Movement of forest?dependent dung beetles through riparian buffers in Bornean oil palm plantations

Threats to tropical forest biodiversity are unprecedented due the unparalleled rates of degradation, fragmentation and conversion agriculture (Barlow et al., 2016; Hansen 2013; Newbold 2014; Phalan 2013). Southeast Asia has some highest loss, with only ~19% its intact forests remaining (Achard Estoque 2019; Lewis 2015; Sodhi 2010). The rapid expansion oil palm (Elaeis guineensis Jacq.) made region world's primary source for vegetable (Fitzherbert 2008; Gaveau Turner 2008). Although loss in plantations is well documented (Meijaard 2020; Meijaard 2018), now an integral part economies Asian countries accounting 3.82% gross domestic product Malaysia alone (~$896 million per year; Mahidin, 2018). Therefore, it important understand how best manage design increasingly common mosaic landscapes that incorporate both natural agriculture, order support region. Remaining patches within often found form strips or restored vegetation by side waterways known as riparian buffers (also called reserves strips; Luke 2019). These riverine areas primarily set aside reduce run-off into streams (Sweeney 2004) but can also improve water quality (Mayer 2007) benefit aquatic forest-dependent terrestrial fauna (Gray 2014, 2019b; Marczak 2010; Ricketts, 2004). In addition, have potential serve movement corridors between fragments continuous (Beier & Noss, 1998; Tewksbury 2002). importance recently resulted their addition a requirement Roundtable on Sustainable Palm Oil (RSPO) certification (Barclay 2017; Lucey 2017). stipulates minimum buffer 5 >200 m each river, width depending river width, placement perceived use However, legal varies across countries; Sabah, 20 rivers >3 (Sabah Water Resources Enactment, 1998), be increased where thought represent wildlife (Environment Protection As become fragmented isolated, persistence species may critically dependent connectivity (Ewers Didham, 2006; Hanski, 1999; Hill, 2012). understanding animal move through landscape separating key consideration conservation management strategies human-modified (Doherty 2021; Gray 2019b). Movement ability influenced species’ behavioural responses habitat boundaries (Jain Kallioniemi 2012), physical costs (Bonte 2012) permeability matrix Scriven Furthermore, species-specific life-history traits impact (Ovaskainen Species most vulnerable taxa (i.e. those need viable populations). typically restricted ranges reluctant cross boundaries, resulting small, isolated populations which suffer local extinctions little prospect recolonisation (Scriven despite planning, relatively few studies investigated behaviour forest-associated (Bouchard Brooks, 2004; Brouwers Newton, 2009; Khazan, 2012; Here, we examine dung beetles (Scarabaeidae: Scarabaeinae) Malaysian Borneo. Tropical good indicators disturbance (e.g. Davis 2001; Gardner Nichols Gardner, 2011), there been several agricultural (Arellano Cultid-Medina da Silva Hernández, 2015). Forest-associated southern Mexico were highly deciduous landscapes, provided connect 2008); while Andean showed interspecific differences patterns distances, associated wing loading preferences (Cultid-Medina Within dominated one study limited ‘spillover’ palm, did not document measure directly 2016). We used mark–release–recapture (MRR) methods, technique insect 1997; Slade 2013) newly developed Joint Modelling approach (JSMM—Ovaskainen JSMM allows species- community-level parameters estimated simultaneously. extend framework account capture process model different land-use change scenarios. these methods ask influence landscapes. particular, test following hypotheses: (a) Dung more likely prefer moving than plantations, reserve buffers. (b) at will differ among species. (c) Conversion limit reducing number individuals captured palm. Our three sites situated Stability Altered Forest Ecosystems (SAFE) Project south-eastern (4.72°N, 117.60°E; Ewers 2011; Figure 1a). At site was forested embedded matrix, connected larger area (2,200 ha) consisting lowland dipterocarp rainforest. >1 ha protected 2011). palms surrounding approximately same age (~8 years). Selection focal (RR03, RR10 RR18) ensure approximate standard configuration structure landscape. Riparian composed remnant old-growth secondary forest. had tall trees (some >40 m), high canopy cover similar mean widths (48, 58 41 RR03, RR18 respectively; Williamson 2020). Minor variations inevitably precise elements (Figure 1b–d). Fieldwork took place November 2016 April 2017. Six beetle selected based previous 2016) span range body sizes (~1–5 cm), include representing two main burial modes (tunnellers rollers), diurnal nocturnal (Table 1). All six chosen they occur commonly observed much lower abundances abundance indicating effective candidates MRR study. males Catharsius renaudpauliani (Ochi Kon, 1996) dayacus (Lansberge, 1886) could distinguished reliably level field, pooled analysis. abundant this 2016), assume majority 17–18 live-capture baited pitfall traps set, spaced 50 apart, sampling 1b–d; Larsen Forsyth, 2005; placed seven eight four reserve. One trap removed from successfully established. Each consisted 1.5 L plastic bottle top inverted funnel (~92 mm diameter). Traps 25 g human faeces, wrapped muslin suspended cm above (Parrett Small holes base allow rainwater drain, handful leaves bottom provide shelter trapped beetles. A Styrofoam plate rain checked re-baited every second day period 14 days all marked. Two pen types mark (Mitsubishi Uni Paint Marker PX-21 Fine Bullet Tip Orange/Green/Pink, Artline 999XF Silver Metallic 0.8 mm). Pilot marks lasted least days, shown no effect longevity (Bates 2006). individual given unique code using series dots elytra Noriega Acosta, After marking, closed bait 24 hr disperse before replaced day. Both marked recaptured recorded released point capture. Similar previously assess statistical analysis conducted r (Version 4.0.3 - R Development Core Team, 2021). applied Ovaskainen al. (2019) analyse capture–recapture datasets over assumed (2004), is, spatially explicit diffusion supplemented mortality selection edges types. For s 1), denoted coefficient (unit m2/day; measuring rate) rate (unit/day). data sufficient estimate separately type, them constant entire area. parameter thus measured habitats relative model, habitat. instantaneous mimics researcher visiting particular attempting during short time period. present originated continuously 24-hr kept active, assumption would poor approximation. extended implementing alternative observation process, active distance centre (see Appendices S2 S3 details implementation). radius circular distinct trap. evaluated fit generating posterior predictive data, locations times first real spatio-temporal variance effort data. compared terms distribution last total moved. To composition movement, generated artificial RR18* 2a), represents modified version actual 1d). RR18*, entirely simulated release 100 site. repeated datasets. movements 2a). proportional caught (oil reserve) reserve). 