Riparian reserves protect butterfly communities in selectively logged tropical forest

Over 400 million hectares of tropical forest have been designated for commercial timber production, an area twice the size Mexico (Blaser et al., 2011), highlighting need to reconcile industry with biodiversity conservation. Selectively logged can retain many species, including a host threatened or endemic species (Fimbel 2001). Nevertheless, extraction valuable tree leads disruption canopy, hotter microclimates in short term, soil compaction and secondary impacts through road networks skid trails, resulting reduction interior specialists increase edge-tolerant generalists (Edwards, Tobias, 2014). Harvesting intensity practices determine severity these environmental changes thus on (Bicknell 2014; Burivalova Understanding which logging reduce degradation maximise retention is key conservation while meeting global demand timber. Riparian zones around streams are often only that legally required be left unlogged within concessions. For instance, Brazilian Forest Code requires protection intact vegetation as ‘areas permanent preservation’ ephemeral streams, springs hilltops protect hydrological ecosystem services, both concessions agricultural lands (Biggs 2019). In Brazil, strip 30 m riparian should protected sides small (<10 wide) 50 (Brasil, 2012; Zimbres 2018), mainland Malaysia Sabah (Malaysian Borneo) buffer 5 20 m, respectively, by law, although regulations poorly enforced (Chappell & Thang, 2007). Research into value has traditionally focused water quality availability landscapes (Luke 2017), revealing even narrow buffers (c. 5–10 m) help regulate hydrology, overall likely affects their benefits landscapes, particularly oil palm plantations, studies shown potential strips safeguarding reducing fragmentation via increased connectivity (Paolino 2018) wider (typically those >80 microclimatic refugia most (Williamson 2020). Species assemblages reserves tend resemble primary more than adjacent plantations cattle ranches, birds (Hawes 2008; Keir 2015; Mitchell some insects (Brito 2017; Gray 2016). Whereas typically present highly dissimilar matrix reserves, little research investigated embedded concessions, but points important contributions selectively subtropical warm-temperate forests Australia, maintained bat activity similar mature (Lloyd 2006). Within intensively managed United States, 30–50 old-growth along support communities birds, amphibians reptiles associated (Guzy 2019), webs grassland South Africa abundance richness dragonflies (Kietzka 2021). However, insect does not always correlate well level disturbance (Bonebrake 2010). loss vertical stratification lead inflated numbers detected understorey ground-based sampling, dung beetles (Davis Sutton, 1998) butterflies (Willot, 2004). contrast, studying community assemblage phylogenetic between habitats point towards ways protecting whole identify clades vulnerable habitat (Burivalova Cardoso Thus, this approach needed accurately assess Here, we effectiveness forest. We do so fruit-feeding butterfly (Family: Nymphalidae) concession south-western Amazonia (Rondônia, Brazil). sampled paired addressed three questions: (a) what forest; (b) counterparts; (c) any across types phylogenetically conserved. The study was set 46,184-hectare Jamari National Forest, Rondônia, Brazil. region's climate hot humid, relatively constant mean annual temperature (25.8°C) precipitation (2,200–2,600 mm/year). There well-defined dry season during Austral winter months from June September. split sections, southern-most 46,184 ha AMATA Brasil SA, company Stewardship Council certified since 2010. They use Reduced-Impact Logging (RIL) strategy, digital mapping all trees >40 cm trunk diameter planned (Mollinari 2019; Montejo-Kovacevich 2018). Timber our took place 2012 2015, yield 10.13 m3/ha (range = 0–35.2 m3/ha). As part RIL perennial were also digitally mapped prior logging. Following law (Zimbres 60 wide kept (30 each side), plus 50-m stream headwater source (grey areas, Figure 1a). 