Effects of forest degradation on Amazonian ferns in a land‐bridge island system as revealed by non‐specialist inventories

Tropical forests have been rapidly deforested and degradation worldwide has outpaced biodiversity field sampling. No study to date assessed the effects of insular habitats induced by hydroelectric dams on Amazonian understory plants. Fern community responses anthropogenic tropical forest islands can be revealed at a faster pace using simple cheap, yet informative, protocols that could applied non-specialists. This seeks both understand drivers fern lycophytes assemblages investigate relative costs effectiveness simplified sampling protocol species were sampled non-specialist who photographed all ferns within seventeen 0.25-ha plots 10 lake Balbina Hydroelectric dam, central Amazonia. Sampling was carried out opportunistically during expedition planned conduct tree inventories. As predictors, we used locally measured or GIS-derived descriptors plot landscape conditions. We multivariate linear models further assess influence predictors patterns richness composition assemblages. A total 286 individual represented least 23 14 genera. The average number taxa per 6.1 in 14.3 mainland. pool found nested subset mainland community. Species positively related island size negatively isolation fire severity. Area, severity also significantly explained variation composition. cost picture-based very modest (only 4% budget), even compared typically low alternative campaigns. conclude structure this archipelago primarily driven size, disturbance. Moreover, show lead inexpensive highly reliable ecological data. opens an avenue for crowdsourcing data collections citizen science approach. florestas tropicais estão sendo desmatadas rapidamente e degradação em todo o mundo ocorre de forma mais rápida que coletas dados campo sobre biodiversidade. Por exemplo, nenhum estudo até momento avaliou os efeitos insulares induzidos por barragens hidrelétricas plantas sub-bosque da Amazônia. respostas comunidade samambaias aos antrópicos ilhas floresta podem ser reveladas um ritmo rápido usando protocolos simples, baratos, ainda assim informativos. Tais aplicados não especialistas acelerando, portanto, nosso entendimento como diversidade espécies é moldado paisagens altamente fragmentadas. Este busca compreender fatores determinantes variações nas assembléias licófitas florestais investigar custos relativos vs. eficácia protocolo amostragem simplificado pode aplicado e, tem potencial para usado estratégias coleta ecológicos coletiva (crowdsourcing). foram amostradas especialista fotografou todas as 17 parcelas 0,25 ha no lago Barragem Hidrelétrica Balbina, na Amazônia central. foi realizada oportunista, durante uma expedição planejada realizar inventário árvores permanentes. Para cada das amostradas, usamos conjunto variáveis preditoras medidas localmente ou derivados SIG escalas parcela, do fragmento florestal paisagem. Usamos Análise Coordenadas Principais (PCoA) Modelos Lineares Mistos Generalizados (GLMMs) avaliar dos padrões variação composição riqueza samambaias. Um indivíduos fotografados, representanto pelo menos gêneros. Em media, observadas 6,1 14,3 continente. encontradas representaram subconjunto maior quanto próxima continente era ilha. diminuiu severidade fogo ilhas. Área, isolamento também explicaram significativamente comunidade. O custo relativo baseado imagens modesto (em caso, apenas orçamento expedição), mesmo comparação ao tipicamente baixo muitas campanhas campo. Concluimos estrutura neste arquipélago impulsionada principalmente tamanho ilha, incêndio. Além disso, demonstramos realizado levar baratos Isso abre caminho ecológicas abordagem ciência-cidadã. degraded worldwide, with more than one-quarter global loss agricultural activities, mining energy infrastructure (Curtis et al., 2018). Specifically Amazon, world's largest rainforest, unprecedented expansion hydropower implemented last decades, resulting wholesale fragmentation pristine (Lees 2016; Winemiller 2016). Large projects established lowland submerge vast areas, creating archipelagos inducing rapid erosion (Benchimol & Peres, 2015b). Consequently, understanding how diversity is shaped fragmented hydro-reservoirs utmost importance identify mechanisms governing persistence newly isolated biotas propose sound conservation strategies. Since publication MarArthur Wilson's book 1967 revealing revolutionary Theory Island Biogeography (TBI), area metrics became widely ecologists evaluate their fragments real world (i.e., embedded human-modified landscapes [Laurance, 2008]). predicted TBI, greater shorter its nearby mainland, higher expected occur, given differential probabilities extinction colonization. Although several limitations TBI