Even small forest patches increase bee visits to flowers in an oil palm plantation landscape

Pollination sustains biodiversity and food security, but pollinators are threatened by habitat degradation, fragmentation, loss. We assessed how remaining forest influenced bee visits to flowers in an oil palm-dominated landscape Borneo. observed six plant species: four crops (Capsicum frutescens L. “chili”; Citrullus lanatus (Thunb.) Matsum & Nakai “watermelon”; Solanum lycopersicum “tomato”; melongena “eggplant”); one native Melastoma malabathricum “melastome”; the exotic Turnera subulata Smith “turnera”. made local grid-based landscape-scale transect-based study spanning 208 2130 m from forest, respectively. recorded 1249 4831 1046 ten-min observation periods. Visit frequency varied among species, ranging 0 S. a mean of 0.62 per flower 10 min C. lanatus. Bee visitation declined with distance both studies, expected decreasing 55% 66% at maximum each study. also tested whether nearest palm patch, 144 m, visitation, found no such associations. Expected was 70%–77% lower for plants close 200 ha fragment compared those near large continuous forests (>400 ha). Our results suggest that, although throughout landscape, bees depend on forests. Larger support more bees, though even 50 has positive contribution. Abstract Indonesian is available online material. Penyerbukan menopang keanekaragaman hayati dan ketahanan pangan, tetapi serangga penyerbuk terancam oleh degradasi, fragmentasi, hilangnya habitat. Kami menilai bagaimana hutan yang tersisa memengaruhi kunjungan lebah ke bunga di lanskap didominasi kelapa sawit Kalimantan. mengamati enam spesies tanaman: empat tanaman budidaya “cabai”; Matsum. “semangka”; “tomat”; “terong”); satu lokal “harendong”; eksotis, “bunga pukul delapan”. membuat studi berbasis kisi-kisi penampang lintang berskala bentang alam masing-masing berjarak 2.130 dari hutan. mencatat 1.249 4.831 dalam 1.046 kali pengamatan berperiode sepuluh menit. Kekerapan bervariasi antar tanaman, mulai menit diamati hingga rata-rata 0,62 menurun semakin jauh jarak kedua studi, dengan ekspektasi kekerapan sebesar 56% pada maksimal setiap studi. juga menguji apakah terdekat, kunjungan, tidak menemukan hubungan seperti itu. Ekspektasi 70–76% lebih rendah untuk dekat fragmen dibandingkan besar kurang terfragmen (> 400 Hasil kami menunjukkan bahwa, bergantung meskipun ditemukan sawit, Hutan mendukung banyak lebah, sebidang tanah seluas pun memberi sumbangan positif. Oil plantations (dominated plantings African Elaeis guineensis Jacq.) covered over 18.7 million hectares (Mha) worldwide 2017 continue increase (Meijaard et al., 2018). Much this expansion been Borneo where least 3.06 Mha species-rich old-growth were converted industrial between 2000 (Gaveau Such dramatic changes land cover have major ecological consequences, highlighting need research that contributes healthy functioning important widespread (Foster 2011). Seeking best ways maintain complexity incorporating other vegetation become issue plantation owners planners 2016; Yahya 2017). One motivation range different taxa, including wild pollinators, within landscape. Although mobile, they can be greatly affected fragmentation. The conversion cultivated potential cause loss feeding sites (sources pollen nectar) suitable nesting (Patrício-Roberto Campos, 2014). requirements complex (e.g., foraging habitats; Antoine Forrest, 2020; Westrich, 1996), which makes them particularly sensitive Most (an estimated 94%) tropical flowering animal-pollinated (Ollerton A decline thus impacts reproduction consequently entire ecosystems (Burkle 2013; IPBES, Potts 2010). Landscape changes, remnant fragmented areas, may genetic variability population stability, potentially leading disappearance populations 2014; Sodhi 2004) having severe effects pollinator services (Potts Many require animal pollination maximize fruit set, size, quality Even primarily pollinated weevils (Elaeidobius kamerunicus