2016 (May–June) at 64 alongside eight six located recently areas two nearby (Figure Two transects four stream: transect 250 away parallel 1c). Points 125 apart other, sufficient ensure spatial independence (Montejo-Kovacevich 2018; Ribeiro Freitas, see Note S1 recapture data, Supplementary Information). may either ephemeral, is, above-ground channel rainy season, perennial, Annual Production Unit (APU) locations, one (the width channels measured 12 varying 2.06 3.38 m; 2.80 m). Each consisted baited traps ~5–10 hung nearest branch available 1 above ground. (Nymphalidae) using Van SomerenRydon cylindrical standardised mixture mashed bananas sugarcane juice had fermenting closed containers sun hr. Traps operated 10 full days cycles 16 (two interiors) time, total 1,920 trap-days. visited every 48 h 06:00 15:00 hr record captured replace bait fresh mixture. photographed, marked released individual same they captured. identifications made photographic collection experienced lepidopterist G.M.K., literature online reference collections. This sampling did require ethical approvals. point, variables relating surrounding environment then summarised values per point. Stream used analyses. number breast height (DBH) ?10 radius trap summed estimated maximum canopy directly trap, openness hemispherical photograph Sigma 4.5 mm circular fisheye lens, following Mollinari al. (2019)), averaged estimates All analyses run R V3.6.1 (R Development Core Team, 2019) graphics generated package ggplot2 (Ginestet, 2011). Packages specified below data scripts carrying out producing results figures publicly open-access repository 2022). compared (i.e. forest, reserve logged) individual-based rarefaction extrapolation curves, iNEXT (Hsieh refer type combination total), aim because find large differences two. Extrapolation obtained doubling initial sample size, (n 1,166). calculated 95% confidence intervals (CI 95%) bootstrapping Changes composition examined level, collated abundances transect, separately 1, n 16). representative avoid autocorrelation issues ordination computed Whittaker beta-diversity index function betadiver() vegan (Oksanen 2013), performed Adonis test (primary forest), Principal Coordinate Analysis (PCoA) visualise clustering habitats. evaluated drivers dissimilarity calculating Bray–Curtis obtaining (Banks-Leite 2014), implemented vegdist() 2013). fitted separate generalised linear mixed models response ‘site’ random effect 1). models, explanatory zero unit variance improve model convergence included cover interacting diameter. Additionally, model, included. These glmmTMB binomial family logit link adjust proportion overdispersion (Brooks 2017). To examine phylogeny, first constructed phylogeny based published sequences genes Nymphalidae genera sample, nine Riodinidae outgroup (accessions Table S1) excluding due lack data. Details reconstruction methods found Supporting Information (Note S2, added polytomies genus-level analysis visualisation, leading 116 species. percentage conspecific individuals (unlogged primary, reserve, forest) least once absolute change (percentage points) % abundance. Abundances account habitat. plotted proportional ggtree (Yu tested signal Abouheif's Moran's I, adephylo (Jombart Dray, collected 2,751 representing 139 points. Point-level highest forests, vary 2a). was, average, 33% higher points, there no significant nor whether (‘riparian logged’) primary’, 2b). Since hereafter ‘primary forest’. recorded (96 24 points), followed (89 (71 include points). Individual-based curves reached asymptote habitats, extrapolations indicating 80% (predicted 120 111, respectively), 85% 84; S1, At significantly different 3a; ADONIS: R2 0.22, p 0.003). strong overlap second PCoA axes, respectively 3a). 