shown, most studies conducted reservoirs powerful and/or wide range biological groups 2021; Storck-Tonon 2017; Yu 2012). Yet, habitat quality good therefore scrutinize landscapes. growing plants reservoirs, mostly focusing trees lianas 2015a; Jones 2017, 2019; Patch-scale attributes (e.g. edge-mediated disturbance) reported main factors affecting structure. However, neotropical understorey Ferns (collectively known pteridophytes) comprise diverse, conspicuous distributed group are well-recognized indicator (Salovaara 2004; Zuquim 2014). In undisturbed forests, soils determinant communities scales (Tuomisto 2003; Zuquim, Costa, 2009) tends increase soil fertility disturbances, including deforestation, fires alter environmental conditions might exert significant islands. disturbance dynamics potential source propagules play major role, perhaps overriding original baseline forests. increasing across Amazon basin, plant inventories still scant geographically biased (Feeley Silman, 2011; Hopkins, 2007; Schulman 2007). Poor availability areas types inhibits unravelling processes, threats. Conservation planners may never sufficient resources required collect enough properly address practice science, especially considering sheer research agenda megadiverse region should include creative solutions efficient acquisition. solution mitigate gap problem outsource collection currently slow accumulation records boost (Dickinson 2010). addition, involving non-academic members, it intrinsic value raising awareness contributing capacity-building education (Schläppy 2017). For charismatic animal groups, division labour between amateur observers experts already produced enormous volume some groups. example, obtained scientists represent 70% 87% mammal bird records, respectively, Global Biodiversity Information Facility (GBIF) (Chandler consequence campaigns occurred around resulted important societal contributions (ElQadi Gonella 2015; Laaksonen Winterton countries, rare, infancy. case systems, ecologically informative collected non-specialists if available. Given straightforward detectability ferns, interesting model promoted aid abundant usually clearly recognize when they see one. Most easily identified from photographs (Suominen 2015) either layman help identification tools (Zuquim 2008, Equally important, excellent properties other floristic (Ruokolainen Sirén 2013; 2019). Therefore, georeferenced photograph valuable provide useful information about prevailing site. Here, first quantitative assessment lycophyte pattern caused mega dam innovative approach done non-biologist patch, characteristics these particular, (i) area, isolation, cover composition, (ii) discuss benefits adopting inventory poorly sites. is, therefore, two-folded intends impacts man-made archipelagic focal demonstrate relevant findings. evidence guidelines opportunistic acquisition optimized target ranging investigated sites well-established vegetation monitoring programs improve our can, turn, contribute enhancing knowledge strategies (hereafter, ferns) Brazilian 4437-km2 reservoir created following impoundment Uatumã river Dam 1987. retains over 3500 variable-sized primary sub-montane dense forest, although large patches successional second growth occur many result post-damming old-growth mortality 2015a). surveys islands, which spaced 2 km apart selected basis degree Landsat ETM+ scenes (230/061 231/061; details Benchimol Fieldwork originally conducting re-census 44 permanent plots. subsequently decided apply but additional financial available such All located adjacent ∼940,000 Biological Reserve, reserve category Brazil. protected against hunting logging, had succumbed ephemeral severe El Niño drought 1997, affected small-sized marked changes mean annual rainfall temperature 2376 mm 28°C, respectively. September October 2018 land-bridge (Figure 1). varied 15 1450 ha. On each island, one four established, depending size. 250 m × m, except those small 125 20 m. 2012 adult (see covering whole Plot dimensions comparable previous continuous surrounding 2012; red square Figure Plots minimum distance 50 nearest margin minimize edge effects, 300 any same plot, quick entirely (RT). person experience fieldwork nor academic training sciences photography skills. He earmarked participate organize food supplies, transport accommodation. Inside individuals recognized taken To become familiar flora, photographic guide 2008) consulted prior onset fieldwork. photographs, morphotypes Canon EOS 600D digital camera coupled common 50-mm lens. specimens recorded then compiled into plot-specific folders digitally forwarded specialist (GZ) level. maximize logistical efficiency, time team plots, time-efficient Permits work Reserva Biológica granted Chico Mendes Institute (ICMBio), took