Faust) set yield (Caudwell, 2001; Zulkefli 2021). Wild particular contribute productivity viability many (Garibaldi 2013) security nutrition (Ellis 2015; Garibaldi 2016). Evidence exists watermelon (Sawe 2020), tomatoes (Cooley Vallejo-Marin, 2021; Neto 2013), chilies (Landaverde Land-use disrupt (see, e.g., Klein 2003; Ricketts 2008). disruptions add declines already seen worldwide, reflecting not just damaging land-use practices pesticide use) climate change While raised incomes living standards Indonesia (Qaim there concerns diet (Food Security Council Ickowitz possible reason difficulty producing highly nutritional, pollination-dependent landscapes insufficient services. recognized better understanding activity improved planning management their beneficial role. Here, we assess West Kalimantan, Indonesia. main aim determine number forest. hypothesize selected species will decrease further our test rates dependent planted patch. expect trees conditions similar provide resources than agricultural fields, fallows, open land. Therefore, when closer non-forested areas. To answer questions, related frequencies while controlling weather time day. knowledge, first This conducted June November concession PT Kayung Agro Lestari (KAL) Kabupaten Ketapang province Borneo, (1°26?00.0?S 110°13?00.0?E; Figure 1a). owned managed Austindo Nusantara Jaya (ANJ), member Roundtable Sustainable Palm (RSPO) (PT Tbk, Before plantation, logged-over natural (~8600 ha) degraded Conversion started 2010 2016, 12,061 out 17,998 had use terms “oil plantation” “plantation” referring area inside “planted palm” sections monoculture plantation. majority grows shallow peat (~63%), less mineral soil (~33%) sands (~4%). palms about 9 apart resulting mostly closed canopy. Intensive maintenance, regular physical clearance ground application herbicides, little understory palms. Sixteen forested areas (20–2333 ha), 21% (3884 concession, identified as High Conservation Value (HCV) regularly monitored company studied angiosperm (Table S1). Four these “chili,” “watermelon,” “tomato,” “eggplant”) common crops. two, introduced Smith, wide distribution All focal non-native except M. malabathricum, colonizing occurs cleared, edges T. method bio-control leaf-eating caterpillars (fire- bagworms), pests (Rashid two studies: (hereafter referred “Grid Study”) systematically planned crop extensively cleared several second transect “Transect Study”). Grid Study spanned up (Figure 2a), Transect edge locations chosen represent sizes distances S2). Forest 1 large, extends beyond boundary; 2 conserved 3 isolated secondary hill surrounded palm. relation 4, classified high conservation value area. 22 July 5 September 2017. In total, obtained 397 plants, growing individual poly bags. Due mortality nursery, individuals 134 lanatus, 108 melongena, 105 lycopersicum, frutescens. placed nursery-grown grid consisting twelve plots, three columns following gradient rows Each plot ~0.5 spacing plants. Plants assigned randomly ensure consisted representative sample all species. scattered naturally subulata, allowing assessment effect using 15 29 October transects (Transects A–F). Transects established 2b), adjacent it located along roadsides planted, starting continuing 300 into Along transect, points. closest designated point one. then walked 100 measured handheld GPS unit, bush point. process repeated points based presence bush. points, S3), due variation subulata. palm, additional 800 (with 824 m) Final satellite imagery, Google Earth. estimate frequencies, above-mentioned define visit making apparent contact stigma or anthers preselected flowers. period lasted min, same observer. During period, recorded, later focused analyzing only data. taxonomic challenges low numbers some morpho-species, analyzed group. Specimens collected photographed (Hessen, visual identification specimens carried John Ascher diagnostic characters documented Soh (2020). varied, depending aggregation (ensuring