3b, 0.65 ± 0.05) magnitude background non-adjacent ‘sites’; 1; 0.6), regardless empty circles) triangles; 0.67 0.03) (paired t-test: t[47] ?2.18, 0.03), distant closest (250 away, 1d). communities. natural distance-decay similarity further show habitat), pairs showed greater sites against 3b: orange lines, 0.05, 54) vice versa, where weaker grey ?0.02, 42). other words, generally were, when demonstrating buffering changes. predictor accounting location site effect; 1), explained 23% variation beta-regression (estimate ?0.25, SE 6.6 × 10?3, < 0.001; 3c). By none (Table S2; 3c), removal distort relationships. structure types, point-by-point major S2). Overall, 72) changed less (36 4a). Some opposite trend respect (29 4a), lower (mean ?14 +27). On difference 8.5% t[64] ?2.5, 0.015; 4c). degree (change reserves: Abouheif I 0.04, 0.3; forest: 0.0006, 0.5; genera, like abundant Morpho spp. majority Chloreuptychia spp., 4a,b, dark shades green). such Pyrrhogyra stratonicius P. edocla, exhibited stronger 4b). When assessing (Abouheif 0.15, 0.01), comparing 0.2). highlights clades, mandated unknown. remnant otherwise landscape diversity undisturbed particular, forests: away; communities; assemblages; (d) others. seven primary-forest completely absent role entire (Edwards larger tracts strips, beyond 60-m-wide laws narrower m. Canopy correlated compositional stressing importance emergent demonstrate Amazonian remain priority sustainable planning. Retention regulatory constraints, ‘set-asides’, case avoidance rocky, steep inaccessible loggers (Putz Together, patches could greatly diminish (Griscom enhance corridors wildlife (Lees Peres, 2008). uniquely virtually forest-affiliated supports findings sub-tropical 2006), broadly native (Mitchell pastures 2008) fast-growing monoculture Butterfly Increased generalist canopy-dwelling benefitting gaps trees, widely reported (e.g. 2012). That retained core comparable indicate playing reservoirs demonstrated vertebrate invertebrate taxa relative responses closely related connected high levels avian (Cardoso southwestern apparently effective lepidopteran taller canopies Logged recover few years after reduced-impact Senior allow well. Butterflies ‘spillover’ distances rainforests proximity dispersal (Lucey Hill, 2012), facilitate movement moths continuous potentially enabling population re-establishment regenerating 2021; preserving high-statured stands re-colonisation depend Long-term critical, especially access enabled roads trails. temperate clear-cut silviculture, selective negatively affected plant (Mäenpää 2020), (Flaspohler 2002), small-scale low-intensity Bolivia effects amphibian year (Fredericksen Fredericksen, set-asides safeguard fully 35 2015). caveats. First, taxonomic group stratum. Fruit-feeding biodiversity, taxonomically resolved, easy manner, tends Neotropics Sampling incursion differed type. thereby open-habitat areas. Second, (1–4 study) recovered fully. burgeoning Neotropical faunal assessed <3 (Azevedo-Ramos 2006; Bicknell França 2016; Third, concession, being fixed Future would forest-like features, Williamson 2020; value. Despite its limitations, represents implications biodiversity. Our clear forests. underscores continuing provide legal ‘Areas Permanent Preservation’ (APPs) landholding Ensuring unexploited maintaining services 2008), seasonally southern Amazonia. Furthermore, inclusion requirement harvest certification schemes, Council, promote Such growing hardwood globally Luke Advances remote-sensing technologies cost-effective accurate way monitor broad regions, which, if correctly, enforcement (Hethcoat enhanced increasing widths, 30-m <10 insufficient mammals fragmented dominated conversion compete setting aside blocks neighbouring land-sparing Gilroy, Preserving might necessary sensitive (Betts Edwards, Borneo, without yields appropriate planning exploit sensitive, inaccessible, A concern tropics cycle, under legislation >50 extracted (Sist Ferreira, act corridors, seen (Gillies Clair, 2017) (Gray turn, accelerate recovery Taken together, careful arthropods. thank providing Márcia Michele de Souza Perdigão others invaluable field assistance. G.M-K funded University Sheffield Masters studentship, Explorers Club Youth Fund British Entomological Natural History Society (Professor Hering Memorial Fund). None. G.M.-K., D.P.E., C.J.M. C.A.P. conceived ideas designed methodology; C.J.M., G.M.-K. S.H.S. data; D.P.E. analysed led writing manuscript. authors contributed critically drafts gave final approval publication. Data Zenodo Repository https://doi.org/10.5281/zenodo.6349414 Please note: publisher responsible content functionality supporting information supplied authors. Any queries (other missing content) directed corresponding author article.