advantage pre-existing database consisted set spatial describing survey site patch 2015a description variables). From full Project (Table S1), pre-selected variables correlated S1). Whenever two (>0.70), only included predictor, according biogeography ecology. follows: (1) hectares (log10(x + 1); AREA); (2) Isolation shortest (Isolation); (3) Percentage perimeter (COVER), estimated multiple-sized buffers (500 1000 buffers). These sizes chosen because Peres 2015a, b) enable non-overlapping landscapes, indicated patch-landscape design; (4) pioneer stems (PPIONEER) among inventoried ≥10 cm DBH (diameter breast height); (5) metric (FIRE), composite ordinal score (0–3) burn (based charred height char marks bole tree) extent three independent researchers surveys. fell 0 (no fire). Only classified 2. statistical inference power, posteriori lump categories 1 (and considered intensity) 3 (high intensity). AREA, Isolation, COVER quantified ArcView 10.1 (ESRI, 2011) Fragstats (McGarigal 2012), whereas PPIONEER FIRE based details). addition database, extracted TM/ETM+ reflectance values imagery 30 scene covers (https://etsin.fairdata.fi/dataset/1b32feb8-e297-4113-b91f-c58fff275039/data; file 60W_2.5S_2.5.tif). entire basin satellite images period 2000–2009 (Van doninck Tuomisto, obtain single band median 60-m buffer centroid plot. separately (Bands 2, 3, 4, 5 7). summarize spectral Principal Component Analysis (PCA) axis predictor (LANDSAT PCA1). multidimensionality reduced Coordinates ordination presence/absence Jaccard similarity index build matrix. PCoA descriptor respond predictors. Considering effect depends scale (i.e. so-called ‘scale effect’, Fahrig, 2013), firstly performed Generalized Linear Models radius 500 m) appropriate showing highest correlation subsequent analyses. chosen, yielded slightly (r = 0.49 r 0.37, respectively). Following leading premises theory (MacArthur Wilson, 1967), examined AREA regression. Mixed (GLMMs). constructed five (AREA, COVER, LANDSAT PC1) possible combinations subsets five. Poisson Gaussian distribution link functions defined random account variation. controlled high levels multicollinearity performing Variation Inflation Factors (VIF) (Dormann 2013) ‘HH’ package (Heiberger, 2020) excluded moderately redundant collinear (VIF < 5). ranked Akaike Criterion (AICc), MuMIn R (Barton, 2018) (ΔAICc 2.0) multimodel AICc (Burnham 2011). Explanatory rescaled modelling, function R. thus model-averaging 2018), combines candidate averaged parameter estimates. test different wildfires, run PERMANOVA (‘adonis’ vegan [Okasanen 2020]) response variables. order surveyed associated calculated modified IndVal (De Cáceres 2010) analysis ‘multipatt’ indicspecies Legendre, included. analyses 3.3.1 environment (R Core Team, https://doi.org/10.5061/dryad.wstqjq2nq (Storck-Tonon 2021). Finally, list regional flora this, published (‘Uatumã’ sites; 2012) 2007 29 quarter-hectare Research Program (PPBio [ppbio.inpa.gov.br]). (GZ), vouchers them INPA herbarium. farthest respectively observed accounting differences (17 mainland). so, (out 29) randomly procedure iteratively times. It note environments same. pictures voucher herbaria, resulting, few cases, resolution. Triplophyllum treated taxon genus level, congeners Lomariopsis prieuriana japurensis lumped together. consistency, harmonized difficult distinguish alone complexes sets. aimed resample reservoir. formed researcher, parataxonomist, boatman assistant RT not busy logistics, he able being out. Thus, practice, working hours allocated his multitasking role expedition. estimate relation expedition, overall person, summing daily allowance, local transportation (boat driver fuel outboard motor) divided participants. Daily allowance market wages 8 h travel time. specifically inventories, multiplied hourly wage protocol. campaign person. Expenses paid Reais (R$), here converted monetary U.S. Dollars (US$) commercial exchange rate (September Central Bank Brazil (https://www.bcb.gov.br/conversao). sets lycophytes, belonging genera S2). ubiquitous Selaginella pedata, (88.2%), Trichomanes pinnatum, Adiantum paraense/tuomistoanum (82.3%) sp. Lindsaea lancea, nine (52.9%). 18 fewer 50% Near half (52.2%) (PCoA1) 48% PCoA1 scores GLMM variable summarizes plot-scale Fire 2), best percentage pioneers PCA1) ‘best’ (as expressed PCoA1) (R2adj 0.45, p 0.001) 0.29, 0.001). When GLMMs models, strongest estimates 3a). intensity ΔAICc 2.0 summary land uninformative explaining richness. There positive semi-log relationship 0.46, 0.001; 3b), negative 0.43, 3c). Community along 4) versus (PERMANOVA: F 4.614; R2 0.21; 0.01). history wildfires sorted gradient Ten exclusively unburnt subjected 13 heavily burnt. Of those, likely indicates rather preference disturbed showed