observer, camera, could adequately monitor simultaneously; range: 1–36, mean: 4.6). sequence transects, plots random. Observations 05:30 18:00 h (mean: 1000 h), most being during morning hours open. rain. used Brinno BCC200 Pro cameras perform observations. T1 Clamp tripod attach camera wooden pole would stand vertically 1b). adjusted focus manually setup. frame rate picture resolution 1280 × 720 pixels 1c). recordings longer keep observations comparable direct observations, treated every separate observation. count viewed videos computer Video Player. Camera studies took place days percentage 0% 60% S3). photographs did allow below order level. For day, temperature, relative humidity Suncare thermo-hydrometer (model 303C). subjectively categorized wind, wetness vegetation, sun exposure S4). data daily rainfall station logger (UA-002 HOBO) fixed record light intensity air temperature 3-hour intervals. meteorological (~50 km site) Pontianak (~188 site; Table Study, periods 28.8°C (23.8–34.0°C) 72.4% (50%–96%). 28.5°C (25–32.4°C) 79.4% (60%–94%). various factors might influence activity. variables five categories: (including humidity, precipitation, sunlight), temporal day year), environmental ID, size type), spatial palm), observation, recording ID method). See S4 details identified. analyses performed R (version 3.5.1 macOS version 10.14.6; Core Team, initial exploration Zuur al. (2010) analyze relationship explanatory variables, generated generalized linear mixed models (GLMMs) Poisson error (log link) distribution. Number response variable included offset models, Reitan Nielsen (2016). Observation model random account overdispersion (Harrison, Other considered include plot, ID. “glmer” function package “lme4” 1.1-15 (Bates 2015) “bobyqa” optimizer. Continuous centered scaled “scale” (R information-theoretic approach identify parsimonious Bayesian information criterion (BIC). Variance inflation (VIFs) “vif” “car” (Fox Weisberg, Dispersion, zero-inflation, uniformity "testDispersion," “testZeroInflation,” “testUniformity” functions “DHARMa” 0.3.2.0 (Hartig, 2020). Confidence intervals calculated Wald “confint.merMod” 2015). Pseudo R2 values (delta method) “r.squaredGLMM” “MuMIn” 1.42.1 (Barto?, Figures 4 created “ggplot” “ggplot2” 3.3.2 (Wickham, Effect predictors “allEffects” “effects” 4.1.0 (Fox, 2003, 2019). field revealed diversity flower-visiting insects frequently visitor 81.4% total S2a,b). 2020) show site genera: Apis sp., Ceratina Geniotrigona Heterotrigona Homotrigona Lasioglossum spp., Lipotriches Nomia Xylocopa spp.. counted 355 2071 723 10-min 32 days. combined 0.23 (0.19 0.31 [Table 1]). 894 2760 323 overall 0.34 (0.52 0.16 As (n = 56 periods), excluded analyses. separately. (Model 1) explaining (R2m 0.520, R2c 0.947) sun, sampling (camera manual observation) 2). contribution explained listed S5a. predictor (based Model held constant S6a. VIF <2. There significant greater 3a), 55.4% detect any frequency. highest frequency, followed 3b,c). positively associated sunlight. Time showed association Temperature correlated thus, unable disentangle variables. play role, model. anticipate visits. However, higher human developed describe Study. 2.1) 483 (and >800 forest; R2m 0.270, 0.445), 2.2) 2-level factor discriminating situated <450 0.259, 0.451). S5b,c. 2.1 S6b,c. 3). 4a). decreased 52.6% 438 m. showing larger (Forests 2) smaller (Forest 3; 4b). 70.3%–76.5% Forests 2, respectively 4c). Visitation negative unexpectedly significantly frequencies. 