotting residuals against explanatory variable models, found differences levels priority, (Figures As transformation sufficient overcome heterogeneity problems, generalized least squares (GLS) model residual variance structure (Zuur 2009). Along categorical (priority size) interaction, initial covariates: (counts/presence/plant cover; account effect abundance), over averaged (to robustness mean fluctuations, ‘learning’ became skilful longer series was) elevation fact tend stable elevations; Morris 2008). preliminary analyses, explored dependency among observations intercept mixed components. GLS negligible (intra-class correlations close 0). outlined Zuur (2009), started saturated without structure. then structures Akaike's criterion (AIC; S1). covariate selection, nested Maximum Likelihood (ML) log-likelihood ratio tests refitted Restricted (REML) reduce bias procedure (Figure Models package nlme (Pinheiro R v.3.6.3 (R Core Team, 2020). Pairwise comparisons done post-hoc Tukey correction multicomp (Hothorn given eight common plants, tested differed paired t-test after checking normality assumptions met. phylogenetically S6), hence no need controlling potentially non-independent due shared evolutionary history taxa (Felsenstein, 1985). Plant’ currently 332 204 (most represented population). They (MUs; 39% 83) frequent counts (74%), followed (Table ranges (total species) 5,500 (presence absence grid cells transects). Monitored mostly (large: 60%, medium: 27%, small: 14%). There is, significant difference (χ2 15.19, df 2, p < 0.001; similar see S4), occur frequently ones. Regarding analysis), values ranged −59.5% +45.5%, (82%) [−10%, +10%]. median positive stability (+0.12%; 2). Allowing variances reduced significantly original AIC interaction species' (likelihood test: L: 0.85, df, 0.655) term dropped model. yearly model, weak did fit tests: 1.36, 0.508, 2.09, 0.149). Model validation problems. Five detected, influence parameter inference interpretation S7). Coefficients indicate Large tended though versus comparison 3); (Post hoc estimate: 3.612, SE: 2.524, Z value: −1.431, 0.3153). Priority surprising PAch, homogeneous (and small-sized populations) 3). case statistically (t 1.48, 7, 0.1833). (log-transformed Pachsd) constant 2; case, smaller count Once survey accounted for, (p 0.001), Likewise, fluctuated Only 8% directly human activities processes, agriculture (4.2%) shifting (2.4%) main proportion fewer (5%) (10%). current scenario changes, double challenge evaluating accurately forecast fate component biodiversity. habitats, huge task administration afford. (AP) addresses involvement collection critical related implementing designs. Thanks compiling standardized way, allowing straightforward Together threats, estimates classically crucial (Bayraktarov dealing frequent, collect unstructured records, systematic presence/absence Martin 2019; Pescott 2015) Barnard AP collects instead, rigorous protocols ad combine responsible tracking MU, minimized. risk estimated (abundance trend fluctuations), ranges, baseline evaluation viability analyses (Morris 2002). been Pereira (2013) (Haase 2018) defined Essential Biodiversity Variables (EBV) prioritize, standardize facilitate comparisons. Hence, methodological proposal enabling us compile well-structured EBVs, challenges. Another feature disseminated regular basis meetings, public webpage project (https://biodiversidadipe.csic.es/ciencia_ciudadana.html). transferred know take actions necessary, avoid declines fulfil mandatory requirements. goal biodiversity-oriented record occurrences. approach becoming popular thanks increasing possibilities offered technological devices, software platforms exchange (smartphones, apps, GBIF). contributed development decade. short-term predictions real modelling future hypothetical scenarios (Ehrlén Morris, 2015). requires adapted fieldwork, mechanisms error. types scarce even studies endangered fact, matrix span short periods (<5 Salguero-Gómez probably effort required subsequent analysis, funding. Methods proposed people previous expertise, generate trajectories evaluate (Elderd 2003). proposal, nevertheless, needs investment, onset (choosing plant, setting personal training mentoring Silva, Long-term species-specific fieldwork ours, continue trustworthy way documenting organisms trajectory vulnerability study, shown how detecting associated essential management. Contrary expectations, results highlight widespread environmentally region, few finding line increase Alps (more decades; Rumpf 2018). remarkable novel neither nor casting doubt classical tenet worse former contrast, do support expectation perform 2004). always thought sized (the cases study). suggests estimating should step vulnerability. another large-scale sporadic volunteers, Lawson (2008) (<50 individuals) isolated performing authors, acknowledged substantial unexplained variation likely measurement source minimized scarcity location covered network. Whatever contrast overwhelming resulting animal Rosenberg Wiens, 2016), accepting conclusions derived vertebrates (Knapp, 2011). Although restricted single region relatively (3–10 years), covering broad biological conditions: 3,000 range, Eurosiberian biogeographical regions, life forms biannuals, perennial herbs, geophytes shrubs. summary, contrary surveillance monitoring, meets components generating long-term, high-quality datasets allow (sensu Bayraktarov Lindenmayer Likens, 2010; presumed demonstrated benefit moving away solely towards comprehensive purposes. Monitoring constitutes core adaptive process, necessary link objectives alternatives (Bakker Doak, 2009; Lahoz-Monfort 2014). requirements delivered approaches. encourage policymakers implement approaches, advocate expansion integration into larger platforms. larger-scale participative would share better challenges conservation. address question-driven pace considering situations central edge locations central–peripheral hypothesis), contrasted types, unprotected areas. Comprehensive involving citizen kinds assist protect while managing structurally communities integrated. work funded Regional Government Aragón, Project RESECOM (LIFE+12 NAT/ES/000180), OAPN (DYNBIO, grant 1656/2015), Research Spanish Agency (VULBIMON, CGL2017-90040-R) Diputación Provincial de Huesca. I.P. supported University Basque Country (ESPDOC18/43). acknowledge staff Aragón Ordesa Monte Perdido National Park, participation D Gómez, M Pizarro, H Miranda J Villellas assisted Lahoz B Valero revision English language. greatest gratitude wonderful army involved (names in: https://biodiversidadipe.csic.es/ciencia_ciudadana.html), whose enthusiasm keeps going despite M.B.G. conceived funding; I.P., J.L.S., P.T. MUs, coordinated, trained mentored validated data; J.L.S. optimized numerical graphical validation; together M.B.G.; wrote manuscript review inputs DIGITAL.CSIC Repository: Plant MONITO 2010–2019. http://dx.doi.org/10.20350/digitalCSIC/13825 (García, 2021). Please note: publisher content functionality any supporting supplied authors. Any queries (other missing content) directed corresponding author article.