8,646 beetles, 355 recaptured, giving overall recapture frequency 4.11% (Appendix S1: Table Onthophagus mulleri 1883) (6.19%) Proagoderus watanabei 2002) lowest (3.03%). Of recaptures, 29% involved released. individuals, 10% release, remainder after multiple (2 12 days). varied (Figures 3 4), obviously linked size mode small such Sisyphus thoracicus (Sharp, 1875) obscurior (Boucomont, 1914) far empirical 102.1 ± SE 5.27 m/day, maximum 220 m/day 2). along buffer, moved 350 (the largest design). Individuals furthest although pattern, varying 4). Based visual inspection trace plots S3), MCMC scheme satisfactory convergence, matched generally 5). predicted fewer very long distances (and hence intermediate distances) 5b; Appendix S4: 8). This indicates heterogeneity and/or rather specific, remain space time. results indicate variation preference difference habitats, rates, probability 6; S3: Four 6a), O. S. showing significant habitat, spp. P. sparsus slight 6a). Only watanabei, 6b). consistent species, corresponding average life c. (expected = 1/mortality rate), reflect becoming inactive leaving area, rubbing off elytra, 6c). except other 6d). 6f). simulation demonstrated 2b). non-release removal result line Similarly, median spp., higher non-released None displayed Habitat barriers dispersal insects 2020), especially 2016, 2017), suggesting bordering Merckx new information functionally insects, adjacent matrix. Low (5% Arellano 18% Colombia—Cultid-Medina 3% Brazil—da 2015) invertebrates (31.6% butterflies—Scriven 33.7% moths—Gray low (due inactive, because rubbed elytra) lead underestimates longer periods, being up later. normally less hr, our greater day: >100 4; S1). standards trapping daily <100 so considered apart independent (Larsen 2005), probably recommend 150–200 assure independence future studies. recommendations suggested South American 500 m, freely others demonstrate strong specificity areas. display constraint obvious decay; however, site-specific down buffer. functional groups see findings highlight significance set-aside growing evidence contiguous maintaining wide invertebrate recaptures largely within-habitat Individual differed specificity. (S. mulleri) medium specificity, respectively, strips, rarely crossed Rollers, thoracicus, particularly affected Previous changes community communities forest, 2014), weak spillover effects Preferential mammal increases availability Deere Alternatively, act example if perceive structural might increase exposure predators 2010), thermal tolerance filter microclimates (Birkett Roslin Edge temperatures drought 300 (Nunes 2021), affecting communities. diversity decrease increasing edge temperature around 80 (Williamson linear further investigation. large (40–60 m) structurally Smaller hotter drier Further investigating affect spatial temporal scales needed. Microclimate extreme, (Jucker 2018; studied, tolerances opportunity penetrate boundary fact, microclimate refugia many Interestingly, reserve, suggests does rely disturbed specialist, occurring numbers 2014). explain why utilise areas, night when cooler. Results simulations demonstrating movement. (riparian reduction emphasising avoiding degradation promoting restoration, important, vertebrate highlighted recent years (Ancrenaz Seaman adds supports Mitchell Mullin While provides do aid scales, scale inter-connecting enable us gene flow Such crucial establish effectiveness promote optimise land planning. supporting retention restoration matrices. legally required Sabah (20 1998) suggest wider 2020) research trade-offs versus yield needed; multi-faceted question vary context (Bicknell Standardising regulation regions contexts, schemes (RSPO), helping (Cole emphasises prioritise corridors. still vastly understudied critical informing planning (Hilty animals quality, length (Mullin determining factors interact ecological inform landscape-level resilient climate change. grateful Biodiversity Council Forestry Department providing permission carry out under access licence JKM/MBS.1000-2/2 JLD.3(159). thank East Rainforest Research Partnership (SEARRP), Project, LOMBOK assistants: Lizzie, Loly, Anis, Zul, Sabidi Noy. you Xin Rui artwork. R.E.J.G., E.M.S. O.T.L. supported Natural Environment Grant NE/K016261/1 Human Modified Forests programme. MOE AcRF Tier 1 (Grant No: RG119/19). L.F.R. PhD fellowship Helsinki University EDUFI. O.O. funded Academy Finland 309581), Jane Aatos Erkko Foundation, Norway Centres Excellence Funding Scheme (223257), European (ERC) Union's Horizon 2020 innovation programme Agreement 856506; ERC-synergy project LIFEPLAN). authors declare conflict interest. conceived idea secured funding collection; R.E.J.G. designed A.Y.C.C.; collected data; L.F.R., O.O., contributed designing analyses; apply case implemented led writing draft, editing final manuscript. shows collaboration countries, including researchers country carried out. joint senior women. engaged diverse perspectives outset. Data available via Zenodo https://doi.org/10.5281/zenodo.3475405 2019a), SAFE website (www.safeproject.net/datasets/view_dataset?id=3475406). Please note: publisher responsible content functionality any supplied authors. Any queries (other missing content) should directed author article.