2.2 (as factor: forest), effects. forests, 66.2% 800m 450 4d). patterns 4). Once accounted for, despite difference scale. indicate Declines habitats systems elsewhere, example, coffee Costa Rica (Ricketts, 2004); mustard radish grasslands Germany (Steffan-Dewenter Tscharntke, 1999); oak woodland chaparral farms California (Kremen 2002). suggests act source forage reside habitats. lack does Flower proximity (the smallest fragment) 70.3% 76.5% forests). Studies elsewhere indicated abundant primary small fragments (Beck 2002; Liow Lucey Hill, 2011; Mayfield, 2005). Despite this, 54 patch boosts visitations surrounding Similar findings affecting landscapes, example 0.24 supporting diverse assemblage (Brosi conserving intact remains crucial (Edwards 2011), damaged should protected practical help taxa (Benedick 2006). Automatic observe (for see [Steen, 2017]). simple, low-cost automatic here valuable increasing unexpected fewer camera. likely height-biased selection (due bushes taller setup), placement, outside peak time. These setup selection. differing levels attraction pollinators. Strikingly, – genus known its pollen-only, nectar-free rely specialized “buzz-pollination” (Vallejo-Marín, 2019) received originates Americas appear effective (Franceschinelli Rosi-Denadai addition, another neotropical Solanaceae, self-pollination (though [Knapp, 2010]) lowest aside lycopersicum. contrast, (originating Africa), mortality, partly result harsh emerged (high temperatures droughts). We, therefore, make conclusions adequacy, since translate events. compare speculate adequacy location. It previously shown 6–8 honey-bee visits, bumble visit, achieve optimum (Adlerz, 1966; Bomfim translates 0.11 open, (0.62) (mean <50 forest: 0.83 flower, >200 0.36 flower). sufficient clearly receives yields (zero study), linked stress dominated requires investigation. mentioned above, dry hot requiring watering twice avoid wilting death. Temperatures reached 34°C average before (max: 17), 930 mm rain 127 mm/day, 64 92 0–4 mm). reflect heat vapor pressure deficit neighboring areas) versus (Hardwick Luskin Potts, Ramdani 2014) increased reduced across island (McAlpine climate-driven small-scale agriculture harder riskier. level spread, declines, global gets warmer predictable ultimately impact holder themselves Pollinator density richness improve pollinator-dependent ecosystems, flower-visitor globally intensification causing functional (Rader Southeast Asian (Liow 2001), butterflies (Koh Sodhi, 2004), moths 2002) experiencing part disturbance (Sodhi 2004). Protecting patches provides variety well wildlife insects. Thus, affect abundance plant–animal interactions (Horner-Devine 2008) landscapes. limitation concern, especially adequate grown sufficiently (Klein 2003). study, fragment, emphasize importance maintaining much around sustain availability. encourage Funding provided USAID’s Forestry Biodiversity Office via CIFOR (Amy Terry Sunderland) NMBU’s TEMNR Field grant. thank Erik Meijaard (Borneo Futures), Yves Laumonier (CIFOR CIRAD), Gusti Hardiansyah, Farah Diba (Tanjungpura University) collaborations thankful everyone Agri (ANJ) who supported stay KAL. KAL Ero Prasetyo, assistance; fire team plants; Johannes Terang providing Nardiyono, Hendriyana Rahman, Viktor assistance setting running studies. Ronny Steen guidance equipment; Miriam van Heist, Lindawaty (Lahuka Services), Dilla Trianti (Indonesian embassy Oslo), Dwi Yoga administrative logistical Nia Atmadja (CIFOR) translating abstract. CP thanks Knut Hessen Lynn Jørgensen scientific support; Lisa Fagerli, Ingvild Asmervik, Ross Whetherbee statistical advice; Luc Le Grand Jonas Anderson constructive comments text; (National University Singapore) identifications. Open Access funding enabled organized Projekt DEAL. project formal cooperation Agri. hosted facilitated work basis freedom publish without interference. agreed advance could, review drafted articles, request us withhold specific judged commercial reasons. completed, requests made. DS conceived CIFOR, Agri, Tanjungpura Futures. AN designed, managed, supervised project. DS, AN, designed described here. wrote original manuscript supervision critical inputs. coauthors reviewed edited final approve version. openly Dryad Digital Repository: https://doi.org/10.5061/dryad.s4mw6m96h (Power Please note: publisher responsible content functionality supplied authors. Any queries (other missing content) directed corresponding author article.